ML20234B594
| ML20234B594 | |
| Person / Time | |
|---|---|
| Site: | Peach Bottom  |
| Issue date: | 04/30/1987 |
| From: | Amos C, Benjamin A, Boyd G, Griesmeyer J, Haskin F, Helton J, Kunsman D, Lewis S, Laura Smith SANDIA NATIONAL LABORATORIES |
| To: | NRC OFFICE OF NUCLEAR REGULATORY RESEARCH (RES) |
| References | |
| CON-FIN-A-1322 NUREG-CR-4551, NUREG-CR-4551-V3-PT1, SAND86-1309, NUDOCS 8707060116 | |
| Download: ML20234B594 (244) | |
Text
{{#Wiki_filter:1 i .. -4 ) r 1 NUREG/CR-4551 DRAFT FOR COMMENT l SAND 86- 1309 Volume 3 (1 of 2) Printed April 1987
'1
)
Evaluation of Severe Accident Risks and the Potential for Risk Reduction: s Peach Bottom, Unit 2 Main Report C. N. Amos, A. S. Benjamin, G. J. Boyd, J. M. Griesmeyer, F. E. Haskin, l J. C. Helton, D. M. Kunsman, S. R. Lewis, L. N. Smith l and Na anal Laboratones Albuquerque, New Mexico 87185 and Lwerrnore, California 94550 for the United States Department of Energy under Contract DE AC04 76DP00789 8707060!.16 870430 PDR ADUCK 05000277 P PDR Prepared for U. S. NUCLEAR REGULATORY COMMISSION i i 1 , 3 1
'1 u_.
s' . . , f' l 1
'l i
l NOTICE ' This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, or any of their employ-ees, makes any warranty, expressed or implied, or assumes any legal liability or responsibility for any third party's use, or the results of such use, of any information, apparatus product or process disclosed in this report, or represents that its use by such i third party would not infringe privately owned rights. 4 Available from Superintendent of Documents U.S. Government Printing Office Post Office Box 37082 j i Washington, D.C. 20013-7082 1 and National Technical Information Service Springfield, VA 22161 j i
- . 's NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) {
l NUREG/CR-4551 SAND 86-1309 .- Volume 3 l EVALUATION OF SEVERE ACCIDENT RISKS AND ! THE POTENTIAL FOR RISK REDUCTION: PEACH BOTTOM, UNIT 2 C. N. Amos* A. S. Benjamin G. J. Boyd" J. M. Griesmeyer j F. E. Haskin J. C. Helton*" D. M. Kunsman S. R. Lewis ** i L. N. Smith"** Printed February 1987 Sandia National Laboratories Albuquerque,NM 87185 l Operated by i Sandia Corporation for the U. S. Department of Energy Prepared for . Division of Reactor System Safety Office of Nuclear Regulatory Research U. S. Nuclear Regulatory Commission Washington,DC 20555 Under Memorandum of Understanding DOE 40-550-75 NRC FIN No. A1322
*Technadyne Engineering Consultants,Inc., Albuquerque, NM
" Safety and Reliability Optimization Services, Inc., Knoxville, TN
"* Arizona State University, Tempe, AZ
"" Science Applications International Corporation, Albuquerque, NM -
PREFACE Because of the time constraints imposed on this work to meet the Nuclear Regulatory Commission's schedule for publication of NUREG-1150, this draft report has not yet received the full level of peer and management review customarily accorded to reports issued by Sandia National Laboratories. The reviews ; i will be completed and corrections made, if necessary, prior to final publication. This report contains the main body of the report only. The appendices will be published separately. l l 1 l 1 l 1 ii
I F RE0 VEST FOR COMMENT This report, NUREG/CR-4551, " Evaluation of Severe Accident Risks and Potential for Risk Reduction," was prepared for the U.S. Nuclear ) Regulatory Commission by the Sandia National Laboratories and its j subcontractors. The methods and results set forth in this report are being used by the NRC to support the development of the Reactor Risk Reference Document (NUREG-1150) and will be used in areas of broad public interest such as probabilistic risk analyses, emergency response planning, siting, NRC safety goal applications, and cost / risk / benefit analyses--indeed, wherever risks to public health need to be considered in regulatory applications. Thus, it is considered imperative that an opportunity for public comment on the results as presented in the report be provided. Comments should be sent to the U.S. Nuclear Regulatory Commission, Washington, D.C. 20555, Attention: Joseph Murphy, Division of Reactor System Safety. These comments will be most useful to the staff if they are received by June 1, 1987. Of particular interest to us is the receipt of comments on the methodology, and results, related to uncertainty analysis. One criticism voiced with respect to the Reactor Safety Study was its lack of an uncertainty analysis. We have included an uncertainty analysis, but we are sure that its nature will be the sub,iect of lively debate. We welcome this, and solicit constructive advice and criticism. The NRC hereby expresses its great appreciation to all participants in 1 this study for their considerable efforts, as well as to all who will take the time and effort to provide it with comments on this report. D. F. Ross, Deputy Director i Office of Nuclear Regulatory Research l
)
i iii L_________--_--__
NUREGCR-4551, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) ACKNOWLEDGMENTS A number of individuals have contributed to this program. In particular, the authors would like to thank the members of the review groups listed in Section 2 who provided valuable comments on the work as it progressed and who participated directly in the collection of input for the uncertainty analysis. The following Sandia personnel also provided analysis and input for use in this report: R. L. Iman, M. J. Shortencarier, and D. C. Williams. ABSTRACT The Severe Accident Risk Reduction Program (SARRP) has completed a rebaselining of the risks to the public from a boiling water reactor with a Mark I containment (Peach Bottom, Unit 2). Emphasis was placed on determining the magnitude and character of the uncertainties, rather than focusing on a point estimate. The risk-reduction potential of a set of proposed safety option backfits was also studied, and their costs and benefits were also evaluated. It was found - that the risks from intemal events are generally low relative to previous studies; for example, most of the uncertainty range is lower than the point estimate of risk for the Peach Bottom plant in the Reactor Safety Study (RSS). However, certain unresolved issues cause the top of the uncertainty band to appear at a level that is comparable with the RSS point estimate. These issues include the modeling of the common-mode failures for the de power system, the likelihood of offsite power recovery versus time during a station blackout, the probability of drywell failure resulting from meltthrough of the drywell shell, the magnitude of the fission product releases during core-concrete interactions, and the decontamination effectiveness of the reactor enclosure building. Most of the l postulated safety options do not appear to be cost effective, although some based on changes to procedures or inexpensive hardware additions may be marginally i cost effective. This draft for comment of the SARRP report for Peach Bottom does not include detailed technical appendices, which are still in preparation. The appendices will be issued under separate cover whcc completed. This work supports the Nuclear Regulatory Commission's assessment of severe accidents in NUREG-1150. iv ,
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY.1987) CONTENTS , Section Eagt
- 1. INTRODUCTI ON. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 - 1 l
1.1 B ack ground and Objectives. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 - 1 1.2 Scope of Analys is. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 - 2 1 1.3 Overview of the Peach Bottom Plant. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 -6 1.4 Organization of the Report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 -7 References for Section 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 - 10
- 2. METHODOLOGY FOR REB ASELINING OF RISK. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1 2.1 Overview of Risk Integration and Review Activities................................. 2-1 4 1
2.1.1 Integration of Pmject Activities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 1 j 2.1.2 Review and Quahty Assurance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 9 i i 2.2 Development of the Central Estimate of Risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 1 1 2.3 Charac terization of Uncertainties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 12 1 2.3.1 Overview of Uncertainty Treatment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 14 2.3.2 Description of the LLH Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 17 2.4 Reassessment of Dominant Core-Damage Sequences.............................. 2-29 2.5 Eval uation of Containment Response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-31 2.5.1 Development and Quantification of the Containment Event Tree......... 2-31 2.5.2 Plant Features Important to Containment Response........................ 2-38 2.5.3 Defini tion of Containment-Release Modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-40 2.6 Assessment of the Radiological Source Term. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-41 2.6.1 Integration With Other Risk-Assessment Tasks............................ 2-42 ! 2.6.2 Source Term Code Pxkage Calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-45
- 2. 6. 3 Overview of RELTRAC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-47 2.6.4 Development of Source Terms for the Central Estimate................... 2-50 2.6.5 Derivation of Source Terms for the LLH Uncertainty Analysis......... 2 53 2.7 Offsite Consequence Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-5 4 References for Section 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
- 3. METHODOLOGY FOR EVALUATION OF RISK-REDUCTION OPTIONS......... 3-1 3.1 Identification of Safety Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3- 1 3.1.1 Generic Risk-Reduction Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-2 3.1.2 Risk-Reduction Measures for Peach Bottom................................. 3-2 3.2 Evaluation of Costs and Other Impacts. . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . .. .. . .. . . 3-9 3.3 Evaluation of Effects on Risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3- 1 3.4 Value/ Impact Assessment. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3- 1 1 3.5 Treatment of Uncertainties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . .
References for Section 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) CONTENTS (CONTINUED) Section Eage
- 4. EVALUATION OF KEY UNCERTAINTY ISSUES. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 1 4.1 Sequence Frequency Issues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-2 4.1.1 Failure to Actuate the Standby Liquid Control System...................... 4-2 4.1.2 Frequency of Dc Power System Common-Mode Failure................... 4-2 4.1.3 Probability of Failure to Vent During an ATWS............................. 4-3 4.1. 4 Power Recovery Uncertainties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-3 l l
4.2 Containment I.oading and Performance Issues. . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . 4-4 4.2.1 Probability of Stuck-Open Vacuum B reaker. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-4 4.2.2 Use of the High Pressure Service Water System Spray as a Recovery... 4-6 4.2.3 Probability that the Operations Staffis Unable to Vent During Station B lackout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-6 4.2.4 Probability that the Operations Staffis Unable to Vent After Ac Power , Reco very . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-7 4.2.5 Level of Suppression Pool Bypass Through a Stuck Open Safety / Relief Valve Vacuum B maker . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-7
- 4. 2. 6 Containment Failum Pmssure and Location. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 8 i 4.2. 7 Containment Failure Size. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . 4- 1 1 4.2.8 Vessel Failure Mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 12 4.2.9 Containment Pressure Prior to Vessel Breach for Station Blackout Sc en ario s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 14 4.2.10 Containment Pressum Rise at Vessel B reach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 16 4.2.1 1 Probability of Drywell S hell Meltthrough. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 18 4.2.12 Probability of Hydrogen Burns in Reactor Building Sufficient to Ca u se Byp as s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 19 4.3 Radiological Source Term Is sues. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-21 4.3.1 In-Ves sel Release from the Fuel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-21 I 4.3.2 Amount of Csl Decomposition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-23 4.3.3 Retention in the Reactor Pressure Vessel.. . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-23 4.3.4 Suppression Pool Decontamination Factor for Aerosols................... 4-25 4.3.5 Suppression Pool Scrubbing of Volatile Iodine Species........... ........ 4-27 4.3.6 Revolatilization Following Vessel Breach (FRVOL)....................... 4-27 4.3.7 Release from the Melt During Core-Concrete Interactions................. 4-29 l 4.3.8 Retention in the Reactor Building and the Refuelin 4.3.9 Late Releases of Iodhe. . . . . . . . . . . . . . . . ........................433 . . . . . . . . . . . . . . . .g B ay. . . . . . . .!
j References for Section 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-34
]
- 5. RESULTS OF RISK REB ASELINING. . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5- 1 l 5.1 Core-Damage Fmquency Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5- 1 5.1.1 Accident Frequencies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5- 1 5.1. 2 Uncertam ty Representation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 -3 5.1.3 Observations Concerning the Core-Damage Frequency..................... 5-8 5.1.4 Comparison to the Reactor Safety Study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-9 vi {
1
1
)
NUREG/CR-4551. VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) i 1 1 i CONTENTS (CONTINUED) j Section P.agc j i 5.2 Containment Analysis Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5- 10 i s 1 5.2.1 Central Estimate of Containment Response. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5- 10 5.2. 2 LLH Containment Analysis Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3- 14 5.2. 3 Comparison to Other S tudies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5- 17 5.3 Results of the Radiological Source Term Analysis ................................. 5-20 5.3.1 Source-Term Results for the Central Estimate.............................. 5-20 5.3.2 LLH Source Term Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-20 5.4 Offsite Consequence Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-29 5.5 Risk Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 -3 7 ; 4 l 5.5.1 Results of Risk for Latent Cancer Fatality and Early Fatality............. 5-37 I 5.5.2 Results for Other Risk Measures. . . .. . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 5-61 5.5.3 Observations Concerning the Risk Results................................. 3-64 5.6 Limitation s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 - 6 8 References for Section 5. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-70
- 6. RESULTS OF RIS K REDUCTION ANALYSIS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6- 1 I 6.1 Effects of Pn:vention Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-2 l 6.2 Effects of Mitigation Options on Containment Response............................ 6-3 ;
6.3 Effects of Safety Options on Risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-5
)1 6.4 Costs of Risk-Reduction Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6- 8 6.5 Comparison of Costs and B enefits. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-9 ,
1
- 7. INSIGHTS AND CONCLUSIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7- 1 i 7.1 Insights and Conclusions from the Rebaselining of Risk............................ 7-1 7.2 Risk-Reduction Insights and Conclusions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7-4 7.3 Limitation s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7' I
i
3
'l NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) i I
j
~
CONTENTS (CONTINUED) Section Eagn APPENDICES A. CONTAINMENT ANALYSIS UNCERTAINTY ISSUES FOR THE LIMITED '
. LATIN HYPERCUB E STUDY OF GRAND GULF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
B. SUPPORTING ANALYSIS FOR THE GRAND GULF SOURCE TERM UNCERTAINTY ASSESSMENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . C. DETATT FD LISTING S OF RISK RESULTS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . D. AS SESSMENT OF RISK REDUCTION MEASURES. . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . E.
SUMMARY
OF REVIEW TEAM COMMENTS CONCERNING THE LIMITED LATIN HYPERCUBE ANALYSIS AS IMPLEMENTED IN SARRP ....................... F. DATA BASE FOR ESTIMATION OF COST AND PERSONNEL DOSE FOR PROPOSED MODIFICATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . J viii
l l NUREG/CR-4531. VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) LIST OF TABLES Table P. age 2-1 Participants in S ARRP Review Tasks for Peach Bottom......... . .................. ... 2-10 2-2 Uncertainty Characteristics of Different Analysis Activities............................ 2-15 2-3 S ummary of CET Events for Peach Bottom. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-32 2-4 Use ofInformation Sources Addressing the Issues in the Containment Event Tree. 2-37 l 2-5 Derivation of the Central Estimate Source Tenn. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-51 I 2-6 Derivation of LLH Source Terms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2--55 3-1 BWR Preventive Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-3 3-2 BWR Mitigative Options. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-4 4-1 Size of Suppression Pool Bypass for a Stuck-Open SRV Tailpipe Vacuum Breaker. 4-8 5-1 Core-Damage Results for Peach Bottom. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-2 5-2 Sensitivity Studies fr.r the Peach Bottom Com-Damage Analysis........... ........... 5-4 5-2 Definition of the Ten Dimension Accident Progression Bins........................... 5-11 5-3 Central Estimate Containment Event Tree R esults. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5- 12 i 1 5-4 Comparison of Containment Failum Modes for Three Studies........................ 5-19 l 5-5 Peach Bottom Central-Estimate Release Fractions for MACCS Consequence . Calculations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 -21 ! 5-6 Peach Bottom Central-Estimate Release Fractions for CRAC2 Conse Calc ulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. .. .. ... .. .. ... .. .. . .. .. ... .5-23 . . . . . . . . . . . .q 5-7 Peach Bottom Cluster Release Fractions for MACCS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-24 5-8 Peach Bottom Cluster Release Fractions for CRAC2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 27 i 1 1 5-9 MACCS Estimates of Mean Consequences for Central Estimate Bins................ 5-31 5-10 CRAC2 Estimates of Mean Consequences for Central Estimate Bins................. 5-32 5-11 MACCS Estimates of Mean Consequences for LLH Clusters......................... 5-33 5-12 CRAC2 Estimates of Mean Consequences for LLH Clusters.......................... 5-35 5-13 Comparison of LLH and Central Risk Estimates by Fractional Contribution of Plant-Damage S tate s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-42 5-14 Comparison of LLH and Central Risk Estimates by Fractional Contribution of Accident Progression B ins: MACCS Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-43 5-15 Defm' ition of the Ten Dimension B ins. . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-44 5-16 Fractional Contribution of LLH Clusters to Risk. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-46 5-17 Results of Chi-Squared Test (MACCS Consequences)................................. 5-57 6-1 Estimated Present Value Total Cost for Risk-Reduction Measures...................... 6-9 ix l __ _ _ _ _ _ _ _ - _ _ _ - _ _ _ _ _ _ _ - _ _ _ _ _ _ _ _ _ - - _ _ _ _ - _ _ _ . _-
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) LIST OF FIGURES Eigurn Eage 1-1 Relationships of Research Programs to SARRP Tasks. . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . 1-4 I l-2 Simplified Cross-Sectional View of the Peach Bottom Containment....................1-8 j 2-1 Overview of SARRP Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2 2-2 Overview of Modularized PRA Methodology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-4 2-3 Overview of the LLH Proces s. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2- 18 2-4 Example of LLH Output for Risk Measums. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-22 2-5 Example of LLH Rank Regression Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-23 2-6 Example of CCDF Risk Display. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2-2 8 3-1 Example of LLH Value/ Impact Display. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3- 15 4-1 Weighting Factors for Containment Capacity for Low Temperature Conditions...... 4-9 42 Range of Reviewer Input for Location of Failum for 138 psig Capacity.............. 4-10 4-3 Weighting Factors for Containment Capacity for High Temperature Conditions.... 4-10 4-4 Range of Reviewer Input for the Probability of Leakage Failure Mode............... 4-12 4-5 Range of Reviewer Input for Probability of Slump-Type Meltthrough................4-14 46 Range of Reviewer Input for Contamment Pmssum Prior to Vessel Breach for S tation B lackout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4- 15 4-7 Range of Reviewer Input for Pressure Rise at Vessel Breach.......................... 4-17 4-8 Range of Reviewer Input for Probability of Drywell Meltthrough..................... 4-19 4-9 Range of Reviewer Input for Probability of Reactor Building Bypass................ 4-20 4-10 Range of Weighting Factors for In-Vessel Release From the Fuel.................... 4-22 4-11 Range of Weightin g Factors for Csl Decomposition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4-24 4-12 Outcomes and Weighting Factors for In-Vessel Retention.............................. 4-25 4-13 Range of Weighting Factors for Scrubbing of Aerosols................................ 4-26 4-14 Range of Weighting Factors for Suppression Pool Scrubbing ofIodine..............4-28 4-15 Range of Weighting Factors for Primary System Revolatilization .................... 4-30 4-16 Range of Weighting Factors for Core-Concrete Releases............................... 4-30 4-17 Range of Weighting Factors for Reactor Building Decontamination Factor.......... 4-32 4 18 Range of Weighting Factors for Refueling Bay Decontamination Factor............. 4-32 4-19 Range of Weighting Factors for Late Iodine Release Levels............................4-33 1 5-1 " Box and Whisker" Display of Uncertainties for Total Core-Damage Frequency..... 5-5 5-2 " Box and Whisker" Plots for Important Peach Bottom Damage States................. 5-6 5-3 Conditional Probability of Containment Response: All Sequences Weighted ] B y The ir Frequ en cies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 - 15 i X
1 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) LIST OF FIGURES (CONTINUED) Figure Eage 5-4 Rank Regression for the Pmbability of Early Containment Failure.................... 5-16 5-5 Conditional Probability of Containment Response: Dr Does Not Occur . . . . . . . . . . . . . . . . . . . . . . . .................................5-18 . . . . . . . . . . . . . . . . . . . . .ywell Meltthr 5-6 Rank Regmssion for the Combined Probability of Early Containment Failure and Vent When Drywell Meltthmugh is Pmcluded . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 5-7 Risk of Latent Cancer Fatalities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3 8 5-8 Risk of Early Fatalities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-3 9 5-9 CCDF for Risk of Latent Cancer Fatalities (MACCS).................................. 5-48 5-10 CCDF for Risk of Early Fatalities (MACCS). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 -48 5-11 Multivariate Rank-Regression Results for Risk of Latent Cancer Fatalities i (MACCS Consequence Calculations) . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . 5-50 , i 5-12 Single Variable Rank Regression Analysis for Risk of Latent Cancer Fatalities 1 (MACCS Consequence Calculations, Conditional on Com Damage) ................ 5-52 j 5-13 Multivariate Rank-Regression Results for Early Fatality Risk 4 (MACCS Consequence Calculations) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-5 3
)
5-14 Single Variable Rank Regmssion Analysis for Early Fatality Risk I (MACCS Consequence Calculations, Conditional on Core Damage) ................ 5-54 5-15 Rank-Regmssion Results for Risk of Latent Cancer By Level......................... 5-55 5-16 Annual Risk of Early Inj ury. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 5-62 5-17 Individual Risk of Fatality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-62 5-18 Multivariate Rank-Regression Analysis for Individual Risk of Fatality............... 5-63 5-19 Results for Risk Measure of Population Dose Per Year................................. 5-63 l 1 5-18 Rank-Regres sion Analy sis for Population Dose. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-64 ; 5-17 Results for Risk in Terms of Offsite Costs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5-65 , 6-1 Illustration of the Impact of Prevention Options on Corte-Damage Frequency........ 6-3 ; 6-2 Comparison of Safety Option Risk of Latent Cancer Fatalities: MACCS Calc ulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i 6-3 Comparison of Safety Option Risk of Latent Cancer Fatalities: CRAC2 Calc ulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6-6-4 Comparison of Safety Option Risk of Early Fatalities: MACCS Calculation........... 6-7 6-5 Comparison of Safety Option Risk of Early Fatalities: Crac2 calculation............... 6-7 6-6 Cost / Benefit Comparison for the Safety Options as Calculated by MACCS.......... 6-10 6-7 Cost / Benefit Comparison for the Safety Options as Calculated by CRAC2.......... 6-11 6-8 Cost / Benefit Comparison for the Mitigative Options, Linear Scale.................... 6-13 xi
NUREG/CR-4551,VOL3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l LISTOF ACRONYMS AND ABBREVIATIONS ADS Automatic depressurization system AE Architect-Engineer : ASEP- Accident Sequence Evaluation Program ATWS Anticipated transient without scram ) BCL Battelle Columbus Laboratories i BNL Brookhaven NationalLaboratory BWR' Boiling water reactor CCDF ' Complementary cumulative distribution function CCI Core-concrete interactions I CDB Coolable debris bed I CET - Containment-event tree CF Containment failure CLWG Containment Loads Working Group CPWG Containment Performance Working Group CRD Contml-rod drive DF Decontamination factor ECCS Emergency-core cooling system ESF Engineered safety feature HIS Hydrogen ignition system HPCS High pressure core spray IDCOR Industry Degraded Core Pmgrant LLH Limited Latin hypercube (uncertainty analysis method) LOCA Loss-of-coolant accident NRC U. S. Nuclear Regulatory Commission NSSS Nuclear steam supply system OCP Optimistic / central / pessimistic (method for uncertainty analysis) PRA Probabilistic risk assessment PRUEP Probabilistic Risk Uncertainty Evaluation Program ) PWR Pressurized water reactor 3 QC Quality control (review group) QUEST Quantitative Uncertainty Estimation of the Source Term Program RAP Regulatory Analysis Program RCIC Reactor core isolation cooling system RPV Reactor pressure vessel RSS Reactor Safety Study , RSSMAP Reactor Safety Study Methodology Applications Program ] SARRP Severe Accident Risk Reduction Program l SASA Severe Accident Sequence Analysis Program l SAUNA Severe Accident Uncertainty Analysis Program j SCG Senior consultant group SLC Standbyliquid control system STCP Source-Term Code Package ! USI Unresolved safety issue I xii
1 I NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBkUARY,1987) I 1 Section 1 INTRODUCTION i This report summarizes the efforts and results of the Severe Accident Risk Reduction Program (SARRP), which was conducted by Sandia National Laboratories with the support of several other contractors to the U. S. Nuclear Regulatory Commission (NRC). The goal of 4 SARRP was to put into a risk perspective the insights that have been generated as a result of recent research into systems behavior and physical phenomena under severe accident conditions in nuclear power plants. In pursuing this goal, these insights have been applied to obtaining a new estimate of risk for several operating plants and to evaluating the potential ) benefits and costs of measures intended to enhance safety. This report will serve as a principal input to the development of the report NUREG-1150, which will summarize the technical position of the NRC on the significance of severe accidents at nuclear power plants. ; This risk perspective was developed through the analysis of reference plants representing a spectrum of different nuclear steam supply systems and containment designs. This report describes the analysis and results for one of the plants, the Peach Bottom Atomic Power Station, Unit 2, which is owned and operated by the Philadelphia Electric Company. Peach Bottom is a boiling water reactor (BWR) using a General Electric-supplief auclear steam supply system from the BWR/4 product line, with a Mark I containment. The reactor is housed in a drywell/ pressure-suppression containment that uses a pool of water to condense steam and thereby mitigate the pressures that might be generated under accident conditions. The Peach Bottom plant was previously the subject of a probabilistic risk assessment (PRA)in the Reactor Safety Study (RSS) [1], and was also one of the reference plants in the Industry Degraded Core (IDCOR) program [2]. l l
1.1 BACKGROUND
AND OBJECTIVES A number of research activities are currently engaged in extensive efforts focused on better characterizing the hazards associated with the operation of nuclear power plants. These efforts typically focus on one or more of the following areas of reactor safety: The likelihood and nature of accidents severe enough to liberate substantial quantities of fission products from the reactor fuel; The challenges such accidents would present to the integrity of the reactor containment; The quantities and physical properties of fission products that could be released to the environment: and 1-1
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987)
- The ultimate consequences of these accidents, both in terms of their effects on public health, and their associated costs to society and to the plant owners.
The first attempt to incorporate the state-of the-art in all of these areas into an integrated perspective on the risk associated with the operation of nuclear power plants culminated in the Reactor Safety Study, which was published in 1975 [1]. Although that study produced the first comprehensive assessment of plant risk, it also focused attention on a number of gaps in the understanding of accident processes. Since that time, many risk assessments of specific plants have been performed. In general, each of these has progressively reflected at least some of the advances that have been made in the ability to predict accident behavior. In order to investigate the significance of more recent developments in a comprehensive fashion, it was concluded that the current efforts of research programs being sponsored by the NRC should be coalesced to produce an updated representation of risk; this led to the formulation of the cunent SARRP objectives. The results of the specific research programs, including their significance with respect to various questions concerning reactor safety, are, of course, documented individually. It is the objective of SARRP to integrate these results and to cast them in the context of their implications with respect to risk. This includes not only an estimate of risk as measured by a number of health effects and financial consequences, but also a thorough understanding of the important sensitivities and uncertainties associated with the state-of-the-art in reactor safety assessment. Many of the critical reviews of former risk studies, particularly the RSS, concluded that the uncertainties in the results had not been fully examined or displayed. A major focus of the SARRP program has been the characterization of these uncertainties, and this emphasis is one of the features that distinguishes this study from many other risk assessment programs. This risk context also provides a mechanism to investigate the degree to which safety might be improved as a result of various modifications to existing plants. Comparisons of the risk reduction that could be achieved and the associated costs incurred as a consequence of such modifications can provide additional information useful to the NRC's decision-making q process. 1.2 SCOPE OF TILE ANALYSIS The technical program embodied in SARRP was implemented through five major j activities: 1-2
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) (1) The formulation of a means to provide a comprehensive delineation of the possible types of containment msponse to severe core-damage accidents and the estimation of their relative frequencies; (2) Development of fission product source terms for the spectrum of accident sequences and resultant containment response end states; (3) Calculation of the offsite consequences of the accidents considered; (4) Integrated calculation of risk using these accident sequences, containment and source term analyses, and offsite consequence calculations; and (5) Investigation of the benefits and costs of measums intended to reduce risk. One of the major inputs to SARRP was the m-evaluation of the frequency of potential core-damage accidents which was accomplished under the Accident Sequence Evaluation Program (ASEP). The details of that reassessment are reported separately [3.] and are only summarized in this report to allow a perspective on overall risk. There was interaction among the SARRP and ASEP analysts to ensure that the accident sequence analysis would integrate properly with the containment and consequence analysis. The relationships of the SARRP tasks with each other and with other programs is illustrated in Figure 1-1. Most of the tasks in the program involved the generation of a central estimate and the evaluation of key sensitivities and uncertainties. The first SARRP activity led to the development of a containment event tree specific to the evaluation of the Peach Bottom containment. The details of the containment analysis are provided in a separate document [4], although the principal inputs and results are repeated in this report. The containment event tree is similar in nature to those used in recent plant-specific risk assessments, but it is generally much more detailed. It provides a mechanism for tracing the accident progression for the different types of core-damtge events, including various phenomena that could affect the likelihood and severity of containment failure, and the performance of the containment safety , features. For the Peach Bottom analysis, the containment event tree was also used to consider actions or events that could arrest the accident after core damage had been initiated. As illustrated in Figure 1-1, this task involved original analysis as well as integration of the results of many other activities. These sources provided information relating to the nature of severe-accident phenomena, the severity of various challenges to containment integrity, and the capacity of the containment to withstand these challenges. The level of detail in the containment event tree was dictated by the desire to dieplay the possible outcomes for a variety of core-damage accidents (recognizing that uncertainties regarding some severe-accident l phenomena prevent precise prediction of the containment response), and the need to provide 1-3 l
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) CLWG SASA BMI-2104 CPWG USI OUEST SAUNA IDCOR IDCOR ASEP BMI-2104 Others Others V V 1P Frequencies of ssionfroM Accident Containment Corewek Phenomena Event Trees
"'#8 Sequences Terms V
Frequencies of Accident
" " Consequences Releases USl Risk AE Estimate Others RAP V
Costs of Risk. Financial Safety Reduction impact of Options Potential Accidents V Value/ impact Assessment EftX ASEP Accident Sequence Evaluation Program CLWG Containment Loads Working Group CPWG Containment Performance Working Group SAUNA Severe Accident Uncertainty Analysis BMI-2104 Battelle assessment of radiological source term SASA Severe Accident Sequence Analysis USI Studies of unresolved safety issues IDCOR Industry Degraded Core Program OUEST Quantitative Uncertainty Estimation of the Source Term AE Architect-engineer studies RAP Regulatory Analysis Program Figure 1-1. Relationships of Research Programs to SARRP Tasks 1-4
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) appropriate discrimination of the conditions associated with the release of fission products to the environment to allow an assessment of the accident consequences. The second major activity also involved drawing from a number of other research and analysis activities (see Figure 1-1). The central estimate source term evaluation for the accident sequences important to risk was based on calculations performed by Battelle Columbus Laboratories [5]. Battelle's calculations were performed with the Source Term Code Package (STCP) (fi], which is basically an integration of the codes used in the analyses previously reported in BMI-2104, Radionuclides Release Under Specific Accident Conditions [7]. Radiological source terms for accident scenarios of significance were calculated at Sandia using the results of the STCP analyses for scenarios with similar characteristics, and from the results of other programs. The uncertainty analysis involved the solicitation of input from experts that was used to modify the model for source-term estimation to reflect the range of uncertainty about important phenomena or parameters. This process included adjustment of the results to account for phenomena believed to be important but not included in current STCP models. i The offsite consequences of the severe accidents were calculated at Sandia using both ) the computer code CRAC2 [1], as well as the improved code recently developed for this purpose, MACCS [2]. The consequence measures considered included offsite health and economic effects. The final activity described above involved the integration of these analyses into a comprehensive evaluation of risk. This process actually proceeded concurrently with the other tasks to ensure efficient transfer ofinformation at the interfaces among the different types of analyses. As already noted, all of the activities in SARRP included an evaluation of the principal j sources of uncertainty. This was accomplished through the limited application of Latin hypercube sampling techniques for the issues that were believed to involve both large uncertainty and the potential to have a significant impact on risk. This limited Latin hypercube (LLH) approach involved using the opinion of a number of experts assembled to review the issues and provide the ranges of outcomes for each issues. A principal benefit of this approach is the ability to provide some indication of the areas in which uncertainties have the largest impact on calculated risk. It was not the intent, however, to provide statistically . meaningful bounds of uncertainty, but rather to develop a range in which risk may reasonably be expected to lie. Following the reassessment of risk, the effects of a number of potential modifications aimed at improving safety were investigated. For each modification, the elements of the risk calculation were repeated, providing an estimate of the change in risk, or benefit, for each 1-5
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) modification. The costs and other impacts associated with each modification were also estimated. Together, these parameters provided a mlatively straightforward comparison of the l benefits and costs of the modifications. I 1 The primary thrust of this report is the pmsentation of the methodology and results of j these analyses and the identification of the conclusions that can be drawn from them. For some of the elements of the analyses, the detailed calculations are provided in other reports, including Analysis of Core Damage Frequency From Internal Events: Peach Bottom, Unit 2 [3]; Containment Event Analysisfor Postulated Severe Accidents: Peach Bottom, Unit 2 L4]; and Radionuclides Release Calculationsfor Selected Severe Accident Scenarios: BWR, Mark i Design (5). Details are summarized in this report only to the extent that they provide a needed understanding and perspective for the results. Areas of analysis not covered in other reports are described in detail in the appendices to this report. 1.3 OVERVIEW OF THE PEACH BOTTOM PLANT This analysis addresses Unit 2 of the Peach Bottom Atomic Power Station. As noted earlier, the Peach Bottom plant is a General Electric BWR/4 boiling water reactor. Features of the plant systems that provide a perspective useful to understanding the results include the following: The station has four diesel generators shared between the two units. Each of the divisions supplies power to one train of the redundant emergency safeguard systems (ESFs). There are two basic trains of emergency de and 120 vac power for each unit, each consisting of a charger, battery, and associated busses. There are also some interties with Unit 2. The plant has a single-train high-pressure injection system (HPCI) which is steam turbine-driven. As with other BWRs, the plant also has a reactor core isolation cooling system which consists of a single-train, steam turbine-driven pump for delivery of water to the core. The plant has two low-pressure core spray pumps delivering flow to the core through spray headers, and two low-pressure injection (LPCI) pumps which provide the same function through injection rather than spray 1m, es. The LPCI function is an operating mode of the residual heat removal l (RHR) system. ' Containment heat removal is accomplished through the RHR system, which consists of the two LPCI pumps with two associated heat exchangers. The heat exchangers are cooled by the high-pressure service water (HPSW) system. 1 1-6
I NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) J
. The HPSW is shared by the two units at the site. This system also has the provision for manual initiation of a flowpath that would allow river water l 1
to be pumped through the HPSW pumps directly to the core, as a last resort m the case of a loss of all other injection functions. As with other BWRs, the plant also has an automatic depressurization ; system (ADS) that can either automatically or manually depressurize the i reactor-pressure vessel to allow injection of water thmugh the low pressure ) systems,if available. ) The emergency service water system serves as the heat-removal medium for the heat exchangers and room coolers which serve the emergency equipment such as the diesel generators and the low pressure injection i pumps. The plant procedures are based on the General Electric Generic Emergency Procedure Guidelines, Rev. 3, soon to be updated to revision 4. The most significant feature of the plant with respect to the mitigation of severe accidents is the suppression pool design for pressure suppression in the containment. Figure 1-2 illustrates a simplified cross-section of the containment building. The primary containment consists of a large steel drywell and a toroidal suppression chamber housed within a secondary ] containment. The secondary containment (or the reactor building) is a large multi-level ) structure which completely encloses the pnmary containment as well as several of the plant I systems. The reactor is situated in the drywell, and steam leaked from the reactor-pressure vessel during accident conditions is condensed by the suppression pool. Vacuum breakers are also provided between the drywell and torus airspace to allow a transfer of air back to the l drywell airspace. The primary containment is inerted during operation. The current emergency procedures call for venting of the containment when the pressure reaches 60 psig, which is just above the design value. 1.4 ORGANIZATION OF THE REPORT The main report summarizes the methods used in the rebaselining of risk (in Section 2) and the investigation of risk-reduction measures (in Section 3). Section 4 presents a definition of the uncertainty issues considered in the LLH study and a brief description of the input from the expert review group for each issue. The results for each major step of the analyses, including the overall risk estimates, are presented in Section 5; the effectiveness of potential risk-reduction measures and the comparisons of costs and benefits are outlined in Section 6. The conclusions that can be drawn from the program regarding the significance and usefulness of the results are provided in Section 7. Supporting information concerning the risk-r:baselining and risk-reduction calculations and results is provided in a separate appendix 1-7
. . l 5
' NUREG/CR.4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987)
/' sa
'1 Blowout Panels 1
Refueling Bay
- a r.e....
R , . . ., Reactor Buildin9 .. N 5 u :=. 4:= - r
=_ ::. . Control Room
---- : := :
Reactor Vessel ::
;g j %*L3 ;: *
.* "i. ....
p:: .. .:.. ,:;. ...* ;::: ...
....:.. ..p...
- ::.. . ... f. :* -
a.. j w. DrywellP.:. 4w-
~
..- l Pedestal ::I
- .: _ . .. .." '.~.
x .....:. . ....
= . Wetwell =. :
- m. . .. . .. .;-..I . . .. . . .
1 m_.
..~~.....:
1
>......:..........~~..~........:.:
. _ . . . . . . . ...._...~...:...
. _... . \ .._.:............._.
.s.._,....._...:..:.....
. ....... ... .. .. ..._.. .. :. ... .... .~. . .~. .. . . .
- . .. .........~..............- ..... .. _. ._ .
.......:........ ,,1 ..
... _. .. . ._.___. _/ ,
._..........;~- :........._......_...
._ . Su ression Poo .. -.-
. . . ....._...~.~.:.....~._._.............
._. ._ .-- . . ..~..........:..._
.. ..... ... .......~.--.._.._._.:.....__..
_.... ... ~_ Figure 1-2. Cross-Sectional View of the Peach Bottom Containment 1-8
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) for each area of the analysis. Appendices A and B discuss the containment and source-term analyses, respectively. A more detailed presentation of the risk results is included in Appendix )
. C, and additional background material concerning the treatment of risk-itduction measures is I provided in Appendix D. Appendix E includes a discussion of the limitations of the LLH approach to the characterization of uncertainty developed from the comments of the expert review group. This discussion is intended to provide additional perspective on the insights of the overall study. Finally, Appendix F outlines the methods and sources of data used to j estimate the costs for the modifications considered. [The appendices of this volume of the drift for comment are being published under separate cover.]
l
)
i j 1-9 L____ _ -_ - - _ _ _ ___ _ - - _ - _ __.----- _ ____---_-___ __ _ _ _ _ _ _ _ ___ _ _ _
NUREG/CR 4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) References for Section 1
- 1. Reactor Safety Study--An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants. U.S. Nuclear Regulatory Commission Report WASH-1400 (NUREG.
75/014), Washington, DC: 1975.
- 2. IDCOR TechnicalKnoxvi Energy Corporation, Summar,) lle, TN: Report. Industry Degraded Core Program, T 1984.
- 3. Kolaczkowski, A. M., et al. Analysis of Core Damage Frequencyfrom Internal Events:
Peach Bottom, Unit 2. U.S. Nuclear Regulatory Commission Report NUREG/CR-4550, Volume 3 of 7 (Draft Report), Sandia National Laboratories, Albuquerque, NM: November,1986. ;
- 4. Amos, C. N. and A. M. Kolaczkowski. Containment Event Analysisfor Postulated Seven: Accidents: Peach Bottom, Unit 2. U.S. Nuclear Regulatory Commission Report NUREG/CR-4700, Volume 3 (Draft), Sandia National Laboratories, Albuquerque, NM:
February,1987.
- 5. Denning, R. S., et al. Radionuclides Release Calculationsfor Selected Severe Accident Scenarios: BWR, Afark I Design. U.S. Nuclear Regulatory Commission Report NUREG/CR-4624, Volume 1, Battelle's Columbus Division, Columbus, OH: July, 1986.
- 6. Gieseke, J. A., et al. Source Term Code Package: A User's Guide (Af0D1). U.S.
Nuclear Regulatory Commission Report NUREG/CR-4587, Battelle Columbus Laboratories, Columbus, OH: July,1986.
- 7. Gieseke, J. A., et al. Radionuclides Release Under Specific Accident Conditions . Battelle :
Columbus Laboratories Report BMI-2104, Columbus, OH: 1984.
- 8. Ritchie, L. T., et al. CRAC2 Afodel Description. U. S. Nuclear Regulatory Commission Report NUREG/CR-2326, Sandia National Laboratories, Albuquerque, NM: 1984.
- 9. Alpert, D. J., et al,"The MELCOR Accident Consequence Code System." Proceedings of the CEC Workshop on biethods ofAssessing the Of-Site Radiological Consequences of Nuclear Accidents. Sandia National Laboratories Report SAND 85-0884C.
Albuquerque,NM: April,1985. l l r 1-10
l l NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Section 2 ) METHODOLOGY FOR REBASELINING OF RISK The overall approach taken in assessing the risk associated with the operation of Peach I l Bottom was similar to that employed in performing a plant-specific probabilistic risk l assessment (PRA). This section describes that process as it was implemented for the SARRP analysis of Peach Bottom. The work outlined in this section represents the first complete j application of an integrated, and fully automated, assessment of uncertainty in PRA. The types ofinformation and results for each of the analysis areas outlined in Figure 1-1 are illustrated in Figure 2-1. Section 2.1 describes the overall integration of the various activities and the review processes intended to enhance the technical quality of the results and to ensure that the developments made by other research and analysis programs have been appropriately considered in SARRP. The general approach to the development of the central estimate of risk is summarized in Section 2.2, and the treatment of uncertainties in SARRP is described in Section 2.3. Sections 2.4 through 2.7 describe the analysis methods for each of the technical activities. Additional supporting information for each task is provided in the appendices. 2.1 OVERVIEW OF RISK-INTEGRATION AND REVIEW ACTIVITIES Accomplishment of the SARRP objectives required the performance of a variety of diverse tasks by individuals from a number of organizations. The process used to integrate all of these activities effectively was therefore critical to the success of the program. The basic integration steps employed in SARRP are described in Section 2.1.1. An equally important activity was the assurance of technical quality through effective review efforts, as discussed in Section 2.1.2. 2.1.1 Integration of Project Activities As illustrated in Figure 2-1, the analysis in each of the principal task areas generated a substantial volume and variety of information. In addition, the Latin hypercube sampling' approach to uncertainty analysis (described in Section 2.3) required that the entire analysis be repeated for many different combinations of assumptions. To accomplish this, an automated approach to PRA has evolved through the SARRP effort. This approach involves dividing the analysis into a number of modules. Within each module is a computer code which performs that particular part of the analysis. Each code has u " post-processor" which provides the interface between modules. The efficacy of this approach rest:: heavily on the careful attention 2-1
y sc e n e n e ~ ic n u s e s io s s s h T s e qt y s s r a ee lt a U ur ea te qF St a w e e r t l ee e e g Se r ee s gic mR H a P htae rone mt s T t uFgg aa ae r f o as Cemm Pd o Pu ad y Do U t nnc aa t nMt q a ne Plic u c n y C O eeDD-d u - n e. d F er la c n ops et ucei str e t iqet n imluin c i si u 'q e p s c r yd o ef f - rEr o n i c eola oaicB DFA h ar PO ASCP PRG F
'v .7 - '
S t e S "luis t ne d c ns a s ees n *f s E Fly a mTsri a r ei u s y C dn n 1 it n ane a ly , P 1 ql ea O aA n C s n e ee t n vA T S nA R Ye r CoE c" o C P TrT mr E m" i* f y s s n ie/ d s ity sol it g loy/ e it e r psi n o e s o tu l ib e mnibe sor ieid t n S o mva a it l a y at ps a iPic t n t p r u it ayoivn t ntclo Co a s D e T g iab l i CSnreui o t e ue yic R U f na a t oDl is nnhSt c t t odVt Por ne n gh y P ee eePt An lopa ync C yR r u mmt nn n e rn laPom o r r ge N lt mind iiemt n i c nin o it si o gogr I eb ae aad pr miota e ol mpo e t sl yeur amco t tnndiu oocqn n espn h iseo t e eom SRHP CCAEO CFDC MDTE
~ot
' l ,l!
NUREG/CR.4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) paid to the interfaces between modules. In general, the post-processor in each module sorts the l results of the analysis performed within that module into groups (or bins) having similar attributes. This grouping is done on the basis of characteristics of the results obtained in a particular module which will have significant impact on the results obtained in subsequent modules. This consolidation into groups imolves some loss of resolution. However, sufficient detail is maintained to provide a level of accuracy in the results that is at least as good as the accuracy with which the results within any given module may be calculated. It should be noted at the outset that this analysis differs little from the approach used in the previous SARRP analyses of Surry [1] and Sequoyah [2]. The principal differences between these analyses and that for Peach Bottom (or Grand Gulf [3]) is that the interfaces between modules have been fully automated. The automation significantly reduces the time required to conduct the analysis. In addition, a greater level of detail can be maintained across the interfaces between modules. This additional detail allows for improved handling of the complex interactions between the nuclear steam supply system and the containment in a BWR. The Peach Bottom analysis represents the furthest evolution in methodology for any of the plants considered by SARRP. As in the Grand Gulf analysis, the uncertainty study was fully automated. However, the Peach Bottom study includes a fully integrated estimation of selected uncertainties associated with the fission-product source term. 1 The overall philosophy of this approach was to develop a point-estimate model for risk 1 by linking together the various modules, and to evaluate uncertainty in risk by studying the mapping from input (data, modeling assumptions) to output (magnitudes of various risk measures) by the point-estimate model. The result obtained from the point-estimate modelis referred to as the central risk estimate (described in Section 2.2). The major uncertainties in this analysis were evaluated by varying the inputs to selected modules in accordance with a Latin hypercube sample of their values. In many instances, these inputs characterize the model rather than being just the inputs to the model. Each member of the sample (150 were used) is a set of values for the selected inputs which results in a different point estimate of risk. Risk is evaluated for each of the 150 members of the Latin hypercube sample. These 150 point estimates of risk are assumed to investigate the mapping of input to output sufficiently that they may be used to characterize the uncertainty in risk. Figure 2-2 shows schematically the interrelationships between modules in the risk model. Each square box represents a computer code. The rectangular boxes represent the l post-processors to the codes. The flow of the analysis is from left to right. The results of one 2-3
a ta Dy laph 2 3 E ca C U C I y k O igrg rC oo o A s S 1 is R lom e I C M r c R oe n eD k,> e te 7 u M 3 q e_ Mg9 s n
, o c C
.. p t )
- m8E R or$ 4ma -
ut s o r lt y b ou s g A pss e n nR j>-- M e u l o io a s o aTrPC s t C i d
- n. tcT A m o R r ht c uS id T L
i e n i romo L E J T- e cP fr ir R c e r M t nd u e os ev mi o A a s ir n s e S R r ie aF D t-i o ns
.i c P P ( es n i d d e e icg u z c q Ar o i
e s r P F r e a u l s s u
- i d
?& E mQ4n<5 e o
- c n
a e lp M f j>-- %N m r m o o a t n f r S w e E e e ie m R P b v n T A a uc r N g r e n V n ep v E d a y O e g
> s L
t o H in t 2-a e n a 2 i e L m*c a*n m e r e n D'q u i a u t n e t n ig a r o F R F C _ mE $ - mK>3~ _ l
- s ity
>MN e
u s C I s l i b A g d a M , n ie E E R T t n o
'7 F r
s n teio S ss u- e t r Cp x E
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) module (the output) constitute some (but not all) of the information (the input) required by the next. The remaining information required by the module is shown in the upper portion of the figure. Inputs to the module which are treated as uncertain in the Latin hypercube study are shown schematically in the lower portion of the figure. These inputs overlay the inputs from the upper portion. Only uncertainties for the three indicated modules were considered. As the diagram suggests, the number of results obtained for a particular module is greater than the ! number of inputs transferred to the subsequent module. In principle, the binning (or collapsing of results) performed by the post processors need not be undertaken. However, the number of calculations required would be ovenvhelming. The four principal modules in the analysis are I described below. Seauence Frecuency Analysis The Top Event Matrix Analysis Code (TEMAC) evaluates the frequency of combinations of events which lead to core damage. These combinations, referred to as cut sets, reflect failures of the plant systems needed to provide core cooling; nearly two thousand i l such cut sets were obtained by the ASEP analysis of Peach Bottom [4]. The TEMAC output is l the frequency of each cut set. The post-processor calculates the frequencies ofindividual plant- ! damage states by summing the frequency of all cut sets belonging to that damage state. A l plant-damage state is a group of cut sets that present an identical set of initial and boundary conditions to the containment event analysis. The thousands of cut sets are grouped into eleven plant-damage states for Peach Bottom. Because of the complex system and phenomenological interactions between the core-cooling functions and the containment response to accidents, it was necessary for the ASEP sequence analysts and the containment-event tree (CET) analysts to work together closely to identify the attributes of the cut-set expressions which determined the grouping into damage states. l l Containment Analvsis The containment analysis in SARRP relied on the evaluation of a very large event tree. l That portion of the analysis was made possible through the development of a new computer code (EVNTRE) which traces, and quantifies the probability of, all possible pathways for the progression of severe accidents [1]. The code was reviewed by the CET review group, which j also reviewed the CET itself for Peach Bottom; the CET constitutes the input to the code [fi,2]. The very large number ofindividual pathways calculated for each plant-damage state prohibited a source-term evaluation being made for each. Thus, the pathways were grouped by the post-processor (POSTSV) into accident-progression bins. The binning was determined on the 2-5
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) basis of attributes of the pathway (e.g., timing of containment failure, degree of suppression- 1 pooi bypass) which were used in the source-term code (RELTRAC) to evaluate the inputs to the fission product release and transport model (which is also incorporated into RELTRAC). The bin attributes are thus specific to the RELTRAC model. Each pathway in a given accident-progression bin has exactly the same source term since each bin defines an identical and unique set of inputs to RELTRAC. It is expected that the source terms for each pathway in a bin would actually be vey similar. This binning is similar to the source-term binning in the Sury and Sequoyah analyses. A distinction is drawn because in those analyses the bin (reflecting the magnitude and timing of the source term) was determined prior to the final CET analysis. For Peach Bottom (and Grand Gulf), only the possible inputs to the source-term model were defined before the final CET analysis. The accident-progression bins are combinations of these inputs which occur with significant frequency. Source terms are calculated by RELTRAC following the CET analysis. Because of the very large number of accident-progression bins resulting from the Peach Bottom CET analysis (more than 3500 were identified), it was expedient to drop from consideration those bins which would not significantly impact risk (due either to low probability or low source-term magnitude). The post-processor was used to discard those bins which had both a frequency of occurrence below a pre-selected cut-off value and a frequency-weighted estimated source term which was less than a second cut-off value. The cut-off values were selected such that the analysts expected no significant reduction in the accuracy of the l results. The total frequency-weighted estimated source term for each member of the limited Latin hypercube (LLH) sample (and for the central estimate) was conserved by POSTSM by adjusting the conditional probabilities of the bins which were kept. The source-term estimate performed by POSTSM employed a simple model similar to those used for estimating source terms in the SARRP analysis of Surry, Sequoyah and Grand Gulf. Source-Term Analvsis A parametric source-term model, implemented within the RELTRAC computer code, was used for source-term evaluations. The Source Term Code Package (STCP) calculations which were performed for Peach Bottom [H] indicated that source terms were generally large, i as compared to those for the other SARRP plants, for the dominant accident scenarios. On the l other hand, the CET analysis indicated that there were several features of the plant, and of the accident progressions themselves, which could reduce the source terms significantly below those calculated by the STCP. It was therefore judged appropriate to utilize a more detailed source-term rwdelin the Peach Bottom analysis than had been used for the other SARRP 2-6 j
DRAFT REPORT FOR COMMENT (FEBRUARY,1987) NUREG/CR-4551, VOL 3: reference plants. The previous models (SURSOR for Surry [1], SEQSOR for Se and GGSOR for Grand Gulf [1]) employ a simple arithmetic expression to evaluate th of fission products from the fuel and their deposition within the primary system an containment. The source-term model implemented within RELTRAC instead solves parametric rate equations for fission product release and transport. Thus, REL time-dependent release histories rather than just integrated releases over the c accident. As in the STCP, the fission-product species are grouped according to similarities i their physical and chemical properties which determine the manner in which th and transported. Every radionuclides within a given radionuclides release group transported and released in the same fraction (of its core inventory at shutdown). I RELTRAC is a purely parametric model for the transport of the nine radionuclides groups which have been considered by SARRP. The values of the parameters i were adjusted by the SARRP analysts such that RELTRAC predicted a distribution products within the plant, and a release to the environment, which was as c:ose that predicted by the STCP for five accident scenarios as simulated by Battelle [H). RELTRAC is believed to produce results which are as close to those that would have been obtained by the STCP as is possible, given the constraints of the analysis. RELTRAC included models for phenomena not considered in the STCP but which were considered b (Further SARRP expert reviewers to be important sources of source-term uncertainty. discussion of the uncertainty analysis is provided in this and succeeding sections). For th purpose of providing central source term estimates, these models were not imp Details of the RELTRAC code and the evaluation of the model parameters are provided Appendix B and in Reference 9. While RELTRAC provides relatively detailed information about the distribution of fission products within the Peach Bottom plant that would be expected for the vario regressions, the result of primary interest was the predicted release to the environm function of time. For the purpose of formulating this information in a manner suitable fo to the consequence calculation, the integral release was divided into two time periods
" puff' release associated with rupture of the containment, and (2) the total relea remainder of the calculation.
For the purpose of providing central-estimate results, this source term information calculated for each accident-progression bin determined (by POSTSM) to be potentially important. For the LLil analysis there were potentially 150 source-term estimat
NUREG/CR-1551, VOL. 3: DRAFF REPORT FOR COMMENT (FEBRUARY,1987) accident-progression bin, corresponding to the different source-term model associated with each member of the sample. Thus, there wem many thousands o associated with the LLH analysis. In order to reduce the number of consequence c to a manageable range, these source terms were grouped (using the CLUSTER co source-term clusters. All source terms assigned to a given cluster were assumed to have identical offsite consequences, although they might arise from very different acc progressions. The CLUSTER code exercises a standard clustering algorithm and uses estimate magnitudes of early and late consequences (based on the magnitude of early and l product releases) to group the source terms. The source term used in the con } calculation for a particular cluster is that for which the estimated conseque approximate the average for all the source terms in the cluster. Thus, the consequ cluster may be linked to a particular accident progression with a specified set o assumptions. In this aspect, the use of clustering differs from that in previous SAR analyses. For the preceding plants, the source term associated with a cluster w in that it was the average of many source terms, and could not be linked to a progression. However, it must be noted that the linkage to an accident prog only for the set of source-term model assumptions that resulted in that source . accident progression may result in entirely different predicted consequences if source-assumptions were used. Consequence Annivsis Offsite consequences were calculated for each accident-progression bin a the central risk estimate and for each source-term cluster using the CR AC2 and M codes. These codes are described elsewhere. The magnitude and timing of fission-p was estimated by RELTRAC as described above. The rate of energy release as the plume was estimated from STCP results based on the timing of containment fai ure and a few key attributes of the accident progression associated with the source term . Other code inputs were obtained from site-specific data and from the STCP mn which wa the particular accident-progression bin being analyzed. In teera tion _ Due to the very large data-manipulation requirements, the integration of all of ' of the study into final risk results was accomplished using a computer co _ __--- - - - - - - - - - - - - - - - - ~ ~
NUREG/CR-4551. VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) i specifically for that purpose. This code, called the Risk Integration, Sensitivity and ) Quantitative Uncertainty Evaluation code, or RISQUE, requires the following input [19]: Frequencies of plant-damage states;
- CET post-processor results in terms of frequencies of accident-progression bins for each plant-damage state; ;
- Source-term cluster assignments for each accident-progression bin; and i
Magnitudes of public-health consequences and cost impacts for each ! accident-progression bin er source-term cluster. j The magnitudes of the release fractions for each source-term bin are not included in the RISQUE code input because they are used directly as the input data for the consequence analyses using the CRAC2 or MACCS codes. The RISQUE code also manipulates the information associated with uncertainty analysis and the risk-reduction evaluation. These other uses are discussed in Section 2.3 and Section 3. 2.1.2 Review and Ouality Assurance l There were a number of internal review activities associated with the task performance ; in SARRP. These activities included the analysts' checks of code input and output, review for reasonableness and consistency, thorough documentation, and peer and management review of i results and documentation. In addition to the diligence and internal review processes aimed at achieving a quality product, important aspects of the study were subjected to review by experts not directly involved in the project analysis. These external reviews covered the methods used and the specific application of those methods. The ASEP evaluation of accident sequences was reviewed by two separate groups: (1) a senior-consultant group (SCG) which focused on the the basic methods and their manner of implementation, and (2) a quality-control (QC) team which considered the details of the analysis of the Peach Bottom plant. The members of the SCG and QC teams for Peach Bottom are listed in Table 2-1. The CETs were reviewed in some detail by the CET review group, which included representatives from other national laboratories and a university (see Table 2-1). The review group set out to answer the following questions:
- Is the CET reasonable?
Are the questions in the tree scrutable, complete, and consistent? Is the level of detail appropriate? 2-9
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRll ARY,1987) d Table 2-1 l PARTICIPANTS IN SARRP REVIEW TASKS FOR PEACH BOTTOM Review Activity Participant Affiliation D. Bley Pickard, Lowe and Garrick, Inc. ASEP Senior M. Bohn Sandia National Laboratories Consultant Group J. Murphy U. S. Nuclear Regulatory Commission W. Vesely Science Applications International Corp. l B. Bell Battelle Columbus Laboratories ASEP Quality G. Kolb Sandia National Laboratories 3 Control Group E. Krantz Idaho National Enginecting Laboratory j G. Parry NUS Corporation 2 A. Payne Sandia National Laboratories M. Corradini University of Wisconsin SARRP CET Review R. Denning Battelle Columbus Laboratories Group (Peach Bottom) S. Hodge Oak Ridge National Laboratory J. Lehner Brookhaven National Laboratory A. Torri Pickard, Lowe, & Garrick H. Asher U.S. Nuclear Regulatory Commission F.E. Haskin Sandia National Laboratories SARRP Review M. Corradini University of Wisconsin Group for R. Denning Battelle Columbus Laboratories Containment LLH G. Greene Brookhaven National Laboratory Issues (Peach Bottom) S. Hodge Oak Ridge National Laboratory C. Hofmayer Brookhaven National Laboratory K. Perkins Brookhaven Natural Laboratory W. Von Riesemann Sandia National Laboratories SARRP Review Group A. Benjamin Sandia NationalLabcratories for Specification of P. Cybulskis Battelle Columbus Laboratories STCP Cases to Be Run R. Denning Battelle Columbus Laboratories (Peach Bottom) F.E. Haskin Sandia National Laboratories D. Williams Sandia National Laboratories C. Amos Technadyne Engineering Consultants SARRP Review Group R. Denning Battelle Columbus Laboratories for Source. Term LLH J. Gieseke Battelle Columbus Laboratories Issues (Peach Bottom) T. Kress Oak Ridge National Laboratory D. Powers Sandia National Laboratories D. Williams Sandia National Laboratories 2 10 __ ._______._._______________________.____._.m__ _ _ _ . _ _ . _ _ _ _ _ _ _ _
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) e Is the CET quantification masonable?
- Are the branch point probabilities traceable?
- Given traceability, are the referenced sources documented, reviewed and complete?
- Is them an underlying methodology for the estimates (i.e., consistency of the approach)?
The CET review group provided a number of recommendations that have been published separately [fi,2]. In addition, limited review was provided by members of the Philadelphia Electric Company engineering staff [11]. Where practical within the schedule constraints of SARRP, the project team attempted to implement these recommendations. ! The radiological source-term calculations performed for Peach Bottom using the STCP
.were reviewed by the SARRP source term review group. In addition, these calculations received careful scrutiny from the SARRP team as a consequence of their efforts to implement RELTRAC and to study the impact of hydrogen combustion on the integrity of the reactor building [12,13]. At the same time of this writing, the RELTRAC code is undergoing intemal -
review at Sandia. Further, external review, is planned. To the extent that RELTRAC shows excellent capability to approximate STCP predictions for the five particular accident progressions, the analysts have reasonable confidence in the validity of RELTRAC source-term - estimates. Detailed results of these benchmark calculations are reported elsewhere [L4]. The validity of the RELTRAC code predictions of source terms for accident progressions which differ markedly from those considered in STCP calculations is yet to be established. The results do, however, agree with the expectations of the analysts. In summary,it is the opinion of the principal investigators that the RELTRAC results are at least as valid as those obtained with the more simple source-term models employed for the other SARRP reference plants. l Each of the task areas also benefited from the additionallevel of review associated with the characterization of uncertainty issues, as discussed in Section 2.3. The members of the ; expert review groups that provided the input for the LLH analyses required a fairly detailed l understanding of the significance of various phenomenu in order to assign meaningful ; weighting factors. This process therefore offered another level of review that led to some l improvements in the models and data. l l 1
-2.2 DEVELOPMENT OF THE CENTRAL ESTIMATE OF RISK j As mentioned previously, the SARRP analysis included evaluation of a " central estimate" of risk which was based on STCP predictions for the radiological source term. The central estimate was not intended to be (nor should it be construed to be) a "best estimate," !
s 1 2-11 o i
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) because the STCP is known to lack models for some of the important source-term phenomena. It was intended instead to provide a comparison of risk results based on the NUREG-0956 [11] source-term methodology (i.e., using the STCP) with a range of risk estimates obtained from the input of the expert reviewers (i.e., the LLH process). The various parts of the central i estimate wen: evaluated as follows: (1) The frequencies of the plant-damage states were obtained from the ASEP 1 point estimates. (2) The containment release-mode probabilities were obtained by quantifying the containment event tree using input judged.to be representative of the median of the reactor-safety community. Characterization of the median was a task performed by the SARRP analysts after considering all the , available analyses and data sources relevant to each of the questions in the j contamment event tree [1(i]. ! (3) The fission-product source terms were obtained from STCP calculations. ; The processes described in Section 2.1.1 were used to define bins for source-term analysis and to estimate source terms for the full set of bins. j In general, these source terms accounted only for phenomena considered in j the STCP; however, adjustments were made to reflect some specific l phenomena, such as direct heating. (4) Consequences were estimated with both the CRAC2 and MACCS codes ! for each source term. l l 2.3 CllARACTERIZATION OF UNCERTAINTIES l l Although there have been significant advances in all areas of risk-assessment technology, there remains significant uncertainty in each of the analysis tasks. The appropriate means by which to characterize this uncertainty remains a topic of substantial debate within the technical community, and there was no attempt in this program to resolve these differences by developing the " correct" representation of uncertainty. The most important results of the analyses reported in this document are engineering and scientific insights that become evident after the completion and integration of each of the steps in the program and thorough review of the results. However, the significance of many of these insights can often be better understood within a quantitative framework. It is therefore essential that a clear presentation be made of the elements considered to be uncertain, and of the potential effects of these uncertainties on the results. The formulation of the uncertainty presentation for the SARRP results therefore had the following objectives: To provide decision-makers with engineering- and/or scientifically-based information that allows them to understand the analysts' treatment of 2-12
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) important issues and the impact on the analysis of the range of viewpoints that experts in the field hold for these issues;
- To develop a quantitative estimate of uncertainty that mflects a cmdible and realistic range in which the analysts have a reasonable confidence that the correct answerlies;
- To identify as completely as possible the key sensitivities and sources of uncertainty for each portion of the analysis, including those that have the most impact on the calculated risk measures;
- To estimate for each part of the analysis a quantitative impact on the results for the uncertainties that were identified; and
. To evaluate the quantitative impact on the risk measures of the uncertainty in each part of the analysis and the combinations of the uncertainties and sensitivities for the different portions of the analysis.
Because many portions of the analysis rely heavily on expen opinion and only limited data from actual experience, it was judged that a direct statistical treatment of all uncertainties was neither practical nor necessarily meaningful. In addition, another NRC-sponsored program, the Phenomenology and Risk Uncertainty Evaluation Program (PRUEP), is investigating a mom compmhensive evaluation of uncertainty [12]. The statement of the SARRP goals relative to uncertainty relies on language that is highly subjective. The question of a " reasonable" assessment is certainly of a subjective nature, and each analyst involved would have different interpretations of its meaning. It was intended that the general outcome of the analysis be such that the uncenainty presentation represent a belief on the part of the analysts that there is roughly a 90% probability that the correct result would lie within the range. Clearly, this is difficult to assess, impossible to test, and analyst-dependent. The reasonableness of the uncertainty characterization depends strongly upon how well the analysts selected the sources of uncertainty which they chose to repmsent in the study. If an important source of uncertainty was omitted, the uncertainty in the risk estimate will :end to be understated, and a reasonable range may not be reflected. In addition, several of the participants in the uncertainty analysis task felt very strongly that the methods by which the input for the uncertainty assessment wem collected and manipulated did not support any probabilistic interpretation, such as a confidence interval. This is discussed further at the end of this section,in the limitations discussions accompanying the results in Section 5, and in Appendix E. Another problem is associated with the use of the terms
" uncertainty" and " sensitivity." The basic method chosen for the uncertainty repmsentation in this study does have some of the features of a sensitivity study, but the results am defined here to represent a measure of uncertainty. There was considerable discussion within the expert-review groups about the use of terminology, and individual interpretations are contained in 2-13
i NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Appendix E. Once again, since universal definitions are not available, these terms are defined here only as they apply to SARRP. The terminology associated with the uncertainty treatment is defined with the discussion of the methodology below. : For all portions of the analysis, the important contributors to uncertainty are discussed in qualitative terms. This is especially useful for some of the areas in which it was not possible to generate a meaningful quantitative estimate of uncertainty. 2.3.1 Overview of Uncertainty Treatment There are many ways to categorize the sources of uncertainty in a risk assessment. One characterization involving three types of uncertainty is presented below: Data Uncertainties arising from incomplete or inconsistent data pertaining primarily to equipment-failure rates. Modeling Uncertainties arising from incomplete or incorrect modeling of either phenomenological processes or engineered systems. Construction Uncertainties arising from variability in plant construction or configuration from that designed and studied, due to both errors and variability in physical properties. Associated with each of the analysis activities are varying degrees of these types of uncertainty; those that have been identified and considered for the major tasks in the risk-rebaselining effort are outlined in Table 2-2. Because of limitations in time, resources, and available methodologies, not all of the uncertainties listed in Table 2-2 were treated in the SARRP analyses, and those that were included were not necessarily considered in the same way The primary focus of the SARRP uncertainty evaluation was on data and modeling uncertainties. Data uncertainties were evaluated primarily for the core-damage frequencies, which were analyzed in ASEP. Data uncertainties that could be reasonably characterized by continuous distributions were propagated via the TEMAC code to obtain a composite uncertainty associated with data variability. Some data uncertainties, however, did not have a sufficiently complete data base to warrant a single continuous distribution. These were treated by ASEP as sensitivity cases, similar to the modeling uncertainties (see Section 5.1). Uncertainties in modeling were considered for the sequence, containment and source-term analyses, but not for the consequence study. Uncertainties arising due to construction were considered only with l respect to the structural capability of the containment (i.e., failure pressure and location of failure). Some of the modeling and data uncertainties considered in the ASEP sensitivi ty i 2-14 i
d NUREG/CR-4551, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) l Table 2 2 UNCERTAINTY CHARACTERISTICS OF DIFFERENT ANALYSIS ACTIVITIES Type of l Analysis Task Uncertainty Comments I Analysis of core. Data Uncertainty in quantification of fault- and event. damage sequences tree models. Modeling Potential for incomplete or incorrect models: system-success criteria or phenomena (e g., behavior of seals for emergency-core cooling systems pumps). Construction - Possibility of system configurations or - capabilities different from design. Containment-event Data Some system-data uncertainties and some analysis interpretation of applicable research data. Modeling Potential for incomplete or incorrect models of phenomenology or plant response or errors in implementation of models. Some phenomena have very different models proposed by different experts. Construction Principal concem is in the containmu.t response relative to predicted failure characteristics. j Source term Data Collection and interpretation of applicable analysis research data. Modeling Potential for incomplete or incorrect models of phenomenology or a lack of any model at all. The modeling uncertainties in this task extend to a fine level of detail concerning the behavior of individual species in a variety of circumstances. Errors in coding or input data are also possible. Construction Not of primary concern for this task. Consequence analysis Data Some uncertainty in demographics, weather, etc., although impacts are rather limited. Modeling Potential for incomplete or incorrect models of health effects, physical phenomena, or - emergency response, j Construction Not applicable. ! l
)
I I 2-15 , i ~. _ _ _ _ _ _ - - _ _ . _ _ _ _ - _ _ _ - _ _ _ _ - -
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) studies were combined in SARRP to provide a composite uncertainty represented by a range of point risk estimates (in the LLH analyses). The ASEP uncertainties associated with data variability are displayed separately in SARRP, and are not combined with the LLH results. Once stated, the means of satisfying the objectives listed above for the uncertainty analysis were not immediately apparent. After lengthy consideration of the available approaches, the SARRP analysts, together with the NRC sponsors, agreed upon the approach described below. The evaluation of data uncertainties for the ASEP estimate of core-damage frequencies I was performed using re'atively standard Latin hypercube (modified Monte Carlo) techniques for sampling from the probability distributions for the data variability. As noted in the previous section, the TEMAC computer code was developed for this purpose. The probability distributions for individual events were propagated through the plant logic model to obtain a 90% confidence interval. ; The principal approach to the characterization of uncertainties other than data variability is the LLH approach;* it is referred to as limited because only a few key sources of uncertainty were considered. Particular issues selected by the SARRP analysts and an expert review group were investigated by determining the range of outcomes that might be expected and sampling the outcomes based on the assignment of weighting factors by the experts. All other inputs to the risk evaluation, including the majority of the data uncertainties considered by ASEP, were held at their mean, or central estimate values. This method is described in Section 2.3.2. As already stated, the construction uncertainty was limited to a consideration of the structural performance of the containment under selected loading conditions. This uncertainty was one of the inputs to the risk evaluation sampled within the LLH approach. The containment performance experts assigned weight to various containment failure pressures and locations,in addition to the size of the leakage resulting from failure given certain loadings.
'An alternative approach, referred to as the optimistic / central / pessimistic (OCP) treatment of uncertainties, was also investigated in some detail. This approach resulted in three separate walkthroughs of the containment and source-term analyses. The assessment for the Surry plant applied both the OCP and LLH approaches, providing an opportunity to gauge the relative merits and weaknesses of each [1]. For Peach Bottom, a partial OCP treatment of uncertainty was carried through only for the CET portion of the analysis. These results are reported only in terms of the relative probability of various containment failure modes [16].
The optimistic and pessimistic walkthroughs of the OCP approach were not developed for the source-term and risk evaluations because of time limitations. 2-16
NUREG/CR-4551 VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 2.3.2 Pescrintion of the LLH Annroach The LLH approach was selected as the principal means of treating modeling uncertainty l because it was judged to provide the best match to the stated objectives, within the other constraints such as schedule and budget. Latin hypercube sampling techniques have been investigated in other programs and found to have very good sampling characteristics when compared to other methods [18]. A full description of Latin hypercube sampling may be found in other sources [12]. The Latin hypercube approach is similar to the better known Monte Carlo techniques, except that the sampling scheme is constrained or stratified. The application of the method and the constraints on the sampling am best explained in terms of the specific process used in SARRP. The basic interest for this assessment is the uncertainty in risk due to the uncertainty in the input information. Risk can be defined by the following equation: RISK k = E i Ej FREQi
- CRMP ,j
- CONSk (FPi j) l where RISK k f
= the risk associated with consequence measure k j FREQi = the frequency of accident sequence i CRMP i,j = probability of containment release mode j, given accident sequence i i
FP i,j = fission product source term for containment release mode j of accident sequence i CONSg = mean magnitude of consequence k, given fission product source term (FP) for sequence i,j There are identifiable uncertainties in each of the inputs to this risk equation. As already described above, a full uncertainty analysis would involve sampling from the range of potential outcomes for every detailed element of every input to the risk equation. The goal for this study was to identify the major uncertainties in each of the inputs and then determine the effect of the uncertainty on the risk measures due to a reasonable variation of those inputs. The LLH approach was the method selected. The LLH process as it was used in this study is represented schematically in Figure 2-3. The first feature which must be recognized when using the results of the study is that the uncertainty analysis was not complete. The word
" limited" in the title refers to the number of issues selected; only a few major uncertainties were addressed. These uncertainties are defined in terms of " issues" which are high level representations of uncertainties in the input. For example, one issue is whether or not the drywell wall is melted through by the core debris following vessel breach. This obviously 2-17
D ES M E EOC U S T S I U O c. t P G E U NS .
. T RO R G I . .
TR 0. 5 0 2 E HO 0 1 1 ts k n GT I C 5, 0 0 0 0 WW EIE la
- s eI'piuso i t E A 2 4, 5 6 9 V
E t r n AnRa i eu l WF 0 0 0 0 0 R e *et rh lc Csf h a Ot oC t N P O - U ET I E 5 3 4 3 3 3 2 1 D NO X TN G N md r n
=t l
e v CVle = ev t l 3
=
t l e v A GR n EEA e e e e P o R S %L L L' L L RW . C ER 0 % % % RE it tc E S RF 5 0 0 % 1 AV I la I DPO > 3 1 5 < SE u
- s E R c e
s R la e S C e c E o r M k s P O S i g C E R d e .n H T U r nt i L U P L O S S .
. . ** g yg& g( $ iyg d jh g:g&N=
R o r s R ia@w l f t
. I s in FO R S-S s n wt oy r
O % % e Cbf a s sn 5 0 1 p E F 7 l oo < b - e GN oi t m p t d Pc a e ' = S A e n r r R e v o g n tee M 3 ty d o loi v s an t E s nl o iur n WIe le U 2 1 3 2 3 n Ditoc i il b M ip r S o e s x oe t p S i ia r iEr e a tccr I a
" le lu b s r an P m
- n S R ia em a e Fo U a S ib E r F Da e let n nC .
DO S E2 ME m U o et eI tc N R , o t l e Re ieo - A OU 1 4 4 2 3 o S r S a E C s P G r r ee t t ao cS - lS )5 e s Ce l nC rS n e s di cr RW uI
- m mg o i
1 ( e V f ose lcn RE o d G r t uo an AV I 1 n l o nV u - nC t i ny SE E a H e t o - L s c a ox i r d e ol r c e R U 1 2 1 4 2 R L ie e mE ao ev S S D R Ain RC DO I 1 2 3 5
' # # s 4
#r #
r r r r e e e e e S b b b b b E m m m m m c s e e e e e I T s n M MM M M te N
- e e io e
- r t ta le le lep le I
A )s T d lu ia f nt e s nim ii on tatar er t r u g lp m a p m a p m a m a p m a eoa nec s i S s S S S Re ty tnt i r ie nsf n es s Er erasu n ect mn gu u n dic o oir tet ye s t e a C d ic mnbio taeptiri r . Nnu pa t no
,iimcsat os t
r r r r r e n tc U H c a ig tat pnr c a et o aef o pe s i p tnt uob anui toe o r e**E e nl r ic r is anrta s ecpliesa t C ( e r co n e gic e aleobe n a e c u s evsr nr la r ne e o u r k s tsinipm I S s s fog e r omvdt sa c t r a r o A e r iz tad e yh es aoe a as et r le e a. aantemteSl l u aCg g B P SHHCGCHMR T( PMHWRA g%
1 I l l l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) involves many other uncertainties (such as the mode and timing of the reactor vessel breach and the heat-transfer processes involved in the fuel debris-concrete interaction) which were not considered individually, but rather in terms of their overall effect. The source-term , uncertainties involve even broader issues; for example, one fission product issue is concerned ) with all uncertainties affecting the release of all radionuclides from the fuel during the core-concrete interaction phase of the accident. It was the intent of the LLH study to include the most important uncertainties. The issues were initially selected by the SARRP team but were then reviewed by the groups of experts in each area of the analysis. It is believed that most of the major uncertainties are addressed, but only more thorough review and comment will verify ) the appropriateness of the issue set selected. Hence, the first limitation in use of the results is i the recognition that the results are limited and not all uncertainties have been addressed. The second step of the analysis was the generation of a range of outcomes for each individualissue. This was intended to represent a reasonable variation of the outcome (e.g. pressure rise) considering all of the uncertainties in the input parameters of the modeling. A second limitation therefore arises from the subjectivity involved in the selection of what l constitutes a " reasonable" range. The ranges were developed by examining the applicable information developed by the reactor safety community. In order to allow efficient manipulation of the information, and to allow expert input in a consistent manner, the ranges l were then discretized and represented by specific outcomes (called levels), usually three to five per issue.* Where possible, the levels represented the results of particular analyses performed by various organizations within the reactor safety community. The levels were originally proposed by the SARRP analysts, but the experts provided additional input and changed the levels where appropriate. l The next step in the process was completed by the expert review groups. This activity was more than a review; a group of experts in each area of the analysis was selected to provide their input as to the selection of the range and the reasonableness of the individual outcomes. This input then formed the entire basis for the subsequent sampling process. In addition to the quantitative irgut, the reviewers were asked to provide information concerning the physical processes they thought govemed the uncertainty. This qualitative information is perhaps more important than the quantitative weighting factors, and substantial effort was expended collecting and reporting this information (see Section 4 and Appendices A and B of this report).
*This discretization can lead to further limitations in statistical interpretation of the results as discussed in Section 5.
2-19
NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) The input was derived separately for the three principal areas covered (systems, containment, j and source term) and is described below following the description of the general methodology. For each level of an issue, each individual reviewer supplied weighting factors to I 1 express his opinion of the relative likelihoods of the outcomes represented by that level. Essentially, the reviewer provided his " degree of belief" for each level, with the constraint that the total weighting factors sum to one for the issue. In terms of the use of these weighting factors in producing the LLH sample, they constitute a discretized probability distribution on the range of outcomes for each issue. However, because of the lack of careful research and analysis which went into assigning these weights, the SARRP analysts and the experts alike do not support this probabilistic interpretation. There is a great deal of subjectivity involved not only in the input but also in the individualinterpretations of the process. The reasons that this subjectivity arises include the following: Reviewers had to judge " reasonable" ranges in terms of their own experiences and views on the uncertainties in the issues, aided by their awareness of prior analyses. The expression of " degrees of belief" when dealing with a range that has been discretized into a limited number of levels also involves interpretation. Then: could be differing views of the scope of individual issues. The process used to elicit the weighting factors from the review team was designed to minimize the potential problems associated with individualinterpretation. The SARRP project team issued to each individual a package that described the background, reference material, and role of the issue in the analysis, and suggested thu levels to represent the range of outcomes. A meeting was then scheduled that was attended by all reviewers. At the meeting the SARRP team provided some explanation of the issue to ensure that all individuals were considering the same scope of problem and a common issue definition. The reviewers were then able to discuss their preliminary opinions and to question each other as well as the SARRP analysts. This interactive process, which is documented in the appendices to this report, helped to identify the principal physical processes contributing to the uncertainty in the issues and to clarify the individual expert's views. There are both advantages and pitfalls to this interactive solicitation of input which are discussed at the end of this section, but schedule constraints did not allow a more rigorous approach to be undertaken. At these meetings, the reviewers provided their weighting factors as well as any adjustments to issue and level definitions (in some cases the reviewers provided some final values after the meeting). 2-20
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) After collection of all of the input, the groups' weighting factors were combined into composites to represent the issue in the LLH sample. The process of creating composites to represent the group depended on the type ofinput for the issue. Some of the issues involved specification of discrete distributions to approximate a continuous variable (e.g., pressure rise from direct heating). For these issues, each reviewer gave his view of the distribution by providing weighting factors on parameter values. The composite represented a distribution comprised of the average of the individual responses. Other issues required probabilistic input for a branch point (e.g., probability of a hydrogen detonation for a given sequence), and each reviewer was asked to provide a single probability estimate. For the latter type of issue, the probability space was discretized into two outcomes (zero and one), and the reviewers' responses were averaged to obtain composite weighting factors for the two outcomes. 4 As illustrated in Figure 2-3, this process resulted in the reduction of the hundreds of individual uncertainties with wide ranges of possible outcomes into a limited number of weighted outcomes for the whole uncertainty analysis (about four levels each for 29 issues). The key to the LLH method is that the individual issue levels are constrained by the weighting factors; the fraction of sample members having a given outcome for a given issue is equal to the weighting factor. For example, if the weighting factor for the largest quantity of core debris participating in direct heating is 0.1, then this outcome will be in 10% of the sample members. In addition, the method allowed for correlations between issues or issue levels. No intentional correlations between issues were introduced for the Peach Bottom analysis. The combination of issue outcomes for the different issues is not constrained; it is a random sampling. l Accidental correlations that can occur in any random sampling were limited by performing the random sampling several times and selecting the input with the least accidental correlation as measured by statistical tests. The sample for Peach Bottom consisted of 150 sample members. Each sample member has one of the outcome levels for each of the issues, with all other inputs not associated with the LLH issues maintained at a central, or best-estimate value determined by the SARRP analysts. A typical result is illustrated in Figure 2-4. The risk for each the sample members is calculated and plotted. The specific combinations of issue outcomes associated with each sample member are also available to enhance understanding of particular sample members. For display and discussion purposes, the plot of the results includes a bar that illustrates the 5th and 95th percentile of the sample. It must be recognized that this bar does not have statistical meaning concerning the uncertainty in risk; it merely applies to the distribution of the sample members. The results of this task are limited to a representation of the uncertainty in the risk 2-21
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) measures attributable to the limited number of specific issues selected, and are further limited to the input of those providing the weighting factors. It is only an approximation to the uncertainty in risk, and a different group of experts could well have established weighting factors with different results. It is hoped that the selection of the review group allowed for broad representation of the industry, but it is known (post-completion) that the review group did not provide ranges for some issues as wide as those that would be indicated by literature on the subject. Nevertheless, it was the intent to provide the NRC with an analysis of reasonable uncertainty in the results; the degree of success will be judged by review of the input and results by other experts. PEACH BOTTOM RISK REi J,TS-EARLY FATALITIES i (EXAMPLE) OUTCOMES / LEVELS
! 4 95tn percentile Sampl 1 e
of the sample Issue #1 #2 #3 Member ! member outcomes ,3 1 1 2 s c-
#2 2 4 1 Etc.
}
#3 1 4 3
[ 4 5th percentile #4 4 2 2 of the sample
#149 member outcoma 1
#5 2 3 3 IM:<icual Sample Members Figure 2-4. Example of LLH Output for Risk Measures Figure 2-5 illustrates another output of the LLH analysis. Rank-regression techniques were used to evaluate the relative importace of specific issues. The importance of an issue, with regard to its impact on risk uncertainty, is measured by the square of the standardized regression coefficient. This value is approximately the fraction of the variance (in the risk ranking of the sample members) which can be represented by the rank-regression model that is attributable to the issue in question. Thus, what is reported herein is not the fraction of total variance which is attributable to a particular issue, but the fraction of the variance which can be modeled by rank regression. These techniques are a good supplement to detailed review of the results; however, the statistics of the sample sizes (with discretized distributions for the issues) used for SARRP do not always allow for meaningful regression analyses. These importance calculations are completed where meaningful, and any limitations associated with the method of 2-22
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) calculation are explained. In all cases, the insights generated by a thorough review of the results by the SARRP team are also included. RISK RESULTS-EARLY FATALITIES 8 RANK REGRESSION ANALYSIS
.E .8 - -
E y 6- - 7
.s E.
B .4 - - c 2 .2 - - 1 2 fS ..
~ , Is3sl h
#4 x E
#5 h#6
#1 #2 #3 lssue Figure 2-5. Example of LLH Rank Regression Analysis -
l For the Peach Bottom study, more detailed statistical analyses were performed than for I the other SARRP plants. The rank-regression analyses performed by the RISQUE code were - supplemented by a more accurate multivariate rank regression that allowed for a more robust determination of the effect of the individual issues on the overall uncertainty. Both sets of rank-regression analyses are discussed in the results and provided in fullin the appendices. In addition, the more detailed analysis allowed the creation of a rank-regression model that could
)
account for approximately 90% of the variance, an improvement over the RISQUE code rank regression which generally can modelless than 50% of the total variance. A comparison of the two methods concluded that the rank-regression output from the RISQUE code is appropriate for a determination of the relative importance of the issues, but less well suited for quantifying ) the degree to which one' issue is more or less important than another. The more detailed rank-regression analysis produces results that do allow a quantitative comparison. The output accurately lists the fraction of the total variance (that is modeled by the rank regression) that can be uniquely attributed to each individual issue. Neither method of rank regression allowed for , an assessment of the importance of issues acting together to produce important effects, although insights concerning some combinations ofissues were developed through sensitivity studies. 2-23 1
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMME!fr (FEBRUARY,1987) Additional statistical tests are also included. The X2 test for " goodness of fit" was performed to help determine the probability of deviation from expected outcomes. Unfortunately, this test requires an expectation of five members in each category, and with the number ofissues and levels in the Peach Bottom sample, only two categories could be defined: below the median risk and at or above the median risk. Thus, this test yields the probability that any grouping of sample members with specific issue outcomes either below or above the median has occurred as a result of chance alone. This test is performed for each issue level and for each issue as a whole, although the latter would rarely be expected to be significant since it would require all of the levels to be grouped either below or above the median (the middle levels for the issue are very likely near the middle of the outcomes by the way the problem was set up). The x 2test can be used to corroborate other tests ofissue significance. The final statistical test of the data included in the analysis package is a STEP multiple l regression which fits all issues to risk (rather than rank of risk) [20]. The poorest fitting issues are dropped, one at a time, until a specified significance level is reached. The issues remaining give the best fit to the risk data within that significance. This test sometimes helps to further , 1 identify the issues driving the results for a particular risk measure, but the setup of the LLH is l l not particularly well suited to this regression analysis, and the results are only useful as a diagnostic tool. In the course of implementing the LLH analysis, a number of observations were made that provide additional perspective in attempting to understand the process and the results obtained. First, one of the factors that led to the selection of the LLH analysis was the desire to pursue an approach that would culminate in the representation of uncertainty in the risk results by bounds representing " reasonable" uncertainty. It was hoped that the analysts participating could say that they had roughly 90% confidence that the correct result would lie in the range calculated. This was viewed as a weakness of the OCP approach, since it was not possible to interpret the optimistic and pe:,simistic values as the bounds of any particular confidence interval. However, because the LLH uncertainty analysis is limited (in that only a relatively small number of areas assumed to be most important were considered) and because it relies completely on the assignment of subjective weights for the individual issues, it is also inappropriate to interpret the LLH results in the context of a probability distribution. It would seem natural to interpret the number of sample members having calculated risks within a certain range as being an indication of the probability that the actual risk is within that range. However, the analysts' opinion is that these results do not support such interpretation. Instead, the LLH results can be considered to represent a range within which the "true" value 2-24 1 L ___ ___ - _ _ _ _ _ - - _ _ _ _ - _ _ - _ _ _ _ - _
NUREG/CR 4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l of risk could reasonably be expected to lie, with no statistical interpretation of" reasonable." ] This statement, however, is subject to the caveat that the analysis was limited in scope; for example, external initiating events and operator errors of commission (as distinguished from errors of omission) were not considered.* Subject to these limitations (which are described 1 further in Section 5), it is hoped that the most significant uncertainty issues were treated, and that the limited number of issues results in a range that is not significantly narrower than that which would result from consideration of mom issues. l l It was the initial intent that the selection of the review-group members provide for representation of the broad spectrum of views within the nuclear-safety community, but m i practice the review groups were comprised primarily of persons engaged in NRC-sponsored research. In particular, the IDCOR program and other industry viewpoints were not well represented. A preferable approach, had it been practical within the constraints of the program, would have been to compile a large list of qualified reviewers, and to assemble the review group based on a random selection from that list. As in the previous discussion, one aspect of the makeup of the groups that provided input on the LLH issues would be a tendency for the ranges to be narrower. However, because the levels for specific issues tended to cover a broad range,it is reasonable to assume that a different composition of the review group would result I in a shift in weighting factors, but not necessarily large changes in the levels considered. This would imply that the range of LLH outcomes might not be appreciably changed, particularly given the methodology used to calculate composites (i.e., arithmetic averaging). The distribution of outcomes might, however, change considerably. Other methods of manipulating input from a wider group of experts might have resulted in somewhat different uncertainty presentations, as demonstrated by some of the OCP results presented in the report for Suny [1]. Furthermore, although the interactions among the group members have been identified as a positive attribute of the process due to the resultant sharing of information and the assurance that all of the reviewers had an equivalent understanding of the issue they were being asked to evaluate, these interactions also raised a concern. Frequently, the interactions appeared to lead to a shift in the positions of some of the reviewers toward consensus. This l tendency did not impact the range of levels considered for an issue. It does, however, cast funher doubt on probabilistic interpretation of the weighting factors assigned. l ' Consideration of external events is currently planned as a follow-on activity. 2-25
NUREG/CR 4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) j The application of the LLH method to each area of the analysis and the types of results I displayed are described below. Section 4 of the main report discusses the actual issues considered in the analysis of Peach Bottom. Core-Damace Freauency - - The uncertainty analysis for the core-damage sequences was completed by the systems l analysts in consultation with other reviewers, and the development of the input did not involve l expert-review meetings equivalent to those held for the containment and source-term areas. j Two quantitative measures of uncertainty were calculated: (1) sensitivity of the results to important phenomenological, data collection or interpretation, or modeling uncertainties, and (2) uncertainty due to data variability. The latter was performed independent of the LLH study and is provided directly for the results of the analysis of core-damage frequency. There is, however, some overlap with the LLH study, since the LLH issues dealt with offsite power non-recovery data and human reliability estimates for certain operator tasks. The outcomes of the sensitivity studies were used in the LLH approach. From the seven sensitivity studies calculated by ASEP, three issues were defm' ed. The sensitivity studies defined the upward and downward sensitivity to the issues relative to the central estimate. The ASEP analysts also provided weighting factors or a probability distribution for the outcomes associated with the sensitivity studies, which were then used directly in the LLH input. Annivsis of Containment Resnonse The CET provided the framework for quantification of the likelihood of variout. ! containment failure modes. The branch points corresponding to the LLH issue outcomes selected for inclusion were structured in terms of the level definitions representing the range of outcomes. The dependencies in the CET were also checked to ensure that the correlations provided by the review group were appropriately modeled. All other branch points were quantified using the central estimate of the CET study [M]. The actual issues are defined in Section 4, and the experts providing the input were listed previously in Table 2-1. Unlike the other SARRP plants, there was considerable overlap between the core-damage frequency issues and the containment-response issues. For instance, the probability of j non-recovery of offsite power was sampled consistently between the cut-set expression and the ! branch points in the CET. In general, great care was taken to assure this kind of consistency across the interfaces between modules in the analysis. l l 2-26
NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1987) The results of the LLH study were also decomposed to allow examination of the uncertainties relative to measures other than risk. The containment parameter of most interest ! with respect to risk is the probability and timing of containment failure given a core-damage accident. Therefore the type of display represented in Figure 2-4 for risk measures was also created for lower-level results, e.g., the probability of early containment failure over all sequences, and for selected individual sequences where useful. The results of other studies are also included for comparison. It should be noted, however, that the issues in the LLH study were selected for their impact on risk, rather than on the probability of early containment failure. Thus, the uncertainty in the probability of early containment failure may not represent a reasonable range. l Source-Term Uncertainties The central estimates of the radiological source terms were calculated by RELTRAC using modeling assumptions (parameter values) consistent with the STCP. The source-term l issues were incorporated into the RELTRAC model in general accordance with the prescriptions agreed to by the review team. RELTRAC exercised these models for the issues in accordance with the levels for each sample member specified by the LLH sample. Because ! the combinations of LLH issue outcomes result in a very large number of different source j terms, some means was required to reduce the the number of explicit source terms in order to limit the time-consuming calculation of consequences for individual source-term combinations. A method was developed to cluster the source terms, resulting in a more limited number of j clusters which could represent all of the source terms without introducing significant inaccuracy. The method for developing the clusters is described in Appendix B. As described previously, each cluster has a unique set of release fractions (and hence consequences) that are defined based on numerical comparison to the full range of outcomes. The clusters are primarily a calculational convenience and can only be related back to specific outcomes of the containment analysis for a particular LLH sample member. Insights concerning the role of source-term phenomena in risk and risk uncertainty can be derived from careful examination of the detailed risk results. These insights are presented in Section 5.3. Uncertainty in the Consequence Annivsis No LLH issues were defined for the offsite-consequence analysis. Stochastic uncertainty associated with plant-specific meteorology is reflected in the complementary cumulative distribution functions (CCDFs). The comparison of CRAC2 and MACCS results 2-27
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) also allows some consideration of modeling differences which provide additional insight, particularly with respect to differences in the modeling.of health effects from radiation doses. The mean risk displays illustrated in Figure 2-4 are supplemented by the CCDFs as illustrated in Figure 2-6. These CCDFs illustrate the probability per year (on the ordinate) that H a given magnitude of consequence (on the abscissa) will be exceeded. As with the displays of mean risk, the results for each sample member are plotted, as are curves representing the 5th and 95th percentile of thEsample. The CCDFs are displays of the weighted average (point-by-point) of 53 CCDFs (one for each source-term cluster). Because CRAC2 does not provide CCDFs for the evacuation assumptions used in the SARRP analysis, all CCDFs in this report correspond to MACCS results. 3 COMPLEMENTARY CUMULATIVE DISTRIBUTION FUNCTION o (EXAMPLE) 8 l h
- 4 -- Individual Sample Members
- g g o a 8 8 : : : o 8 8 i i i
- l Yi : : :
o i i i ! I* D o j : lg 495th Percentile of 5 . j the Sample Outcomes 8 s a f 8 o 3 Sth PIrcentile of l the Sample Outcomes I CONSEQUENCE SEVERITY Figure 2-6. Example of CCDF Risk Display Presentation of LLH Risk Results The RISQUE code includes provisions for calculating a great deal of information concerning the risk results. Due to sheer volume, not all of the results are printed, or even calculated; rather, the risk results can be investigated through careful selection of more detailed calculations to support particular uses. The most important measures and breakdowns of the risk results are summarized in Section 5 and provided in detail in Appendix C. The full complement of available output for any given risk measure includes the following:
- Risk conditional on core melt, '
2-28
NUREG/CR-4551, VOL. 3: DR AFT REPORT FOR COMMENT (FEBRUARY,1987)
- Risk for any given individual or set of plant-damage states (groups of core-damage cut sets), i
+ Fractional contribution of each plant-damage state to mean risk, l
- Fractional contribution of each source-term bin to mean risk, I
- Rank regression for allI1H issue outcomes,
. Step multiple regression for all LLH issue outcomes,
- x 2test of uniformity of risk for all issue outcomes, and )
- Detailed analysis of individual sample members, including fractional j contributions of damage states, source-term bins, and clusters. j j
The output for Peach Bottom includes only a subset of this information; it was necessary for the analysts to select the parameters they felt would provide the information I necessary to understand the important risk issues, without generating an unmanageable volume i of output. 1 I 2.4 REASSESSMENT OF DOMINANT CORE-DAMAGE SEQUENCES The process of identifying and quantifying the core-damage sequences involved , performing an update by the Accident Sequence Evaluation Pmgram (ASEP) of the original I analysis performed for the Reactor Safety Study (RSS) [21]. In order to obtain maximum I benefit from the limited resources available for accomplishing the update, detailed analysis was performed only for areas that appeared to be particularly important to risk; the RSS and other relevant analyses were relied upon otherwise [4]. The ASEP analysis began with a review of the existing PRA and a visit by the analysis team to Peach Bottom. This plant visit was valuable in developing in a relatively short timt information important to a number of areas. Following the plant visit, new system event trees were constructed to delineate the core-damage sequences. Because of the extensive interrelationship between the systems needed to maintain core cooling, conditions in the containment, and systems providing containment-protection functions, close interactions between the containment and system analysts was a critical part of the event-tree development. These interactions led to the development of a set of plant-damage states reflecting explicit consideration of the sequence conditions that could produce unique effects on the likelihood of various containment-failure modes. The characteristics of the damage states include the following: 2-29
1 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRU ARYl1987) 1
- The status of the pressure boundary of the nuclear steam supply system, ;
including the size of any breach and the pressure in the rea:: tor pressure 1 vessel; -{
. The status of the containment pressure boundary, including the size of any induced leakage; ;
. The level of heat addition to the containment;
. The timing of the success or failure of various core-cooling functions;
. Whether or not the core is vulnerable to melting due to the status of the :
containment; and l l
. The status of the residual-heat removal system. l l
System fault-tree models were then developed for ecch of the events in the core-damage event trees. The fault trees were constructed only to the level of detail necessary to account properly for dependencies among different systems (for example on electric power or other l supporting services) and to provide for the efficient application of available reliability data. Together with the event trees, these fault trees provided an integrated model of plant behavior. The data base used to quantify the events in the system fault trees was developed both from plant-specific data and from industry-wide experience. The ASEP generic data base [22) was derived primarily from that originally produced by the Interim Reliability Evaluation Program (IREP) [2B, with data from other studies also incorporated. Common-cause failures were quantified based on information reported in an assessment of industry experience performed by the Electric Power Research Institute [24]. The assessment of human reliability, , including the estimation of the likelihood of various actions taken to recover lost safety functions during an accident, was performed using a combination of available methods and engineeringjudgment on the part of the systems analysts and human reliability specialists. l The models and data were combined to quantify the sequences by obtaining sequence-level minimal cut sets. These cut sets are minimum combinations of initiating events (i.e., various LOCAs or transients, such as loss of the power conversion system) and hardware faults or human actions leading to failure of the systems needed to prevent core damage. At < this level of detail, the sequence results can be reviewed carefully for the specific aspects of 1 plant operation that contribute most to core-damage frequency and risk, which in turn allows ! I for the definition of potential safety improvements that are most likely to be cost effective. The i final assessment of accident-sequence cut sets included the consideration of appropriate recovery actions that would likely be attempted by the plant staff. Thus, the final sequence cut sets are intended to represent a realistic estimate of plant accident sequences and their frequencies. 2-30
1 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) j 2.5 EVALUATION OF CONTAINMENT RESPONSE . l A major focus in SARRP was the development of a containment-event tree (CET) for each of the plant types represented by the reference plants. These event trees delineated the pathways that could lead to releases of fission products as a consequence of core-damage accidents. The approach taken in the development and application of the CET for Peach Bottom is summarized in this section. The results of the containment-event analysis are ) presented in Section 5, and additional detail is provided in Appendix A. The full analysis is 1 documented in a separate report [M]. 1 2.5.1 Develcoment and Ounntification of the Containment-Event Tree l l The CET for Peach Bottom was developed by identifying the types of containment ) responses that might be expected to impact risk and the various events and conditions that j could affect those responses, at a level of detail that could reasonably be supported by the l information currently available. This led to the construction of an event tree that is significantly expanded beyond those previously used in PRAs; the CET for Peach Bottom contains 107 questions, many of which have more than two branching options. Because of the number of possible branching points, the event tree is too large to depict graphically. The questions are listed in Table 2-3. The event tree is comprised of the following types of questions: (1) Entry states. What are the conditions in the reactor pressure vessel and containment prior to melting of the core that could influence the accident progression? These are provided primarily by the definitions of the plant-damage states, as outlined in Section 2.4. (2) Phenomenological events. What are the phenomena that could affect the progression of severe accidents, at what points during the accident timeline do they occur, and what are their subsequent effects on the accident progression? (3) Considerations pertaining to breach of reactor vessel. What is the condition of the reactor core at the time of breach (if any) of the reactor pressure vessel and what is the pressure? (4) Survivability of containment systems. Does the pressure-suppression system survive the conditions occurring during severe accidents which exceed its design bases? (5) Containment-failure modes. What are the loads that challenge containment, does containment survive these loads, what is the nature of the failure (approximate size and location), and what is the subsequent pathway for release of fission products to the environment? The questions are posed in ways that require the answers to be expressed in terms of likelihoods. For a station-blackout accident, for example, the likdihood that core degradation 2-31 L_ -_-_
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 2-3
SUMMARY
OF CET EVENTS FOR PEACH BOTTOM i Dependencies on l Question Posed by the Event Tree Prior Questions !
- 1. What is the initiating event? None
- 2. What is the initial break location? None
- 3. What is the level of pre-existing leakage or isolation failure None
- 4. Is there a loss of offsite power? None
- 5. Is there a station blackout (diesel generators fail)? None
- 6. Is de power not available? None
- 7. For TC, does SLC fail to inject? None
- 8. Does a S/RV stick open early? None
- 9. Do the HPCI and RCIC systems fall to inject? None
- 10. Does the CRD hydraulic system inject? Nonc
!!. Do the LPCS and LPCI systems fail? None
- 12. Do the RHR systems fail? None
- 13. Does the condensate system fail? None
- 14. Does HPSW fail in a mode which would preclude injection? None ;
- 15. Is there a failure which precludes drywell cooler operation? None
- 16. Is ADS blocked or not called upon? None
- 17. Does the RPV remain pressurized? 1,5,6,8,9, 10, 11, 12, 16
- 18. For TC, is inadequate level maintained with LPl? 1, 7, 16, 17
- 19. What type of sequence is this (summary of plant damage)? 1, 5, 9, 10, 11, 12, 14, 17, 18
- 20. Is there an LP system break induced by power cycling (ATWS)? 1, 16, 19
- 21. Do low-pressure ECC systems continue /begin to inject? 11, 17, 19
- 22. Do containment sprays fail before core degradation? 1, 11, 12, 14, 19, 21
- 23. Is the containment vented before core degradation? 1, 19, 20
- 24. Does the RPV repressurize before core damage? 8, 17, 19, 20, 23 ;
- 25. Does (do) any S/RV-tailpipe vacuum breaker (s) stick open? 8, 17, 19, 24
- 26. Is there a repressurization break of low-pressure systems? 24
- 27. Does containment fail before core degradation? 19, 20, 23, 26
- 28. What is location of early containment leakage? 19, 20, 23, 26, 27
- 29. For drywell failures, is the failure at the head seal?
{ 19, 20, 26, 28
- 30. What is the containment leakage level before core degradation? 1, 19, 20, 23, 26, 27,28,29
- 31. Is the suppression pool drained before core degradation? 27, 28, 30 i
2-32
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 2-3 (Continued) i
SUMMARY
OF CET EVENTS FOR PEACH BOTf0M Dependencies on Question Posed by the Event Tree Prior Questions
- 32. Does ac power remain lost during core degradation? 5, 6, 19, 24
- 33. What is the status of CRD? 5, 10, 23,'30, 32
- 34. What is the status of HPSW7 5, 14, 23, 30, 32 j
- 35. Does the core melt? 13,'17, 19, 24, 26 30, 33, 34
- 36. - What is containment pressure before core degradation? 1, 9, 12, 19, 20 4 22, 26, 30
- 37. Will the suppression pool flash following containment vent or rupture? 1, 8, 9, 12, 19,21,24
- 38. Is the RPV depressurized during core degradation? 5, 17, 19, 20, 21, 23, 24, 26, 30, 32, 37
- 39. Is there injection during core degradation? 13, 21, 30, 31, 32 33, 34, 37, 38
- 40. What is the status of containment sprays? 21, 22, 30, 31, 32, 37
- 41. What is the level of flow to the drywell during core degradation? 1, 20, 25, 26, 38,39,41
- 42. What is the hydrogen generation and pressure rise during core degradation? -1, 17, 19, 20, 23, 24, 26, 30, 33, 38, 39, 41
- 43. Does at least one drywell vacuum breaker stick open? 1, 17, 19, 20, 24 25, 26, 28, 42
- 44. Does hydrogen discharge lead to suppression pool bypass? ' 17, 24, 42
- 45. Does hydrogen discharge to suppression pool induce wetwell failure? 17, 24, 44
- 46. - Is the vent threshold reached during core degradation? 45 -
- 47. Does containment venting not occur during core degradation? 5, 12, 19, 23, 30, 32,40,42,46
- 48. Does the RPV repressurize during core degradation? 1, 8, 38, 45, 47 !
- 49. Is there a repressurization break during core degradation? 48
- 50. Does the containment fail by pressure during core degradation? 19,29,30,45,47,49
- 51. What is location of containment leakage prior to vessel breach? 28, 30, 45, 47, 49,50
- 52. For drywe!! failures, is the failure at the head seal? 29, 30, 47, 51
- 53. What is the level of containment leakage before vessel breach? 19, 29, 30, 45, 47, 49,50,51,52
- 54. What is the status of CRD7 5, 23, 32, 33, 50, 53 f 5. What is the status of HPSW7 5, 13, 21, 23, 30, 31, 34, 37, 39, 47, 53-
- 56. Is MSIV leakage control or condenser vacuum maintained? 32 2-33
NUREG/CR-4551, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) Table 2-3 (Continued)
SUMMARY
OF CET EVENTS FOR PEACH BOITOM Dependencies on Question Posed by the Event 'lree Prior Questions
- 57. Is the suppression pool drained before vessel breach? 27, 28, 30, 31, 45, 50,51,53
- 58. What is the level of suppression peo! bypass at vessel breach? 20, 26, 43, 44, 49, 51,53,57
- 59. What is level of RB breach / bypass before vessel breach without burn? 5, 20, 23, 26, 32, 47, 49, 52, 53
- 60. Are the fire sprays actuated before vessel breach without burn? 37, 53, 58, 59
- 61. Does SGTS fail before vessel breach without burn? 23, 32, 47, 60
- 62. Does hydrogen burn in RB before vessel breach? 20, 26, 42, 49, 53,59,61
- 63. What is level of RB breach / bypass before vessel breach? $2,59,62
- 64. Are fire systems actuated before vessel breach? 52,60,62 j l
- 65. Does SGTS operate until vessel breach? 61, 62, 64
- 66. What is containment pressure after containment failure or vent *during meltdown? 53
- 67. Is there injection to the RPV up to the time of vessel breach? 13, 21, 30, 31, 37, 39 48, 53, 54, 55, 57
- 68. What is the nature of the meltdown? 35, 67
- 69. What is the mode of vessel breach? 19, 35, 67, 68
- 70. Is the RPV blown down before significant melt ejection? 38, 48, 53, 68, 69
- 71. Does direct heating occur? 68, 70
- 72. Does an ex-vessel steam explosion occur? 68, 69
- 73. What is pressure rise due to blowdown (effective vapor suppression and reactor 38, 71 cavity dry)7
- 74. What is the pressure rise due to debris quenching? 1, 22, 40, 67
- 75. Is the suppression pool bypassed or are vents not cleared at vessel breach 37, 38, 53, 58, 71, (effective vapor suppression) 72, 74
- 76. Does pressurization fail containment at vessel breach? 45, 47, 50, 51, 53, 69, 75
- 77. What is the base pressure before vessel breach? 45, 47, 50, 53, 69
- 78. What is the pressure rise at vessel breach? 45, 47, 50, 51, 53,69,75
- 79. Does direct melt-structure attack fail containment at vessel breach? 68, 69, 70, 74
- 80. What is the location of containment failure after vessel breach? 51, 69, 75, 76, 79
- 81. For drywell failures, is the failure at the head scal? 52,79,80
- 82. What is the containment leakage level after vessel breach? 53, 69, 75, 76 79,80,81
- 83. Is the suppression pool drained following vessel breach? 57,76,80 2-34
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 2-3 (Continued)
SUMMARY
OF CET EVENTS FOR PEACH BOT'IOM Dependencies on Question Posed by the Event Tree Prior Questions
- 84. Is ac power not available late? 6, 19, 24, 32
- 85. Is residual heat removal not operating late? 12, 32, 37, 69, 82, 83, 84 g
- 86. Do containment sprays not operate following vessel breach? 22, 37, 40, 82, 83, 84
- 87. Is service water sprayed following vessel breach? 22, 55, 84, 86 '*
- 88. Is water not supplied to the debris late? 13, 21, 32, 37, 38, 53, -
54, 55, 57, 67, 69, 82, 83, 84, 86, 87 =
- 89. Is the debris not coolable without water? 69, 71, 72
- 90. Is the debris coolable with water? 37, 55, 67, 68, 69, 71, 72, 82, 83, 89 .
- 91. Is the debris not cooled? 69, 88, 89, 90
- 92. Is the containment not vented late? 32, 38, 46, 48, 69, 82, 84, 85, 91
- 93. Does the containment fail late due to temperature? 69, 80, 82, 88, 89, 92
- 94. Does the containment fail late? 69, 82, 85, 88, 91, 92, 93
- 95. What is the location of late containment leakage? 58, 80, 82, 88, 92, 93, 94 -
- 96. For drywell failures, is the failure at the head seal? 81, 88, 94, 95 97, What is the level of late containment leakage? 82, 88, 92, 94, 95, 96 98, is the suppression pool drained late? 83, 94, 95, 97
- 99. What is the level of late suppression pool bypass? 20, 25, 26, 38, 41, 49, 58, 69, 95, 98 100. Do drywell sprays continue? 37, 55, 86, 87, 94, 98 101. What is the level of late RB bypass without a burn? 59, 63, 69, 79, 96, 97 102. Are fire systems operating late without a late hydrogen burn? 37, 64, 96, 97, 99, 101 103. Does standby gas treatment not work late without late hydrogen burn? 65, 84, 102 104. Does hydrogen burn in the RB after vessel breach? 91, 97, 101, 103 '
105. What is the level of late RB bypass? 97, 101, 103, 104 106. Are fire systems operating late? 96, 102, 104 107. Does standby gas treatment work late? 103, 106 2-35
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) occurs under high- or low-pressure conditions in the reactor pressure vessel is considered. This has an effect on the energy released upon breach of the reactor pressure vessel, which in turn could affect the mode of containment failure. Answers to such questions require information about the reactor design, plant-operating procedures, the accident progression, the phenomenology of reactor accidents, and the capabilities of containment. It should be noted
} that one difference between the CETs for Peach Bottom and Grand Gulf and those for Surry and Sequoyah is that, for the BWRs, the potential for arresting core damage prior to breach of the reactor vessel was considered.
The estimation of these likelihoods is often a highly uncertain process, and requires that information from a large number of sources be examined and incorporated in order to reflect the current level of understanding most appropriately. Table 2-4 lists many of the key issues represented by the questions in the event tree, and identifies some of the major research and I analysis programs from which information was drawn to evaluate the corresponding questions. Among the sources ofinformation used in the analysis are the following:
, (1) Containment Loads Working Group (CLWG)[25,26];
(2) Battelle calculations for the Accident Source Term Project Office, and subsequent calculations in support of SARRP [H,22]; (4) Quantitative Uncertainty Estimation for the Source Term (QUEST) program [21]; (5) Industry Degraded Core (IDCOR) program [22]; (6) Severe Accident Sequence Analysis (SASA) program [3D-32]; (7) Severe Accident Uncertainty Analysis (SAUNA) Program [23]; (8) Accident Sequence Evaluation Program (ASEP) [4]; (9) Steam Explosion Review Group [3A]; (10) High-pressure ejection test series [25]; (11) Available probabilistic risk assessments [21,36]; (12) The Final Safety Analysis Report [32]; (13) Architect-engineer and other estimates of containment failure pressure [31]; and ) (14) Others [32,4Q]. The S ARRP analysts, in consultation with NRC, developed a set of guidelines with respect tc the priorities to place upon the use of this variety of sources in evaluating the events in the containment event trees. To the extent possible, STCP [3] calculations form the basis of i the evaluations. The information developed by the CLWG and CPWG was also considered. Where additionai mformation was required, other analyses performed or sponsored by NRC 2-36
I! II AgmMC>Wd G" .- i [ytOdNkt$ 4Op n gy>3' WM8xH VOW GO55 _ i i r s te t X X X X X X - - X - X - - - O _ E - - - - - - - - X - - - - - A A _ R X - - - X X X - X - X - X - _ P _ S E P I - - - - - - X X - - - - - - E I I R 1 . W F G _ E R - - - - X - - - - - - - - - V E S E _ T _ _ N _ _ E F _ M T - - - - - - X - X - - - - - _ _ A _ _ N I sP e A c r T u N SoA O nN - - X X - - X - C iUo - X X - - - _ A _ E ta S _ I m r 1 1 f o n N I ISAA S - - - - - X X - - - - - X - E S U S S I R U 41E C D X - X - X X X X X X X - X - 21 1 I le G b aN T TI S S S E - - - - - - X - - - - - - - E U R Q D D A I N M - - X - - X X - - - - - X - I B S E C _ R G ". _ U W - - - - X O P X - - - - - - X - -
"n S C a y
N b O I G = T W - - - - - - A L
- X X X - - X X -
M R C e r a FO e u s I N i s e r F s ht a e e s l O r e L 1r g g u u k e r r n e o s u c E a e n a l a i s s e i r o i t r in n io S e
. i a a r n m r r y v ir no p U g b a a i o u e r
r ip e b t a a v a t m t is s o b u w k g a c u u al op n s s s o s s de s y t a . c a n I c c xL lp x e g e tm e r c u a r l o mt a i t a p. r t a m e c e o p d p o i v e r a s n s p t c s u e v mss m- d t nh e r ; a i l a ip o ae a y ec a h u ieg no r b r v e t tev e h me t l te e e a . i lp f u ; i sr sr ts f o nr o . f a k _ nt s ist m d a s a a o l o et c le' e ib a g hc l ae e e en e e r n . d le ur a ;ad s* t l . V ta s s a s d u tl n ae r si rl ya t e oe e oc es s u t s r r s ee e< n s c gc g R e p v r v- ce i t ne iwle b r t c ee roc uc c n tcm cc t u mp f eca n / - e n . ev u r sa o s its S o ix h t a t n inh fl t inh ftl gg un hth t tcs es r ts inr a o oh <er ieu f e ds r r t e c r a oi a oi a ni fi ow ohtg ty tyt u
< t lp cf ef dl i d ir e dv re d i l i n g o on c f e rt o
ow aiiu bw o e l n s e p da oope f o o nu ld f g ii f n o ab e br lam iat on ct a t t f o ho ty .h
. m.g h gr tniac f r oe ivd ivd n a
t i. mi oi
.g
.u no mwi ta s a lekc h c .u it go eo vr e vr es e et c o c c o nc n mkii immp um um i
iS b Lts A Oe 0e nTr eMa n ua ass txll Ef o u e tx to o m r o r S. t S. i t S. t i E. P. c f n
- 9. 0 1 2 3 4 1 2 3 4 5 6 7 8 1 1 1 1 1
- I ll !lll
r-3. NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) were used. These sources were not sufficient to provide full treatment of some issues raised in the CET. In some instances, calculations were performed by SARRP pensortnel or other NRC 4 contractors specifically in support of the CET quantification. Other information sources, including the opinions of the SARRP expert review group and analyses sponsored by the 1 utility industry, were used. Because the information provided by these sources often either did not70 vide a direct means to quantify particular branch points, or provided divergent indications of the appropriate probabilities, the initial quantification of the Peach Bottom CET was performed ,ujing the OCP methodology outlined for Surry in Section 2.3 [1]. That is, three separate estimates of the probabilities for each question were propagated through the tree. The one labeled " pessimistic" tended to provide higher probabilities for the pathways that lead to higher source terms and lower probabilities for lower source-term pathways. The ones labekd " central" and
" optimistic" can be interpreted in an analogous manner. For the reasons cited earlier, the 4 optimistic and pessimistic walkthroughs were carried out only for the CET analysis; the central ,
values were retained for the LLH analysis. l In developing the central estimates for the event tree, the branch points were often { assessed first in subjective, qualitative terms (e.g., an event might be "likely" or " remotely possible"). Values were then assigned to these verbal descriptors. The semiitivity of the release-mode probabilities to the assignment of alternative quantitative values for>the verbal descriptors was examined, and it was found that the variances were generally srohil rdanw to . the uncertainties in the phenomena themselves, for example, as indicated by thd differences among the results for the three walkthroughs. - 2.5.2 Plant Features Imoortant to Containment Resnonse The branch points and probabilities for the Peach Bottom containment event tree reflected consideration of a number of plant-specific features that could have important effects on the progression of a core-damage accident. The most important feature of the Peach Bottom Mark I containment relative to its ' response to severe accidents is its small volume compared to PWR and Mark III BWR containments. The total free volume (drywell and wetwell airspace) is almost one-tenth of that for a PWR large dry containment. While the ultimate pressare capability is larger than other power reactor containments, its capacity to hold non-condensible gases is between one-third and one-quarter of that for the other designs (excluding the BWR Mark II, which is similar to the Mark Iin this respect). The capacity of the containment to retain non-condensibb gases has 2-38
~ NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) a la ge impact on its ability to retain its integrity during sevem accidents. The concentration of oxygen within the containment is maintained below 5% during power operation by nitrogen inening (this is true for all domestic BWRs with Mark I and II containments.) This inerting precludes the combustion of hydrogen produced during severe accidents from being a threat to the integrity of the primary containment, as it may be for larger, non-inert, containments A second, related, aspect of the small containment design is the relative proximity of the containment pressure boundary to the reactor vessel. At Peach Bottom, the point at which the drywell shell contacts the drywell floor is a little over 3 m (about 11 ft) from the centerline of me reactor pressure vessel (RPV). This geometry gives rise to the possibility that core debris may spread over the drywell floor (following meltthrough of the RPV) and contact the drywell shell [H]. This mechanism for loss of containment integrity may have a significant impact on the severe accident performance of Mark I containments. The pressure suppression pool is a key aspect of most BWR containments. For loss-of-coolant accidents (LOCAs), the pool provides a pressure-suppression function (through steam condensation) which allows for the low-volume containment design. Under severe accident conditions, the pool has a demonstrated ability to retain fission product aerosols i [M,M]. However, the most recent analysis of the structural capability of Mark I containments [M] suggests that the most likely location for overpressure failure is in the drywell. Since I drywell failure results in bypass of the suppmssion pool, the retention capability of the pool would then be moot. The reactor building surrounds the Mark I containment. While the volume of the reactor building is large (approximately five times that of the containment), its pressure capability is low (2 psig). Blowout panels are provided to protect the building structure from moderate pressurization rates. The reactor building is equipped with a standby gas treatment i system (SGTS) which is designed to filter fission products (primarily iodine) from the building atmosphere. However , this system would most likely be inoperable during severe-accident conditions at Peach Bottom because it would be isolated by temperature-sensitive dampers. Even so, analyses performed with the STCP [3], by IDCOR [22], and in connection with the SARRP program indicate that the building may retain a substantial fraction of the fission-( product aerosch which could be released from the primary containment. The retention is primarily a result of aerosol deposition via sedimentation, on surfaces within the building. l 1 However, hydrogen combustion within the building (which is not inerted) will reduce this deposition, or, in the case that the combustion is rapid enough to result in structural failure, may eliminate it completely. 2-39
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) j The available core cooling systems for BWRs are numerous and diverse. The Peach Bottom configuration includes two systems for injection to the RPV under high pressure (not including the feedwater system) whose pumps are driven by steam turbines and are independent of ac power. The low pressure injection systems are ac-powered. The low pressure emergency core cooling system (ECCS) consists of four pumps. Two pumps provide both injection to the RPV or containment (drywell) spray and/cr suppression pool cooling (containment heat removal). These four pumps take suction from the suppression pool. However, their net positive suction head (NPSH) requirements are such that their operation is doubtful if the suppression pool is at the saturation temperature corresponding to the containment pressure. Such conditions are a likely consequence of many severe accidents. Injection to the RPV can be provided by other systems (condensate and high pressure service water). Drywell spray can also be provided with high pmssure service water, but use of the system in this mode is not called for by current procedures. The BWR core support, which provides individual support for each group of four fuel bundles from the vessel bottom head, is judged to minimize the probability of pressurized ejection of core debris into the drywell. In comparison, the core in a PWR is supported by the core plate, which in turn is supported by the core barrel; the entim core is essentially hung from the top of the pressure vessel. In the BWR design, slumping of relatively small quantities of core debris (due to localized failure of the core-support castings or meltthrough of adjacent portions of the core plate) is anticipated to be likely to melt through a lower-head penetration in the reactor vessel, thereby depressurizing the vessel. This is expected to occur before a significant fraction of the debris which has collected in the bottom head could reheat to a molten condition. Thui., significant direct heating is less likely for BWR designs than for PWR designs. 2.5.3 Definition of Containment Release Modes The outcomes of the CET form a set of discriminated accident pathways. Since the number of such pathways was quite large, it was necessary to group them into a smaller set of bins that could be used for source-term analysis. The information reflected by these accident-progression bins includes the modes by which the containment fails, as well as other details that could influence the source term, such as whether or not core-concrete interactions sufficient to generate a vaporization release have occurred. For Peach Bottom, the containment release mod:s are discriminated by the following features: (1) Whether or not the reactor pressure vessel is breached; 2-40
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) . (2) Whether or not injection to the core is recovered prior to vessel breach (but following core damage); i (3) Whether or not the reactor pressure vessel is at high pressure when ) meltthrough occuni; l (4) The a proximate time (if any) at which water is supplied to the core debris f on the drywell floor; ] (5) Whether or not there am com-concrete interactions; j (6) The location and timing of wetwell failure (if any), including intentional ) venting; (7) The location and timing of drywell failure (if any) (8) Whether or not the drywell sprays operate, and when (or if) they fail; (9) The extent to which the suppression poolis bypassed both before and after vessel breach; (10) The extent to which the reactor building is bypassed 'up until core-concrete ! I mleases may begin; (11) The extent to which the reactor building is bypassed following the time at which core-concrete releases could begin; and (12) Whether or not direct containment heating occurred. I This degree of discrimination among containment release modes is approximately. consistent wia the level of modeling detail in RELTRAC. In practice, RELTRAC obtained all of the information required about the accident-progression pathways for which source terms were to be calculated directly from the definition of the accident-progression bin. Thus, the bin specifies the features of the pathway in slightly greater detail than is indicated by the above list of features. However, this level of detail is generally only pertinent to formulating the input to j the source-term calculation. -; The likelihood of the release modes was calculated for each of the plant-damage states l assessed by ASEP to have non-negligible frequencies. This generated the frequencies for each of the accident-progression bins, for integration into the risk calculation. ! 2.6 ASSESSMENT OF THE RADIOLOGICAL SOURCE TERM The next step in the assessment of the public risk associated with Peach Bottom was the estimation of the radiological release to the environment. The science of fission-product release j and transport has been the subject of considerable research in recent years, resulting in increased understanding and new estimates of accident consequences. The understanding is unfonunately not to the point at which definitive results agreed to by the whole of the research community are available, but the analyses summarized in this report include the advances that 2-41
I NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) have been made to date and are intended to allow an interim perspective and an improved understanding over the previous comprehensive evaluation of public risk in the Reactor Safety Study [21]. The second objective of this analysis is to act as a catalyst for the scientific interaction required to achieve the goal of developing a definitive understanding of the radiological effects of severe nuclear power plant accidents. This report summarizes the methods and results of an effort that integrated a number of research activities. The goal of this summary is to afford the reader sufficient understar. ding to review the integrated risk analysis and to allow efficient refemnce to the mom detailed msearch. More specific information on the source-term task is included in Appendix B. It was the intent of this program to estimate the plant-specific source terms using the state-of-the-art methods embodied in the Source Term Code Package (STCP) [f5]. The STCP was a coordinated effort by the NRC and its contractors to upgrade the agency's analytical procedures for, and understanding of, the radiological releases due to reactor accidents. The NRC staffs review of the STCP and their conclusions concerning its viability are presented in NUREG-0956 [15]. The NRC report is based in part on results and insights described in Radionuclides Release Under Specific LWR Accident Conditions (referred to hereafter as BMI-2104) [21]. Since the publication of BMI-2104, there has been additional work to refine various portions of the suite of codes that make up the STCP, as well as extension of the plant-specific analyses to both additional accident scenarios and additional reference plants [_46). This section includes an overview of the source-term task and the interface of this task with the other project activities. The methods used to produce the central estimate source terms as well as the source terms for the LLH uncertainty analysis are also described. 2.6.1 Integration with Other Risk-Assessment Tasks The goal of the source-term analysis was the estimation, for each unique combination l of core and containment conditions following core melt, of the amount of each radioactive species that would be released to the environment and would therefore be of concern for public health risk. Due to the large number of both potential accident conditions and different fission products, a method was needed to make the calculation tractable. This involved the grouping of results and inputs to reduced the number of actual calculations required to make reasonable source-term estimates for the full spectrum of accidents. Three principal groupings form the basis of the overall source-term methodology: 1 (1) Core and containment conditions and ultimate post core-melt conditions l were grouped into categories, called accident-progression bins. The l l 2-42 _______.___.________._______m ___m____.___ - - - - . _ _ _ _ _ _ _ _ _
I l NUP.m/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) i' scenarios assigned to a bin each produce similar response in terms of release of the radioactive species to the environment. (2) The radioactive species were grouped into classes that behave similarly. f (3) The source-term estimates for the LLH were collected into clusters so that a j manageable number of consequence calculations could be performed. i l There were many other simplifications involved in the methodology. For example, the 3 source term for each accident-progression bin was based on a ven limited number of STCP analyses. The simplifications associated with the methodology are discussed wherever appropriate in this section and in Appendix B. Accident-Pro <2ression Bins q l The first step in the assessment of radiological source terms was the development of I categories, called accident-progression bins, that could be used to group the different 1 l combinations of core and containment end states into categories that could be assessed together 1 in terms of the source term. These categories were developed for Peach Bottom by delineating l the conditions that might be determinants of the radioactive release. The characteristics I considered in the bin development included far more detail than has been considered in the source-term bins for Surry and Sequoyah [1,2] and the accident progression bins for Grand Gulf [3]. This detail allows more complete treatment of the uncertainty issues. The features of the accident progression which are specified as attributes of the accident-progression bin are dictated by the input requirements of RELTRAC. The features are as follows: The plant damage state which initiates the accident progression (determining timing variables for the RELTRAC integration); Whether the decay heat generation on the core is high or low at the onset of fuel damage and whether or not there is injection to the RPV from the control rod drive (CRD) hydraulic system. Whether or not the RPV was breached, and if it was breached, whether injection was recovered beforehand; a The RPV pressure at vessel breach; The time at which water was supplied to core debris on the dywell floor;
- Whether or not core-concrete interaction occurs; The timing and location of containment leakage (leakage assumed not to depressurize containment);
The timing of wetwell rupture (7 ft 2opening assumed) or venting; The timing of drywell rupture (7 ft 2opening assumed); The time at which dywell spray begins; 2-43
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Whether or not drywell spray continues following containment leakage or l failure;
- The flow pathway of fission products from the RPV prior to vessel breach;
+ The time at which the suppression pool is bypassed;
- The level of reactor building bypass up until the time that core debris is deposited on the drywell floor;
- The level of mactor building bypass following the time at which com debris is deposited on the drywell floor;
. Whether or not the fire protection system sprays in the mactor building are activated (only relevant for safety option calculations);
The containment pressure immediately prior to vessel breach;
- The pressure rise in the containment associated with vessel breach; Whether or not the reactor core was assumed to collapse into the RPV lower plenum prior to vessel breach; Whether or not direct containment heating occurred; and Whether or not offsite power was recovered prior to the time that the major portion of the vaporization release associated with core-conente reaction has occurred.
Thus, the accident-progression bins group the accident pathways both by containment response and timing of that response. There is some redundancy among the above features (i.e., they are not all independent), but this redundancy is required to support the inherent flexibility of RELTRAC. Each bin (combination of the features listed above) provides RELTRAC a unique set of inputs. However, depending on the levels of the source-term issues . which correspond to a particular LLH sample member, the resu;ts calculated by RELTRAC for two different bins may be similar. Some of the above features of the bins were used to estimate the heat of release based on STCP results, as described in Appendix B. Grounine of Radionuclides The second simplification was the grouping of radionuclides that have similar properties such that their release and deposition phenomena can be treated together. The definitions of the radionuclides groups are similar to those used in the RSS. These basic groups were retained for the SARRP radionuclides group definition both because they continue to provide a fairly effective means of grouping the radioactive species and because they have j served as the standard for offsite consequence analysis, facilitating comparison to other studies. Two changes were made to better discriminate certain types of releases: groups 5 and 7 were each divided into two groups, as defined below. The release groups used for Peach Bottom in SARRP are as follows: 2-44
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Group Kev Radionuclides Release Tvoe 1 Xc,Kr Noble gas release 2 I,Br Halogens (the SARRP bin includes J organic iodine) ! 3 Cs,Rb Alkali metals l 4 Te Telludum, also includes Se and Sb i 5 Sr Strontium 6 Ru Noble metals, also includes Mo, Pd, Rh, and Tc 7 12 Lanthanum, also includes Nd, Nb, Eu, Y, Pr, Pm, Sm, and Zr 8 Ce Cerium, also includes Np and Pu (part of group 7 in RSS) 9 Ba Barium (part of group 5 in RSS) f l 2.6.2 Source Term Code Package Calculations j l The selection of STCP runs which were used as the basis for quantifying parameters in j the RELTRAC code was based on a preliminary assessment of which accident progressions (sequences) would dominate risk. Some consideration was given to critical features of the j accident progmssion, and STCP runs were selected so as to demonstrate the effects of a few of l these (such as containment venting). The primary procedure used in the selection of STCP I l runs was a meeting of the review group (listed previously in Table 2-1). Because this meetmg I was held prior to the completion of the containment event tree (timing of the meeting was i dictated by schedule constraints), the STCP runs selected do not fully cover the risk-dominant accident progressions. The computer code runs do, however, supply sufficient information to enable the SARRP analysts to evaluate RELTRAC parameters such that RELTRAC provides reasonable extrapolations of the STCP results which were obtained to accident progressions identified as being important subsequent to the execution of the STCP calculations. The results presented here are thus primarily based on analyses performed by Battelle for this program [H]. Some insights from the BMI-2104 results were also used in formulating RELTRAC models. As described in the Battelle report, there have been a number of changes to the STCP since the publication of BMI-2104. Most of the changes involve the logistics of code usage and are merely improvements in the interfaces. However, there have been several improvements in individual models that am summarized below and described more fully in the Battelle report. 2-45 L __-_ - _ __ . _ _ _ _
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987)
. A new version of the CORSOR code, CORSOR-M, has been incorporated into MARCH 3. It allows a more consistent treatment of the release of fission products and the transport of sources of decay heat from the fuel.
- The STCP has been updated to include CORCON-Mod 2 to improve calculation of the thermal-hydraulic loads on containment associated with core-concrete interactions and as input to the VANESA code, which calculates the mleases from these interactions.
The coupling of the MERGE and TRAP-MELT codes has been improved since BMI-2104, with potentially significant impact. Any companson of j this study with BMI-2104 must consider the following: The decay heat contribution to the thermal-hydraulics is now considered; The nodalization in the fission-product transport calculations was-improved; Gas properties in TRAP now account for hydrogen, and the heat transfer coefficients have also been changed in TRAP; Aerosol modeling has been changed somewhat; and , The treatment of chemisorption on walls was improved, with potential effects on the highly reactive species (Te). While there have been severalimprovements to the codes, the STCP cannot at this time be considered verified and error free. An on-going review by Brookhaven National Laboratories is intended to provide this assurance. The STCP runs that served as the basis for this analysis are discussed in a separate report by Battelle [3]. These analyses served as the basis for the central estimates, although the SARRP study generated many combinations of accident conditions not specifically studied that were derived from the five basic scenarios covered in the Battelle work. These scenarios are described briefly below. TC1 Seauence. This sequence involves a scenario in which drywell failure precedes core damage. This failure results in the loss ofinjection to the vessel and core damage ensues. This is an ATWS sequence, and successful vessel depressurization and initial water level control is assumed. TC2 Secuence. This scenario also involves ATWS. However, vessel depressurization is not accomplished. High pressum injection is assumed to fail as a result of elevated suppression pool temperature. Core damage ensues in an intact containment. Drywell failure occurs immediately following vessel breach. TC3 Secuence. This scenario is similar to TC2, described above. However, the containment is successfully vented, through the suppression pool (wetwell) prior to vessel breach. There is never leakage from the drywell. TB1 Secuence. This is a station-blackout scenario in which turbine-driven pumps initially supply core cooling. Depletion of the batteries causes these i 2-46
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) pumps to fail, and the vessel remains at high pressure up until vessel breach. The drywell fails late as the result of slow pressurization. TB2 Sequence. This sequence is similar to TB1, described above. However, drywell failure is assumed to result from the containment pressurization which ! accompanies vessel breach. ! l Battelle also analyzed an interfacing-systems LOCA scenario using the STCP. It was, however, the concensus of the ASEP and SARRP analysts, with agreement fmm at least some of the review team members, that this sequence was not sufficiently likely to warrant further consideration. Thus, these STCP results were not included in the current work. The Battelle report provides the release fractions to the drywell or wetwell for three phases of the accident: the in-vessel release period, a puff release associated with a blowdown following vessel breach, and a core-concrete interaction release. Information concerning timing and other physical aspects of the accident are also provided. As described in Appendix B, the Battelle work was the starting point for both the central estimate and the LLH source-term analyses which were performed with RELTRAC. As stated previously, it was the intent of the program to have STCP runs for all l dominant accidents. The Battelle report provides source terms for accidents similar to the J dominant scenarios, but the SARRP analysis included key variations in the accident I progressions which warranted funher evaluation. 2.6.3 Overview of RELTR AC ! At the time that the Battelle analyses were commissioned, it was thought that most blackout sequences would involve operation of the turbine-driven injection pumps for several hours prior to core damage. However, further evaluation by ASEP resulted in the conclusion that a large fraction of the blackout-induced core damage sequences involved loss of all injection, followed by relatively prompt core degradation. These latter sequences have the potential for greater deposition of fission product aerosols in the containment. Deposition in the former type of sequence is severely limited due to the large steam flow from the boiling suppression pool which " sweeps out" the containment. In addition, the results of the containment analysis, which are summarized in Section 5.2, indicated important aspects of the accident progressions (e.g., drywell head seal leakage, venting in blackout sequences, operation of drywell sprays) which had also not been considered in the Battelle analyses. Because STCP results were not available to cover these risk-significant scenarios, RELTRAC was employed to perform reasonable extrapolation of the results which were available. l 2-47
NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) The philosophy in the implementation of RELTRAC for this analysis is similar to that for using SURSOR, SEQSOR, and GGSOR in the analysis of other SARRP reference plants. These previous codes are also parametric source-term models. However, they are parametric in integral fission-product release. RELTRAC is parametric in the rate equations that govern release. The form of the rate equations for each of the processes which RELTRAC models is intrinsic to the code. These forms wem determined by the analysts after camful myiew of the STCP results both for Peach Bottom and Grand Gulf. _ in addition, BMI-2104 and the msults ofindependent calculations, with CORCON and other portions of the STCP, wem reviewed. The RELTRAC rate equations include a number of constants which form a part of the code input. RELTRAC was " tuned" such that its predictions closely match those of the STCP, for five sequences described in the previous section, by adjusting these input constants. In this j manner RELTRAC was made to approximate not only environmental release fractions predicted by the STCP, but also the distribution of fission products within the containment. j This ability to represent intermediate results gives greater confidence that the effects of deviations of the accident progression from that modeled by the STCP are more accurately represented than was possible using the integral models for the other plants. RELTRAC is essentially a control-volume or compartment model. The movements of fission-product species between compartments are represented by a system of ordinary differential equations. Integration of this system of equations provides a prediction of the distribution of fission products among the compartments as a function of time, The compartments used to represent the Peach Bottom containment are the following: (1) The fuel (6) The suppression pool - i (2) The vessel atmosphere (7) The reactor building atmosphere i (3) The vessel surfaces (8) The reactor building surfaces (4) The drywell atmosphere (9) The environment ! (5) The drywell surfaces The complexity of the rate equations varies from a simple decontamination factor representation, such as for the suppression pool retention and deposition on surfaces in the reactor building, in which retention is proportional to the rate ofinflow, to the solution of the MAEROS aerosol equations [42] for deposition in the drywell. Integration of the rate equations does not proceed over discontinuities. These discontinuities.1 result from events in the
.l accident progression, such as vessel breach or rapid depressurizatica of the containment when a rupture occurs. Discontinuities such as these are modeled in F.ELTRAC as instantaneous redistributions of the fission products among the compartm.mts. For instance, if the i
2-48 ! i 1
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) containment ruptures in the drywell, essentially all fission products which are airborne in the drywell are instantaneously transported to the reactor building (further redistribution is subsequently calculated before integration of the rate equations resumes). RELTRAC is an integral and fully automated part of the modular methodology outlined in Figure 2-2. The definition of the accident-progression bins as determined by the EVNTRE post-processor (POSTSM) uniquely determines values of the RELTRAC parameters which vary according to the accident progmssion (e.g., mactor building decontamination factor) and the timing of discontinuities (e.g., containment failure and vessel breach). RELTRAC output consists of the fraction of the initialinventory of each fission product group released, and the length of time over which that release occurs, for both the period immediately following containment failure and for the succeeding period, up to the end of the calculation. These results are in a form suitable for input to CRAC2 and MACCS. Other output, for input to CLUSTER and for diagnostic purposes, is also produced. As discussed in Section 2.1.1, source terms were not calculated for every possible accident progression. Source terms were calculated only for those accident progressions which were expected to be important to risk or which made up a significant fraction of the core damage frequency (greater than 107o). The importance of each bin in terms of risk was estimated based on an approximate calculation of the source term. The approximate source term was expressed as the equivalent cesium release fraction (the release fraction for the cesium I group which the MACCS code would predict to induce similar late effects as for the set of release fractions for nine groups considered). The equivalent release fraction of cesium was assumed to be proportional to offsite consequence of release. A weighted fractional frequency for each bin was calculated by multiplying the conditional probability of the bin (probability of the bin given the occurrence of the associated plant-damage state) by the relative contribution to core-damage frequency of the associated damage state and by the equivalent release fraction. The bins contributing to the central estimate were then ordered by decreasing weighted fractional frequency. The weighted fractional frequencies were summed. The bins were then taken in order until the total weighted fractional frequency for the bins taken equalled at least 907o of the original total. Bins with conditional probability greater than 107o were than added to the list of those taken if they were not already present. The risk which was associated with the bins not taken (which total less than 107o of the total weighted fractional frequency) was represented by adjusting the conditional probability of the remaining bins such that the total weighted fractional frequencies of those bins equalled the original total. The adjustment was made in proportion to the weighted fractional frequency of the bin (i.e., such that the bin with 2-49
NUREG/CR-4551, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) the highest risk would represent the largest portion of the risk which would have been neglected had the bins assessed to have a low contribution been simply neglected). This procedure was implemented by POSTSM, the EVNTRE post-processor. Approximate source terms for each bin were estimated by PBSOR, the integral parametric source-term model which had originally been developed to approximate STCP results for Peach Bottom. PBSOR was superceded by RELTRAC when the need for more accurate source-term assessments became apparent. i The efficacy of the above method for reducing the number of source term and { consequence calculations required was clearly demonstrated in the course of the analysis. The central estimate CET analysis showed several hundred bins as being potential progressions for severe accidents. Application of the collapsing process described above resulted in the identification of 36 bins as potentially significant to central estimate risk. Thus, RELTRAC was used to calculate central-estimate source terms only for those bins. The accuracy of this approach has not been fully investigated. However, preliminary comparisons between the source-term weight calculated by PBSOR and the equivalent cesium release fraction based on RELTRAC results gives the analysts reasonable confidence that the error introduced is relatively small (compared to other sources of inaccuracy in the analysis). 2.6.4 Development of Source Terms for the Central Estimate Since practical limitations precluded calculation with the STCP for all combinations of sequence and containment phenomena represented by the accident-progression bins, the source terms for the central-estimate bins were calculated using relatively simple parimetric models of the release processes as represented in RELTRAC. RELTRAC allowed the source terms for different scenarios to be calculated from available STCP runs with alterations to account for bin-specific conditions. As much as possible, the alterations were set to mimic the STCP treatment, although in some cases the sequence conditions did not correspond too closely to any available run, and considerable analyst judgment was involved in formulating a parametric rate equation to represent the process involved. As described in the previous section, the results of the containment analysis were sorted into accident-progression bins (or bin attributes) to allow direct transition to the source-term analysis. The general parameters defm' ing the bins are listed in Table 2-5. A more detailed representation of these attributes of the accident progression (described in Appendix B) was used to establish the input to RELTRAC for the calculation of source terms for each bin. g Release timing information is obtained directly from RELTRAC. The detailed bin attributes l I 1 1 2-50
NUREG/CR-4551,'VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 2-5 1 DERIVATION OF THE CENTRAL ESTIMATE SOURCE TERM .i 2 Bin Parameter Outcome Pdncipal Effect on Source Terrn
- 1. Sequence- Fast station blackout! Similar to STCP TB sequence but suppression pool subcooled initiation Slow station blackout 2 resulting in increased residence time of fission products in containment.
Slow ATWS Based on STCP TC2 sequence. Fast ATWS Similar to STCP TC2 sequence, but suppression pool subcooled.
- 2. Vessel Yes Normal STCP calculation. j No Only' the in-vessel portion of f.he release is considered. ]
- 3. Power recovery No~ Normal STCP calculation.
After vessel breach Impacts timing of late containment failure / venting. Before vessel breach Reduced in-vessel retention. Lupacts timing of early containment failure / venting. Power always available Normal STCP calculation.
- 4. CCI Yes Normal STCP calculation.
Wet Water delivered to debris bed after CCI begins. Limited Core debris quenched after short period of CCI releases. No No CCI releases. ,
- 5. Early Drywell to environment No retention of fission products in the reactor building. )
containment I failure Drywell leak to Increased residence time for fission products in primary environment containment. No retention in reactor building. Drywell to refueling bay Reduced retention in reactor building, per STCP sensitivity study [1]. Drywell leak to refueling increased residence time in primary containment. Reduced bay retention in reactor building. Drywell to reactor building Normal STCP calculation. Drywell leak to reactor Increased residence time in primary containment. building Wetwell to environment All releases subject to pool scrubbing (including those from stuck-open S/RV vacuum breaker).3 No retention in the reactor building. Wetwell leak to All releases subject to pool scrubbing.3 lacreased residence - environment time in primary containment. No retention in the reactor building. Wetwell to reactor building All releases subject to pool scrubbing.3 Wetwell leak to reactor All releases subject to pool scrubbing.3 Increased residence building time in primary containment. No Early Failure No early release from containment.
- 6. Containment Yes impacts timing of late release. Puff release at vessel breach failure at like TB2 or TC2 sequence for STCP.
vessel breach No No puff release. Like TB1 for STCP. I i 2-51
l 1
)
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 2-5 (Continued) DERIVATION OF THE CENTRAL ESTIMATE SOURCE TERM Bin Parameter Outcome Principal Effect on Source Term ,
- 7. Late Drywell to environment See comment for Parameter 5.
containment failure Drywell leak to See comment for Parameter 5. environment Drywell to refueling bay See comment for Parameter 5. Drywell leak to refueling See comment for Parameter 5. bay Drywell to reactor building See comment for Parameter 5. Drywell leak to reactor See comment for Parameter 5. building Wetwell to environment All CCI releases subject to pool scrubbing.3 No retention in reactor building. Wetwell leak to All CCI releases subject to pool scrubbing.3 Increased l environment residence time in primary containment. No retention in reactor building. Wetwell to reactor building All CCI releases subject to pool scrubbing.3 Wetwell lesi to reactor All CCI releases subject to pool scrubbing.3 Increased building residence time in primary containment. No Failure Nominal (design) leakage assumed.
- 8. Drywell spray Early Increased deposition of all particulate in drywell.
Late Increased deposition of CCI releases in drywell. Early only Increased deposition of particulate releases to drywell prior to containment failure. None Normal STCP calculation.
- 9. Direct Yes Enhanced release of refractory fission products, principally Ru containment group.
heating No Normal STCP calculation. j l
- 10. Suppression Early and late Minimal retention of fission products in suppression pool. l pool bypass i Late Normal STCP calculation.
None CCI releases subject to suppression pool scrubbing. I lmmediate failure of vessel injection. 2 Failure of vessel injection delayed 6 to 8 hours. 3 Subject to parameter 10. I i l 2-52
h NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l 1 were also used in the estimation of heat of release for all scenarios based on specific STCP . results (described in Appendix B). As illustrated in Section 5, the core-damage profile involved relatively few types of sequences, dominated by scenarios involving the loss of I electric power. The details of the actual calculations for the central source term estimates are provided in Appendix B. As stated previously,it was the intent of the calculation that the n:sults would correspond to what would be obtained if a STCP run had been completed for each of the bins. Although some uncertainty is introduced by the approximations of the methodology, it is believed that this uncertainty is fairly small relative to the phenomenological uncenainty ranges acknowledged in the LLH uncertainty analysis. Thus, the approximations involved here probably do not add greatly to the overall uncertainty, nor do they-seriously affect the comparisons between the central estimate and the LLH results (discussed in Section 5). 2.6.5 Derivation of Source Terms for the LLH Uncertainty Analyses The discussion above refers to the central estimates of the source terms. The LLH uncertainty analysis described in Section 2.3 required additional source terms to represent the uncertainty in this part of the analysis. The selection ofissues for the LLH study is discussed -; in Section 4 and Appendix B. Source-term uncertainty was considered in terms of ten issues that describe phenomena that are both uncertain and that could significantly impact accident source terms: (1) Release fractions from fuelin-vessel, (2) Fraction of in-vessel Cs and I release which is released in elemental form (rather than CsI), j (3) Retention of in-vessel deposition revolatilized after vessel breach, l (4) Suppression pool decontamination factor for aerosols, (5) Suppression pool decontamination factor for volatile forms ofiodine, I (6) Fraction ofin-vessel deposition revolatilized after vessel breach, (7) Magnitude of CCI releases, (8) Average reactor building decontamination factor, (9) Average refueling bay decontamination factor, and l (10) Late releases of volatile forms ofiodine from the suppression pool. 2-53
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) In addition, a few of the uncertainty issues concerning containment loading and performance impact source terms. Leakage through a stuck-open S/RV vacuum-breaker and the containment pressure inen:ase resulting from vessel breach are the most notable examples. Several of the issues consider phenomena that are not included in the STCP but that are thought to be important, and all-of the issues involve ranges that are beyond what the STCP would predict. It was therefore necessary to generate source terms that could account for all the combinations ofissue outcomes proposed by the review group. This was fairly simple to accomplish within the RELTRAC framework. Because the description of the model used to calculate the source terms requires detailed explanation ofindividual parameters, it is described fully in Appendix B, and only the approach is summarized here. The approach used to construct source terms from issue specifications was a purely parametric one. Representation of source-term uncertainty issues in RELTRAC is entirely consistent with the overall approach to the RELTRAC model. The majority ofissues impact only the values of constants in the rate equations. Other issues required the parametric rate equations for the postulated phenomena be developed and incorporated into the code. Provision was make in the development of RELTRAC for the representation of containment issues which impact source terms. An overview of the approach for estimating the source terms is presented in Table 2-6. The entries in the table describe the treatment of the source-term issues, although the actual calculation is more complex, as described in Appendix B. As delineated in the table, the source-term estimate considers the impact of the accident progression bin as well as the LLH issues on specific inputs to the overall source term estimate. More complete description of the RELTRAC treatment is also provided in the appendix and in Reference 9. It is not practical to calculate consequences for each of the source terms generated by RELTRAC. Thus, an additional simplification was required for the LLH: source-term clustering. This involves the grouping of the thousands of source terms from RELTRAC into l a lesser number of clusters that can approximately characterize many of the individual source j terms. Consequences are then calculated for the clusters only, in this case reducing the ! required consequence evaluations to 54. The clustering process is described in Appendix B. 2.7 OFFSITE CONSEQUENCE ANALYSIS l I The estimation of offsite consequences due to reactor accidents for the assessment of
]
Peach Bottom was performed using the CRAC2 and MACCS computer codes. CRAC2 [43] is I 2-54 I; - _ _ _ _ _ _ - _ _ - - _ _ _ - _ _ _ _ - _ - _ - _ - - _ l
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 2-6 DERIVATION OF LLH SOURCE TERMS Source-Term Issues Tn atment in RELTRAC Release from the fuelin-vessel Four levels of release rate constants sampled for each of the nine fission-product groups. Fraction of Csl decomposition Four fractions sampled for percent of in-vessel iodine release appearing as volatile iodine.1 Fraction of in-vessel fuel release that Four levels of vessel deposition-rate constants sampled for each of is released from the vessel the eight (non-gaseous) fission-product groups. l Pool decontamination for aerosols Two decontamination factors, one each for scrubbing of releases through S/RV quenches and through drywell vents. Five pairs of I decontamination-factor values (high/high, medium / medium, etc.) sampled, assuming same value for each fission-product group. l Pool decontamination for volatile Five decontamination factors sampled.1 forms ofiodine Revolatilization of Cs and I from Four values of the re evolution rate constants sampled the vessel (only for accident progressions without recovery of injection to vessel). Rate constants were set such that total re-evolution at the , end of the calculation equalled integral value specified by the issue- i level definition. Magnitude of CCI releases Four levels of release-rate constants sampled for each of six fission-product groups (Te, Sr, Tu, La, Ce, Ba). No effect if CCI is prevented.2 Reactor building decontamination Four decontamination factors sampled. No impact if reactor building bypassed.2 Refueling bay decontamination Four decontamination factors sampled. Only has impact if release is through refueling bay.2 Late release of volatile forms of Release rate assumed proportional to quantity of iodine in pool. iodine from the suppression pool Four levels of the proportionality constant sampled. Constant set such that total release at end of calculation equalled integral value specified by the issue level definition, assuming central estimate value of total iodine in pool. 2 Volatile iodine is modeled as a tenth radionuclides group within RELTRAC. It is added to the total iodine group for output. 2 Determined by accident-progression bin definition, Table 2-5. an improved version of the CRAC code, which was developed for and used to estimate the offsite consequences reported in the Reactor Safety Study [21]. The MACCS code (MELCOR Accident Consequence Code System) [42] is a much improved version of CRAC2, so much improved that it was decided not to call it CRAC3. The principal improvements include the following: (1) an improved architecture that facilitates development of intermediate results and the performance of uncertainty studies, (2) a multi-plume dispersion model that includes a 2-55
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) multi-step crosswind concentration profile, (3) a dry deposition velocity that depends on particle size, (4) an improved wet deposition model that does not overpredict ground concentrations produced by rainout, and (5) expanded sets of environmental-transport, j dosimetry, health-effects, and economic-cost models, all supported by critical reviews ofinput data. Not all of the models have yet been verified by thorough peer review. The consequence analysis, using either the CRAC2 or MACCS code, requires the q following input:
- Source-term release fractions for radionuclides groups, along with timing of release and the sensible heat associated with the release; The inventory at reactor scram of all isotopes important to offsite consequences;
- The population distribution around the reactor site; ;
Emergency response parameters and assumptions; and ) Weather, land-use, and economic data for the region around the reactor site. Given these inputs, the consequence codes predict the following: The downwind transport, dispersion, and deposition of the radioactive materials released from the failed containment;
+
The radiation doses received by the exposed populations via direct (cloudshine, inhalation, groundshine) and indirect (ingestion, inhalation) pathways; The mitigation of those doses by emergency response actions (evacuation, ! sheltering, and relocation of people), interdiction of milk and crops, and decontamination or interdiction of land and buildings; The early fatalities and early injuries expected to occur within one year of the accident, and the latent cancer fatalities expected to occur over the lifetime of the exposed individuals; The total population dose received by the people living within some distance (50 miles) of the plant and the early fatality risk to persons living near the I plant (within one mile); and The offsite costs of emergency response actions, and of the interdiction and decontamination of land, buildings, milk and crops. By performing calculations for all possible combinations of representative sets of source terms, weather sequences, and exposed populations, statistical distributions of consequence measures are developed that depict the range and probability of consequences. If desired, the uncertainties associated with offsite consequence estimates can be developed by a variation of input parameter values using structured Monte Carlo sampling techniques [10), but such sensitivity studies were not performed ss part of this study. 2-56
i NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l Site-specific offsite consequence calculations require as input a population distribution for the region around the reactor and a one-year file of hourly meteorological data that is characteristic of the region. The Peach Bottom population distribution was constructed from 1980 census data on a polar grid, which has 16 angular sectors and 20 radial intervals that ; extend outward to 500 miles. (The radial intervals used were 1,2,3,4,5,10,15,20,25,30, 40,50,60,70,85,100,150,200,350, and 500 miles.) The census data was also corrected for average daily transient population. A representative set of weather sequences was constructed from a file of meteorological data as follows. First, the 8760 (the number of hours in a year) weather sequences in the meteorology file were sorted into 29 bins using wind speed, atmospheric stability, and precipitation as sorting criteria. Then four sequences were randomly selected from each bin that contained sequences (28), which produced a stratified sample of 110 representative l weather sequences. The Peach Bottom analysis was based on one year of meteorological data gathered at the Peach Bottom site. The land usage and economic data were collected for the 15 states within 500 miles of the Peach Bottom site. Each of the 320 elements of the polar-coordinate grid (consisting of 16 l sectors and 20 radial intervals) was then associated with one of the 15 states, or the ocean if applicable. The habitable land fraction of each element v.as estimated by inspection of maps. State average values were used for the fraction of land in each element that is devoted to farming, the identity of the crops grown, and their worth [1].]. National average data was used I for the worth of public and private urban land, buildings, equipment, inventories and possessions (expressed as total worth of tangible public and private assets per person). l The following emergency-response assumptions were used in the Peach Bottom CRAC2 calculations: Of the population located within 10 miles of the site,95% are evacuated radially at a speed of 2.0 m/sec. The evacuation begins one and one-quarter hours after the warning time, and once the evacuees reach a radial distance of 15 miles they are assumed not to receive additional dose. People who do not evacuate continue their normal activities (they are not directed to shelter). The non-evacuees who are projected to receive a groundshine dose to bone marrow within seven days of the accident that equals or exceeds 25 rem are relocated at the end of one day. All other non-evacuees are relocated in seven days if interdiction and decontamination cannot reduce projected groundshine or inhalation doses to bone marrow to less than 25 rem in 30 years. Radiation shielding factors for evacuees and for non-evacuees as follows: 2-57
NUREG/CR.4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Non-Evacuees Evacuees Cloud dose 0.75 1.0 Ground dose 0.33 0.5 Inhalation dose 0.75 1.0 The CRAC2 calculations used a dry deposition velocity of 0.01 m/sec for airborne particles (1 mean particle diameter), and assumed that the washout coefficient (A) in the rainout model equals 0.0001 times the rainfall rate. The MACCS calculations used dry a deposition velocities of 0.002,0.01 and 0.05 m/sec respectively for particles with diameters less than 0.7 , between 0.7 and 5.1 , and greater than 5.1 p. The MACCS washout coefficient was assumed to be equal to 0.00012 times the square root of the rainfall rate. Release coefficients for the central estimate and the cluster source terms were developed as described in the previous section. The source-term analysis also assigned a characteristic STCP sequence (TB1, TB2, TC, TBS or TBR with or without core-concrete interactions delayed ten hours) to each source-term estimate. For the central-estimate source terms, the release times, release durations, sensible heat release rates, and particle size distributions were assumed to be the same as the assigned STCP sequence [H]. For the clusters, the release times, durations, elevations, warning times, and sensible heat-release rates wem calculated as weighted sums of the central source term values for those parameters. For example, the release duration Ati and the rate of release of sensible heat Ei for cluster i were calculated with the following equations: Ati= Q_njdj EAii= QJijEjdj Ej nj I jnj where Atj, and Ej are the release duration and the release rate of sensible heat of central estimate source term j, respectively, and nj is the number of variants of central source term j that contributed to cluster source term i. The central estimate and cluster source terms produced by these procedures were used directly in the MACCS consequence calculations. Because CRAC2 cannot treat a multi-puff release, the multi-puff MACCS source terms had to be modified for CRAC2 calculations. Thus, CRAC2 source terms were created from MACCS multi-puff source terms by increasing the duration of the first MACCS puff by one hour and releasing the entire source term in that extended puff. The actual source terms used in the consequence evaluation and the results of the CRAC2 and MACCS consequences are provided in Section 5. 2-58
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1987) l l References for Section 2
- 1. Benjamin, A. S., et al. Evaluation of Severe Accident Risks and the Potentialfor Risk Reduction: Surry Power Station, Unit 1. U.S. Nuclear Regulaton Commission Report NUREG/CR-4551, Volume 1 (Draft Report for Comment), Sandia National Laboratories, Albuquerque,NM: February,1987,
- 2. Benjamin, A. S., et al. Evaluation of Severe Accident Risks and the Potentialfor Risk Reduction: Sequoyah Power Station, Unit 1. U.S. Nuclear Regulatory Commission l Report NUREG/CR-4551, Volume 2 (Draft Report for Comment), Sandia National Laboratories, Albuquerque,NM: Februay,1987.
- 3. Amos, C. N., et al. Evaluation of Severe Accident Risks and the Potentialfor Risk Reduction: Grand Gulf, Unit 1. U.S. Nuclear Regulatory Commission Report NUREG/CR-4551, Volume 4 (Draft Report for Comment), Sandia National Laboratories, Albuquerque,NM: February,1987.
- 4. Kolaczkowski, A. M., et al. Analysis of Core Damage Frequencyfrom Internal Events:
Peach Bottom, Unit 2. U.S. Nuclear Regulatory Commission Report NUREG/CR-4550, Volume 3 of 7 (Draft Report), Sandia National Laboratories, Albuquerque, NM: November,1986.
- 5. Griesmeyer,3. M. Users' Guidefor the EVNTRE Computer Code. Sandia National l Laboratories, Albuquerque,NM: (to be published).
- 6. Corradini, M. L., et al. A Review of the SARRP Containment Event Trees, U.S.
l Nuclear Regulatory Commission Report NUREG/CR-4569, University of Wisconsin, Madison,WI: May,1986.
- 7. Corradini, M. L. " Additional Comments on the Peach Bottom and Grand Gulf Containment Event Trees," letter report to M. Cunningham, U.S. Nuclear Regulatory Commission, Madison, WI: October 13,1986.
- 8. Denning, R. S., et al. Radionuclides Release Calculationsfor Selected Severe Accident Scenariost BWR, Mark I Design. U.S. Nuclear Regulatory Commission Report
, NUREG/CR-4624, Volume 1, Battelle's Columbus Division, Columbus, OH: July, l 1986.
- 9. Helton, J. C. and J. M. Griesmeyer. Users' Guidefor the RELTRAC Computer Code.
Sandia National Laboratories, Albuqueque, NM: (to be published).
- 10. Murfin, W. B., et al. Users' Guidefor the RISQUE Computer Code. Sandia National Laboratories, Albuquerque,NM: (to be published).
- 11. " Comments on Peach Bottom Containment Event Tree," letter from V. Boyer, Philadelphia Electric Company, to B. Morris, U.S. Nuclear Regulatory Commission, September 19,1986.
- 12. Dingman, S. E., et al. " Analysis of Peach Bottom Station Blackout with MELCOR,"
Proceedings of the 14th Light Water Reactor Research Safety Information Meeting. U.S. Nuclear Regulatory Commission, Washington, DC: 1986.
- 13. Greene, S. R. "The Role of BWR Mark I Secondary Containments in Severe Accident Mitigation," Proceedings of the 14th Light Water Reactor Research Safety Information Meeting. U.S. Nuclear Regulatory Commission, Washington, DC: 1986.
2-59
I l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987)
- 14. Helton, J. C., et al. Incorporation of Uncertainties into Reactor Accident Source-Term Estimates. U.S. Nuclear Regulatory Commission Report, Sandia National Laboratories, Albuquerque, NM: (to be published). )
- 15. Reassessment of the Technical Bases for Estimating Source Terms. U.S. Nuclear Regulatory Commission Report NUREG-0956, Washington, DC: August,1985.
- 16. Amos, C. N. and A. M. Kolaczkowski. Containment Event Analysisfor Postulated Severe Accidents: Peach Bottom, Unit 2. U.S. Nuclear Regulatory Commission Report l NUREG/CR-4700, Volume 3 (Draft), Sandia National Laboratories, Albuquerque, NM:
February,1987.
- 17. Phenomenology and Risk Uncertainty Evaluation Program (PRUEP). U.S. Nuclear Regulatory Commission Project, FIN #A-1593, Sandia National Laboratories, Albuquerque, NM.
- 18. Iman, R. L. and J. C. Helton. A Comparison of Uncertainty and Sensitivity Analysis Techniques for Computer Models. U.S. Nuclear Regulatory Commission Report NUREG/CR-3904, Sandia National Laboratories, Albuquerque, NM: March,1985.
- 19. Iman, R. L. and W. J. Conover. Sensitivity Analysis Techniques: Self-Teaching )
Curriculum. U.S. Nuclear Regulatory Commission Report NUREG/CR-2350, Sandia NationalLaboratories, Albuquerque,NM: June,1982.
- 20. Iman, R. L. et al. Stepwise Regression with PRESS and Rank Regression.. Sandia National Laboratories Report S AND79-1472, Albuquerque, NM: 1980.
- 21. Reactor Safety Study--An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants. U.S. Nuclear Regulatory Commission Report WASH-1400 (NUREG-75/014), Washington, DC: 1975.
- 22. Harper, F. T. "ASEP Data Reevaluation." Memorandum to Senior Consultant Group, et al., Sandia National Laboratories, Albuquerque, NM: March 15,1985.
- 23. Carlson, D. D. Interim Reliability Evaluation Program Procedures Guide. U.S. Nuclear Regulatory Commission Report NUREG/CR-2728, Sandia National Laboratories, Albuquerque, NM: 1983.
- 24. Fleming, K. N., et al. Classification and Analysis of Reactor Operating Experience involving Dependent Events. Electric Power Research Institute Report NP-3967 (Interim Report), Palo Alto, CA: June,1985.
- 25. Estimates of Early Containment Loads from Core Melt Accidents, Report of the Containment Loads Working Group. U.S. Nuclear Regulatory Commission Report NUREG-1079 (Draft for Comment), Washington, DC: December,1985.
- 26. Containment Loads Working Group. " Consensus Summaries for Standard Problems 1 through 6." Letter Reports to J. Telford, U.S. Nuclear Regulatory Commission, May-June,1984.
- 27. Gieseke, J. A., et al. Radionuclides Release Under Specific Accident Conditions.
Battelle Columbus Laboratories Report BMI-2104, Columbus, OH: 1984.
- 28. Lipinski, R. J., et al. Uncertainty in Radionuclides Release Under Specific LWR Accident Conditions. Sandia National Laboratories Report SAND 84-0410, Albuquerque, NM: February,1985.
i 29. IDCOR Task 23.1: Integrated Containment Analysis. Technology for Energy Corporation, Knoxville, TN: 1984. 2-60
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FFBRUARY,1987) l
- 30. Cook, D. H., et al. Loss ofDecay Heat Removal Sequences at Browns Ferry Unit One -
Accident Sequence Analysis. U.S. Nuclear Regulatory Commission Report NUREG/CR-2973, Oak Ridge National Laboratory, Oak Ridge, TN: May,1983.
- 31. Harrington, R. M., et al. ATWS at Browns Ferry Unit One - Accident Sequence Analysis. U.S. Nuclear Regulatory Commission Report NUREG/CR-3470, Oak Ridge '
National Laboratory, Oak Ridge, TN: July,1984.
- 32. Harrington, R. M., et al. Station Blackout at Browns Ferry Unit One - Accident Sequence Analysis. U.S. Nuclear Regulatory Commission Report NUREG/CR-3470, Oak Ridge National Laboratory, Oak Ridge, TN: July,1984.
- 33. Rivard, J. B., et al. Identification of Severe Accident Uncertainties. U.S. Nuclear Regulatory Commission Report NUREG/CR-3440, Sandia National Laboratories, Albuquerque,NM: September,1984.
- 34. Steam Explosion Review Group. A Review of the Current Understanding of the Potentialfor Containment Failure Arising from In-Vessel Steam Explosions. U.S.
Nuclear Regulatory Commission Report NUREG-1116, Washington, DC: February, 1985.
- 35. Tarbell, W., et al. High-Pressure Melt Streaming (HIPS) Program Plan. U.S. Nuclear Regulatory Commission Report NUREG/CR-3025. Sandia National Laboratories, 4 Albuquerque, NM: August,1984.
- 36. Hatch, S. W., et al. Reactor Safety Study Methodology Applications Program: Grand Gulf #1 BWR Power Plant. U.S. Nuclear Regulatory Commission Report NUREG/CR-1659, Volume 4, Sandia National Laboratories, Albuquerque, NM: October,1981.
l 37. Final Safety Analysis Reportfor Peach Bottom, Unit 2 . Philadelphia Electric Company, Philadelphia, PA: 1974.
- 38. Greimann, L., et al. Final Report, Containment Analysis Techniques, a State-of-the-Art Summary. U.S. Nuclear Regulatory Commission Report NUREG/CR-3653, Ames Laboratories, Ames,IA: March,1984.
- 39. Weinstein, M. B., et al. " Primary Containment Leakage Integrity: Availability and Review of Failure Experience," Nuclear Safetyt 21,1980. i
- 40. McClymont, A. S. and B. W. Poehlman. Loss of Off-Site Power at Nuclear Power i Plantst Data and Analysis. Electric Power Research Institute Report NP-2301 (Interim l Report),Palo Alto, CA: March,1982.
- 41. Green, G. A., et al. " Mark I Containment Drywell: Impact of CCI on Containment Integrity and Failure of Drywell Liner," Proceedings of the International Symposium on Source Term Evaluationfor Severe Accident Conditions, IAEA-SM-281/36, : October, 1985.
- 42. Marble, W. J. Preliminary Report on Fission Product Scrubbing. General Electric Company Report NEDO-XXXX, San Jose, CA: 1983.
- 43. Cunnane, J. C., et al. "The Scrubbing of Fission Product Aerosols in LWR Water Pools Under Severe Accident Conditions - Experimental Results," Proceedings of the Topical Meeting on Fission Product Behavior and Source Term Research. American Nuclear Society, Snowbird, UT: July,1984.
- 44. Greimann, L. G. et al. Reliability Analysis of Steel Containment Strength. U.S.
Nuclear Regulatory Commission Report NUREG/CR-2181, Ames Laboratories, Ames, IA: August,1982. 2-61
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987)
'45. Gieseke, J. A., et al. Source Term Code Package: A User's Guide (Modi). U.S.
. Nuclear Regulatory Commission Report NUREG/CR-4587, Battelle Columbus
. Laboratories, Columbus, OH: July,1986.
- 46. Gieseke, J.' A., et al. Informal Report on Source Term Predictions for Various Containment Failure Assumptions. Battelle Columbus Laboratories, Columbus, OH:
August,1984. ,
- 47. Gelbard, F., and J. H. Seinfeld. " Simulation of Multi-Component Aerosol Dynamics,"
Journal of Colloid Interface Science: 78,p.485,1980.
- 48. Ritchie, L. T., et al. CRAC2 Model Description. ' U. S. Nuclear Regulatory Commission Report NUREG/CR-2326, Sandia National Laboratories, Albuquerque, ,
NM: 1984. l
- 49. Alpert, D. J., et al, "The MELCOR Accident Consequence Code System." Proceedings of the CEC Workshop on Methods ofAssessing the Off-Site Radiological Consequences ofNuclear Accide.nts. SAND 85-0884C, April 1985.
- 50. Alpert, D. J. and J. C. Helton, " Uncertainty and Sensitivity Analysis for Reactor Accident Consequence Models." Proceedings of the CEC Workshop _on Methods of Assessing the Off-Site Radiological Consequences of Nuclear Accidents.
SAND 85-0885C, April 1985. l
- 51. Aldrich, D. C., et al. Technical Guidance for Siting Criteria Development.. U. S.
Nuclear Regulatory Commission Report NUREG/CR-2239, Sandia National Laboratories, Albuquerque,NM: December,1982. i i 2-62
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) i l l Section 3 METIIODOLOGY FOR EVALUATION OF RISK-REDUCTION OPTIONS A second major effort in SARRP was the assessment of the reduction in risk that could be achieved by various postulated plant modifications. The risk-reduction measures considered included a number that have been proposed in previous studies and several aimed at the specific features of Peach Bottom that appeared to contribute most to both the frequency of core damage and to risk. It must be noted, however, that many of the generic safety options proposed in other studies were eliminated from consideration without detailed evaluation because it was clear that the more expensive options could not be cost-effective with respect to the reductions ) in risk that could potentially be achieved. This was particularly true for options aimed at preventing core damage, since the fmquency of core dam. age was assessed to be relatively low for Peach Bottom. The risk-reduction measures include modifications to plant systems, the l addition of new equipment, and changes to operating procedures and practices. The proposed l options were developed in an attempt to address the risk associated with two types of consequence measures: the health and safety of the public and the potential for financiallosses, both offsite and onsite. The list of options for specific consideration for Peach Bottom therefore included measures aimed at reducing the frequency of core damage (referred to in this l report as preventive options) and at decmasing the consequences of core-damage events (called mitigative options). The benefits that could be realized by the implementation of each of these options were also compared to estimates of their costs and other potential negative impacts. The following sections describe the manner in which the safety options were identified and evaluated to determine their value/ impact attributes. The results of the risk-reduction assessment are provided in Section 6; a more detailed description of the treatment of the options is provided in Appendix D. 3.1 IDENTIFICATION OF SAFETY OPTIONS Safety optiom. for consideration in this study were identified both from a number of sources that have postulated generic modifications for BWRs, and from an assessment of the specific accident sequences and features associated with containment response that were found to contribute most to risk for Peach Bottom. 3-1
NUREG/CR4551. VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 3.1.1 Generic Risk-Reduction Onlions Many measures intended to reduce risk have been proposed as a result of safety insights gained through various research and analysis activities. Some of these proposals have l been in response to particular safety issues that have arisen relative to operating plants, while j others are intended to address key uncertainties or sensitivities that appear to have significant l risk implications. Other NRC-sponsored research programs are currently underway or have been completed to evaluate similar measures, and SARRP drew from these programs as well. The generic options myiewed for applicability to Peach Bottom are listed in Table 3-1 (for preventive measums) and Table 3-2 (for mitigative options). One option that is not listed but that has been suggested in a number of studies as a valuable risk-reduction measure is the overall improvement of plant operations through the implementation of reliability-enhancement programs. While it is anticipated that such a program would indeed offer positive benefits,it does not lend uself readily to evaluation of the reduction in risk that could be achieved. Nevertheless, the results of this assessment can be used to imply the degree to which any measure can be effective in reducing risk. 3.1.2 Risk-Reduction Measures for Peach Bottom The generic safety options were considered in the context of the design and risk features for Peach Bottom, and were either specified in more detail or were eliminated from further consideration. Some of the generic options were not included in the analysis for Peach Bottom because they would clearly have a limited potential to reduce risk or because they were not applicable to the Peach Bottom design. In addition to considering the generic options, the Peach Bottom results were reviewed in an attempt to propose changes addressed to specific and unique aspects of the accident sequences. Frequently, such options offer the most promise of being cost-effective. The options considered for Peach Bottom are described in the following discussions. Description of Preventive Ootions The options intended to improve safety through the reduction of the frequency of core-damage sequences are described briefly below. These options were identified primarily by l examining the sequence-level cut sets highest in frequency. A more detailed description of each of the options and the manner in which it was assessed is provided in Appendix D. Option P1- Reduced Probability for Common-Cause Failure of De Power. The dominant contributors to the frequency of core-damage for Peach Bottom involve scenarios in 3-2
NUREG/CR-4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRU ARY,1987) l l l Table 3-1 BWR PREVENTIVE OPTIONS Proposed Option Impact on Sequences Reasons forInitial Proposal Impmved reactor- Decrease frequency of sequences . Past PRAs have shown pmtection system involving failure to scram (ATWS) importance of ATWS for BWRs Alternate rod-insertion Decrease frequency of(ATWS) - Past PRAs have shown system sequences importance of ATWS for BWRs Modification oflogic Increase number of systems that - Past PRAs have shown for automatic can respond to transients importance of failure to depressurization system depressurize Concern regarding failure of operators to depressurize manually Higher air pressure for Increase availability of automatic
- Past PRAs have shown i SRV solenoids depressurization system importance of failure to j (Mark I &II depressurize l containments) l 1
Independent cooling for Increase availability of com- - Importance of this failum l HPCI/HPCS pumps cooling systems mode for HPCI and HPCS Improved suppression Decrease frequency of sequences . Importance ofloss of pool cooling with containment failure leading containment cooling to core damage sequences in past PRAs
+
High consequences of core melt with containment failed ) Improved containment Decrease frequency of sequences
- Imponance ofloss of I venting with containment failure leading containment cooling to core damage (also a mitigation sequences in past PRAs option) -
High consequences of core melt with containment failed Improved ac/dc power Decrease frequency of station- + Importance of station :
& RCIC reliability blackout sequences blackout in past PRAs Improved HPCI/HPCS Increase availability of core- -
Importance of these systems reliability cooling systems in respondi" to transients Procedures for inter-Decrease frequency of V - Risk significance of facing systems LOCA sequence containment bypass 3-3
I NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l Table 3-2 BWR MITIGATIVE OIrrIONS Proposed Option Impact on Sequences Reasons forInitial Proposal 1 Missile shield Prevent containment failum due
- Remaining uncertainty in to in-vessel steam explosion steam explosion potential Ac-independent Reduced potential for hydrogen . Importance of hydrogen as a hydrogen igniters burns to fail containment and cause of containment failum (Mark III containments) elimination as along-term threat . Importance of station-blackout sequences Hydrogen burn Eliminate potentialimpact on
- Importance of hydrogen as a prevention (Mark III) containment of hydrogen burns cause of containment failure Drywell flooding Decontamination of releases from .
Importance of core-concrete core-concrete interactions interactions if suppression poolis bypassed Core retainer Eliminate com-concrete - Importance of release interactions from core-concrete interactions if suppression pool is bypassed Drywell curb Reduced potential for com . Potential for early drywell (MarkI) debris to attack drywell shell failum can lead to higher consequences Improved drywell Reduced potential for late - Importance oflate failure i cooling capability failure of drywell in past PRAs l Wetwell vent Reduced potential forlate
- Importance oflate overpres-l (filtered / unfiltered) overpressurization of containment surization in past PRAs (also a preventive option)
Improved reactor Reduced radioactive release to = Many BWR containments building spray environment are enclosed in the reactor building Removal of dampers Increased operability / capability of . Some BWR SGTS isolate from standby gas reactor building to reduce fission- for temperature conditions treatment system (SGTS) product releases found in accidents (Some plants only) Improved drywell Decmased source term and + Importance of sprays for sprays drywellloading containment phenomena Significance of sprays to radionuclides release Additional containment Reduced containment failure . Reduction in likelihood of strength or volume probability offsite release Alternate reactor Changes high-pressure core-melt . Potential for decmasing , depressurization system sequences to low pressure early containmentloading i i
)
3-4
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) which de power is unavailable following a loss of offsite power due to common-cause failure of the batteries. In the ASEP assessment of Peach Bottom,it was determined that the testing j and maintenance activities for the four trains of emergency batteries (trains A and B for Unit 2 and trains C and D for Unit 3) are subcontracted, and that administrative procedures do not , preclude same-day testing or maintenance of multiple trains. Therefore, the potential for an error to be made on more than one train exists. The safety option that was evaluated involved the implementation of controls that would require that testing and maintenance activities be staggered. It was assumed that the train A and train D batteries would be subjected to mutine testing together on a quarterly basis, with testing of train B the following month and train C the month after that. The discharge tests would all be done separately. These steps would tend to f minimize the chances that an error could be duplicated on more than one train of batteries. Option P2: Use of the Fifth Diesel Generator. The emergency ac power system at i Peach Bottom is comprised of four diesel generators. However, a fifth, lower-capacity diesel generator exists at the site. Although it could not be used to recover all of the functions lost if one of the emergency diesel generators were to fail, it could be used to supply power to a control-rod drive (CRD) pump, which could provide injection to the reactor vessel. The safety option that was considered therefore involved extending current procedures and technical , specifications to cover the use and maintenance of the fifth diesel generator as well. It is assumed that the administrative controls would be structured in such a way as to minimize the ) potential for the fifth generator to be affected by failure causes common to the others. It would also be necessary to establish connections from the generator to the CRD pump. l Because of the relatively low flow capacity of the CRD pump, this option would not affect short-term station-blackout scenarios. It could, however, reduce the frequency of longer-term scenarios, in which eventual depletion of the station batteries would othenvise lead to loss of the ac-independent cooling systems. In order for this option to be effective in the long term, it would also be necessary to establish venting of the wetwell, providing a path for j the rejection of decay heat in an open-cycle mode. This would entail changing the procedure l for containment venting, and might require modification of the hardware associated with the i venting in order to ensure its availability under the long-term blackout conditions of interest. I Ontion P3: Use of the Diesel-Driven Fire Pump for Iniection. Peach Bottom has a diesel-driven pump for the fire-protection system which has a capacity of 2500 gpm with a , shutoff head of 125 psi, and which could be used for injection of cooling water to the core. Current procedures call for its use under certain conditions, and blind flanges and hosing connections exist for injecting into the the low-pressure injection lines. However, the 3-5
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) procedures do not addmss the accident scenarios for which this pump could be most beneficial, notably those involving station blackout. Option P3 therefore examined the effects of revising the procedures to provide explicit instructions to the operators regarding how to align the valving to use the fire pump for injection when ac and/or de power is unavailable. As was the case for option P2,it would also be necessary to modify the proeures for venting and make any necessary changes to the venting hardware. In addition to the long-term station-blackout sequences addressed by option P2, the fim j pump could be used to maintain core cooling under short-term blackout conditions as well, by using the automs depressurization system (ADS) to mduce reactor-vessel pressure to below the pump's shutoff head. To enable the ADS to be used for cases in which de power is unavailable, either early or because oflong-term battery depletion, it would also be necessary to modify the operators for the ADS valves. This could be accomplished by connecting a manually-controlled nitrogen supply to the actuators on the ADS SRVs. Three other points should be noted with regard to the consideration of preventive options for Peach Bottom. Fint, although it was not explicitly considered as a part of any of the safety options, it would be advisable to provide at least some containment-pressure instrumentation with de power; currently, it is understood that all such instrumentation requires ac power. For long-term station blackout sequences, having de-powered instrumentation ; available could provide information that would improve the operaton' ability to understand the situation and take appropriate measures in the six hours between the initiation of the accident and the depletion of the batteries. Second, because the sequences involving depletion of the batteries are important, extending the capacity of the batteries was examined as en option. It was discarded, however, because it would not be particularly effective in reducing core-damage frequency. The limited effectiveness results from the potential for the RCIC and HPCI pumps to fail due to high turbine-exhaust pressure or high pressure suppression pool temperature, conditions that would occur shortly after the assessed time of battery depletion in the current analysis. Extending the life of the batteries would change the cause of failure of the injection pumps, but would not eliminate the sequence. The third point concerns the emergency service-water system. At Peach Bottom, this system provides cooling to the four diesel generators. The pumps for this system receive power from diesel-generators 2 and 3 (from Units 2 and 3, respectively). The service-water system has a backup pump which is powered from diesel-generator 4 (for Unit 3). However, in the original ASEP analysis, it was assumed that the use of one of two booster pumps 3-6
l NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987)
)
(powered by diesel-generators 2 and 3) would be required for success of the backup pump. Therefore, the loss of these two generators would still result in failure of all of the diesel- ; generator cooling, and consequently all of the generators would be lost. However, Peach I Bottom personnel have tested the backup cooling-water pump and determined that it can l function successfully without the support of the booster pumps. Therefore, the analysis was revised, and no safety option associated with modifying this configuration was assessed. However, it would still appear to be prudent to change the power supply for one of the booster
)
pumps to diesel-generator 4 to provide added assurance of the availability of the backup pump. j l Ducriotion of Mitigative Ontions i A number of safety options intended to reduce the likelihood of containment failure and/or the severity of releases resulting from core damage were identified. Again, additional l detail can be found in Appendix D. Option M1: ADS without De Power. As noted for option P3, the ability to depressurize the reactor vessel without de power is a potentially desirable option to consider. I For option P3, such an ability was considered as part of a low-pressure core-cooling alternative for its potential to prevent core damage. Option M1 considers the capability to depressurize l without de power for its potential to mitigate the severity of a core-damage accident. The loss of de power (either from battery failure or depletion) characterizes the majority of plant-damage states which contribute to the calculated risk for Peach Bottom. Because operation of the SRVs in relief mode requires de power, core degradation and subsequent breach of the reactor vessel would occur at high pressure. Ejection of the core debris would be mor energetic than for a low-pressure core melt and early loading of the containment would also be greater due to the ejection of hot gases under pressure. Direct heating of the containment and the possibility that debris ejected under pressure could melt the drywell shell if contact were made all increase the probability of containment failure induced by RPV meltthrough at high pressure. Option M1 considers a direct connection between high-pressure nitrogen bottles and some of the SRV actuators. The connection would be made such that manual opening a valve in the connecting line would pressurize the actuators causing the SRV to open Option M2: Drywell curb. ADS without Dc. and Imoroved Venting. Option M2 represents a combination of features aimed at various aspects of the accident progression. In addition to adding the capability to operate the ADS valves without de power, as in option M1, a concrete curb would be constructed around the inner circumference of the drywell wall. The 3-7
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) intent of the curb is to prevent attack of the drywell shell by the core material as it is ejected from the reactor vessel. The SARRP containment loading expert reviewers indicated that the ejection of the core debris under high pressure could lead to melting of the drywell shell because a jet or (jets) of debris would be directed from the RPV pedestal toward the drywell shell. The proposed curb could not prevent this postulated containment failure mechanism, but could prevent drywell meltthrough in the event of a low-pressure ejection of core debris from the RPV. The modification of the ADS would address many of the sequences that might otherwise proceed at high pressum. This option also includes improving the capability to vent the wetwell. For station-blackout sequences,it is difficult to accomplish venting. This option would include modifying the power supplies to the valves needed (and changing the l procedures) in order to enhance the ability to perform venting under blackout conditions. Ootion M3: Divwell Curb. This option considers the separate impact of only installing the drywell curb, as described for option M2. Option M4: Imoroved Drywell Sorays. The drywell spray system at Peach Bottom is dependent on the availability of ac power. Spray is provided by the RHR pumps which take suction from the suppression pool. Thus if either ac power is failed, or the suppression pool is I saturated, the normal spray system is inoperable. Since the dominant core-damage sequences involve station blackout, the sprays are generally not available as a mitigative measum. Option M4 is intended to examine the effect of adding the capability for spray operation for station-blackout sequences. This could be accomplished either by the installation of an entirely separate spray train, or by utilizing the diesel-driven fire pump described for option P3. Ootion MS: Imoroved Venting. Option M5 exa nines the effectiveness of the improved capability for venting of the wetwell, described for option M2, as a separate modification. Ootion M6: Area-Wide Fire Sorays for the Reactor Building. After egress from the containment, the fission products released by melting of the core must generally still pass through the reactor building. In portions of the containment event analysis, the failure of the drywell or wetwell is sufficiently severe to cause extensive damage to the reactor building, limiting its ability to retain fission products. However, not all failures of containment are that severe. At Peach Bottom, the reactor building fire sprays are designed only to limit the spread of fire from one unit to another. The spray pattern is therefore only a localized curtain between units. This safety option examines the effect ofincreasing retention of the radionuclides in the reactor building by extending the fire sprays to cover the building more fully (it should be noted that the reactor buildings for some other BWRs, e.g., Browns Ferry, already have area-wide fire-suppression systems). 3-8 1
)
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,19tt7) 3.2 EVALUATION OF COSTS AND OTHER IMPACTS An impact assessment was performed for all of the options identified for Peach Bottom. The types ofimpacts considemd included the following:-
- The one-time costs of installing or implementing the modification;
. Recurring costs due to operation and maintenance activities;
-
- Radiation exposure of personnel during installation of the modification; and
- Radiation exposure due to continuing maintenance.
The scope of the SARRP risk-reduction task did not include detailed design development for all of the proposed modifications, although some of the key mitigative features were the subject of a detailed design and cost estimation effort performed specifically for SARRP [1]. Several others have been the subject of previous investigations [23A). The-scope of SARRP did'not allow for a detailed investigation of the remaining safety improvements, and the impacts for these options were estimated based on comparisons to available information. The data used for this study are summarized in Appendix F. Each proposed risk reduction option was described in terms ofits specific application. This includes a description of design parameters and delineation of the major equipment needs, if any. The options were then compared to those included in the data base in Appendix F to generate an estimate of the costs and other impacts. While this process is not ideal, it does allow for the preliminary assessment of the cost-effectiveness of the options. Any options that appear to offer the potential to meet value/ impact criteria can then be subjected to rnore rigorous analysis. It should also be noted that the installation costs for many of the options could be dominated by the costs associated with replacement power if Peach Bottom had to be shut i down during implementation of the modification. This is a difficult parameter to estimate because the cost is dependent on a number of factors, including the specific schedule mandated for implementation, the status of other projects in progress at the utility,- the financing L arrangements of the utility, and the availability of power at the time of the modification. Several previous studies have attempted to reflect the costs of replacement power by considering the total level of effort required to implement a modification to determine if it could be performed in successive shutdowns without causing an increase in the shutdown length. A basic assumption in these analyses is that there is'no regulatcry pressure driving the implementation schedule, and that very large-work crews couio be employed. These 3 - - -_ __ _ _____-__- -_- - _ _ _ _ _ _ - _ _
l NUREGCR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) assumptions are probably not realistic for some of the options, and past experience indicates that unexpected delays in implementation due to a variety of causes has been common. Due to the large variability in costs that could be calized if replacement power costs were included, the basic value/ impact ratios were calculated without reflecting them. This allowed for the formulation of preliminary conclusions and enabled a paring of options to those that had a potential for sufficient net benefit to merit further consideration. Replacement power was a consideration in drawing the conclusions for these options. In support of this activity, a discussion of insights concerning the impacts associated with replacement power costs is included in Appendix D. 3.3 EVALUATION OF EFFECTS ON RISK The value of each option was assessed by recalculating all of the risk measures, accounting for the integrated effect of the change. For safety improvements that resulted only in a change in the frequencies of plant-damage states, this involved a straightforward requantification of the risk parameters. For options that had an impact on containment or source-term analyses, appropriate model changes and requantification was performed, and the I pieces of the analysis were re-integrated to provide new estimates of risk. For the preventive options, or for mitigative options that also had a preventive effect, the affected sequence cut sets were re-quantified to reflect the change. The revised sequence frequencies were then incorporated into the frequencies for the plant-damage states, and the risk calculations were repeated. Similarly, mitigative options were treated by making the appropriate changes to the containment-event tree, by revising the branch-point probabilities and recalculating the frequencies of the accident-progression bins. Again, the risk calculations were repeated after the changes were incorporated. The evaluation of the specific effects of each option involved a number of assumptions that are documented in Appendix D. These assumptions included such aspects as estimates of reliability of the option once implemented and the capability of the system in terms of ability to prevent or mitigate accident conditions. Important sensitivities to these assumptions are discussed in the results of the risk-reduction analysis (Section 6). Any plant change that is directed at reducing core-damage frequency or risk also has the potential for introducing competing risks that could result in an overall adverse impact. These competing risks can be difficult to recognize because, for example, even a measure aimed at reducing core-damage frequency can increase risk if there is a redistribution of frequency into 3-10 L_-- - - _ _ _ - - _ - - - - _ - _ _ - - - - - - - - - - - - - - - - - - - _ - - -
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) more severe damage states, and because very subtle interactions could be introduced in the actualimplementation of the design. The scope of this study did not allow for a full evaluation of competing risks, since the designs were generally defined in only a limited, scoping manner. j However, in order to evaluate a plant change appropriately, an attempt was made to postulate competing risks, and the description of each alternative was amended to include restrictions intended to reduce the potential for adverse effects. As indicated by the results described in Section 6, the potential for adverse effects to outweigh the possible benefits was important for several options. The impact of the proposed safety options were only considered within the range of severe core-damage accidents addressed in other areas of this report. Plant performance for other potential accidents was not evaluated. Such consideration must be made before serious consideration can be given to the implementation of any of these proposals. 3.4 VALUE/ IMPACT ASSESSMENT Procedures and assumptions used in value/ impact analyses can vary widely with both the analyst and the purpose of the study. The NRC has previously considered this issue and has identified specific features that are needed for regulatory application of the results of any value/ impact analysis [5]. While it was not the intent of this program to perform the regulatory analysis to determine specific rulemaking requirements, the procedure adopted for the value/ impact analysis in SARRP is consistent with NRC regulatory uses and therefore facilitates review and comparison. The methods used for the value/ impact assessment also drew upon the Handbookfor Value Impact Assessment [5]. In the terminology of that report, the analysis in SARRP is a
" limited effort value/ impact" assessment. However, in some cases there is additional information presented that is not outlined in the handbook. These two references do not specify all aspects of the analysis, and other studies were reviewed for particular psrameters
[2,1]. l The value/ impact assessment was performed through direct comparison of the ; recalculated risk measures expressed as averted accident costs to the total impact calculated as a j present-value cost. These cost / benefit comparisons were generated for different measures of l averted costs to facilitate the use of the results in value/ impact assessments performed under different criteria and with different assumptions. The comparisons were developed separately for each of the following elements: (1) Total offsite costs as calculated by the consequence codes; (2) The sum of these offsite costs plus the onsite costs due to the accident; 3-11 j i
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) (3) Offsite costs calculated by assuming an equivalent cost of $1000 per person-rem for exposure to the population out to a radius of 50 miles from the plant; and (4) The sum of offsite costs calculated as in (3) plus onsite costs. The offsite costs calculated by CRAC2 and MACCS include the costs of health effects as well as relocation expenses, lost wages, decontamination costs, lost public and private property, and the value ofinterdicted land and crops. The estimated onsite impacts of the accidents reflected consideration of the costs associated with lost capital, cleanup and decommissioning, power replacement, and onsite health effects. The cost estimates were made using the methods reported in NUREG/CR-3673 [2]. The replacement power required as a consequence of the accident was assumed to be supplied by already-installed spare capacity. It was assumed that the equivalent of the rated power of the plant, adjusted for a 65% capacity factor, required replacement. A discount rate of 4% was applied, and because the spare capacity would almost certainly be supplied by a l fossil plant, an escalation rate of 6% was also used in the calculation. It was assumed that replacement power would be required for ten years, after which time a new plant would supply power. A correction was included to account for the replacement power if the accident occurred in the last 10 years of plant life. The capital cost of the loss of the utility's investment was calculated, with the prescat worth being derived with a 4% discount rate. This calculation assumed a 40-year plant life (accounting for the actual remaining life of the facility at the time of core damage), and also assumed a uniform rate of depreciation over the 40 years. The 1985 cost of construction was used as a base ($3000/kwe installed). The remaining onsite costs were taken from NUREG/CR-3673 [2]: cleanup cost was assumed to be $1.7 x 109; decommissioning cost was taken as $1.0 x 108; and onsite health costs were estimated to total $6.5 x 106, In the NRC regulatory analysis, the value is calculated as a reduction in the total population dose out to fifty miles, evaluated at $1000 per person-rem. Other values for averted cost calculated in SARRP are based on the reduction in consequence measures as determined by CRAC2 and MACCS. For these calculated costs, offsite property damage costs were used directly; while averted health effects were valued at $1 million per early fatality, $100,000 per latent cancer fatality and $100,000 per early illness. These values were obtained from an I investigation of available statistics pertaining to the amount which society has been willing to expend to avert fatalities and injuries [E]. The specific values selected are somewhat unimportant because offsite costs are almost always dominated by property damage rather than 1 3-12
NUREG/CR4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) health effects. All costs in the value/ impact analysis wem treated in the same manner as the costs for the rebaselined risk; all values were discounted (4%) to obtain presem worth, i The algorithm for calculating averted accident costs did not include costs tha+ are commonly referred to as secondary costs, i.e., the economy-wide costs of a severe reactor accident due to considerations such as adverse public opinion, enhanced concern in the financial community, and, particularly, regulatory constraints placed on the industry after an accident. Inclusion of these costs might significantly increase the benefit of some safety options. As noted above, this value/ impact analysis obtained a number of different measures for assessing the overall value of the proposed safety options. These results permitted a detailed review of contributing factors and led to additional conclusions that might not have been drawn if a single ratio using one particular formula had been used. Conclusions regarding the merit of i an option or the need for additional information to assess the option more fully were derived from these results. 3.5 TREATMENT OF UNCERTAINTIES i Most of the inputs that were needed for the value/ impact analysis (construction costs, maintenance costs, etc.) were estimated using a three-point uncertainty range of high, central and low values based on a review of the available data. These values were then used in the value/ impact assessment for each option in the LLH analysis. The impact of the preventive options was calculated by amending the core-melt expression generated by ASEP to reflect the effects of the option, and then requantifying the mean frequencies of the plant-damage states for each LLH sample member. The TEMAC code [2] was used to evaluate these frequencies, and its output was used directly by the RISQUE code. The risk was then recalculated, and the results are presented in Section 6. For the mitigative options, the containment event tree code was used to determine the impact of the mitigative options for each of the combinations ofissue levels determined in the LLH sampling method. Modifications were made, either to branch-point probabilities or to the structure of the event tree, which would approximately reflect the impact of the particular mitigation option under consideration. The new results were calculated using a portion of the RISQUE code developed specifically for this purpose As with the basic risk results, the RISQUE code includes provisions for calculating a number of parameters ofinterest with respect to gaining a better understanding of the risk-3-13 1
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) reduction results. The most important measures and breakdowns of the risk-reduction results are summarized in Section 6 and are provided in detail in Appendix D. The full complement of available output for any given risk measure includes the following:
- Risk reduction for any given individual or set of plant-damage states (groups of core-damage sequences);
- Fractional contribution of each plant-damage state to mean risk, including the new risk for safety options;
- Fractional contribution of each accident-progression bin to mean risk; ]
- Fractional contribution of each cluster to mean risk; and
. Detailed analysis of individual risk sample members, including fractional contributions of damage states, accident-progression bins, and clusters.
Figure 3-1 illustrates the value/ impact analysis that is also performed as part of the risk-reduction task. The averted accident costs are plotted with the cost of the option. The example provided in Figure 3-1 is simplified; for the full results presentation, all four of the measures of averted accident costs listed in Section 3.4 are presented for each safety option. As noted above, the averted accident costs are determined for each LLH sample member by considering the reduction in health effects and property damage in monetary terms, with the costs of the options represented by a three-point uncertainty range. The sample members are reordered and plotted in terms of total averted cost, with a bar used to represent the range between the 5th-and 95th-percentile sample members. A bar is also used to represent the range of costs for the . option. Areas where the bars overlap indicate potential cost effectiveness. Options that initially indicated a significant potential to be cost-effective were subjected to further investigation of , i competing risks and other cost impacts not included in the original assessment (such as the cost ; of replacement power). ; i 3-14 l
)
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) . 1 I ILLUSTRATION OF VALUE/ IMPACT ASSESSMENT l Range of cost of option g W OPTION 1 l -lt w pl Partially effective OPTION 2 l!?l' M, ; Effed/e for all M bwer estimate of risk OPTION 3 l 0 l lq > tl Not effective etc. l l AVERTED RISK (EXPRESSED AS AVERTED COSTS) --> I Figure 3-1. Example of LLH Value/ Impact Display 1 i 3-15 t
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) References for Section 3
- 1. Cherdack, R, et al. Study of Severe Accident Mitigation Systems. U.S. Nuclear Regulatory Commission Report NUREG/CR-4781, Sandia National Laboratories, Albuquerque,NM: February,1987.
- 2. Berry, D. L., and G. A. Sanders. Study of the Value/ impact ofAlternative Decay Heat Removal Concepts for Light Water Reactors. U.S. Nuclear Regulatory Commission Report NUREG/CR-2883, Sandia National Laboratories, Albuquerque, NM: 1983,
- 3. Hatch, S. W., et al. Shutdown Decay Heat Removal Analysis - General Electric i BWR3/ Mark 1 Case Study. U.S. Nuclear Regulatory Commission Report 1 NUREG/CR-4448, Sandia National Laboratories, Albuquerque, NM: (to be published). l
- 4. Benjamin, A. S., and F. T. Harper. Value Impact Investigation of Filtered-Vented Containment Systems and Other Safety Optionsfor a BWR Mark I Containment . U.S.
Nuclear Regulatory Commission Report NUREG/CR-4065 (Draft Report), Sandia NationalLaboratories, Albuquerque,NM: September,1984.
- 5. Denton, H. R. (Memorandum). "NRR Office Letter No.16, Revision 2--Regulatory Analysis Guidelines." U.S. Nuclear Regulatory Commission, Washington, DC:
October 30,1984.
- 6. Heaberlin, S. W, Handbookfor Value-Impact Assessment. U.S. Nuclear Regulatory Commission Repon NUREG/CR-3568, Pacific Northwest Laboratory, Richland, WA:
1983.
- 7. Burke, R. P., and D. C. Aldrich. Economic Risks of Nuclear Power Accidents. U.S.
Nuclear Regulatory Commission Report NUREG/CR-3673, Sandia National Laboratories, Albuquerque, NM: 1984.
- 8. Strip, D. R., Estimates of the Financial Consequences of Nuclear Power Reactor Accidents. U.S. Nuclear Regulatory Commission Report NUREG/CR-2723, Sandia National Laboratories, Albuquerque, NM: September 1982.
- 9. Iman, R. L. and M. J. Shortencarier. A User's Guide for the Top Event Matrit Analysis Code (TEMAC). U.S. Nuclear Regulatory Commission Report NUREG/CR-4598, Sandia National Laboratories, Albuquerque, NM: August,1986.
1 4 3-16
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) j l Section 4 EVALUATION OF KEY UNCERTAINTY ISSUES As described in Section 2.2, the Limited Latin Hypercube uncertainty approach relied j on the selection of key uncertainty issues that had a significant impact on the estimated risk of the Peach Bottom plant. These issues are defined in this section, along with a discussion of the range of uncertainty associated with each. The issues were considered in three categories: , 1 (1) Sequence frequency issues, (2) Containment loading and performance issues, and j (3) Radiological source term issues. l For the first category, sequence frequency issues, SARRP and ASEP analysts reviewed the ASEP accident core-melt sequence results and the basic plant models to define the key issues with large uncertainties or a significant impact on core-damage frequency. The ASEP study included sensitivity analyses which were used to help identify the LLH issues [1]. The ASEP analysts then developed a weighting function to represent their degree of beliefin the possible outcomes for those issues based on the current body of knowledge. For the latter j two categories listed above, similar criteria were used to establish the list of issues, but the issues often involved greater uncertainties due to lack of any available research or large disagreement within the reactor safety community as to both the methods of evaluation and the potential outcomes. It was decided that for these issues, mostly phenomenological in nature, l the best representation of the range of uncertainty would be obtained through the input of an expert review group. This review group broadened the base ofinformation for both selection f and evaluation of the issues. As described in Section 2.2, the experts were called on to provide j l their individual views on a range of distinct outcomes for each issue by selecting weighting ) factors for each outcome. These weighting factors determined the relative frequency of each O outcome in the LLH sample. The purpose of this section is to provide the reader with an introduction to all of the I uncertainty issues included in the LLH sampling for Peach Bottom. The details of the i evaluation of these issues are included in the appendices of this report while their significance in terms of risk is discussed in Section 5. The ranges of uncertainty provided by the review group are also provided in this section. The actualimplementation of the LLH method is also f described in the appropriate appendix for each area of the analysis. Because the appendices l provide the detailed discussion, the information here has not been heavily referenced to source material. I
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 4.1 SEQUENCE FREQUENCY ISSUES Four issues were grouped into the category termed " front-end issues". These issues address the some key sensitivities associated with the frequency or characteristics of the important core-damage scenarios. It should be noted that the ASEP study included uncertainties due to the reliability data as well as the sensitivity of the result to some other issues. The LLH issues focused on key sensitivities affecting the accident sequence results and j include uncertainties in both modeling assumptions and application of data when information is very limited. Each issue selected for inclusion in the LLH sampling is defined separately below and the range of weighting factors for each outcome is also listed. Weighting factors for the front-end issues were supplied as continuous distributions by the ASEP analysts. Continuous distributions are consistent with the representation of data uncertainties in the TEMAC code. For the purposes of the SARRP LLH, these distributions were discretized by taking the 5th,30th, 70th, and 95th percentiles of the distributions as representing issue
" levels". Additionalinformation conceming uncertainties in ihe assessment of the core-damage frequency is included in the ASEP report for Peach Bottom [1].
4.1.1 Failure to Actuate the Standby Liould Control System The probability that the operations staff fails to actuate the standby liquid control system in time (up to 5 minutes) following an anticipated transient without scram (ATWS) was considered as one uncertainty with a potentially significant impact. The ASEP sequence analysis included this uncertainty thru the use of a log-uniform distribution on the failure probability. For the LLH study in SARRP, the distribution was discretized into four probabilities, with weighting factors derived directly from the ASEP failure distribution: Probability of Failure to Actuate SLC .0007 .003 .034 .15 Weighting Factors .1 .4 .4 .1 4.1.2 Freauency of De Power System Common Mode Failure The most important ASEP accident scenario in the ASEP evaluation involves the complete loss of de power due to a common-mode failure. This is recognized as being an uncertain failure mode with limited data for quantitative evaluation. Because of the large uncertainty and the importance of this type of accident both in frequency and consequences this was selected as an LLH issue. A log. uniform distribution was assigned to the beta factor associated with battery common-mode failure. (The beta factor is the probability that a second : battery fails, due to common mode, givtn the failure of the first battery.) In the ASEP analysis 4-2
t NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) -[ of Peach Bottom, the beta factor for more than two batteries is the same as for two batteries. Thus the value sampled here is the probability that all de power fails given a failure of the first battery. Beta factor for Battery Failure 1.7 x 10-5 2.4 x 104 1.7 x 10-2 2.4 x 10-1 Weighting Factors .1 .4 .4 .1 4.1.3 Probability of Failure to Vent, Durine an ATWS The failure of the operations staff to successfully vent the containment prior to core damage during an ATWS scenario was considered as another LLH uncertainty issue. Given that the operators have vented containment, the probability that injection will not be maintained is sufficiently small that sequences involving venting with subsequent core damage can be ignored. Thus this issue impacts directly on the frequency of ATWS sequences. The ASEP j analysts assigned a mean failure probability of 0.9 for this event. A maximum entry distribution was assumed to develop the LLH levels and weights. The discrete values and - j weighting factors are listed below: ) Probability of Failure to Vent: ATWS 0.7 0.88 0.93 0.99 Weighting Factors 0.1 0.4~ 0.4 0.1 As illustrated by the values above, there was actually a very limited uncertainty assigned to this event, principally because of the very limited amount of time available to perform the steps - required to establish venting during an ATWS. 1 4.1.4 Power Recovery Uncertain 1jn Accidents involving loss of ac power were the most significant contributors to the core-damage frequency. In addition to their frequency importance, these accidents are also quite important to the assessment of containment response and consequences since the availability of electric power can greatly change the progression of the accident once initiated.' One of the most important aspects of these accidents is therefore the probability of power recovery at various times. Although there is data available for assessing power recovery probabilities, the actual parameter remains uncertain due to uncertainties in the data itself, in its interpretation, and in the application of generic data to a panicular plant. The uncertainty in this probability was assessed directly from the raw data collected from U S nuclear plants. (The data used is the same as the input to the study described in NUREG 1032 [2].) A correction was applied to the function used to fit the data such that the mean curve (nonrecovery probability versus time) 4-3
NUREG/CR4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) . approximated the curve reported in NUREG-1032.' The function was then used to produce curves for the non-recovery probability as indicated by the 5th,30th,70th, and 05th percentiles of the data. These four curves represented the four levels of this issue, and were weighted as follows: Percentile on Recovery - ! Curve 5% 30 % 70 % 95 %
-Weighting Factor' .1 .4 .4 .1 This issue impacts both the power non-recovery factors used in the sequence frequency - -
assessment as well as the power recovery probabilities in the containment event tree. Values are taken directly from the curves for the' specific times called for in the accident sequence cut ' set or in the containment event tree. 4.2 CONTAINMENT LOADING AND PERFORMANCE ISSUES The following containment loading and response issues were included in the LLH - study for Peach Bottom:
. (1) Probability of a stuck-open tailpipe vacuum breaker (2) Use of the high pressure service water system spray as a recovery (3) Probability that the operations staff fails to vent containment during a station blackout, (4) Probability that the operations staff fails to vent containment following L the recovery of ac power, (5) Size of bypass due to stuck-open tailpipe vacuum breaker, !
(6) Containment failure pressure and location of failure, (7) Containment failure size, (8) Failure mode of the reactor pressure vessel, (9) Containment pressure prior to vessel creach for station blackout, (10) Pressure rise at vessel breach, (11) Probability of drywell meltthrough, and (12) Probability of hydrogen burns sufficient to cause reactor building bypass. Each of these issues is defined in the following sections. It should be noted that the consideration of severe.1 of the issues involves a level of detail considerably greater than that' l 4-4 I i l
1 l I NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) I presented here, and that Appendix A defines the actualimplementation of the LLH input for the { containment analysis. There are certainly other containment issues that involve significant ! i uncertainty but the consensus opinion of the SARRP analysts and reviewers was that the issues J listed above were the most important relative to the assessment of the Peach Bottom plant. 4.2.1 Probability of a Stuck-Onen Vacuum Breaker Examination of operating history for BWRs has indicated a potentially significant problem relative to partial suppression pool bypass for steam released through the safety / relief valves (SRVs). Each SRV tailpipe is equipped with vacuum breakers to prevent suppression pool water from being drawn back into the pipe as the steam remaining in the tailpipe : condenses following an SRV demand. Some plants have experienced significant reliability problems with these vacuum breakers to the point where the valves would fail to reclose { following nearly every demand. This problem appean to result from the large force applied to the valve as it opens after an SRV demand. Physical damage to the valve flapper and the hinge pin have been seen. This problem has been addressed through redesign, both at Browns Ferry and Peach Bottom, but there has not been enough operating experience to show that the design change has been fully effective, particularly under the conditions of repeated demands as might be the case with many of the core-melt accident sequences. The result of this failure would be , a partial bypass of the pool--a phenomenon that is actually more important to the source term calculation. The issue is maintained as a containment issue since the failure mechanism is related to containment performance more than the phenomenology of source terms. The event is treated in the containment event tree because it changes the progression of the physical effects of the accident such as pressurization of containment, and because it has a potentially large impact on the source term since it can lead to pool bypass or a reduced scrubbing effectiveness. This event is only important as a containment issue, it has no impact on the probability of core damage. However, the uncertainty in this issue was treated similar to the front-end issues because the data values are derived from ASEP sources. The discrete values used to represent th~ uncertainty in the LLH analysis are listed below: Probability of a Stuck-Open Tailpipe Vacuum Breaker .29 .62 .94 1.0 Weighting Factor .1 .4 .4 .1 l l Once again, the weighting factors are taken directly from the ASEP failure distribution rather than being based on reviewer input. The relatively high values for this event come from operating experience which has indicated a real reliability problem with these valves at some facilities. Most plants (including Peach Bottom) have since performed modifications to reduce 4-5
NUREG/CR455t, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) the probability of a stuck-open vacuum breaker, but the data is too weak to allow firm ; conclusions regarding the effectiveness of the modifications. It should be noted that this data ; accounts for all of the vacuum breakers that are challenged, as well as the increased probability associated with multiple demands. 4.2.2 Use of the High Pressure Service Water System Sorav as a Recovery One of the potential recoveries considered in the containment event tree model was the use of the high pressure service water (HPSW) system as a means of providing drywell spray, Since this is a recovery option that involves operator action outside the written procedures, the assessment of a probability of success or failure for this action is subject to large uncertainty. To be consistent with the quantification of other human-error events the probability for this event was derived from ASEP, and the ASEP uncertainty for this event was discretized into the following values: I Probability of Using HPSW for Spray 0.13 0.3 0.66 0.94 Weighting Factor .1 .4 .4 .1 4.2.3 Probability that the Onerations Staff is Unable to Vent Durine Station Blackout i In the absence of some postulated drywell failure modes (described in Section 4.2.11), venting of the containment may serve to reduce the quantity of fission products released from containment. Depending upon the effectiveness of suppression pool scrubbing (discussed in Section 4.3.4) this reduction in release may be large. Thus the uncertainty in the probability of venting has a potentially large impact on the uncertainty in predicted risk. Procedures have been established for venting the Peach Bottom containment during a station blackout. Review of these procedures by the ASEP and S ARRP analysts resulted in the ; conclusion that their successful implementation was viRely. This assessment is based on the requirement that plant personnel enter high radia<on and high temperature areas of the reactor building to manually open valves. There is mcertainty in this sonclusion that venting is unlikely, because anticipation of the need to vent could allow actions to be taken before physical limitations were too great and venting co dd become feasible. The assessment of the ASEP analysts was that the mean failure probability for venting , during a station blackout is 0.9. A maximum entropy distribution (lower bound 0.5 and upper band 1.0) was assumed. This distribution was discretized as follows: 1 4-6
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1987) - Station blackout l with no ac power 0.7 0.88 0.96 0.99 Weighting Factors 0.1 0.4 0.4 0.1 Due to the high mean value of failure, the assessed uncertainty in this issue is actually very small, limiting the impact of this issue in the LLH sample. 4.2.4 Probability that the Ooerations Staff is Unable to Vent After Ac Power Recovery This issue involves the same concerns as the previous issue, but the boundary conditions are different. It is potentially important for the same reasons as discussed above. Given that ac power is available and that the containment pressure exceeds the threshold for venting, the procedure calls for venting of containment using ac-powered valves. This procedure is executed from the control room. The assessment of the ASEP analysts was that the mean failure probability for venting under these conditions is 0.01. A log-uniform distribution was assumed (with an error factor of ten). The distribution was discretized as l a follows-
]'
Station blackout with ac mcovery 0.0015 0.0042 .012 .028 Weighting Factors 0,1 0.4 0.4 0.1 4.2.5 Level of Sunoression Pool Bvonss Through a Stuck Onen Safetv/ Relief Valve Vacuum Breaker 1 This issue is a different facet of the uncertainty associated with the first containment ; issue, the probability of a stuck open vacuum breaker. Because information on the exact failure mode is lacking, the size of containment bypass resulting from this event was also selected as a containment uncertainty issue for the LLH study of Peach Bottom. Two levels of typass were considered, small and large. Small leakage implies that the failure mode of the vacuum breakerinvolves a failure to fully reseat, perhaps due to physical distortion of the valve flapper. The second outcome would correspond to the valve being completely stuck open, the equivalent flow area would therefore be somewhere between forty and eighty square inches. The probabilities assigned to these outcomes are indicated in Table 4-1. The reviewers were fairly consistent and assigned a probability of approximately .8 to the small leakage case, citing the limited evidence which suggested damage to the valve rather than a completely stuck open valve. Several reviewers also noted that repeated demands of the SRV would also create forces on the valve that could help to force it to reclose, thus lowering the probability of a 4-7
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) completely stuck open vacuum breaker. The composite was used directly in the LLH: 84%'of the sample members had a small leakage if this failure mode occurred, while the other 16% had large leakage assigned to this case. Table 4-1 SIZE OF SUPPRESSION POOL BYPASS FOR A STUCK OPEN SRV TAILPIPE VACUUM BREAKER Levelof Probability i Bypass Reviewer SARRP 1 2 3 4 5 6 7 Composite
- 1. Small(.005.022 ft2) .9 .9 .8 .9 .8 .7 .8 .9 .84
- 2. Large (.27 .55 ft2 ) .1 .1 .2 .1 .2 .3 .2 .1 .16 4.2.6 Containment Failure Pressure and Location ;
The containment failure pressure is an uncertain parameter that can significantly affect risk for sequences involving overpressurization. There is also uncertainty in the location and size of the failure. The predicted location of failum may be correlated to the failure pressure and it is included in this issue. The size of the failure may also be correlated to both the location and the ultimate failure pressure; and this uncertainty is treated as issue 3 described in the next section. The uncertainty in this parameter was represented by five levels for the LLH study. Level 2 mpresents the msults of an Ames laboratory study of Peach Bottom [3], while level 4 represents the RSS value. The other levels allow a reasonable spectrum around the two Peach Bottom specific values. The failure location is important because of the impact on the source term. Three locations are distinguished; the drywell head, the drywell knuckle (in this analysis the knuckle is generally assumed to be any drywell location below the head seal), and the wetwell. The RSS assumed that the failure would occur in the wetwell. Subsequent studies (based on the Ames results) have predicted a drywell locations--either the knuckle between the cylindrical and spherical sections or at the head. The drywell head is of significance because release through this path would be directly to the refueling bay which has the least holdup and the potential for a direct path to the atmosphere through the refueling bay blowout panels. Another study of the failure pressure by Chicago Bridge and Iron is currently in progmss. Because of the dependencies on the boundary conditions, the reviewers were asked to provide their input for two cases as defined below: 48
l 1
-d NUREG/CR-4551, VOL,3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) a q
l Case 1. Prior to or immediately after vessel breach, with containment temperatures less than 500 F.
]
j l Case 2. - After vessel failure, with corium on the floor and high drywell temperatures of 800-1200 F. 1 To consider correlations, reviewers were asked to provide weighting factors for each combination of three failure locations and five failure pressure levels. The reviewers inputs ; for case 1 are illustrated in Figure 4-1. The results indicate that the level 3 value of failure pressure,138 psig, was weighted most heavily, which corresponds to an increase over the Ames work to account for conservatism in that study. The range of values weighted -l corresponded to the Ames study as a lower limit and the RSS pressure as the upper limit. j Since the information concerning location is too detailed for this section (weighting factors were provided for three locations for five failure pressures), only the results for predicted location of failure at 138 psi are presented in Figure 4-2. There was variation in the
)
expected location of failure, although the figure illustrates two trends: (1) the drywell location I is weighted significantly more heavily than the wetwell and (2) the two types of drywell failure, head leakage versus the other locations, were weighted approximately equally by most ., reviewers. The details of the reviewer input on this subject are contained in Appendix A. l l 0.7 - Case 1: T < 500* F 0.6 - 0.5 - x 0.4 - Weighting Factor x 0.3 - 0.2
- 0.1 -
0 X X 84 psi 117 psi 138 psi 160 psi 225 psi , Containment Failure Pressure (psig) Figure 4-1. Weighting Factors for Containment Capacity for Low Temperature Conditions *
*On all figures pmsented in this section of the report, the range of reviewer input is illustrated by a solid bar, while the mean of the individual inputs is represented by an "X".
4-9
NUREG/CR 4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 0.6 - 0.5 - 0.4 - Weighting 0.3 - Factor 0.2 - 0.1 0 Wetwell Drywell Head Drywell Knuckle Location of Failure for 138 psi Failure Pressure Figure 4-2. Range of Reviewer Input for Location of Failure for 138 psig Capacity ** The containment failure pressure weighting factors for case 2 are illustrated in Figure 4-
- 3. Comparison to the case 1 results indicates a decrease in estimated capacity due to the temperature increase.
0.6 - Case 2: T > 800* F 0.5 - X 0.4 x Weighting Factor 0.3 - I 0.2 - 0.1 x ; O K 84 psi 117 psi 138 psi 160 psi 225 psi Containment Failure Pressure (psig) Figure 4-3 Weighting Factors for Containment Capacity for High Temperature Conditions
**No mean values are indicated because the format of reviewer input did not lend itself to extraction of a mean for the conditions illustrated. See Appendix A. for details.
4-10
NUREG/CR-4551, VOL 3: DRAFT REPORT FO'R COMMENT (FEBRUARY,1987) - 4.2.7 Containment Failure- Size Another aspect of containment failure that is of critical concern is the size of the failure. Many~ previous assessments have assumed that failure implies a rapid depressurization but IDCOR has suggested that a temperature-induced creep failure of the drywell shell that could result in a leak that would depressurize the containment at a much slower rate. This issue is clearly correlated with the previous issue. The review group decided to use two levels: leak and rupture. The leak corresponds to a hole which does not depressurize the containment within two hours but which does prevent further pressure loading (this would be in the range of.03 sq ft to .25 sq ft). The rupture size refers to an opening which allows full depressurization of the containment in a short time period (greater than .7 sq ft). In order to provide the correlation with previous issues, several cases required reviewer input: Case 1. Wetwell failure due to pressurization (most likely rapid) with a l low torus temperature ( < 350 F). 4 Case ~ 2. Drywell failure (excluding head failure) due to a rapid f pressurization with a Jow torus temperature (< 500 F). Case 3. Drywell failure (excluding head failure) due to a slow pressurization with a high torus temperature ( > 800 F). I Case 4 Drywell head failure. The reviewers' inputs are illustrated in Figure 4-4. Case 4 is not shown in the figure because it was unanimously decided that the drywell head failures.would always be of the leakage type. For cases land 2, the reviewers judged there to be a very high likelihood of a rupture failure, and the same composite probability of rupture failure was used for each. For the high temperature drywell knuckle failure, there was an even split on assignment to leak versus rupture. Several reviewers cited the increased leakage at sites such as the equipment hatch that would start to contribute significantly at high temperature as the basis for their , I weightings for case 3. The composite values were used to examine this uncertainty. Cases 1 and 2 were combined in the LLH analysis such that the failure size was independent of location. For example, the rapid pressurization cases were assigned a rupture-type failure in 90% of the sample members, a leakage-type failure in the other 10%.
.4-11 o - - - - - - _ - _ _ _ __-- - - - - - - )
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 0.8 " 0.7 - 0.6 - Probability Leakage 0.4 - 1 4 Failure Mode 0.3 0.2 - 0.1 x X 0 Case 1 Case 2 Case 3 Containment Conditions (See Text) Figure 4-4. Range of Reviewer Input for the Probability of Leakage Failure Mode 4.2.8 Vessel Failure Mode There have been several failum modes postulated for the reactor pressure vessel (RPV) for a BWR. The Peach Bottom containment event tree considers four failure modes: (1) An in-vessel steam explosion which propels a water slug and core debris upward through the vessel to fail the upper closure head. This mode was postulated to fail containment (alpha mode) in WASH 1400. (2) An in-vessel steam explosion which directly fails the bottom head of the RPV. (3) Gross meltthrough of the RPV bottom head following slumping of the molten core debris into the lower plenum. This mode is modeled in the , STCP. l (4) Localized meltthrough of a penetration weld in the RPV bottom head as a , result of the slow accumulation of partially molten core debris in the lower I plenum. This mode is modeled in the IDCOR Modular Accident Analysis l Program (MAAP). l The first two modes were judged by the SARRP analysts and expert reviewers to be sufficiently unlikely as to have a negligible impact on any risk measure for Peach Bottom. BWR vessels have features which would tend to mitigate the effects of a steam explosion. Structures above the core (e.g., steam dyers and separators) would tend to absorb the impact 4-12 l l (
i l i i NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) . of an upward-directed steam explosion. The vessel supports (a lower head skirt supported by a pedestal) would tend to minimize the likelihood of a lower-head failure, j l However, the last two modes can have very different impacts on the containment' ] loading at vessel breach. If the RPV is at high pressure, the slump failure mode (#3 above) implies that a significant quantity of molten core debris is ejected under pressure into the drywell. The pressurized ejection of core debris may give rise to direct heating of the drywell atmosphere. Direct heating is discussed in detailin Appendix A. The basic premise is that the l dispersal of finely divided, hot core debris throughout the atmosphere will lead to oxidation of 'l the debris (zirconium and iron components) and enhanced heat transfer to the atmosphere. Pressure increase due to heating and hydrogen generation will result. Direct heating was assumed to be precluded by the localized failure mode (#4 above), since the vessel is assumed to depressurize before a significant quantity of molten debris is ejected for that failure scenario. Another issue (see section 4.2.10 below) addresses the containment pressurization that results from the vessel breach. Unlike the other SARRP plant studies, direct heating has been I divided into two issues. This issue addresses the conditions necessary for direct heating, and issue 6 treats the magnitude of direct heating, should it occur. If the RPV is at low pressure, the failure mode affects the quantity of debris available flow across the drywell floor and contact the drywell shell; possibly resulting in melting of the steel wall. This failure mode is discussed below in Section 4.2.12. Some of the experts held that if a slump failure occurred, it was virtually assured that the drywell shell would be melted when debris spread across the floor to the point of embedment (where the edge of the floor contacts the drywell wall). Some also observed that, in the flow melt scenario, the debris is at or below the liquidus temperature when it contacts the drywell floor and thus it will not spread to the wall immediately. In addition, the flow mode presumes that the core debris comes out of the vessel gradually. Thus the core-concrete interaction may erode the cavity at a a rate sufficient to retain the debris as it falls from the vessel; precluding contact between the debris ; and the drywell shell. Other experts felt that the likelihood of this containment failure mode would be more or less independent of the mode of vessel breach. It is acknowledged that the l RPV failure mode may impact this containment failure, although the weighting factors did not reflect such dependence. The uncertainty in the vessel failure mode arises from the lack of experimental data indicating how a reactor core would melt. There is no consensus among experts in' severe accident phenomena as to how rapidly core debris accumulates in the lower plenum. In addition, the quantity of debris which must accumulate before the vessel pressure boundary is 4-13
NUREG/CR455t, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) breached is not known. The complex geometry within the BWR vessel makes this issue more uncertain than for PWR's. Figure 4-5 shows the probabilities provided by the reviewers. The weighting factors indicate an approximately even split between the two meltthrough modes of vessel breach, although a number of individual reviewers thought that the slump mode (allowing high-pressure ejection) was less likely, as indicated by weighting factors in the 0.1 to 0.3 range. l l As with the other issues that required a probability as input, the issue was treated with a ! binary outcome: one case always has a slump meltthrough typified by the STCP, while the other case is a localized meltthrough typified by the MAAP model. The mean of the reviewen' responses was used, resulting in the assignment of 0.5 as a fraction of sample members in which a slump melt would occur. 1 0.7 - e 0.6 e e l 0.5 - e l Probability of STCP Type 0.4 - Melt-Through l 0.3 - e e 1 l l 0.2 e 0.1 e 1 2 3 4 5 6 7 8 Reviewer Figure 4-5. Range of Reviewer Input for Probability of Slump-Type Meltthrough 4.2.9 Containment Pressure Prior to Vessel Brecch t'or Station Blackout i Scenarios The preliminary ASEP work on Peach Bottom core-melt scenarios identified the accidents involving a complete loss of ac power, the station blackout sequences, as the ovenvhelming contributor to core-melt frequency. Since the containment safeguards are not available for this scenario, the level of containment pressurization that occurs prior to the breach of the reactor vessel is important to risk since it will contribute to the total pressurization immediately after vessel breach that could lead to containment f1ilure. There am a number of 4-14
i NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) uncertainties in the prediction of this pressure, including total heat generation, amount of hydrogen generation, the degassing of the drywell concrete and the timing of the sequence. Figure 4-6 delineates the range of weighting factor assignments on this issue for the long-term blackout accidents. It was generally concluded that the low value represented by level I was unrealistic due to the very low values (in the opinion of this review group) for all causes of pressure increase such as hydrogen generation values used in the IDCOR analysis (which was the basis for level 1). The 10-20% weight for this case accounts for the uncertainty in hydrogen that would lead to pressure values between 20 and 38 psig. Most of the reviewers assigned their largest weighting factor to the BMI-2104 result (38 psi). It was also a general conclusion that values higher than 62 psig would not be expected, and no calculations applicable to this case would point to pressures that high. 1-0.9-0.8 - 0.7 - 0.6 - i Weighting 0.5 - N Factor 0.4 - l A 0.3 0.2 i I g 0.1 x
; I i 0 x i 20 psi 38 psi 62 psi 80 psi l Containment Pressere Figure 4-6. Range of Reviewer Input for Containment Pressure Prior to Vessel Breach for Station Blackout This issue was only considered for long-term station blackout accidents (TB damage states). For the short-term accidents, heating of the containment and concrete degassing are not significant contributors to pressurization. The uncertainties in hydrogen generation were assessed by the SARRP analysts to have only a minor impact on the containment failure probability for the short-term blackouts, and thus the containment pressure prior to vessel breach was not sampled for these damage states. (On the basis of STCP calculations, the value was set at 35 psig.) In the initial ASEP assessrwnt the TB damage state was a significant 4-15 l
1 1 NUREG/CR-455t, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l- contributor to the total core-damage frequency. In the reevaluation used as the basis for this ,i o report (which uses a modified success criterion for the emergency service water system), the 'l TB damage state is far less significant. l 4.2.10 Containment Pressure Rise at Vessel Breach There are a number of uncertainties associated with the estimated pressure rise in the i containment which msults from meltthrough of the RPV when it is at high pressum. The most significant of these are: (1) The temperature of the gases expelled from the RPV; (2) The extent of direct containment heating by the dispersal; I (3) The quantity of hydrogen in the vessel (related to the in-vessel hydrogen generation rate); (4) The efficacy of the suppression pool in cooling the ejected gases; and (5) The quantity of steam and hydrogen generated from quenching of the core debris in water on the drywell floor.* There are also uncertainties in the containment pressure rise resulting from meltthrough of the RPV when it is at low pressure and there is water on the drywell floor (such as would result from the operation of the drywell sprays). For a low pressum meltthrough, the fifth area of uncertainty listed above is predominant. As originally posed to the expert reviewers, this issue considered six cases which dealt with the five areas of uncertainty in combinations that were relevant to the Peach Bottom containment analysis. These cases are presented, along with the experts' weighting factors on the various level in Appendix A. Only one of these cases was sampled in the limited issue sampling for this study. The other cases proved to have relatively minor impact on the containment performance (given the large uncertainty attributable to the issue of drywell meltthrough discussed below) or because the postulated preconditions were not likely. Only the first case of this issue was retained. This case assumes that the RPV is at high I pressure, the drywell floor is dry, and that the drywell vent downcomers am still submerged in the suppression pool. In the central estimate quantification of the CET,it is this case which is [ the predominant cause of containment overpressurization failure at vessel breach. The l weighting factors assigned by the reviewers for this case are presented in Figure 4-7.
'The pressure rise due to the interaction of core debris with water in the'drywell sumps was treated as a separate issue.
4-16
1 NUREG/CR-4551, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) i The uncertainty in pressure rise shown in the figure is not large in comparison to the uncertainty in containment failure pressure. In fact, the entire range represented is less than one-half of the range of uncertainty in containment failure pressure (about 70 psi). Thus it is
]
not surprising that this issue, as it was sampled, has little impact on the uncertainty in risk. j However, calculations performed subsequent to the experts' determination of weighting factors i indicate that the range is probably too small [4]. A case which presumed direct containment heating was not included. The primary reason for its exclusion was that the preponderance of outcomes lead to containment failure at vessel breach. It was thus decided to modify the containment event tree such that direct heating, if it occurs, always results in containment failure. Therefore the slump-type melt versus the flow type melt issue (Section 4.2.8), which determines the preconditions for dimet heating, becomes a surrogate issue for the importance of direct heating. The relative ; importance of direct heating to risk can be inferred from the results presented in Section 5 relative to the importance of the mode of vessel breach. i 0.9 - Case 1: High RCS Pressure 0.8 " No Direct Heating { 0.7 No Pool Bypass i Drywell Floor Dry i Weighting 0.5 -
- x Factor o,4 ,
0.3 - 0.2 - 0.1 0 l x x 30 psi 45 psi 60 psi 75 psi 120 psi Pressure Rise at Vessel Breach Figure 4-7. Range of ReviewerInput for Pressure Rise at Vessel Breach As noted previously, this issue involved several other cases which are described in l Appendix A. The reviewers also provided input on the pressure increase associated with the steam and hydrogen generation due to quenching of the core debris in water in the sump. The reviewers were fairly consistent in allowing for an additional 5-15 psi pressure increment due 1 1-17 \ _ _ _ _ _
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) to steam generation from a one-half full sump (500 gallons)--the actual input is provided in Appendix A. 4.2.11 Probability of Derwell Shell Melf through This issue results from a suggestion by researchers at Oak Ridge and Brookhaven National Laboratories (ORNL and BNL) which proposes that the there is a potential for a large amount of molten metal to be in direct contact with the steel shell after vessel bn:ach [1]. Their calculations indicate that this contact would result in failure of the steel shell of the drywell. In ORNL calculations, a large ponion of the core is metallic at 2300 K and this molten material i flows on the drywell floor in a six inch depth pool. This issue addresses the probability that l the drywell integrity will be lost by such a phenomenon after vessel breach. IDCOR calculations show a much lower temperature at the shell, but the assumed temperature of the melt was very low compared to many other estimates [6]. l The review group had differing views on the issue, some believing that the core material would quickly solidify while another had calculated a fairly constant temperature for the first hour due to oxidation of the nrconium in the melt. Others felt that the pressurized ejection would disperse the material well and it could cool on surfaces in the cavity and outside over the whole containment. In order to accommodate the boundary conditions that might effect this issue, the reviewer input was collected for four cases: 1) a slump-type melt as the STCP predicts,2) a flow-type melt into a dry cavity,3) a flow melt into a wet cavity and 4) a high pressure ejection scenario. The results for the review of this issue are shown in Figure 4-8. There was limited variation in the issue, for exarnple the first case was assigned probabilities of meltthrough from 0.5 to 1.0. Most of the reviewers thought that shell failure was a relatively likely for all cases. l One reviewer felt that this failure mode was cenain in all cases. All of the reviewers except two thought that drywell spray, or other supplies of water to the drywell floor might reduce the probability of meltthrough, but none believed that meltthrough failure would be precluded by the availability of water. The actual failure could be due to either melt or creep. Because this case called for probabilities rather than weighting factors, a composite was used to determine the fraction of sample members to be assigned drywell failure. Since the differences between the cases were not large, this uncertainty was represented by one issue and one-half of the sample members assume failure while the other half preclude drywell meltthrough.this. This simplified approach ense. red that a significant fraction of the sample members (fifty percenti did ! 4-18
NUREG/CR4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1987) not involve a drywell meltthrough, allowing risk and ec atainment failure probability to be evaluated and compared with and without this failure mode.
.1.0 "
0.9 - 0.8- x 0.7 - Probability 0.6 - x of Drywell 0.5- x Meltthrough o,4 x l 0.3 - 0.2 - 0.1 - - l 0.0 - l Slump Melt Flow Melt, Flow Melt, High Cy Cavity Wet Cavity Pressure Election Figure 4-8. Range of Reviewer Input for Probability of Drywell Meltthrough 4.2.12 Probability of Hvdrocen Burns in Reactor Buildine Sufficient to Cause Bvonss After containment failure, combustible gases (hydrogen and carbon monoxide) are released to the reactor building where they will burn when combustible mixtures with oxygen are obtained. Computer code calculations (using MARCH, HECTR and CONTAIN) performed as part of the SARRP program, indicated that combustion of these gases within the reactor building could produce pressure loadings several times greater than the design pressure. Since the reactor building is important relative to fission product retention, principally via deposition, an energetic event large enough to cause structural failure and direct paths to the environment is risk significant. The experts cited two principal sources of uncertainty associated with this issue: 1) uncertainties to the rate of combustion (flame speed) of the gases and 2) uncertainty in the structural response following the burn. IDCOR calculations do not - predict burning because of the low estimated rates of hydrogen and carbon monoxide production which do not result in the accumulation of combustible mixtures. The SARRP expert reviewers, however, believed that a combustible mixture would always form. Battelle calculations indicate many hydrogen burns that increase the average building leak rate to the 4-19
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) atmosphere and thus lower the decontamination effect. Recent analyses at ORNL show hydrogen burns with little associated pressure increase. Only two levels were assigned for the uncertainty analysis, making this an issue for selection of probabilities rather than weighting factors. The two levels correspond to no bypass and bypass significant enough to limit the residence time such that there is essentially l no decontamination. The reviewers provided input for two sequence types, station blackout and ATWS. The resulting probability assignments are illustrated in Figure 4-9. The results indicate considerable differences among the reviewers. Those who did allow for significant bypass based their results on the possibility of a hydrogen buildup. One of the reviewers found the potential for major building damage to be nil. Since the inputs were not really distinct for the two cases, this issue was treated as a single issue with the sampling based on the composite of the reviewers' inputs. As with other issues involving probabilities, the representation of the uncertainties was maximized by having some of the sample members have no bypass and others have complete bypass. Thus, the probabilities assigned were used to assign complete bypass for 20% of the sample members. l 0.6 x 0.5 0.4 x Probability X Station Blackout of Reactor 0.5 Building Bypass A ATWS 0.2 A 0.1 3 3 0 s
#1 #2 #3 #4 #5 Reviewer Figure 4-9. Range of Reviewer Input for Probability of Reactor Building Bypass i
4-20 l 4 I l
a NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 4.3 RADIOLOGICAL SOURCE TERM ISSUES The analysis of the radiological source term was accomplished using the STCP as the principal tool, but not all of the accident progression bins could be simulated due to time and budget constraints. In addition, the models of the STCP are known to involve both considerable uncenainty and limitations relative to modeling some specific phenomena. The uncertainties in sourte term assessment were more difficult to evaluate within the constraints of the LLH study due to the level of detail needed to properly consider individual effects. The
" issues" are defined more generally and include many more detailed considerations which are described mom fully in Appendix B.
The following radiological source term issues were included in the LLH analysis of Peach Bottom: (1) Magnitude of in-vessel miease from the fuel, (2) Amount of Csl decomposition in the RPV, (3) Amount of RPV retention, 1 (4) Suppression pool decontamination factor for aerosols, (5) Suppression pool scrubbing of volatile iodine, (6) Revolatilization of iodine and cesium from vessel following vessel breach, 1 l (7) Releases fmm the melt during cere-concrete interactions, { 1 (8) Reactor building and refueling bay decontammation factors, and I (9) Late releases of iodine from the suppression pool. Each of these issues is defined briefly below. The input of the expert review group is also provided for most of the issues to illustrate the range ofinputs assessed in the LLH study. For each issue, one of the distinct levels, or sets of outcomes, was constrained to be that which would be obtained using the STCP,if applicable. Thus the weighting factors help to illustrate the view of the experts concerning the uncertainty in the STCP in modeling the particular issues. 4.3.1 In-Vessel Release from the Fuel This issue is essentially the same as studied for the other SARRP reference plants, although the individual uncenainties are affected by the specific plant. There are a number of 4-21
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) uncertainties associated with the estimated releases from the fuel during the melting of the core in the reactor vessel. The models of the actual radionuclides releases are know to be uncertain due to a lack of experimental evidence and data. The factors most affecting the release such as core thermal history also cannot be assessed with great accuracy since they depend on the in-vessel melt progression. The levels and the assigned weighting factors are illustrated in Figum 4-10. As noted above, the levels include discrete outcomes for each of the source term groups. l The reviewers weighted the STCP result most heavily (labeled base case on the figure) but also I assigned significant weighting factors for the levels on eitner side of the base case. The most obvious result illustrated in the figum is that it was essentially a consensus that the high release fractions associated with level 4 were unrealistic. None of the calculations done for Peach Bottom would indicate the potential for release that high, and several of the myiewers indicated that there were reasons for expecting lower releases than those represented by level 4. It should be noted that higher release for this issue generally result in lower overall releases to the environment, because the in-vessel releases usually end up on deposited on vessel surfaces or in the suppression pool. WASH 1400 LEVEL 1 LEVEL 2 LEVEL 3 LEVEL 4 GROUP 1.gw Baig ligh Hich-Hiah 1-3: Xe ,Kr, I, Cs 0.5 0.89 1.0 1.0 4: Te 0.05 0.38 0.76 0.9 Sb: Ba 0.001 0.02 0.10 0.49 Sa: Sr 5.5 x 10-6 5.9 x 10-4 5.8 x 10-3 o,49 6: Ru 7.6 x 10-8 7,8xjo-7 8.2 x 10-6 c,11 7a: La 7.6 x t 0-9 8.5 x 10-8 8.3 x 10-7 0.0077 0.6 - 0.5 - 0.4 - Weighting Factor 0.3 - x >: 0.2 -. 0.1 - X 0
- Low Base High High-High in-Vessel Release from Fuel Figure 4-10. Range of Weighting Factors for In-Vessel Release from the Fuel l
4-22 1
1 i NUREG/CR-4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) 4.3.2 Amount of Cs1 Decomposition This issue, which applies to all of the plants being studied in SARRP, is a specific concem relative to iodine releases from the vessel. The STCP assumes that iodine will always i be present as Csl. Recent experimental data indicates that, in the presence of steam and j 1 radiation (as in the RPV), Csl may tend to decompose to more volatile forms such as elemental j iodine. The more volatile forms could then be released directly without retention in the vessel, l with a potentially significant impact on public health risk measures. At this time, the evidence ! is not conclusive and four outcomes representing a uniform distribution between 0.0 and 1.0 as the fraction of iodine that is released directly without retention due to conversion to a more vclatile state was used in the LLH study. It should be noted that this issue does not include the rather large uncertainties associated with the retention of the iodine that is in the form of CsI-- ! this is part ofissue 3. .j l The review group thought that there was great uncertainty in this issue as illustrated by their weighting factors shown in Figure 4-11. One of the reviewers thought that the issue was less uncertain for Peach Bottom; he stated that the BWR case is very different due to chemistry effects associated with the borated water that would very likely cause high reaction rates of Csl with the boron and result in substantial conversion of the CsI. His weighting factors indicate 1 that the conversion would be very effective, although probably not complete. The other ] reviewers weighting factors were driven by a lack of evidence to support any number, although this was expressed somewhat differently among the reviewers, with some giving very low weights to either end of the range. None of the reviewers felt that this issue could be ! discounted entirely (weighting factor on level I was never unity). l 4.3.3 Retention in the Reactor Pressure Vessel i There are significant uncertainties in our understanding of the physical processes ofin-vessel retention, and there are also limitations in the current modeling capability which may introduce very significant uncertainty in the results. Representation of this issue as an uncertainty parameter is very complex because it would be desirable to acknowledge the dependencies on the accident sequence, the individual fission product group, and the fuel release, while further recognizing that the resulting correlations could be complex and unique for individual species. Sensitivity studies for Peach Bottom have shown that retention is fairly substantial for the TB sequences, and fairly small for the others. Because of this, the LLH i analysis input was developed for two cases, one representing the TB sequences and the second 4-23
NUREG/CR 4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,19S7) 0.8 - 0.7 - 0.6 - 0.5 Weighting Factor 0.4 - g 3 0.3 - 0.2 - x j 0.1 . x 0
- 1. (0.0) 2. (.33) 3. (.67) 4. (1.0)
Csl Deconiposition Level Figure 4-11. Range of Weighting Factors for CsI Decomposition including all others. In the sampling process, it was assumed that the TB case was applicable for accident progressions in which there was no flow to the RPV from the control rod drive system and the decay power was relatively low (corresponding to a long-term station blackout.) The most significant code limitation that affects this issue is the fact that the current models do not adequately account for heat and mass transport due to natural convectic*n. The impact of this has been studied in QUEST and considered within the SARRP program. This issue was defined in terms of the fraction of a radionuclides group which was releared from the fuel which was deposited in 'the RPV at least until the time of vessel breach. Reevolution of this material following vessel breach is addressed by a subsequent issue (Section 4.3.6). The outcomes for two of the sequence groups are illustrated in Figure 4-12, as are the ranges of weighting factors assigned by the reviewers. The level structure was defined to allow ! maximum flexibility in assigning weighting factors that differ for the two sequence groups, but still retain the considerations as one LLH issue. For that reason, it was necessary to repeat a l level for each of the two sequence types. Levels 2-4 are derived from Peach Bottom STCP
)
msults. Recognizing the potential for values significantly different than any of those predicted ) by the code package, two additional levels were added to the range. The level 5 values were simply selected to represent very low retention factors, while the level 1 values were selected 4-24
NUREG/CR 4551, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) based on QUEST results and the Surry analysis. As indicated, the reviewers weighted the lower levels of retention more heavily. GROUPS LEVEL 1 LEVEL 2 LEVEL 3 LEVEL 4 LEVEL 5 FVES FOR TB SEQUENCES 2 0.04 0.24 0.24 0.90 0.97 3 0.04 0.16 0.16 0.82 0.97 4 0.02 0.09 0.09 0.22 0.9 5,6 & 7 0.02 0.14 0.14 0.70 0.9 FVES FOR TC SEQUENCES 2 0.04 0.24 0.90 0.90 0.97 3 0.04 0.16 0.82 0.82 0.97 4 0.02 0.09 0.22 0.22 0.9 5,6 & 7 0.02 0.14 0.70 0.70 0.9 0.7 - O.6 - . 1 O.5 - 0.4 - Weighting , Factor 0.3 - X I X 0.2 - 0.1 - x g 0 I --- f 1 l Lev.1 Lev.2 Lev.3 Lov.4 Lev.5 j in-Vessel Retention Level ! l Figure 4-12. Outcomes and Weighting Factors for In-Vessel Retention , 1 4.3.4 Sunoression Pool Decontamination Factor for Aerosols 1 There is significant uncertainty associated with the scrubbing phenomena associated l with the suppression pool for the aerosol species. Battelle has recently used the SPARC code I to perform a sensitivity study for Peach Bottom to investigate this issue, and has confirmed that the uncertainties are large: the resultant decontamination factors (DFs), expressed as the ratio 1 of aerosol masses entering and leaving the pool, ranged from slightly less than a factor of two I up to essentially complete scrubbing. Review of the results indicates no basis for eliminating either of the extremes. Additional uncertainty relative to SPARC predictions is indicated by available experimental evidence which illustrates decontamination factors as much as as order ! of magnitude higher than the comparable SPARC results [2]. Even for a constant set of input 4-25
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1987) assumptions, the decontamination varied widely with time, ranging from very large to a factor of two for the same sequence different times into the accident. Although the ultimate DFs are quite different for the vessel releases (DFVpa) versus the Jore-concrete interaction releases j (DFCpa), the factors governing uncertainty are similar and issue 4 treats both release types together. The difference in scrubbing for these two types of mieases result from the geometry differences between the drywell vents and the T-quenchers, as well as due to the different distribution on particle size, the vessel releases tending to be larger aerosols. The scrubbing of the core-conente species only applies to sequences like TC3 which involve wetwell venting. There is other evidence available relative to pool scrubbing effectiveness besides the Battelle sensitivity study. In general, other models tend to assign higher scrubbing values than those predicted by SPARC. In addition, recent experimental evidence also suggests very substantial decontamination factors [3]. The DFs for discharge through a quencher-type geometry with a five foot depths wem found to be in the 2000-5000 range. The decontamination levels and weighting factors selected by each of the reviewers are shown in Figure 4-13. The reviewers were nearly unanimous in their belief that the scrubbing of the ia-vessel release would be very large, with most of the weight assigned to level 2. The reasons for individual reviewer's weighting factors on this issue are somewhat complicated by the two parts of the issue, and are described fully in Appendix B. Levei_1 Leve! 2 Level 3 Leve!4 Level 5 DFVpa- 5000 200 40 10 1.8 Vessel Release OFCps-- 100 22 5 1.8 1.8 Core Concrete Rel. 0.8 ~ 0.7 0.6 - 0.5 x Weighting 0.4 Factor 0.3 0.2 m X g [ x 0.1 x I x 0 ' Lev.1 Lev.2 Lev.3 Lev.4 Lev.5 Level of Aerosol Scrubbing Figure 4-13. Range of Weighting Factors for Scrubbing of Aerosols 4-26
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 4.3.5 Suonression Pool Scrubbing of Volatile Iodine Snecies This issue is similar to issue 4 above, but is concerned with the scrubbing of volatile iodine species rather than aerosols. This issue is of importance if the CsI undergoes transformation to more volatile forms (see Section 4.3.2). It should also be noted that this scrubbing only applies to the early sources ofiodine and does not include consideration of the j late releases ofiodine that are treated within source term issue 9, discussed below. If some part of the Cslinventory is converted to elemental iodine the conditions would not favor high DFs: (1) the pool is likely to be at or near saturation, (2) there are no additives (such as boron used in PWR containment sprays) to enhance iodine retention, (3) continued sparging of the pool due to releases of hot gases and steam from the RPV could re-release absorbed iodine, and (4) boiling due to decay heat could also re-release some iodine. These same factors would be considerably less important if the volatile species was HI. Finally; there is uncertainty associated with the boundary conditions in containment, such as containment pressure, which can have significant impact on the scrubbing of the volatile iodine species. Releases of vapor iodine species am not treated in the STCP and a detailed assessment of this issue was not included in the scope of SARRP. The SARRP analysts suggested to the l reviewers a range of decontamination factors to essentially cover the cases of no scrubbing up to very effective scrubbing. The only evidence available on this issue when weighting factors were assigned was preliminary calculations done by Battelle Northwest that indicated very l effective scrubbing. These results were incomplete; for example, the sensitivity to pH of the water was not fully investigated. Values of decontamination factors are illustrated in Figure 4-14, along with the we ghting factors provided by the reviewers. The reviewers cited a lack of evidence as the reason for the weighting factors assigned, however, all believed that significant scrubbing was very likely. Several reviewers also believed that there was little difference between the first two levels since the scrubbing would j be very effective in either case. 4.3.6 Revolatilization Following Vessel Breach (FRVOL) This issue, which is analogous to the similar issue for Surry and the other plants, refers to the phenomena that could result in releases of volatile species from the RPV following vessel breach which are are not modeled in the STCP. If a substantial amount of these species are retained in the RPV, decay heating could eventually result in their revolatilization and addition i to the source term in containment. This issue has two components. The first is whether or not l 1 \ 4-27 l l l
1 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 1 l 1 0.5 - 0.45 - 0.4 - g 0.35 - l 0.3 - X X Weighting 0.25 - Factor x 0.2 - g 0.15- l
- l 0.1 0.05 -
l x l 0 I 1.(500) 2. (50) 3. (10) 4.(3) 5. (1) lodine Pool Scrubbing (Decontamination Factor) Figure 4-14. Range of Weighting Factors for Suppmssion Pool Scrubbing of Iodine whether or not temperatures in the RPV reach levels which are high enough to revolatilize the deposited fission products (the deposited chemical forms may have low vapor pressums). The second is whether there is sufficient flow from the RPV to the containment to sweep out the revolatilized fission products. For the BWR, very high drywell temperatures are predicted late in many accident sequences, and this heating could promote substantial volatilization. This issue has the potential to be very rick-significant, particularly if the inventory of Cs, I and Te predicted to be retained up until vessel breach is large, because the revolatilized inventory released after vessel breach may not undergo scrubbing in the suppression pool (or at least scrubbing may be reduced relative to the earlier releases). The phenomenology of revolatilization, including both the degree and the timing, is very uncertain. The IDCOR l results for Peach Bottom, derived with the MAAP code, predict very high levels of revolatilization (on the order of 90%). The MAAP results are driven by its prediction of very high drywell temperatures, and while the STCP does not include revolatilization phenomena, it does predict high drywell temperatures (700-1000 K), although these temperatures are less than those predicted in MAAP. In any case, the STCP temperature predictions would point to some revolatilization. Tellurium, Cs and I could all be subject to revolatilization, but for this issue the Cs and I are considered only, and these are considered as one issue with the same 4-28 j
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1987) levels. There is some dependence on the actual accident progression that is also not fully I modeled since the drywell temperatures reached are dependent on the accident scenario. Levels were derived for this issue to span this uncertainty, from no revolatilization to the very high levels predicted by the MAAP code. The levels are listed in Figure 4-15 along with the display of reviewers' weighting factors (FRVOL refers to the revolatilization fraction). Once again the weighting factors point to intermediate levels of revolatilization, levels 2 and 3, l with the two extremes, levels 1 and 4, weighted less. J 4.3.7 Release from the Melt Durine Core-Concrete Interactions ; { As currently modeled, the species released from the melt during core-concrete interactions (CCI) as calculated by CORCON-VANESA drive the central estimate of Peach Bottom risk. There are significant uncertainties in these release predictions that make this issue a high priority for the LLH study. Some insights into the range of uncertainty associated with l I this issue have been obtained from review of calculated releases as predicted by a number of l different studies. The results of these studies indicate that the CORCON-VANESA l 1 calculations for Peach Bottom have fairly high releases with fairly narrow ranges for Te and J Sr. The same studies also predict smaller, but significant, releases of La and Ce with moderate ranges across sequence differences and code input sensitivities. There are a number of CORCON modeling assumptions that were not investigated in these sensitivity studies, and there was no consideration of the uncertainties in the VANESA code. IDCOR predictions of l CCI release are significantly lower for Peach Bottom than those obtained from CORCON- l VAKESA calculations. l l 1he IDCOR results and the uninvestigated uncertainties in VANESA and CORCON lend support to treatment of the variations in this issue as an LLH sensitivity issue. The I SARRP analyst established four levels of release as illustrated in Figure 4-16. Levels 2 and 4 - ) l are essentially derived from the Peach Bottom sensitivity studies. Since the information was incomplete on the releases of Group 6, Ru, the RSS high value was used to represent chemical l uncertainties not yet addressed for these species. Level 3 represents the lower range of the TB sensitivity study also reported above. Level I was defined to represent the uncertainties that could lead to lower release fractions. The weighting factors assigned to the different levels are also illustrated in Figure 4-16. With one exception, the two central levels (2 and 3) were weighted with approximately 80% of the total: one reviewer weighted the lower values of levels 1 and 2 more heavily. 4-29
. i. .
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,' 1987) 0.7 - 0.6 - 0.5 - x 0.4 - Weighting Factor x 0.3 - 0.2 - X l 0.1 - l j X
' I 0
- 1. (0.0) 2. (.15) 3. (.5) 4. (.9)
Revolitization Level from RCS (FRVOL) Figure 4-15. Range of Weighting Factors for Primary System'Revolatilization VALUES OF FCCl410-FCOR) 3 SPECIES Level 1 Level 2 Level 3 Level 4 Xe, Kr, Cs, i 1.0 1.0 1.0 1.0 Te 0.1 0.5 0.75 0.93 Sr 0.01 0.1 0.73 0.73 , Ru <10 5 <$o 5 <3 o-5 0.05 I La 1.0 x 10-4 0.006 0.015 0.065 Ce 1.<0 x 104 0.006 0,028 0.12 Ba . 0.61 0.054 0.54 0.54 j l 0.6 - 0.5 - I I 0.4 - X ! l x , Weighting 0'3 - Factor O.2 - x l 0.1 I x I I 0-Lev.1 Lev.2 Lev.3 Lev.4 Core Concrete Releases Figure 4-16. Range of Weighting Factors for Core-Concrete Releases 4-30 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . - _ _ _ _ _ _ _ _ _ _ . - _ - - . _ _ .1
1 i NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1937) 4.3.8 Retention in the Reactor Building and the Refueling Bay Releases from the containment are frequently subject to holdup and retention within the ! reactor building. The effective decontamination factors associated with these phenomena (DFse) were selected as an issue for the LLH study due to their direct impact on risk. The , i calculated results (STCP) for Peach Bottom include this effect for nearly all releases, and while j the decontamination is not particularly large (DF = 1.5 - 2.5) it does play a role in the outcome. The biggest determinant of the decontamination effect is the role of recurrent hydrogen and carbon monoxide burns in the secondary containment. There is a great difference between the STCP methodology and the IDCOR ! methodology. The retention factors calculated in IDCOR are critical to risk because they are the sole reason that the large revolatilization values are not important--the releases are reduced typically by more than an order of magnitude due to secondary containment retention, i Phenomena that generate large DFsc values include: enhanced deposition due to large recirculation flows in the secondary containment, longer residence times due to containment failure by leakage rather than more catastrophic modes, the smaller combustible gas generation a rates modeled in IDCOR, and the methodology differences between IDCOR and the STCP in modeling aerosols. The levels for the the LLH study reflect this difference between the STCP results and 4 the IDCOR results. The first two levels reptesent the variability in STCP results, levels 3 and 4 represent IDCOR results in a relatively smooth continuum. Because of the large variations between these different calculations, se.luence-to4equence variation was not included and the l same levels were proposed for all seque nces. ) This issue was divided into two parts, one dealing with the releases to the reactor building and the other dealing with releases direct to the refueling bay (those associated failure of the drywell head). The weighting factors for DFsc values for release to the reactor building are illustrated in Figure 4-17, while the releases directly to the refueling bay were weighted with the lower DF values indicated in Figure 4-18. For the former, there was fairly good agreement that the two ends of the range should have 10-20% weighting factors. Although the l individual's inputs varied somewhat, the two middle ranges, levels 2 and 3, were each ( weighted about equally on the average. l
- Releases directly to the refueling bay would be expected to have substantially different effective decontamination factors and the many of the previous code predictions do not really apply. New values for the possible decontamination factors -were selected to span the relatively 4-31
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) . 0.7 - 0.6 - 0.5 - 0.4 - X Weightin9 Factor x 0.3 - 0.2 - x l 0.1 l x
)
0
- 1. (15) 2.(3) 3. (1.5) . 4. (1.0)
Reactor Building Decontamination (DF) Figure 4-17. Range of Weighting Factors for Reactor Building Decontamination Factor 0.8 - 0.7 - 0.6 - 0.5-Weighting 0.4 - X Factor 0.3 - x 0.2 - x 0.1 x O i 1.(5.0) 2. (3.0) 3. (1.5) 4. (1.0) Refueling Bay Decontamination (DF) Figure 4-18. Range of Weighting Factors for Refueling Bay Decontamination Factor narrow range of values believed reasonable by the review group. . The levels and results are indicated in Figure 4-18. The weighting factors selected by the individual reviewers indicated 4-32
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1937) i J considerable differences on this issue, although the range of decontamination factors with j significant weight is small: decontamination factors from a value of 1.0 to a factor of 3.0.
]
l 4.3.9 Lnte Releases of Iodine This issue accounts for potential sources ofiodine at relatively late times in the accident. j Revolatilization from the suppression was the primary consideration of this issue. For ] simplicity, the late sources estimated for this issue are treated as organic iodine, and no I decontamination by the pool or deposition within the containment structure is relevant. It should be noted that this simplified representation of late sources does not address the full i scope of this issue, particularly late sources of other species. Revolatilization (or resuspension) of iodine from the suppression pool is postulated to j result from three possible causes: 1) Pool flashing at containment failure,2) Pool boiling as a l l result of decay heating,3) Changes in the chemical form of fission products in the suppression j 3 pool (e.g., due to radiolysis). Most of the discussion at the review meeting focused on the j l third issue. l l The weighting factors for each level of release are illustrated in Figure 4-19. As j illustrated, it was a majority opinion that some of these mechanisms would result in late sources, although the fairly large release fraction oflevel 4 received only a 10-20% weight. 0.6 - i I 0.5 - X 0.4 -- Weighting 0.3 -.
- Factor 0.2 -
g X 0.1 -. X <' 1 O
- 1. (0.0) 2. (.05) 3. (.15) 4. (.5)
Late lodine Release Level (Release Fraction) l l Figure 4-19. Range of Weighting Factors for late lodine Release Levels l 4-33
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1087) l 1 References for Section 4 i
- 1. Kolaczkowski, A. M., et al. Analysis of Core Damage Frequency From Internal Eventst Peach Bottom, Unit 2. U.S. Nuclear Regulatory Commission Report NUREG/CR-4550, Volume 4, Sandia National Laboratories, Albuquerque, NM:
October,1986.
- 2. U.S. Nuclear Regulatory Commission. Evaluation of Station Blackout Accidents at Nuclear Power Plants, NUREG-1032, Washington, DC: May,1985.
- 3. Gwimann, L., et al. FinalReport, Containment Analysis Techniques, A State-of the-Art Summary, Ames Laboratories. U.S. Nuclear Regulatory Commission Report NUREG/CR-3653, Ames,IA: March,1984.
- 4. Shaffer, C. "MELCOR 1.6 Calculations for Vessel Depressurization in Mark I .,
Containments" Proceedings of the 14th Light Water Reactor Research Safety i Information Meeting, U.S. Nuclear Regulatory Commission, Washington, DC: 1986
- 5. Green, G. A., et al. " Mark I Containment Drywell: Impact of CCI on Containment Integrity and Failure of the Drywell Liner" Proceedings of the International Symposium on Source Term Evaluationfor Accident Conditions, IAEA-SM-281/36, Columbus, OH: October,1985.
- 6. IDCOR Task 17.S. Technology for Energy Corporation, Knoxville, TN: 1986.
- 7. Amos, C. N., et al. An Appraisal of Offsite Doses Resulting From Severe Accidents at Typical BWRs. General Electric Company Report NEDO-3088, San Jose, CA:
January,1985.
- 8. Cunnane, J. C., et al. "The Scrubbing of Fission Product Aerosols in LWR Water Pools Under Sevem Accident Conditions - Experimental Results," American Nuclear Society Topical Meeting on Fission Product Behavior and Source Term Research, Snowbird, UT: July,1984.
l 4-34 i L____________ _ _ - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - _ _ _ - _ _ _ _ - _ _
1 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 1 Section 5 RESULTS OF RISK REBASELINING The results of the SARRP study of Peach Bottom risk are discussed in this section. > Brief summaries of the results of each analysis area (core-damage frequency, containment response, radionuclides release and consequences) are included in Sections 5.1-5.4, i respectively. Section 5.5 contains the integrated calculation of risk. The results of the risk-reduction evaluation are discussed in Section 6. 5.1 CORE DAMAGE FREQUENCY RESULTS 5.1.1 Seauence Frequencies The ASEP study of the Peach Bottom plant resulted in a new estimate of the total frequency of core damage and of the individual accident sequences contributing to that total [1]. The total mean core-damage frequency reported by ASEP was 8.2 x 10-6 per year. It should be noted that for the SARRP analysis of BWRs, core damage is not synonymous with breach of the vessel by melted core material as was the case for the PWR analyses. The Peach Bottom containment event tree includes explicit modeling of events, particularly power recovery, which l may occur after core damage is initiated but which lead to an arresting of the accident in-vessel. i The dominant accident sequences are listed by plant-dsmage states in Table 5-1. The ! plant-damage states refer to accidents similar in terms of the safety functions that have failed and which also have similar initial conditions in containment. As described in Section 2, the plant-damage states were defined to be of sufficient discrimination to serve directly in the interface between the sequence frequency and the containment analyses. As illustrated in the table, two of the damage states (TBUX and TB) account for 80% of the total core-damage frequency. These sequences are both station blackout accidents--they involve failure of all on-site emergency ac power after a loss of offsite power. The sequences differ primarily in timing, with TBUX referring to an early loss of all cooling due to a complete failure of de power and TB involving initial injection success followed by failure after six to eight hours due to battery depletion. When all other station blackout scenarios are added in, the total I contribution due to this accident type is approximately 86%, clearly the dominant contributor to Peach Bottom core-damage frequency. The accidents traditionally known as ATWS scenarios (failure of the reactor to shutdown after a trip) represent about 12% of the core-damage frequency. Additional detail concerning the accident sequences calculated to be important to Peach Bottom is included in Appendix A and in the ASEP report for this plant (1]. 5-1
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-1 DOMINANT PLANT-DAMAGE STATES FOR PEACH BOTTOM Damage Description Mean Annual Statel Frequency TBUX Loss of offsite power, either as an initiating event or following a trip due to 4.2 x 10-6 i another cause, with coincident failure of all de power such that all core cooling fails and vessel depressurization is precluded. Power is assumed not recoverable and core damage ensues. TB As with the sequence above, the sequence is initiated by a loss of offsite power 2.3 x 10-6 as an initiator or after a trip. The core is cooled successfully (RCIC or HPCI) for about 7 hours until the batteries are depleted. The vessel repressurizes after. battery failure due to loss of ADS control. l TCUX Any initiator followed by failure of the reactor to trip (ATWS). The standby 4.8 x 10-7 j liquid control system (SLC) is initiated and works, but HPCI fails and the ! vesselis not depressurized. TCSRX2 An ATWS with failure of the SLC. Core cooling is initially maintained by 2.4 x 10-7 the HPCI which then fails due to high pool temperatures. The containment fails due to overpressurization (prior to core damage) but by a leakage failure mode that does not depressurize the containment. Because of high containment pressure the SRVs cannot operate in the relief mode, the vessel remains at high pressure and no injection is possible. TBUP Loss of offsite power, either as an initiating event or following a trip due to 2.0 x 10-7 (TBU) another cause, with coincident failure of all core cooling (either randomly or due to de power failures). One relief valve sticks open leading to a slow depressurization of the ves::el. Power is not reccvered in 35 minutes and core i damage ensues. ! TCSX Any initiator followed by failure of the reactor to trip (ATWS). The standby 1.7 x 10-7 liquid control system (SLC) fails or is not initiated, and HPCI fails later (-35 min) due to high pool temperature. The vessel is not depressurized successfully. TBU Short term station blackout with failur: of all injection, and power is not 1.6 x 10-7 restored in 35 minutes. For this sequence, the de power failure of core cooling is not included, and primary system depressurization is likely. TBP The sequence is initiated by a loss of offsite power as an initiator or after a 1.2 x 10-7 (TB) trip. A relief valve sticks open. The core is cooled successfully (RCIC or HPCI) for about 7 hours until the batteries are depleted. TCSRV23 An ATWS with failure of the SLC. Core cooling is initially maintained by 2.5 x 10-8 the HPCI which then fails due to high pool temperatures. The containment fails catastrophic due to overpressurization prior to core damage. The low pressure systems fail due to the saturated pool conditions and neither the H.PSW nor the condensate is initiated in time to prevent core damage.
)
Total (96% of total frequency) 7.9 x 10-6 l in some cases the plant-damage states were combined in the containment analysis. These are indicated in l parentheses. I 5-2
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 5.1.2 Uncertainty Representation l The results discussed above are mean frequencies. It was also a goal of the ASEP re-evaluation to estimate the important uncertainties. As described in Section 2.2, there were two types of uncertainty representation. The first uncertainty is that due to variation in the data used for the basic event frequencies and probabilities. Considering the uncertainty in the results due just to this data uncertainty, the results of ASEP with respect to core-damage frequency can be ! t represented by a distribution with the following parameters: J Upper 95% confidence bound 2.4 x 10-5 Mean value 8.2 x 10-6 Median value 5.1 x 10-6 Lower 5% confidence bound 1.3 x 10-6 As indicated above, there is about a factor of 20 uncertainty in the core-damage frequency associated with the input data. This degree of uncertainty, of course, applies to the total core- ! 1 damage frequency. The uncertainty due to data associated with individual sequences could be 1 1 greater or less. j l The second unce tainty characterization was carried out in ASEP as a sensitivity study. Some key uncertainties in modeling were examined through sensitivity studies which looked at the impact on the results if different assumptions had been made. These issues represent uncertainties in the phenomena that could lead to core melt including equipment success criteria, alternative uses of reliability data, and reevaluation of operator actions. The issues l l studied are listed below (other sensitivities were also investigated but not formally evaluated): Alternative Common-Cause Characterizations. The common-cause events were recognized as being an area of uncertainty in methodology that could have a significant impact on the results. Two alternatives were modeled: 1) i elimination of beta factors and 2) use of values from EPRI NP-3967 ! directly, rather than use of the means from that report as 95% bounds--the l base case for the study. Alternative Maintenance Frequency for ESWS Valve MOV-0498. It was l determined early in the study that the reliability of a single valve could impact the results. The value used in the base case (3.0 x 10-5) was changed to the ASEP generic value (8.0 x 104) for this sensitivity. Alternative ESWS Model and Success Criterion. The failure modes of the ' ESWS were important to the the potential for failures of diesel generators. The initial model required successful operation of the booster pumps, but later information indicated that the booster pumps are not required. The sensitivity study was used to investigate the effect of the new success criterion. [The SARRP risk study used the newer success criterion (no booster pumps needed) based on the recommendation of the ASEP analysts.] 5-3
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987)
- Ooerator Action Sensitivities for ATWS. This sensitivity study identified the effect of requiring SLC initiation within two minutes (instead of the four minutes assumed in the base case) and at the sane time increased the base case failure probability for the error of maintaining the level in the vessel too low.
. Combination of ESWS Model and ATWS Human Errors. The two studies listed previously wem considered in combination.
Based on these issues, a set of six sensitivity cases aimed at investigating their significance was developed as indicated in Table 5-2. As listed in the table, most of the sensitivities had relatively minor impact on the total core-damage frequency, and the largest . l effect was an approximate factor of three decmase if the beta factor representation of common- i mode failures was removed from the base case. 1 Table 5-2 -l SENSITIVITY STUDIES FOR CORE-DAMAGE ANALYSIS Total Core-Damage No. Issue Description of Sensitivity Study Fmquency--Mean
-- Core-damage fmquency for the bar.: case 8.2 x 104 _
- 1. Co nmon-cause Eliminate beta factor common-mode failures 3.4 x 10-6 failure rates
- 2. Common-cause Use alternative (more pessimistic) interpretation of beta factors 8.8 x 10-6 ,
failure rates for common cause failure rates
- 3. ESWS Maintenance Raise base case to ASEP generic value 1.2 x 10-5 4
- 4. ESWS Success Remove booster pump requirement from models 6.9 x 10-6
- 5. ATWS Actions Require SLC in two minutes and raise failure probability for level 9.8 x 10-6 control.
- 6. Combine 4 and 5 Realistic models of ESWS and ATWS. 8.5 x 10-6 The probability distributions (data uncertainties) for the individual events in the sequence cut sets were propagated both for the base-case analysis and for each of the six sensitivity cases. The result was the development of a " box-and-whisker" display for the uncertainties in the core-damage frequencies, as illustrated by the last display on the right in Figure 5-1. The bounds of the box represent the highest and lowest mean values from the sensitivity studies. The figure also includes the mean,5th percentile, and 95th percentile values for the base-case analysis, and the highest 95th-percentile and lowest 5th percentile values from the sensitivity cases. As illustrated, the variation in mean frequency of core-5-4
i l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) i damage associated with all of the sensitivity cases is limited to about a factor of three _ The 5th and 95th percentiles of the base case resulting from data uncertainties account for a factor of twenty uncertainty. Combining the worst combinations of uncertainties from the sensitivity and data uncertainty analyses yields an uncertainty of approximately two orders of magnitude. The results of the base case and uncertainty representations (sensitivity studies and data) are illustrated in Figure 5-2 for the most frequent damage states. The effect of the sensitivity studies are more readily seen at this level of detail. The TBUX damage state was 1 E .3 MEY. UPPER 95% ME DIAN LOW 5% 1E.- 1 2.e s m:.r
. ... c, . yo e,- y- - --
.3;y .
$-f '"'J'A'.T'" # '
"" { {6
- 3,;y m..
1..- .
. . . . s.r.
' mar e,_
tE.- il!
.ris
.na .
s!!ilati is!i!!!! l aiselcass ,i ,,l l .i .i .i causen4'e assuu
-.w._.
Figure 5-1. " Box-and-Whisker" Display of Uncertainties for Total Core-Damage Frequency 5-5 _ _ _ - - _ - - _ _ _ _ _ _ _ - - - - _ - _ _ . - - . _ l
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1907) only affected by sensitivity study one, elimination of the beta representation of common-cause failure. As described earlier, this damage state is dominated by a common-cause failure of the de power system, and the mean of the sensitivity fell below 1 x 10-8 when the common-cause event was eliminated. The damage state TB was less affected by the sensitivity cases, although cases three and four which affected the ESWS model did have a total impact that resulted in a factor of two increase and decrease in the mean damage-state frequency. The mean of damage state TCUX was not affected by most of the sensitivity studies because one of the sequences in the group which was insensitive to the issues dominated the mean frequency. However, the range of uncertainty is almost three orders of magnitude for this damage state because of the uncertainties involving the probability of the control rods failing to insert due to mechanical causes. Group TCSR was most affected by sensitivity case five which specifically involved the ATWS scenarios, while the sequence groups TBUP and TBU were most affected by the sensitivities involving the ESWS system. This latter sensitivity is important due to the dominance of station blackout scenarios which include a sensitivity of the diesel generators to the availability of ESWS cooling water. 10" g C 2 79
] TDUX ,,
4 30 :::= .. Z ..
~
I *R Tesx Teux _
- T TBUP TBU go 4 g
-'C;= ,
g y
=
ca
' -=r-l, [; -
. Ot:.
~
$mDU
'"~'
] ..g-.
= T "
- - L~.' ".,
- w. -_
Figure 5-2. Box and Whisker Plots for Important Peach Bottom Damage States 5-6
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) In keeping with the LLH methodology described in Section 2, the SARRP analysts drew on ASEP results and insights to suggest a number ofimportant uncertainty issues for the LLH study (see Section 4.1). The issues considered are listed below, along with the weighting l factors and outcomes used in the LLH: Issue Outcome - Weighting Factor -
- 1. Probability of Failure to P = .0007. 0.1 l
' Actuate the SIL P = .003 - 0.4 P = .034 (Base Case) 0.4 P = .15 0.1
- 2. Probability of Dc Power P = .000017 0.1 Common Mode P = .00024 0.4 P = .017 0.4 f P = .24 0.1 l'
- 3. Probability of Failure to Vent P = .7 0.1 Containment P =.88 0.4 P = .96 (.9 is Base Case) 0.4 l 2
P = .99 0.1
- 4. Probability of Non-Recovery P = 5% 0.1 of Electric Power P = 30% 0.4 (Values shown are the percentiles P = 70% 0.4 l of the probability distribution P = 95% 0.1 j for power non-recovery after 1 a given time period.)
The issues selected were chosen based bcth on their impact on the core-damage I frequency as well as their impact on the most risk significant scenarios. The impact of these issues on risk is discussed in Section 5.5. The results of the LLH computation of core-damage j frequency are listed below, illustrating the total impact of the four issues considered- l Highest Sample Member 4.2 x 10-5 ; 95th percentile 3.8 x 10-5 l Mean 6.8 x 10-6* ; Median 3.8 x 10-6 5 5th Percentile 1.2 x 10-6 ; Lowest Sample Member 8.2 x 10-7 l The total impact of all the issues is approximately a factor of thirty uncertainty in the core-damage frequency. As with the ASEP sensitivity cases, these issues had varying impact on the individual damage states, a feature important to risk since the risk profile i~ s highly j dependent on the makeup of of the core-damage profile.
- Note the agreement between the LLH mean and .ASEP sensitivity case 4 (which is the base !
case for the SARRP work). .j 5-7 1
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 5.1.3 Observations Concerning the Core-Damage Frecuency The Peach Bottom results are dominated by station-blackout sequences. The most important sequences involve failure or depletion of the de power systems, either a common-mode failure of both batteries following the loss of offsite power, or the expected depletion of batteries at about 6 hours into a total power loss. Other important contributors to these sequences are diesel generator common-mode failures and service water failures, particularly as related to the diesel cooling function. This finding is not surprising since the redundancy of core-cooling systems helps to limit the importance of most types of random combinations of failures, except for the common-cause contribution. The importance of this latter item, service ) water diesel cooling, may be overstated in the ASEP result. Recent interactions with the plant have indicated that the success criteria used in the models (the need for booster pumps) is incorrect, based on actual tests at the plant. Changing this success criteria would change this ! i vulnerability, although the total impact is not too great as indicated by the results of sensitivity 1 case 4 presented in the previous section. Once again, the SARRP risk analyses did incorporate this finding, and sensitivity case 4 is actually the base case for SARRP. The other important features of these dominant station blackout sequences include the offsite power loss frequency, the offsite power recovery frequencies for various times after the initiator, and the probability ofloss of offsite power as the result of a reactor trip. The ATWS sequences were not found to be particularly important to core-damage frequency (12% of the frequency), although there were two parts of the analysis considered quite uncertain and important to the frequency of ATWS scenarios. The fint is the probability of failure to actuate the standby liquid control system following an ATWS, although the uncertainty here is probably more controlled by the time available for action than by the operator failure rate, given that a prescribed time is available. The failure of the mechanical portion of the control rods is also important since it leads to non-recoverable failure of the scram system, but the data conceming this failure mode is very weak. It should probably also be noted that although the results are driven by station-blackout scenarios, the total frequency of core damage is relatively low. For example, the diesels are important to the blackout sequences in a relative sense, but it should be noted that Philadelphia Electric has placed great emphasis on diesel reliability and the diesels were found to be approximately an order of magnitude more reliable than the industry average. Thus the dominance of station blackout does not imply a plant vulnerability--it is just the most important sequence in a com-damage profile that is relatively low in frequency. The battery common-5-8
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) mode failure potential was found to be important, but the analysts believe that some minor procedural changes would help lower the probability of this potential common mode. When core-damage frequencies are assessed to be as low as these msults suggest, it is prudent to inject some words of caution regarding the conclusions to be drawn from the { quantitative msults. The details of the ASEP evaluation, limitations in scope and methodology, , I and discussion of the sensitivity analyses should be reviewed prior to the formulation of i conclusions regarding core damage frequency. The most obvous caveat is that the scope of the study reported here did not include tmatment of external initiating events (primarilly fire, flood, l l and earthquake), although current plans call for inclusion of external events at a later date. It l seems likely, given the low core-damage frequency for internal events, that the external events I could have a substantialimpact on the core-damage frequency. A second observation is that 1 the results are Peach Bottom-specific, a plant which has been evaluated several times with j respect to severe accident capability. These studies have probably resulted in improvements, such as the diesel reliability improvements, which might not be expected in a plant that had not been studied previously. 5.1.4 Comnarison to the Reactor Safety Study ' The Reactor Safety Study estimated a core-melt frequency for Peach Bottom of 8.2 x 10-5 (2), compared to the updated ASEP value of 8.2 x 10-6. It is difficult to compare the studies directly, since the RSS value is the sum of median values while the value reported here is a mean, but there are insights to be drawn from an analysis of contributors. The differences in the results are caused by two features: 1) plant modifications, including some in msponse to the RSS and as a result of continuing ar.alysis at the utility and 2) changes in PRA analysis techniques that cause focus on additional subtleties while attempting to be less conservative (more realistic) in the overall risk characterization. Some specific insights are listed below: Transients involving long-term loss of decay heat removal were dominant in WASH-1400 as well as in more recent BWR studies. Containment venting has largely eliminated the potential for containment failure from these types of sequences, and more analysis of equipment effects followmg containment failure indicates that most core-cooling systems am likely to survive. ATWS sequences are now assessed to be an order of magnitude less likely, due mostly to regulatory changes and increased understanding and modeling accuracy concerning the phenomenology of the event. The station blackout sequences are assessed to be two orders of magnitude more likely here than m the RSS, but due to changes in modeling and understanding rather than plant changes. The Peach Bottom diesels are actually quite reliable, but new nMeling techniques (in greater detail than in 5-9
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) RSS) identify common-mode failures, de power faults, battery depletion, and diesel cooling interactions which were not identified or modeled at the time of the RSS. The basic conclusions that transients and not LOCAs dominate the frequency appears to hold, although the character of the transients is different as described above Overall, the analysts were impressed with the fact that plant changes incorporated by the utility had addressed the concerns previously identified and this resulted in a lower core-damage frequency in spite of the capability of PRA to identify more subtle intenctions such as battery depletion which are now important to the results. 5.2 CONTAINMENT ANALYSIS RESULTS A brief myiew of the containment response results (which am intermediate results of the risk analysis) are presented below to allow a perspective on the predicted containment performance. Section 5.2.1 summarizes the results of the central estimate of the containment response. Section 5.2.2 includes the results of the LLH containment event tree analysis. The presentation of the results of the LLH analysis of containment response are considerably abridged. A wealth of information is available, but only a summary is provided here out of a concem for being concise and cogent. Additional perspective on the impodance of containment phenomena to the risk measures is provided in Section 5.5 of this report. 5.2.1 Central Estimate of Containment Resoonse The central estimate results of the containment analysis are documented in a separate report [2]. The results are summarized here to allow perspective concerning the role of various containment response endstates in the final risk calculation. The containment event tree outcomes can be soned into accident progression bins which describe the results in terms of the occunence and non-occurrence of key phenomena. One set of these accident progmssion bins is defined in Table 5-2. The letters that make up an accident progression bin definition-correspond to the letters defined in the table for each of the ten characters in the name. As . illustrated in the table, these definitions provide insight into the status of the containment as well as information concerning the type of phenomena any radionuclides releases will be exposed to. The central estimate of the containment response is listed in Table 5-3. For the TB damage state there are many accident progression bins which contribute to the overall result. About 40% of these bins are primarily defm' ed by the containment failure mode--an early i drywell rupture to the reactor building. Most of these outcomes result from the drywell shell 5-10
NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-2 DEFINITION OF THE TEN DIMENSION ACCIDENT PROGRESSION BINS Bin Parameter Outcomes 1 Sequence Type A TB, fast C TC, slow B TB, slow D TC, fast 2 Vesu:lBreach A Yes B No 3 Ac Power Availability A Not available B Recoveredlate C Recoveredearly D Always available 4 Core <oncrete Interactions A Yes, dry drywell floor B Yes, wet drywell floor C Initially, but arrested late in the accident D None i 5 Early Containment Failure A Drywell rupture to the environr:wnt I B Drywellleak to the environtrat l C Drywell rupture to the zefuelig bay i D Drywellleak to the refueling baj E Drywell rupture to the reactor building F D ywellleak to the reactor busting G Wetwell rupture or vent to the environment H Wetwellleak to the environment I I Wetwell ruptere or vent to the reactor building J Wetwellleak to the reactor building K No early containment failure 6 Containment Failure at A Yes Vessel Brexh B No 7 Late Containment Failure A Drywell rupture to the environment B Drywellleak to the environment C Drywell nipture to the refueling bay D Drywellleak to the refueling bay E Drywell rupture to the reactor building : F Drywellleak to the teactor building j G Wetwell rupture or vent to the environment H Wetwellleak to the environment ; I Wetwell rupture or vent to the reactor building J Wetwellleak to the reactor building K No late containment failure 8 DrywellSpray Availability A Sprays available from the start B Late spray availability only J C Early sprays only D No sprays 9 Direct Heating A Yes B No 10 Suppression Pool Bypass A Pool bypass both early andlate B Late pool bypass only C None 5-11 i
1 l
)
NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) ; i Table 5-3 d 1 CENTRAL ESTIMATE CONTAINMENT EVENT TREE RESULTS Accident Plant Dnmace State Progression Bin
- TB TBU TBUX TCUX TCSX TCSRX2 TCSRV23 AAAAEAADBB - 0.06 0.1 - - - -
AAAAEAEDAB - - 0.04 - - - - - AAAAEAEDBB - 0.22 0.35 - - - - AAAAIGBDBB - 0.08 0.05 - - - - AAAAKBEDBB - 0.06 0.04 - - - - AAAAKBFDBB - 0.06- 0.04 - - - - AABBEAEBBB - 0.04 0.02 - - - - AABBKBIBBC - 0.04 - - - - - AABCEAEBBB - 0.04 0.03 - - -- - l AABCKBKBBC - 0.04 - - - - - AACBEAEABB - 0.05 0.02 - - -- - AACBKBIABC - 0.06 0.04 - - - - AACDKBKABC - 0.03 - - - - - ABCDKBKABC - 0.14 0.08 - - - - BAAAEAADBB 0.06 - - - - BAAAEAEDAB 0.03 - - - - - BAAAEAEDBB 0.23 - - - - - - BAAAFABDBB U.03 - - - - -- - BACBEAEABB 0.05 - - - - - - BACBIBIABC 0.02 - - - - BACBKBIABC 0.04 - - - - BACDEAEABB 0.04 - - - - BACDIBIABC 0.03 - - - - - BACDKBKABC 0.02 -- - - - _ BBCDGBGABC 0.03 - - - - - _ BBCDIBIABC 0.09 - -- - - - BBCDKB:MBC 0.12 - - - - - CADBABAABB - -- - - - -- 0.03 CADBABADBB - -- - - - - 0.03 CADBEAAABB - - - - 0.02 - - CADBEAACBB - - - - 0.02 -- - CADBEAEABB - -- - - 0.08 0.1 - CADBEAECBB - - - - 0.07 - CADBEAEDBB - - - - - 0.09 - CADBEBAABB - -- - - - 0.03 CADBEBADBB - - - - - - 0.03 CADBEBEABB - - - - - - 0.09 CADBEBEDBB - - - - - -- 0.09 5-12 E____--_--__________ _ _ _ . _ _ _ _ _ - - _ _ _ _ _ _
l i J NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 1 i Table 5-3 (continued) CENTRAL ESTIMATE CONTAINMENT EVENT TPIE RESULTS Accident Phnt Dampe State ) Progression Bin TB TBU TBUX TCUX TCSX TCSRX2 TCSRV23 CADBIBIABC - - - - 0.03 0.03 - CADBIBICBC - -- - - 0.03 - CADBIBIDBC - - - - - 0.02 - CADBKBIABC - - - - 0.04 - - CADDABAABB - - - - - - 0.03 CADDABADBB - - - - - -- 0.03 i CADDDBDABB - - - - - 0.03 - i CADDDBDDBB - - - - - 0.03 - { CADDEAEABB - -- - - 0.08 0.12 -
)
CADDEAECBB - - - - 0.08 - -
)
CADDEAEDBB - - - - - 0.11 - CADDEBEABB - - - - - - 0.1 CADDEBEDBB - - - - - - 0.09 CADDIBIABC - .- - - 0.04 0.03 - l CADDIBICBC - -- - - 0.03 -- - 1 CADDIBIDBC - - - - - 0.03 - ! CADDKBKABC - - - - 0.02 - - CBDDEBEABC - -- - - - - 0.06 CBDDEBEDBC - - - - - -- 0.06 CBDDIBIABC - - - - 0.03 - 0.02 CBDDIBICBC - -- - - 0.03 - - CBDDIBIDBC - - - - - - 0.02 i CBDDKBKABC - - - - 0.14 -- - DADBEAAABB - - - 0.06 - -- - DADBEAEABB - - - 0.22 - - - DADBFACABC - -- - 0.03 - - - DADBKBIABC - - - 0.1 - -- - DADDEAEAAB - - - 0.02 - - - DADDEAEABB - -- - 0.21 - -- - DADDFACABC - - - 0.03 - - - DADDKBKABC - - - 0.04 - - - DBDDKBKABC = = = 1 21 = = = Approximate Total 0.79 0.90 0.81 0.93 0.74 0.57 0.72
*The central estimate bins are also assigned numbers as designators in addition to the listing of dimensions by letter as in this table. The key to the numbering system and the mapping of the twenty-two dimension bins into the ten dimension bins is provided in Appendix A.
l l 5-13
NUREG/CR4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) meltthrough failure mode (see Section 4 for more discussion of this issue). Overall, about one-half of the outcomes involve a bypass of the suppression pool late in the accident. For this damage state them is a 12% chance of no vessel breacn and themfore no containment failure. The TBU and TBUX damage state results parallel those discussed above for TB (the first letter is different due to the difference in the time evolution of the scenario). The drywell liner meltthrough has the biggest impact. For the TBUX damage state however, the probability of no vessel breach is minimal due to the nature of the accident (loss of all de power) which greatly limits recovery options. The damage states starting with a TC are all ATWS scenarios. The TCUX and TCSX damage states have very similar results. For these accidents there is roughly a 20% chance of arresting the accident prior to vessel breach. Once again the drywell failure to the reactor
^
building is the dominant response (more than 50% contribution). Some contribution from venting the containment is also seen. l The TCSRX2 and TSCRV23 damage states are both defined by containment failure prior to core damage. TCSRX2 involves a smaller failure that relieves further pressure buildup but which does not depressurize containment while TCSRV23 involves a catastrophic failure prior to core damage. These damage states are made up of many accident progression bins with no single bins accounting for more than about 12% of the total. 5.2.2 LLH Containment Analysis Results It is not practical to examine the results of the containment portion of the analysis alone for all of the 150 risk samples, hence it was decided to provide additionalinsight through one of the more familiar determinants of of the containment response--conditional probability of containment response at key times.* The term "early" is taken here to include all cases for which significant radionuclides releases to the environment occur at the time of reactor vessel l breach. Meltthrough of the drywell has been modeled as an early failure for the results presented in this report. There is a significant uncertainty with regard to the timing of this failure mode which has not been explicitly considered. Figure 5-3 illustrates the probability of containment response when considering all accident sequences (the sequences are weighted by their frequencies). The results are shown for the early and late time frames, and for failure ; versus vent and for the no failure outcome. It should be noted that terminology can be i somewhat imprecise with regards to containment response. In general," failure" is used here to l
' Probabilities of containment response presented in this section are conditional on core damage (i.e., given severe core damage has occurred).
l 5-14
- - - _ - - _ _ _ - - _ - - _ - - - - _ _ _ _ _ _ - - - - - - . - - - - - - - - --- 1
l 1 NUREG/CR-4531, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) ; i j 1.0 '
- LJte t
~'T-
! + 1 ig l -r-a i e 1 : I !
0.1 l e x ! l Late - E - a
- Failure s b : 8 -
~
$
- E 5
. No 0.01. Failure Early Failure ,
Early vent %. %. zem zmo 0.001 + + Figure 5-3. Conditional Probability of Containment Response: All Sequences Weighted by Their Frequencies mean any loss of pressure integrity, including leakage but excluding intentional venting. Cases involving containment venting with subsequent containment failure are classified as failure for this display, whereas containment venting without failure is classified as a vent release. As indicated, there is approximately two orders of magnitude uncertainty in the probability of early loss of integrity (failure / leak or venting). However, when these two early containment responses are combined to give an overall probability of early release, the uncertainty range becomes an order of magnitude, approximately extending from a probability of 0.1 to 0.9. I (This combination of early failure and vent is not illustrated.) The late containment failure range is large because there am cases which involve no late ; failure at all--these correspond to the 75 sample members which have a probability of unity of a drywell meltthrough (assigned to the early failure category), given that there is a vessel breach. It should be noted that in the central estimate the difference in timing between early and late containment failure is not large, the early failure occurs at approximately 2 hours while the late j failure occurs at approximately 4 hours. In the LLH analysis, the time delay before late failure i a occurs can be greater, depending on the containment failure pressure and the pmssure rise at 1 5-15 l
l 1 l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEuRUARY,1987) . l l l vessel breach which correspond to the sample member in question. Even the two hour delay 1 which was calculated for the central estimate is significant in terms of radionuclides release since ! l in late failures the containment is intact at the time of peak aerosol generation from core-concmte interaction. The "no containment failure" result provides some indication of the uncertainty in total containment response. There is approximately an order of magnitude uncertainty in the probability of no containment failure, given that them is com damage. The results of the rank-regmssion analysis of the probability of early containment failure is illustrated in Figure 5-4. The key issues affecting the LLH result is the probability of drywell meltthrough. Fifty percent of the se.mple members have drywell shell meltthrough as a l failure mode whereas the other fifty percent do not. This uncertainty therefore dominates, as illustrated in the figure. Due to the time constraints associated with this program, it was not possible to model this particular failure mode as fully as necessary to characterize its actual impact. In particular, the drywell meltthrough was treated as an early containment failure, as if the containment failure was concurrent with vessel breech. In reality, the failure would occur l at some later time and the initial releases from the core-concrete interactions would have been attenuated by sedimentation and scrubbing by the suppmssion pool up to the time of the actual meltthrough. 0.80 Statistically Significant 0.70 Not Statistically Significant 0.60 - o w ? c> . . man eME $
$ 0 50 ' 8 g $I$@g@g$$ .^ E $. ",; .$ $ i E E&
e @m y8@oRH Tlj s32 ? e 16 g 0.40 - E e - e e $ 9 "8 8e $ l E - E 3 a g83 $ .8 ese m 5 5
$ gE <$ $ ! k i d E $. M 23..8 j n E5 $$ ] 3
~
0.30 -
,a e ! I $
lS $o sSSE E8 $ $ $ j 5_mdea $e s.m 5 5M. 0.20 - e se So e s a . . _ % _% _m _ 0.10: oE3&
;9 g ? 5 mgm<o u.
a an3e m H I u. u. iiiE 8 m .u o o o o m o o o o >@e
$ $@m-eo e o o
- c. m, a. c.
0.00 q" - - " Issues Figure 5-4. Rank Regression for the Probability of Early Containment Failum In order to examine the role of other uncertainties in the estimate of the probability of containment failure, the results were recalculated assuming that the drywell meltthrough does 5-16
l NUREG/CR 4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) not occur. These results are depicted by the display in Figure 5-5, while the rank regression for this case is illustrated in Figure 5-6. (The rank regression for this case was actually j developed for the combination of early loss of integrity and early vent, although the results woud be similar for the early loss of integrity category alone.) The first conclusion is that while the drywell meltthrough was the largest uncertainty in early containment failure - probability, the removal of the failure mode does not change the overall result for early l containment failure significantly. The top of the uncertainty range for early failure changes ) from approximately 0.8 to 0.5, while the overall range does not otherwise change greatly. I There are however, more sample members with early containment failure less than 0.5. It is also interesting to not that the probability of the no contait ment failure is approximately the same with or without consideration of the drywell meltthrough. The rank regression indicates l 1 that other uncertainties result in an uncertainty range similar to the base case, the de power ! common mode uncertainty and the mode of vessel breach being the most important. The l sequence frequency issues are important to the uncertainty in containment failure frequency since those issues change the relative likelihood of damage states which have different containment response. In particular, any issue which changes the relatively likelihood of the j TB damage state can change the LLH result because this damage state includes an issue which samples the base pressure in containment as an LLH issue. As discussed in Section 4, the mode of vessel breach actually acts as a surrogate for the importance of direct heating in these results. The containment failure pressure and the pressure rise due to interaction with sump water are also significant. Therefom it appears that in the absence of drywell meltthrough, the uncertainties involving the occurrence of dimet heating become most important, but that other uncertainties concerning pressurization and containment capability am also important. 5.2.3 Comnarison to Other Studies Comparison of these results to other studies to of the Peach Bottom plant allows f additional insight into the changes in predicted behavior due to advances in the technology, and helps to illustrate uncertainties since different contemporary studies have differing conclusions. The results of the containment analysis are illustrated in Table 5-4 for this study compared to the R < valuation and the IDCOR study of Peach Bottom L4]. As illustrated there are very large differences in the predicted containment response. The RSS values are dominated by wetwell overpressure due to the assumption at that time that the weakest point would be in the wetwell. It is generally accepted that the RSS evaluation was incomplete and a more recent study has pointed to a much higher probability of a drywell failure mode (although this study also has uncertainties and areas of incompleteness. The IDCOR result is also quite different l l ! 5-17 u___-_--_- l
;{,jll!!li <
2EmQ5= .
2 g l g# n@b$ qO@>*<: ~gm . 6 * ~ l I l l ~ = o- ' y * " = t ** . i il . " - = b b a r* - o r . P " . . zo _ ?=ge oo ' 7 mEx ?E* ms* <i~ obo 8 *TE 1 3ny Y'r 9 a5 k gr c b o - g' " eE f ge
- Zo~ O8g oin mEag-# W9= = go~
z2 eE h%* MS =a l oin. t h n . c y 0 ah e ecm r e d 0 0 s .s s a e s er re r oh. e v s 0 8 er e r b e r s a e d r e c o we vs a5 < > p p , s b n g y r c h p c n e o k e o o py T T- d w e e i u b l a ry c pc a mS i a t e e br -. n ap o v s re o g t op ob . ud o T yr r W r a a o r ls e e se us hr n r cosaee s - d o r l b pno -a po. e c zs z m e arse elt iuid i t c A e ms rp ois t f v t a m - v l n . t u - .u n eco i o m t u l le t n n t l e b e p mb t oio. e c v r o e c a y w e v e v p r p e er r e er ru s, s re e or s i l l r u re o e r v u u ia l e r t r e d oo slia u u u ia l lia l i r e a wtc u pwry ae l ia o t w e w t t a f p r f p r o u p i p W ee r r o lp S lu lu ejtn n nn t d s f f . f. f. f l e t . s d r _ oL .d_ t c l ia c i a P iia a i z o o o o s e n o .ser mf u o o f S D F A T H FF S C CC C V C P$P P o ] )) . ] l {h i g 3<nc3 t>4 N W gwm &-@ 9gr{ s grE'o'ng e 1 QoE.cgaR o g.cN r Ea <9 dr3 4& K e:r." n= l l1!t!lIjll;!lJfil!jl <ll l;l.I l NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) from the SARRP evaluation for two reasons: 1) their containment evaluations indicated a greater potential for containment failure due to thermally-induced failure and 2) the drywell melt-through was not considered likely in the IDCOR analysis. In general, the IDCOR loadings also are smaller than those predicted in SARRP. These comparisons illustrate the uncertainty in the containment response, with the drywell meltthrough being the single most important uncertainty. Table 5-4 COMPARISON OF CONTAINMENT FAILURE MODES FOR THREE STUDIES Failure Mode Percent Contribution SARRP Central RSS IDCOR Station Blackout Sequences (TB) l Drywell Melt through 40 % -- -- Wetwell Vent Followed by Drywell Melt-through 15 % -- -- Drywell Overpressure 20 % -- -- Wetwell Vent 15 % -- -- Late DrywellOvertemperatum 2% -- 28 % Wetwell Overpressure 2% 20 % -- In-Vessel Steam Explosion -- 1% -- Wetwell Overpressure Before Com Melt -- 79 % -- No Containment Failure 6% -- 72 % l ATWS Sequences (TC) Drywell Melt-through 3% -- -- Wetwell Vent Followed by Drywell Melt-through 15 % -- -- Early Drywell Overpressum 44 % -- 1% Late Drywell Overpressure 44 % -- 12 % Wetwell Vent 35 % -- 78 % Late Drywell Ovenemperature -- -- Wetwell Vent and Late Drywell Overtemperature -- -- 9% Wetwell Overpressure 3% 99 % -- In-Vessel Steam Explosion -- 1% -- Wetwell Overpmssure Before Com Melt -- -- No Containment Failure -- -- 72 % 5-19 NUREG/CR-4551, VOL.3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 5.3 RESULTS OF THE RADIOLOGICAL SOURCE-TERM ANALYSIS This section describes the results of the source-term analyses for Peach Bottom in terms of the release fractions and characteristics of each accident-progression bin and cluster. The methods used to obtain these results are. discussed in Section 2.5 and Appendix B. Insights concerning the source term are obtained through the examination'of the risk results in combination with the source term magnitudes provided here, and discussion of the significance ; of the source terms with mspect to risk is provided in Section 5.5. ] 5.3.1' Source-Term Results for the Central Estimnte The bin definitions in Section 2.5 describe the characteristics of the source-term bins' and the technique used to derive the release fractions for the central estimate. The results of the calculations in terms of the fraction of initial core inventory that is released to the environment ! are presented in Table 5-5. The entries in the table include the release fractions for the nine l groups of radionuclides as well as the important boundary conditions used in the consequence analysis. The MACCS code includes provisions for calculations of releases at different times and those bins with releases better represented by two mleases (e.g., a puff release followed by another release over a longer time period) include two lines in the table, one for each time period. The central estimates were generally derived from STCP results for Peach Bottom, with some extrapolations to cover scenarios not analyzed directly through a STCP run. It should also be noted that the source terms provided in Table 5-5 were developed for use in the MACCS code. Unlike MACCS, the CRAC2 code is unable to treat releases involving multiple . j puffs. Therefore, the MACCS source terms were consolidated into single-puff releases for the CRAC2 code; the results ate provided in Table 5-6. f 5.3.2 LLH Source-Term Results As described in Section 2.5, the RELTRAC code was used to calculate the LLH source terms, and since practical limitations preclude the calculation of consequences for such a large number of source terms, they were clustered as described in Appendix B. The release fractions for the 54 clusters that were used are provided in Tables 5-7 and 5-8 for the MACCS and l CRAC2 calculations, respectively. The results in Section 5.5 present the fractional . contribution to risk for the clusters, and those values can be used to identify the source terms l contributing to the various risk measures; insights relative to risk are provided in that section. 1 I 5-20 l . _ _ _ _ . _ _ . _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ . _ _________.________________.____1______.____.____________.___._ NUREG/CR-4551, VOL3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l Table 5-5 PEACH BOTTOM CENTRAL ESTIMATE RELEASE FRACTIONS FOR I FOR THE MACCS CONSEQUENCE ESTIMATE l J Rel? Evac.2 Release Fraction By Group . Cond.1 Time Dus2 Elev2 Time Heat 2 3 4 6 7 1 5 8 9 ) Bin Prob. (hr) (hr) (m) (hr) (cal /s) Kr-Xe I Cs-Rb . Te Sr Ru La Ce Ba j 13 .04 3.1 0.01 30 3.1 9.3+7' 8.5-1 6.4-6 7.5-6 4.9-6 0.0 0.0 0.0 0.0 2.8-8 3.1 13.6 30 9.0+3 1.5- 1 3.4-4 3.1-4 4.7-3 1,4-2 6.1-8 4.8-4 9.5-4 1.0-2 2 .06 3.1 0.0 30 3.1 9.3 +7 8.5-1 6.4-6 7.5-6 4.9-6 0.0 0.0 0.0 0.0 2.8-8 3.1 13.6 30 4.5 + 5 1.5-1 9.8-4 8.6-4 7.1-3 2.3-2 8.6-8 7.2-4 1.4-3 1.6-2 3 .006 2.9 0.1 30 3.1 5.8+8 8.4- 1 4.3-7 2.37, 4.1-6 0.0 0.0 0.0 0.0 9.9-8 2.9 13.8 30 9.0+5 1.6-1 5.3-3 5.4-3 5.2-1 1.6-1 6.8-7 5.3-3 1.1-2 1.1-1 5 .03 16.8 0.0 30 3.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 9 .02 3.4 0.1 30 3.1 4.7+7 8.41 3.9-7 2.3-7 2.1-6 0.0 0.0 0.0 0.0 1.1-7 3.4 13.6 30 9.0+3 1.2-1 1.6-3 1.8-3 5.6-3 0.0 0.0 0.0 0.0 1.6-5 12 .01 3.4 0.1 30 3.1 2.4 + 9 8.4-1 5.8-6 1.7-5 1.6-4 0.0 0.0 0.0 0.0 1.8-6 3.4 13.6 30 1.3 + 6 1.6-1 6.8-3 7.0-3 5.7-2 1.6 1 6.8-7 5.3-3 1.1-2 1.1-1 14 .002 2.3 0.25 30 3.1 1.0+6 2.5-1 9.8-5 1.9-4 5.5-4 0.0 0.0 0.0 0.0 5.0-5 2.5 14.5 30 8.5 + 3 7.1-1 4.1-3 4.6-3 1.6-2 0.0 0.0 0.0 0.0 1.1 4 15 .001 2.3 0.47 30 3.1 2.9+7 9.0-1 3.4-4 6.6-4 4.0-3 0.0 0.0 0.0 . 0.0 1.0-4 2.7 14.2 30 1.2 + 6 9.8-2 9.0-3 9.4-3 6.4-2 1.6-1 6.8-7 5.3-3 1.1-2 1.1-1 22 .02 3.9 1.1 30 3.1 3.2 +5 9.3- 1 6.5-3 1.5 -2 2.1-2 6.8-2 7.8-8 3.2-3 5.1-3 3.2-2 5.0 12.0 30 1.3 + 5 6.6-2 1.8-2 1.3 2 2.4-1 3.9-1 8.0-7 3.2-2 4.7 2 3.0-1 26 .03 3.9 0.0 30 3.1 1.3 +8 8.1-1 1.2-4 1.3-4 3.1-5 0.0 0.0 0.0 0.0 9.7-7 3.9 13.1 30 9.3 +3 1.3- 1 6.6-3 1.6-2 2.2-2 7.2-2 8.2-8 3.4-3 5.4-3 3.4-2 ) 29 .02 3.9 0.0 30 3.1 1.3 + 8 8.1-1 1.2-4 1.3-4 3.1-5 0.0 0.0 0.0 0.0 9.7-7 3.9 13.1 30 9.3+3 1.9-1 7.4-3 1.6-2 3.3-2 9.0-2 1.2-7 4.9-3 7.6-3 4.8 2 l 32 .03 10.0 0.5 30 3.1 0.0 1.4-1 2.7-6 2.8 9 6.8-4 1.0-5 1.4-9 3.4-5 3.2-5 1.7-4 10.5 6.5 30 2.5 +5 7.7 1 6.1 6 2.9-9 2.8-3 1.5-5 3.9-9 9.9-5 8.0-5 4.0-4 l 1 33 .04 10.0 0.5 30 3.1 0.0 1.41 8.0-6 7.8-9 1.0-3 1.6 5 2.0-9 5.1-5 4.7.-5 2.7-4 l 10.5 6.5 30 1.2 +6 7.7 1 1.8-5 8.0-9 4.2-3 2.3-5 5.5-9 1.5-4 1.2-4 6.4-4 44 .03 3.9 0.89 30 3.1 4.1 + 5 9.3-1 3.5-3 7.8-3 1.3-2 3.6-2 2.9-3 1.7-3 2.7-3 1.7-2 4.8 12.2 30 1.3 +5 7.1-2 1.9-2 1.8-2 2.31 4.2-1 1.2-4 3.3-2 5.0-2 3.1-1 45 .30 3.9 1.1 30 3.1 3.2 + 5 9.4- 1 6.4-3 1.5-2 2.1-2 6.8-2 7.8 8 3.2-3 5.1-3 3.2-2 5.0 12.0 30 1.3 + 5 6.42 1.8-2 1.3-2 2.4-1 3.9-1 B.0-7 3.2-2 4.7-2 3.0-1 48 .08 3.9 1.1 30 3.1 3.2+5 9.4- 1 1.8-2 4.2-2 3.1 2 1.1-1 1.1-7 4.8 3 7.6-3 5.2-2 5.0 12.0 30 6.6+5 6.4-2 5.2 2 3.6-2 3.6-1 6.2-1 1.1-6 4.8 2 7.1-2 4.8-1 50 .03 10.2 0.1 30 3.1 6.5+7 8.9-1 1.6-5 1.6-8 3.7-3 6.0-5 7.6-9 1.9-4 1.8-4 9.4-4 10.2 6.8 30 2.3+5 1.1-1 8.5-6 2.4-9 4.6-3 1.6-5 5.7-9 1.5-4 1.2-4 5.8-4 56 .06 17.0 0.0 30 3.1 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 5-21 NUREGCR-4551, VOL.3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-5 (Continued) PEACH BOTTOM CENTRAL ESTIMATE RFI EASE FRACTIONS FOR FOR THE MACCS CONSEQUENCE ESTIMATE Rel2 Evac.2 Release Fraction By Group Cond.! Time Dur2 Elev2 Time Heat 1 2 3 4 5 6 7 8 9 Bin Prob. (hr) (hr) (m) (hr) (cal /s) Kr-Xe 1 Cs-Rb Tc Sr Ru La Ce Ba 58 .01 7.3 1.0 30 5.0 0.0 5.1-1 1.9-2 1.7 2 1.1-2 2.3-2 3.2 8 1.4-3 2.2-3 1.5-2 8.3 12.0 30 2.1 +6 4.9-1 3.8-2 4.3-2 l'6-1 . 3.1-1 5.6-7 2.'4 2 3.5-2 2.31 ! 63 .008 7.0 1.2 30 5.0 3.9 +6 9.31 1.1 2 9.8-3 1.4 2 4.2-2 4.3-4. 2.7-3 4.2-3 2.7-2 8.3 12.0 30 1.1 +6 7.4 2 2.3-2 2.8-2 2.0-1 4.4-1 8.0-7 3.4-2 5.0-2 3.1-1 ) 64 .08 7.0 1.2 30 5.0 3.9 + 6 9.2-1 1.2-2 1.1-2 1.5-2 4.1-2 6.4-8 2.6-3 4.2 3 2.6 2 8.3 12.0 30 1.1 + 6 8.2-2 2.5-2 3.1 2 2.3-1 4.4-1 8.5-7 3.4-2 5.0-2 3.1-1 66 .02 7.0 1.2 30 5.0 3.9+6 9.2-1 3.5-2 2.9-2 2.3-2 6.6-2 9.0-84.0-3 6.3-3 4.2 2 8.3 12.0 30 5.7 +6 8.2-2 7.3-2 8.8-2 3.4-1 7.0-1 1.2-6 5.1-2 7.6-2 5.0 1 72 .01 10.2 0.1 30 5.0 4.7+7 8.3-1 .7.1-7 9.9-7 5.8-5 5.0-5 2.0-9 3.7-6 5.6-6 3.5-5 10.2 10.1 30 1.2 +4 1.7-1 1.2-6 1.6-6 2.4-4 9.95 6.6-9 1.2-5 1.6-5 9.4-5 75 .02 6.8 0.1 30 5.0 4.7+7 1.0 7.1-5 1.4-4 9.7-4 1.4-6 0.0 0.0 0.0 2.9-5 76 .007 20.3 0.0 30 5.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 ' 77 .03 20.3 0.0 30 5.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 I This is the conditional probability of the bin given com damage. The source terms are only provided for bins with non-zero probability in the central estimate. 'Ihe other bins are needed for the analysis of safety options. A full listing of source terms for all the bins is provided in Appendix B. 2 Release time is the time between shutdown and release, duration is the time during which the release occurs, elevation is the height at which the release occurs, evacuation time is the time at which evacuation starts. 3 The key describing the link between bin numbers and the definitions of the bins is included in Appendix A. 4 Values in this table are written in shorthand notation: 9.3+7 - 9.3 x 107 l ) 1 i 5-22 NUREG/CR-4551, VOL.3: DRAFT REPORT FOR COMME!G (FEBRUARY,1987) Table 5-6 PEACH BOTTOM CENTRAL ESTIMATE RELEASE FRACTIONS FOR 'Il-IE CRAC2 CONSEQUENCE ESTIMATE Rel2 Warn 2 Release Fraction By Group 2 Cond. Time Dur Elev2 Per. Heat 1 2 3 4 5 6 7 8 9 Bin Prob.1 (br) (br) (m) (hr) (cal /s) Kr-Xe I Cs-Rb Te Sr Ru La Ce Ba 13 .04 3.1 1.0 30 1.5 9.3+74 1.0 3.5-4 3.2-4 4.7 3 1.4-2 6.1-8 4.8-4 9.5-4 1.0 2 2 .06 16.8 0 30 15 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 3 .006 2.9 1.0 30 1.4 5.8 + 8 1.0 5.3 3 5.4-3 5.2-2 ~ 1.6-1 6.8-7 5.3-3 1.1-2 1.1-1 5 .03 16.8 0 30 15 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 9 .02 3.4 1.0 30 1.8 4.7+7 9.6 1 1.6-3 1.8-3 5.6 3 0.0 0.0 0.0 0.0 1.6-5 12 .01 3.4 1.0 30 1.8 2.4+9 1.0 6.8-3 7.0-3 5.7 2 1.6 1 6.8-7 5.3-3 1.1-2 1.1 -1 14 .002 2.3 1.2 30 0.7 1.1 + 6 9.6-1 4.2-3 4.8-3 1.7-2 0.0 0.0 0.0 0.0 1.6-4 15 .001 2.3 1.5 30 0.7 2.9+7 1.0 9.3-3 1.0-2 6.8-2 1.6-1 6.8-7 5.3-3 1.1-2 1.1-1 22 .02 3.9 2.1 30 2.3 3.2 + 5 1.0 2.4-2 2.8-2 2.6-1 4.6-1 8.8-7 3.5-2 5.2-2 3.3-1 26 .03 3.9 1.0 30 2.3 1.3 + 8 9.4-1 6.7-3 1.6-2 2.2-2 7.2-2 8.2-8 3.4-3 5.4-3 3.42 29 .02 3,9 1.0 30 2.3 1.3 + 8 1.0 7.5-3 1.6-2 3.3-2 9.0-2 1.2-7 4.9-3 7.6-3 4.8-2 32 .03 10 1.5 30 8.4 0.0 9.1-1 8.8-6 5.7-9 3.5-3 2.5 5 5.3 9 1.3-4 1.1-4 5.7-4 33 .04 10 1.5 30 8.4 0.0 9.1-1 2.6-5 1.6-8 5.2-3 3.9-5 7.5-9 2.0-4 1.7-4 9.1-4 44 .03 3.9 1.9 30 2.3 4.1 +5 1.0 2.3-2 2.6-2 2.41 4.6-1 3.0-3 3.5-2 ' 5.3-2 3.3-1 45 .30 3.9 2.1 30 2.3 3.2 + 5 1.0 2.4-2 2.8-2 2.6-1 4.6-1 8.8-7 3.5-2 5.2-2 3.3-1 48 .08 3.9 2.1 30 2.3 3.2+5 1.0 7.0-2 7.8-2 3.9-1 7.3-1 1.2-6 5.3-2 7.9-2 5.3-1 50 .03 10.2 1.0 30 8.6 6.6+7 1.0 2.4-5 1.8-8 8.3-3 7.6-5 1.3 8 3.4-4 3.0-4 1.5-3 56 .06 17 0 30 15 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 58 .01 7.3 2.0 30 3.8 0.0 1.0 5.7-2 6.0 2 1.7-1 3.3-1 5.9-7 2.5-2 3.7-2 2.5-1 63 .008 7 2.2 30 3.5 3.9+6 1.0 3.4-2 3.8-2 2.1-1 4.8-1 4.3-4 3.7-2 5.4-2 3.4-1 64 .08 7 2.2 30 3.5 3.9+6 1.0 3.7-2 4.2-2 2.51 4.8 1 9.1-7 3.7-2 5.4-2 3.4-1 66 .02 7 2.2 30 3.5 3.9 + 6 1.0 1.1-1 1.2-1 3.6-1 7.7-1 1.3-6 5.5-2 8.2-2 5.4-1 72 .01 10.2 1.0 30 6.7 4.7+7 1.0 1.9-6 2.6-6 3.0-4 1.5-4 8.6-9 1.6-5 2.2-5 1.3-4 75 .02 6.8 0.01 30 3.3 4.77 1.0 7.1-5 1.4-4 9.7-4 1.4-6 0.0 0.0 0.0 2.9-5 76 .007 20.3 0.0 30 16.8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 77 .03 20.3 0.0 30 16.8 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1 This is the conditional probability of the bin given core damage. The source terms are only provided for bins with non-zero probability in the central estimate. The other bins are needed for the analysis of safety options. A full listing of source terms for all the bins is provided in Appendix B. 2 Release time is the time between shutdown and release, duration is the time during which the release occurs, elevation is the height of the release, warning period is the time between notification and the release. 3 The key describing the link between bin numbers and the definitions of the bins is included in Appendix A. 4 Values in this table are written in shorthand notation: 9.3+7 - 9.3 x 107 5-23 l NUREG/CR-4551, VOL3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) a . Table 5-7 PEACH BOTTOM CLUSTER RELEASE FRACTIONS FOR MACCS . Rel2 Evac.2 - Release Fraction By Group Usage! Time Dur2 Elev2 Time Heat 1 2 3 4' 5. 6. 7 8 E9 ; Clus Fac. (hr) (hr) (m) (hr) (cal /s) Kr-Xe ! Cs-Rb Te Sr Ru L.a Ce Ba 1 .01 6.8 - 0.5 30 3.1 0.0 7.7 23 7.3-5 8.0-14 1.1-5 6.2-5 1.7-11 2.6-8 3.2-8 4.9-6 7.3 9.5. 30 6.5+4 5.7-1 7.5-4. 4.9-14 1.9-5 1.3-4 .5.612 8.5-9.'9.2 2.9-6 ; 2 .05 ' 6 .01 -30 3.1 4.7 +7 8.4-1 1.4-5 8.7-6 2.7-4 3.3 4 1.9-9 2.9-64.7-6 1.8-4 I 6 11 30 1.1 +4. - 1.6-1 9.0 3 1.1-5 7.5-4 1.1-3 1.6-9 2.4-6 3.7-6 2.2-4' 3 .10 6.7 .01 30 3.1 6.5+7 8.7-1 2.0-41.4-4 1.5-4 7.0-6 1.3-8 1.3-7 1.4-71,2-5 6.7 11 30 1.6 +5 1.3-1 4.4-2 3.6-3 3.1-3 7.3-6 3.5-8 3.3 7 2.8-7 : 2.4-5 4 .11 13 .01 30 5 9.8 + 9 1.0 9.0-3 1.7-4 1.6-2 4.8-3 2,4-7 ' 3.1-4 4.6 8.0-3 13 7.4 30 3.4 +6 2.7-4 9.4-3 6.6-5 3.2-2 2.9-3 3.5-7 4.1-4 5.0-4 8.2-3 5 .11 5.5 .05 30 5 9.8 + 6 3.8-1 4.4-4 1.7-5 3.4-6 7.9-8 . 0.0 0.0 0.0 . 1.7-6 6 14 30 9.7 +5 6.2-1 4.1 2 8.2-3 6.1-2 2.9-2 2.9-6 1.7-3 1.7-3 1.4-2 6 .07 7 0.01d 30 5 3.9+9 . 9.0-1 4.7-4 2.0-4 4.4-4 6.3-6 3.1-4 6.3-6 6.3-6 8.4-6 ) 7 13 30 9.3 + 3 9.9 2 1.7 1 6.5-2 8.8 2 2.3 2 '2.0-6 1.3-3 1.3-3 1.2-2 7 .08 7 0.01 30 5 1.7+9 9.2-1 2.3-6 6.3-6 1.1-5 3.3-7 1.4-5 2.9-7 2.9-7 7.2-7 7 13 30 5.2 +6 8.4 2 1.41 1.2-1 2.0-1 8.6-2 2.3-3 5.2-3 5.213 4.2-2 8 .06 5.5 .75 30 5 3.3 +7 6.5-1 2.9-2 5.5-2 1.1-2 2.5-4 0.0 0,0 0.0 5.2 3 6.3 14 30 4.9+6 3.5-1 2.3-2 6.9-2 1.9-1 3.5-1 4.7-6 7.2.1 1.3-2 2.4 9 ,06 7 0.0 30 5 8.7+9 1.0 3.72 5.6-2 4.0-2 1.1-3 4.1-2 8.2-4 8.2-4 5.2 3 7 13 30 5.2+6 0.0 1.5-2 8.1-2 3.7-1 6.9 .1 9.0-6 1.4-2 2.7-2 4.6 1 1 10 .003 3.1 0.0 30 3.1 9.3 +7 4.5 1 4.2-6 4.0-6 3.6-7 0.0 0.0 0.0 - 0.0 '4.0-9 11 0.0 3.4 .84 30 3.1 2.6+7 1.1 1 9.7-6 1.0-4 5.6-5 2.1-5 2.6-9 1.5-6 . 1.6-6 1.3 5 4.3 13 30 8.0+5 8.2-1 5.9-4 1.1-4 9.8-4 3.0-4 2.9-8 1.8-5 1.8-5 1.6-4 12 .02 2.5 .25 30 3.1 0.0 2.5-1 2.8-3 4.1-5 7.7-6 1.8-7 0.0 0.0 0.0 3.2-6 . 2.8 14 30 2.5 +6 3.9-1 5.7-3 6.1-5 1.2-5 3.1-7. 0.0 0.0 0.0 5.2 13 .0009 3.4 .84 30 3.1 2.6 +7 1.0-1 1.1-4 1.4-4 7.7-5 2.1-4 3.5-9 5.4-6 1.0-5 1.8-4 4.3 13 30 P. 0+5 8.1-1 2.8-4 1.5-4 1.4-3 3.0-3 4.1-8 6.2-5 1.1-4 2.2-3 14 .0005 3.6 .71 30 3.1 7.3 +5 8.5-1 8.1-4 6.7-4 5.3-4 2.2-4 6.6-6 1.6-5 1.7-5 1.3-4 4.3 12 30 2.0+4 1.5-1 8.5-2 5.0-4 5.3-3 1.8 1.7-7 1.1-4 1.1-4 .9.7-4 15 .003 2 .46 30 3.1 0.0 1.9-1 4.5-3 3.8-3 7.1-4 1.6-5 0.0 ' O.0 0.0 3.0-4 2.5 14 30 2.5 +6 4.1-1 1.2-2 8.3-3 1.6-3 3.9-5 0.0 0.0 0.0 6.8-4 16 .0009 2.5 .25 30 3.1 0.0 1.3-1 8.6-2 3.4-4 8.5-6 0.0 0.0 0.0 0.0 4.8-7 2.8 .14 30 2.5 + 6 1.9-1 1.3-1 5.1-4 1.4-5 0.0 0.0 0.0 0.0 9.2-7 17 .0005 2.3 .21 30 3.1 1.2 +6 9.8 2 1.6-7 6.2-8 1.3-8 1.5-11 0.0 0.0 0.0. 1.1 9 2.5 14 30 8.6 + 3 7.9-1 1.4 3 1.2-5 4.4-6 2.3-8 0.0 0.0 0.0 6.4 18 .003 2.5 .25 30 3.1 0.0 1.3-1. 3.0-4 1.2-6 8.0-8 0.0 0.0 0.0 0.0' 1.2-8 2.8 .14 30 4.4 +4 7.6-1 1.7 3 8.9-4 1.5-4 2.4-3 5.6-6 5.0-5 9.4-5 1.8-3 19 .007 2 .46 30 3.1 0.0 1.9-1 6.0-4 5.0-4 9.3-5 2.1-6 0.0 0.0 0.0 3.8 2.5 14 30 4.3 +4 7.8-1 8.8-4 9.0-5 5.1-4 4.3-4 7.2-6 2.65'2.6-5 2.3-4 j 20 .02 3.1 0.01 30 3.1 9.3 +7 7.8-1 6.4-6 7.3-6 2.7-6 1,4-8 0.0 0.0 0.0 3.9-8 3.1 14 30 9.0+3 2.2-1 1.1-3 7.9-4 3.8-3 1.2-3 1.2-7 7.2-5 7.1-5 6.44 5-24 l NUREG/CR-4551, VOL3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-7 (Continued) l PEACH BOTTOM CLUSTER RFMASE FRACTIONS FOR MACCS Rel2 Evac.2 Release Fraction By Group 3 2 Usage! Time Dur2 Elev Time Heat i 2 3 4 5 6 7 8 9 Clus Fac, (hr) (hr) (m) (br) (cal /s) Kr-Xe . 1 Cs-Rb Te Sr Ru La Ce Ba 21 .006 3.1 .01 30 3.1 4.7 +7 4.5-1 2.3-5 1.4-7 2.2-8 0.0 0.0 0.0 0.0 1.6-9 ' 3.1 14 30 9.0 +3 5.5-1 2.8-2 1.1-2 1.8-2 3.6-3 3.5-7 2.1-4 2.1 4 2.0-3 l 22 .03 3.1 0.01 30 .3.1 9.3 +7 7.9-1 7.5-6 7.2-6 2.6-6 1.4-8 0.0 0.0 0.0 4.1-8 3.1 14 30 4.5+4 2.1-1 7.7-3 2.4-3 1.1-2 3.6-3 3.5-7 2.1-4 2.1-4 1.9-9 23 .006 3.1 0.01 30 3.1 2.3 + 8 4.6-1 3.3-5 7.2-7 5.6-8 0.0 0.0 0.0 0.0 8.2-9 3.1 14 30 4.5+4 5.4 'I 3.3-2 1.1-2 1.8-2 2.6-2 3.5-7 5.4-4 1.0-3 2.0-2 1 24 .04 3.4 0.01 30 3.1 4.7 + 7 7.9-1 4.5 5 8.2-8 5.3- 8 4.1-19 0.0 0.0 0.0 4.6-9 ! 3.4 14 30 9.0 +3 2.1-1 5.4-2 1.8-3 8.1-3 8.6-3 4.0-4 7.8-4 1.4-4 6.4-3 l 25 .0001 3.1 0.01 30 3.1 9.3 +7 5.5-1 4.2 2 4.2-2 7.8-2 8.2-4 4.1-2 8.2-4 8.2-4 9.8-4 3.1 14 30 9.0+3 0.0 4.4-4 0.0 0.0 0.0 0.0 0.0 0.0 0.0 26 .0006 3.1 0.01 30 3.1 1.2 + 9 5.0-1 7.5-6 8.2-6 6.8-7 0.0 0.0 0.0 0.0 6.0-10 3.1 14 30 9.2 + 5 5.0-1 1.4-1 7.5 2 7.2-2 1.5-2 1.5-6 9.1-4 9.1-4 8.2-3 27 .0008 2.5 2.3 30 3.1 1.6 + 5 6.1-1 2.2 2 4.3-2 1.1-2 1.7-2 8.6-4 9.2-4 1.8-4 8.1-3 4.8 12 30 1.3 +5 3.9-1 1.7-1 1.6-1 2.3-1 1.9-1 1,3-2 1.8-2 3.3-3 1.5-1 28 .02 3.1 0.01 30 3.1 1.2 + 9 8.9-1 1.4-5 1.5-5 5.3-6 2.3-20 0.0 0.0 0.0 7.0-8 3.1 14 30 9.2 + 5 1.1-1 1,2-1 1.7-2 4.8-2 1.6-2 1.5-6 9.4-4 9.4-4 8.2-3 29 .008 2.3 .25 30 3.1 0.0 1.5-1 1.5-2 6.5-4 2.7-4 1.4-4 3.2-5 2.3-6 2.3-6 1.4-4 l 2.5 14 30 2.5 +6 8.1-1 1.0-1 2.3-3 2.0-3 7.4-3 1.3-4 2.5-4 4.6-4 5.1-3 30 .001 2.3 .75 30 3.1 4.9 +5 6.5-1 3.0-6 2.0-6 4.7-7 3.9 10 0.0 0.0 0.0 2.6-8 l 3 14 30 1.1 + 5 3.5 1 2.0-1 1.9-1 1.5-1 2.9-2 2.2-4 1.8-3 1.8-3 1.5-2 ; I 31 .005 3.1 0.01 30 3.1 5.8+8 9.2-1 1.1-4 3.2-6 1.8-6 2.9-8 0.0 0.0 0.0 8.2-7 3.1 14 30 9.1 a s.5-1 1.2-1 2.2-4 1.6-2 9.6-2 1.3-6 2.0-3 3.7-3 6.5-2 32 .0002 2.3 2.5 30 3.1 1.5 +5 9.0-1 1.5-2 2.2-2 9.5-3 3.3-2 2.7-7 4.2-4 8.3-4 1.6-2 4.8 12 30 1.3 + 5 9.7-2 1.5-1 1.0-1 2.0-1 3.9 1 5.5-6 8.3-3 1.6-2 2.9-1 33 .003 3.1 0.01 30 3.1 1.2+9 1.0 1.1 3 2.6-4 5.0-3 2.0-4 1.0-2 2.0-4 2.0-4 1.9-4 3.1 14 30 4.6+6 0.0 3.3 2 1.5-7 5.1-2 4.7-2 4.4-6 2.8-3 2.8-3 2.3-2 34 .003 3.4 .65 30 3.1 1.0 + 8 8.8-1 6.5-3 7.0-3 5.3-3 2.2-3 3.7-8 5.6-5 1.1-4 1.9-3 4 13 30 6.7+6 1.2-1 4.0-2 3.9-2 1.2-1 2.9-1 3.9-6 5.9-3 1.1-2 /.21 35 0.0 3.4 1.2 30 3.1 5.8+7 9.2-1 7.3-2 8.4-2 4.5-2 6.4-1 1.1-6 1.6-3 3.0-3 5.3-2 4.5 13 30 7.0+6 7.6-2 3.8-1 3.4-1 3.9-1 5.8 1 7.6-6 1.1-2 2.1-2 4.1 1 36 .002 2.3 .55 30 3.1 3.3 + 6 8.7-1 1.0-2 5.6-3 3.5-3 2.6-5 0.0 0.0 0.0 1.2-4 2.8 14 30 5.6+5 1.3 1 1.1-1 6.2-2 9.0-2 1.3- 1 8.9-3 1.1-2 2.1-3 8.5-2 37 .003 2.3 .25 30 3.1 5.2+6 2.6-1 1.8-2 2.3-4 9.3 5 1.1-6 0.0 0.0 0.0 2.8-5 2.5 15 30 4.2 +4 7.4-1 1.7-1 1.2-2 2.4-2 2.6-2 3.5-7 5.4-4 1.0-3 1.8-2 38 .0002 3.1 .87 30 3.1 3.7+6 9.2-1 2.5-2 3.1-2 2.5 2 3.5-2 5.8-7 8.8-4 1.7-3 2.9 2 4 13 30 9.8 + 5 8.3-2 1.7-1 1.1-1 3.0 1 4.3-1 5.7-6 8.6-3 1.6 2 3.1-1 39 .0008 2.3 .88 30 3.1 0.0 6.4 1 3.5-2 6.7-3 2.7-3 3.2-5 0.0 0.0 0.0 5.6-4 3.1 14 30 6.4 +6 3.5-1 2.8-1 1.6-1 1.6-1 1.6-1 7.9-3 2.9-3 5.3-3 9.2 2 40 0.0 2.8 1.3 30 3.1 1.1 +7 9.1-1 5.7-2 7.0-2 3.4-2 4.;-2 6.9-7 1.0-3 2.0-3 1.3 1 4 13 30 6.9 + 6 8.8-2 4.7-1 5.0-1 4.7-1 6.3-1 8.4-6 1.3-2 2.4-2 4.5-1 5-25 NUREG/CR-4551, VOL3r DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-7 (Continued) PEACH BO'ITOM CLUSTER RET RASE FRACTIONS FOR MACCS Evac.2 Release Fraction By Group 3 ' Reiz Usage! Time Dur2 Elev2 Time Heat 1 2 3- 4 5 6 7 8 9 Clus Fac. (hr) - (hr) (m) . hr) ( . (cal /s) Kr-Xe I Cs-Rb Tc Sr Ru La Ce Ba '41 .002 2.3 .89 30 3.1 3.0+7 9.0-1 1.2-2 4.2-3 2.1-2 3.3-4 1.7-2 3.3-4 3.3-4 3.3-4 3.1 14 30 2.6+6 9.9 2 1.6-1 7.0-2 1.0-1 ; 3.0-2 2.9-6 1.8-3 1.8-3 1.6 2 ; 42 .003 3.4 0.0 30 3.1 - 4.8 +9 8.9 1 3.1-1 3.8-3 1.3-3 8.1-7 0.0 0.0 0.0' 6.4-5 3.4 14 30 1.3 +6 1.1-1 8.2-2 3.43 9.6-33.1-3 ~ 3.1-7 1.9-4 1.9-4 1.6-3 43 .0004 2.3 ' .5 30 3.1 0.0 3.8-1 4.3-2 3.0-4 1.2 4 1.3-6 0.0 0.0 0.0 -2.4-5 : 2.8 14 30 6.3 + 6 6.1-1. 1.5-1 6.5-2 8.2-2 1.7-1 8.3-3 . 3.0-3 5.5-3 9.5-2 l 44 .0001 2.3 2.0 30 3.1 9.1 +5 8.9 1 9.9-2 1.2-2 4.4-3 5.0-1 4.1 8 6.45-.1.3-4 2.5-3 ] 4.3 13- 30 6.3 +5 1.1-1 1.3-1 1.1-1 3.1 1 6.3-1 8.6-6 1.3-2 2.42 4.6- 1 . 45 .0008 2.5 .25 30 3.1 0.0 1.3 1 8.6-2 3.4-4 8.5-6 0.0 0.0 0.0 0.0 4.8-7 2.8 14 30 6.3 +6 7.6-1 2.7 1 2.8-2 ' 3.8-2 5.7-2 2.0 1.1 3 1.9-3 3.8 + 1 46 .0003 2.3 .75 30 3.1 2.4 +6 3.9-1 1.3-1 1.6-3 9.4-5 0.0 0.0 0.0 0.0 2.7-6 3 14 30 5.7+5 6.1-1 5.1-1 4.3-1 4.2-1 8.7-2 8.7 6 5.3-3 5.3-3 4.7-2 47 .001 2.3 .88 30 - 3.1 3.0+7 1.0 5.9-2 8.4-3 2.0-2 7.1-4 3.3-2 6.7-4 6.7-4 1.3 3.1 14 30 2.6+6 3.6-7 1.9-1 8.9 8 3.32'2.2-1 2.8-6 4.5-3 8.5-3 1.5-1 48 .0009 2 46 30 3.1 0.0 1.7-1 1.1-1 3.4-3 3.6-4 8.2-7 0.0 0.0 0.0 '5.9-5 2.5 14 30 2.5+6 8.0-1 4.8-1 1.8-2 3.3-2 7.4-2 8.9-7 1.3-3 2.5 4.9-2 49 0.0 2.3 .88 30 3.1 1.5 +7 3.2-1 1.11 2.0-4 1.1-5 0.0 0.0 0.0 - 0.0 3.2-7 3.1 - 14 30 6.4 +6 6.8-1 5.0-1 4.0-1 4.0-1 8.9-2 1.0 5.6-3 5.6-3 4.8-2 50 .0009 3.4 .01 30 3.1 1.2+ 10 7.8 1 5.2-1 2.1-3 4.0-4 5.4-7 0.0 0.0 0.0 3.8-5 3.4 14 30 6.4 +6 2.2-1 1.1-1 3.5 2 8.4 2 1.3-1 1.7-6 2.6-3 4.9-3 9.3-2 51 .0001 2 .46 30 3.1 0.0 1.7 1 1.1-13.4-3 3.6-4 8.2-7 0.0 0.0 0.0 5.9-5 1 2.5 14 30 2.5 +6 8.3-1 5.3 1 8.3-2 1.7 1 1.9-1 2.1-6 3.2-3 6.0 1.2 1 52 .0001 3.4 0.01 30 3.1 2.4+10 8.9 1 6.0-1 2.2-2 1.2-2 3.0 6 0.0 0.0 - 0.0 3.2-4 ) 3.4 14 30 6.4 + 6 1.1-1 6.1-2 4.9-2 1.4-1 3.3-1 4.5-6 6.8-3.1.3-22.4-1 i 53 0.0 - 3.4 0.01 30 3.1 2,4+10 8.9 1 6.0-1 2.2-2 1.2 2 3.0-6 0.0 0.0 0.0 3.2-4 3.4 14 30 6.4 + 6 1.1-1 1.3-1 1.8-1 4.4-1 6.8-1 9.2-6 1,4-2 2.6-2 4.9-1 54 .15 - - - -. 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 - 0.0 0.0 I This is the the frequency of usage for each cluster, given core damage. 2 Release time is the time between shutdown and release, duration is the time during which the release occurs, . evacuation time is the time at which evacuation starts. 3 Values in this table are written in shorthand notation: 2.7+6 - 2.7 x 106 4 Some of the durations in the release are very small, on the order of a minute. The RELTRAC code includes a step at vessel breach which includes an i.tstantaneous relocation of the radionuclides. For some calculations the duration was based on this part of the code with the result that the puff release is assigned a duration that is artificially short. This could result in some underprediction of early effects since the short, high-energy release results in more loft into the atmosphere. This calculational difficulty will be resolved before final publication of these results. 5 26 _ ___ _ _ _ __ _ _ _ __ _ --_ _._ _ __ - _ ._. - I NUREG/CR-4551, VOL.3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-8 PEACH BOTTOM CLUSTER REI EASE FRACTIONS FOR CRAC2 Re1 Warn 2 Release Fraction By Group Usage Time Dur2 Flev Per. Heat 2 1 2 3 4 5 6 7 8 9 Clus Faci (hr) (hr) (m) (hr) (cal /s) Kr-Xe ! Cs-Rb Te Sr Ru La Ce Ba 1 .01 6.8 1.5 30 5.2 0.0 6.5-13 8.24 1,3-13 3.0-5 1.9-4 2.3-11 3.5-8 4.1-8 7.8-6 2 .05 6 1.0 30 4.4 4.7+7 1.0 9.0-3 2.0-5 1.0-3 1.4-3 3.5-9 5.3-6 8.4-6 4.0-4 3 .10 6.7 1.0 30 5.1 6.6+7 1.0 4.4 2 3.7-3 3.3-3 1.4-5 4.8-8 4.6-7 4.2-7 3.6-5 4 .11 2.9 1.0 30 9.4 9.8 + 9 1.0 1.8-2 2.4-4 4.8 2 7.7-3 5.9-7 7.2-4 9.6-4 1.6-2 5 .11 5.5 1.5 30 2 9.8 +6 1.0 4.1-2 8.2-3 6.1-2 2.9-2 2.9-6 1.7-3 1.7-3 1.42 6 .07 7 1.0 30 3.5 3.9+9 1.0 1.7 1 6.5-2 8.8-2 2.3-2 3.1-4 1.3-3 1.3-3 1.2-2 7 .08 7 1.0 30 3.5 1.8 +9 1.0 1.4-1 1.2-2 2.0-1 8.6 2 2.3-3 5.2-3 5.2-3 4.2-2 8 .06 5.5 1.8 30 2 3.3 +7 1.0 5.2 2 1.2-2 2.0-1 3.5-1 4.7-6 7.2-3 1.3-2 2.5-1 9 .06 7 1.0 30 3.5 8.8 +9 1.0 5.2-2 1.4-2 4.1-1 6.9-1 4.1-2 1.5-2 2.8-2 4.7-1 l 10 .003 3.1 0.01 30 1.5 9.3 +7 4.5-1 4.2-6 4.0-6 3.6-7 0.0 0.0 0.0 0.0 4.0-9 l 11 0.0 3.4 1.8 30 1.8 2.6+7 9.3-1 6.9-4 2.1-4 1.0 3 3.2-4 3.2-8 1.9-5 2.0-5 1.7-4 l 12 .02 2.5 1.3 30 0.9 0.0 6.4-1 8.5-3 1.0-4 2.0-5 4.9-7 0.0 0.0 0.0 8.4-6 l 13 .0009 3.4 3.4 30 1.8 2.6+7 9.1 1 3.9-4 2.9-4 1.5-3 3.2-3 4.4-8 6.7-5 1.2-4 2.4-3 14 .0005 3.6 3.6 30 2 7.3+5 1.0 8.6-2 1.2-3 5.8-3 2.0-3 6.8-6 1.3-4 1.3-4 1.1-3 n 15 .003 2 2.0 30 0.4 0.0 6.0-1 1.6-2 1.2-2 2.3-3 5.5-5 0.0 0.0 0.0 9.8-4 l 16 .0009 2.5 2.5 30 0.9 0.0 3.2-1 2.2-2 8.5-4 2.3-5 0.0 0.0 0.0 0.0 1.4-6 17 .0005 2.3 2.3 30 0.7 1.2+6 8.9-1 1.4 3 1.2-5 4.4-6 2.3-8 0.0 0.0 0.0 6.5-8 18 .003 2.5 2.5 30 0.9 0.0 8.9-1 2.0-3 8.9-4 1.5-3 2.4-3 5.6-6 5.0-5 9.4-5 1.8-3 19 .007 2 2.0 30 0.4 0.0 9.7-1 1.5-3 5.9-4 6.0-4 4.3-4 7.2-6 2.6-5 2.6-5 2.7-4 20 .02 3.1 3.1 30 1.5 9.3 +7 1.0 1.1-3 8.0-4 3.8-3 1.2-3 1.2-7 7.2-5 7.1-5 6.4-4 21 .006 3.1 3.1 30 1.5 4.7+7 1.0 2.8-2 1.1 2 1.8-2 3.6-3 3.5-7 2.1-4 2.1-4 2.0 3 22 .03 3.1 3.1 30 1.5 9.3+7 1.0 7.7-3 2.4-3 1.1-2 3.6-3 3.5-7 2.1-4 2.1-4 1.9-3 23 .006 3.1 3.1 30 1.5 2.3 + 8 1.0 3.3-2 1.1-2 1.8 2 2.6-2 3.5-7 5.4-4 1.0-3 2.0-2 4 24 .04 3.4 3.4 30 1.8 4.7+7 1.0 5.4-2 1.8 3 8.1-3 8.6-3 6.0-4 7.8-4 1.4-4 6.4-3 25 .0001 3.1 3.1 30 1.5 9.3 + 7 5.5 1 4.3-2 4.2-2 7.8-2 E.2-4 4.1-2 8.2-4 8.2-4 8.2-4 26 .0006 3.1 3.1 30 1.5 1.2 + 9 1.0 1.4-1 7.5-2 7.2-2 1.5-2 1.5-6 9.1-4 9.1-4 8.2-3 l 27 .0008 2.5 2.5 30 b.9 1.6+5 1.0 1.9-1 2.0-1 2.41 2.1-1 1.4-2 1.9-2 3.5-3 1.6-1 28 .02 3.1 3.1 30 1.5 1.2+9 1.0 1.2-1 1.7-2 4.8-2 1.6-2 1.5-6 9.4-4 9.4-4 8.3-3 29 .008 2.3 2.3 30 0.7 0.0 9.6-1 1.2 1 3.0-3 2.3-3 7.5-3 1.6-4 2.5-4 4.6-4 5.2-3 { 30 .001 2.3 2.3 30 0.7 4.9+ 5 1.0 2.0-1 1.9-1 1.5-1 2.9-2 2.2-4 1.8-3 1.8-3 1.5-2 i 31 .005 3.1 3.1 30 1.5 5.8+8 1.0 1.2-1 2.2-4 1.6-2 9.6-2 1.3-6 2.0-3 3.7-3 6.5-2 l 32 .0002 2.3 2.3 30 0.7 1.5 +5 1.0 1.7 1 1.2-1 2.1-1 4.2-1 5.8-6 8.7-3 1.7-2 3.1-1 33 .003 3.1 3.1 30 1.5 1.2 + 9 1.0 3.4-2 2.6-4 5.6-2 4.7-2 1.0 2 3.0-3 3.0-3 2.3-2 34 .003 3.4 3.4 30 1.8 1.0+ 8 1.0 4.7-2 4.6-2 1,3-1 2.9-1 3.9-6 6.0-3 1.1-2 2.1-1 5-27 l _ ___ _ - - . - . - I NUREG/CR-4551, VOL3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-8 (Continued) PEACH BOTTOM CLUSTER RELEASE FRACTIONS FOR CRAC2 Ret Wam2 , Release Fraction By Group Usage Time2 Dur. 2 Elev Per. Heat 1 2 3 4 5 6 7 8- 9 Clus Fac1 (hr) (hr) (m) (hr) - (cal /s) Kr-Xe -I Cs-Rb Te Sr Ru La Ce Ba-35 0.0 - 3.4 3.4 30 1.8 5.8 +7 1.0 4.5-1 4.2-1 4.41 6.4-1 8.7-6 1.3-22.4-24.61 36~ .002 2.3 2.3 30 0.7 3.3+6 1.0 1.2-1 6.9 2 9.4-2 1.3-1 8.9-3 1.1 2 2.1-3 8.52 37 .003 2.3 2.3 30 0.7 5.2+6 1.0 1.9-1 1.2-2 2.4-2 2.6-2 3.5 7 5.4 1.0-3 1.8-2 38 .0002 3.1 3.1 30 1.5 3.7+6 1.0 1.9-1 1.4-1 3.2-1 4.7-1 6.3-6 9.5 3 1.8-2 3.4 -l ' -) 39 .0008 -2.3 2.3 30 0.7 0.0 9.9-1 3.1 1 1.7-1 1.6-1 1.6-1 7.9-3 2.93 5.3-3 9.3-2 I 40 0.0 2.8 2.8 30 1.2 1.1 +7 1.0 5.3- 1 5.7-1 5.0-1 6.71 9.1-6 1.4-2 2.6-2 4.8-1 41 .002 2.3 2.3 30 0.7 3.0+7 1.0 1.7-1 7.4-2 1.2-1 3.0-2 1.7 2 2.1-3 2,1-3 1.6-2 42 -.003 3.4 3.4 30 1.8 4.8+9 1.0 3.9-1 7.2-3 1.1-2 3.1 2 3.1-7 1.9-4 1.9-4 1.7-3 43 .0004 2.3 2.3 30 0.7 0.0 9.91 1.9-1 6.5-2 8.2-2 1.7-1 8.3-3 3.0-3 5.5-3 9.5-2 44 .0001 2.3 2.3 30 0.7 9.1 + 5 1.0 2.3 1 1.2-1 3.1-1 6.3-1 8.6-6 1,3-2 2.4-2 4.6-1 45 .0008 2.5 2.5 30 0.9 0.0 8.9-1 3.6 1 2.8-2 3.8-2 5.7-2 2.0-3 1.1 3 1.9-3 3.8-2 46 .0003 2.3 2.3 30 0.7 2.4+6 1.0 6.4-1 4.3 1 4.2 1 8.7-2 8.7-6 5.3-3 5.3-3 4.7-2 l 47 .001 2.3 2.3 30 0.7 3.0+7 1.0 2.5-1 8.4-3 5.3-2 2.2-1 3.3-2 5.23 9.2-3 1.5-1 48 .0009 .2 2 30 0.4 0.0 9.6-1 5.9-1 2.1-2 3.3-2 7.4-2 8.9-7 1.3-32.5-3 4.9-2 49 0.0 2.3 2.3 30 0.7 1.5 +7 1.0 6.1-1 4.0-1 4.0-1 8.9-2 1.0-2 5.6 3 5.6-3 4.8-2 50 .0009 3.4 3.4 30 1.8 1.2+ 10 1.0 6.3-1 3.7-2 8.4-2 1.31 1.7-6 2.6-3 4.9-3 . 9.3-2 51 .0001 2 2 30 0.4 0.0 1.0 6.4 1 8.6 2 1.7-1 1.9-1 2.1-6 3.2-3 6.0-3 1.2-1 52 .0001 3.4 3.4 30 1.8 2.4 + 10 1.0 6.6-1 7.1-2 1.5-1 3.3- 1 4.5-6 6.8-3 1,3-2 2,41 53 0.0 3.4 3.4 30 1.8 2.4 + 10 1.0 7.3- 1 2.0-1 4.5-1 6.8-1 9.2-6 1.4-2 2.6-2 4.9-1 54 .15 - - - - 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 j I This is the the frequency of usage for each cluster, given core damage. -) 2 Release time is the time between shutdown and release, duration is the time duririg which the release occurs, l evacuation time is the time at which evacuation starts. 3 Values in this table are written in shorthand notation: 2.7+6 - 2.7 x 106 4 Some of the durations in the release are very small, on the order of a minute. The RELTRAC code includes a step at vessel breach which includes an instantaneous relocation of the radionuclides. For some calculations the duration was based on this part of the code with the result that the puff release is assigned a duration that is artificially short. This could result in some underprediction of early effects since the short, high-energy release results in more loft into the atmosphere. This calculational difficulty will be resolved before final publication .;' of these results. 5-28 . - _ _ _ _ - ____-_-___-__--________-___-_________-____A ) 1 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 5.4 OFFSITE CONSEQUENCE RESULTS q ) For each of the source terms listed in Tables 5-5 through 5-8, offsite consequences were calculated using the MACCS and the CRAC2 codes. Mean results for six consequence measures (early fatalities, early injuries, individual risk of death, latent-cancer fatalities, l 1 population dose within 50 miles of the site, and total offsite costs) are presented in Tables 5-9 j (MACCS results) and Table 5-10 (CRAC2 results) for the central estimate bins.* The results for the clustemd source terms are delineated in Tables 5-11 and 5-12 for MACCS and CRAC2 respectively. The clusters are useful for understanding the risk results presented in the next section. Inspection of the results presented in all of the tables reveals the following: ; (1) The mean predicted number of early fatalities is fairly small,less than one j early fatahty for most of the source terms. The MACCS code generally { predicts values about a factor of two to three greater than CRAC2. The l maximum number of early fatalities is 32 with the MACCS code and 11 l with CRAC2. Comparison of the clusters to the central estimate accident ! progression bins is quite difficult since there is no direct correspondence l between the two. However, the numbers of predicted early fatalities are I
- very low for the clusters also, very often a fraction of one fatality and j always less than 12 fatalities.
(2) The early injury results are quite similar overall, although for individual bins sometimes MACCS predicts higher risk than CRAC2 while for other I bins the opposite is true. (3) MACCS generally predicts about three times more latent-cancer fatalities l than CRAC2. j (4) MACCS estimates of population dose within 50 miles of the site exceed those from CRAC2 by factors of two to ten, with the largest differences associated with the smaller consequence bins.. (5) Offsite costs predicted by MACCS generally exceed those predicted by MACCS by factors from three to thirty. Identification of the actual reasons for differences between CRAC2 and MACCS results is not straightforward, since there are many different variables which must be compared. Studies are currently underway to compare the code results systematically to determine the causes for any major differences. At this time, the analysts have arrived at certain hypotheses regarding the factors that contribute most to the diffemnces, but it is premature to attempt to draw firm conclusions. There are, however, some known uncertainty issues which could be affecting these results. These include the following: *The central estimate results are listed by bin number. The cross-reference between the bin number and the bin defm' ition is provided in Appendix A. 5-29 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Dose Thresholds. Because early fatalities and injuries both have dose thresholds, their occurrence is very sensitive to meteorological conditions and effectiveness of emergency response. Consequently, because many weather scenarios produce no early effects, the value of the mean result is very sensitive to the magnitudes of the few non-zero results. Therefore, mean early fatality and injury values are quite uncertain, which suggests that the CRAC2 and MACCS results are not very different. In any case, the numbers of early fatalities predicted by either code for most releases is very low. It is probable that the early injury results are affected by dose thresholds. The different results between MACCS and CRAC2 showed no consistent trend, suggesting the influence of thmsholds on early effects. Early Fatality Results. The MACCS results are generally somewhat larger than the CRAC2 results. The MACCS calculations are sensitive to the dose threshold for the lung model, which is significantly lower than the threshold used in the CRAC2 lung model. The amount that this may contribute to the differences between the calculations is not yet known. Pooulation Dose and Latent-Cancer Fatalities. There has been a significant amount of investigation conceming the differences for these two risk measures. It is known that the crop pathways are very irnportant in the MACCS results. While all the differences between CRAC2 and MACCS have not been fully investigated, there are some known differences: -- In MACCS, multistep cross-wind plume concentration profiles and multiplume releases lead to larger coverage at lower concentrations, and some of these lower concentration releases are not interdicted. More recent cancer risk factors are used in MACCS, many of which am five times greater than those used in CRAC2 and the ingestion cancer risk factor is a factor of 20 greater. l MACCS crop models transfer more dose than CRAC2, despite a more stringent interdiction criteria. -- Dose concversion factors also differ by tens of percents between the two codes. l Total Costs. The MACCS estimates of total offsite costs exceed those of CRAC2 because the total dose is also calculated to be higher and because decontamination is assumed to be more costly in MACCS than in CRAC2. MACCS also calculates emergency response costs but CRAC2 does not. In addition, MACCS includes another criterion forinterdiction, the inhalation dose, which can increase the amount ofinterdiction needed, and hence its cost. The Peach Bottom results could show a larger difference between CRAC2 and MACCS estimates of costs than the other plants because of the makeup of the releases,in this case the La nad Sr releases are quite high. The CRAC2 interdiction criteria do not directly include these species whereas the MACCS do possibly leading to higher costs in MACCS. Thus, there are several differences between the CRAC2 and MACCS results. Detailed investigation of these differences is continuing. The risk implications of these differences are discussed in the following section. 5-30 NUREG/CR.4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-9 MACCS ESTIMATES OF MEAN CONSEQUENCES FOR CENTRAL-ESTIMATE BINSI Individual Latent Population Offsite Cond. Early Early Risk Cancer Dose--50 Miles Costs Bin Prob.1 Fatalities Injuries (0-1 Miles) Fatalities (Person-Rem) ($) 1 .04 2.4 32 3.6-2 3.1-5 3.2+2 3.1+6 1.2+8 2 .06 0.0 0.0 0.0 0.0 0.0 0.0 3 .006 2.3-1 1.3 1.1-3 2.7 +3 1.2+7 2.8+9 5 .03 0.0 0.0 0.0 0.0 0.0 0.0 9 .02 4.3-6 6.3-3 6.6-8 1.9+2 4.4 +5 2,9 +7 j 12 .01 2.2-1 1.1 1.1-3 2.8 + 3 1.2+7 2.8 +9 14 .002 7.6-3 1.4-1 9.7-5 4.0+2 9.1 +5 6.7+7 15 .001 2.8-1 1.7 1.2-3 2.9+3 1.2+7 2.8 +9 22 .02 5.9 1.5+1 4.9-3 6.0 +3 2.0+7 1.1 +10 l 26 .02 2.0-1 7.9-1 1.4-3 1.7+3 8.4+6 9.6+8 29 .02 3.7-1 1.3 2.0-3 1.9+3 9.0+6 1.3 +9 32 .03 0.0 3.2-4 0.0 2.6+ 1 1.3 +5 4.9 +6 33 .04 0.0 3.4-4 0.0 4.3+1 2.2+5 7.8 +6 44 .03 6.2 1.7+1 4.7-3 6.0+3 2.0+7 1.1 +10 45 .30 5.9 1.5 +1 4.9-3 6.0+3 2.0+7 1.1+10 48 .09 1.7+ 1 1.2+2 4.2-3 9.7+3 2.8 +7 1.9 +10 50 .03 0.0 4.5-4 0.0 8.2 + 1 3.5+5 5.1+7 56 .06 0.0 0.0 0.0 0.0 0.0 0.0 58 .01 1.5 2.2+1 2.1-3 6.4+3 1.9+7 1.1+10 63 .008 6.3 6.0 + 1 2.3-3 7.6+3 22 +7 1.5 + 10 64 .08 6.8 6.3 +1 2.4-3 7.7 +3 2.2+7 1.5+10 66 .02 3.2+ 1 2.8 +2 1.7-3 1.2 +4 3.1+7 2.4+10 72 .01 0.0 0.0 0.0 1.0 + 1 1.1 +5 1.2+6 75 .02 0.0 0.0 0.0 3.7 + 1 7.8+4 3.7+6 76 .007 0.0 0.0 0.0 0.0 0.0 0.0 77 .03 0.0 0.0 0.0 0.0 0.0 0.0 I This is the conditional probability of the bin given core damage. The results are listed only for bins with non-zero consequences in the central estimate. The other bins are used for analysis of safety options. The consequences for all bins are provided in Appendix B. The cross-reference key between central estimate bin numbers and the bin definitions are included in Appendix A. 2 Values in this table are written in shorthand notation: 2,4 2.4 x 10-3 5-31 NUREG/CR.4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-10 CRAC2 ESTIMATES OF MEAN CONSEQUENCES FOR CENTRAL-ESTIMATE BINS Individual Latent Population Offsite Cond. Early Early Risk Cancer Dose--50 Miles Costs Bin Prob.1 Fatalities Injuries (0-1 Miles) Fatalities (Person-Rem) ($) 1 .04 0.0 2.2 32 0.0 9.2+1 6.8+5 8.7+6 2 .06 0.0 0.0 0.0 0.0 0.0 0.0 3 .006 4.1-2 7.9-1 2.4-4 7.9+2 4.0+6 7.0+7 5 .03 0.0 0.0 0.0 0.0 0.0 0.0 9 .02 0.0 3.9-4 0.0 6.9 + 1 4.3+5 5.3+6 12 .01 4.7-2 8.21 2.4-4 7.8 +2 3.5+6 7.9+7 14 .002 0.0 8.9-3 0.0 1.4 +2 1.2+6 1.8+7 15 .001 5.5-2 9.1-1 2.7-4 1.1 +3 7.0+6 1.0+8 22 .02 4.7 3.2+1 2.4-3 2.7+3 1.8+7 2.5 +8 26 .02 4.7-3 2.1 1 5.4-5 6.7+2 3.2+6 1.0+8 29 .02 1.5-2 4.0-1 1.3-4 7.8 +2 3.7+6 1.1+8 32 .03 0.0 9.7-3 0.0 8.2 7.3+4 1.6+6 33 .04 0.0 1.6-2 0.0 1.1 + 1 9.8+4 1.6+6 44 .03 4.3 4.5 + 1 1.6-3 2.8 +3 1.8+7 2.4 + 8 45 .30 4.7 3.2 + 1 2.4-3 2.7+3 1.8+7 2.5+8 48 .08 1.1 + 1 2.2+2 3.4-3 4.1 +3 2.3+7 5.2+8 50 .03 0.0 3.8-4 0.0 2.3 + 1 1.6+5 1.6+6 56 .06 0.0 0.0 0.0 0.0 0.0 0.0 58 .01 2.7 5.7 3.2-3 1.9+3 1.2+7 2.9 +8 63 .008 1.4 1.2 +2 5.3-4 3.6+3 2.1+7 3.8+8 64 .08 1.6 1.5 +2 5.34 3.6+3 2.1+7 3.9 +8 66 .02 2.7 6.1 +2 5.3-4 5.4+3 2.6+7 1.0+9 72 .01 0.0 0.0 0.0 3.2 2.3 +4 1.6+6 75 .02 0.0 0.0 0.0 8.2 5.4+4 1.7+6 76 .007 0.0 0.0 0.0 0.0 0.0 0.0 77 .03 0.0 0.0 0.0 0.0 0.0 0.0 1 This is the conditional probability of the bin given core damage. The results are listed only for bins with non-zero consequences in the central estimate. The other bins are used for analysis of safety options. The consequences for all bins are provided in Appendix B. The cross-reference key between central estimate bin numbers and the bin definitions are included in Appendix A. 2 Values in this table are written in shorthand notation: 2.2 2.2 x 10-3 5-32 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-11 MACCS ESTIMATES OF MEAN CONSEQUENCES FOR PEACH BO'ITOM LLH CLUSTERS Cluster Usagel Early Early Individual Latent Population Offsite Factor Fatalities - Injuries Risk Cancer Dos--50 Miles Costs (0-1 Miles) Fatalities - (Person-Rem) ( $ ). I' .01 0.0 1.2-42 0.0 5.9 9.9+4 9.1+6 2 .05 0.0 1.2-3 0.0 4.1+ 1 5.5+5 5.5+7 3' .10 3.2-4 1.7-2 .4.2-6 5.2+2 1.3 +6 1.2+8 4 .11 2.6-3 3.7-2 3.5 5 5.1 +2 2.8+6 3.1+8 5 .11 2.0-2 2.6 1 1.6-4 1.2+3 6.7+6 4.5 +8 6 .07 3.9-1 1.6 2.2-3 3.3 +3 8.4 +6 6.2+8 7 .08 8.0-2 8.0-2 4.9-4 7.3 +3 1.7+7 1.9+9 8 .06 2.9-1 6.0 8.4-4 8.2+3 2.1+7 4.4 +9 9 .06 2.9 5.7+1 1.4-3 1.0+4 2.5+7 - 1.2+10 10 .003 0.0 0.0 0.0 2.4 6.8 +3 1.8+5 11 0.0 7.4-6 4.3-3 1.1 7 6.1 + 1 3.1 +5 - 1.2+7 . 12 .02 1.0-3 4.7-2 1.5-5 3.4+1 1.1+5 7.0+7 ' 13 .0009 7.2-6 4.0-3 1.1-7 1.4 +2 1.2+6 2.8 +7 14 .0005 5.1 2 4.5-1 4.9-4 2.3 +2 1.2 +6 1.4+8 j 15 .003 4.7-3 1.41 6.7 5 1.4 +3 2.7+6 1.6+8 16 .0009 8.4-1 3.9 5.1 3 3.5 +2 9.4+5 2.0+8 17 .0005 1.2 3 4.3-2 1.7-5 7.6 2.8 +4 1.6+7 18 .003 1.8 3 5.5 2 2.5-5 1.4+2 9.9+5 3.2 +7 .; 19 .007 3.2-3 1.2-1 4.3-5 9.7 +1 '4.4 + 5 1.9+7 20 .02 9.3-6 6.33 1.4-7 1.3+2 7.2+5 2.1+7 21 .006 2.1-2 2.3 1 2.5-4 9.4 +2 2.8 +6 1.4 + 8 22 .03 1.1 3 3.5-2 1.5-5 3.2+2 1.6+6 6.8 +7 23 .006 6.9-2 4.41 7.1-4 1.0+3 5.3+6 2.5+8 24 .04 3.8-2 3.2-1 4.2-4 3.5+2 2.6+6 1.6+8 25 .0001 2.2 1 1.4 3.3-4 2.9+3 2.9+5 3.4+9 . 26 .0006 1.3-1 1.0 8.9-4 4.4+3 9.8 +6 8.7+8 27 .0008 2.9 9.4 4.4-3 7.6+3 2.0+7 2.4+9 28 .02 6.0-2 5.71 5.3-4 1.7+3 6.3+6 3.7+8 29 .008 8.2-2 7.4-1 7.8-4 6.2 +2 3.2+6 2.6+8 30 .001 1.10 4.2 3.1 3 6.5 +3 1.3+7 1.7+9 1 -l 5-33 -l 1 NUREG/CR.4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-11 (continued) Cluster Usagel Early Early Individual Latent Population Offsite Factor Fatalities Injuries Risk Cancer - Dos--50 Miles Costs (0-1 Miles) Fatalities (Person-Rem) ($) 31 .005 1.3 1 9.0-1 8.8-4 1.6+3 9.9+6 1.1 +9 32 .0002 3.6 1.2+ 1 4.6-3 6.4+3 2.0+7 3.9+9 33 .003 7.0-3 1.0-1 8.5 5 1.4 +3 7.4 +6 1.5+9 34 .003 1.4-1 1.5 5.5-4 5.7+3 1.8 +7 3.5+9 35 0.0 6.5 1.3 +2 1.8-3 1.6+4 3.2+7 9.1+9 36 .002 8.0-1 4.0 2.3 3 4.5+3 1.5+7 1.3+9 37 .003 5.2-1 2.3 2.5-3 1.2+3 5.7+6 3.3+8 38 .0002 3.2 3.1 + 1 2.7-3 8.4 +3 2.2+7 5.5+9 39 .0008 6.1-1 5.8 2.1-3 8.7+3 2.0+7 2.9+9 40 0.0 1.2+ 1 2.0+2 2.13 1.8 +4 3.3+7 1.1 +10 41 .002 1.6-1 1.1 7.2-4 5.2+3 1.1 +7 1.5+9 42 .003 8.2-1 5.6 6.2-4 1.4 + 3 3.5 +6 2.8 +8 43 .0004 5.5-1 3.2 2.8-3 5.4+3 1.7+7 2.2+9 44 .0001 5.6 4.3+1 4.4-3 8.4+3 2.6+7 6.4 +9 45 .0008 1.1 4.9 5.4 3 3.1+3 1.1 +7 8.5+8 46 .0003 4.90 3.8 + 1 4.4 3 1.2+4 2.0+7 5.0+9 47 .001 4.2-1 5.3 1.2-3 4.1 +3 1.5 +7 4.6+9 48 .0009 2.7 1.3+1 1.4-2 2.9 +3 1.2+7 9.2+8 49 0.0 1.3 5.5+1 1.8-3 1.4 +4 2.3 +7 6.0+9 50 .0009 1.7 3.8 + 1 8.8-4 4.6+3 1.5+7 1.8+9 51 .0001 3.6 3.0+ 1 1.42 6.5 +3 2.0+7 2.0+9 52 .0001 2.5 8.0 + 1 1.1-3 7.4 + 3 2.1 +7 4.3 +9 53 0.0 9.1 2.6+2 1.9-3 1.2+4 2.9+7 9.7 +9 54 .15 0.0 0.0 0.0 0.0 0.0 0.0 I This is the frequency of usage for each cluster, given core damage. 2Some values in this table are written in shorthand notation: 1.2 1.2 x 104 l I l 5-34 i l NUREG/CR-455!, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) i I Table 5-12 CRAC2 ESTIMATES OF MEAN CONSEQUENCES FOR PEACH BOTTOM LLH CLUSTERS i Cluster Usage l Early Early Individual Latent Population Offsite i Factor Fatalities Injuries Risic Cancer Dc:-50 Miles Costs (0-1 Miles) Fatalities (Person-Rem) ($) 1 .01 0.0 4.5-32 0.0 2.8 2.8 +4 1.8 +6 2 .05 0.0 6.5-4 0.0 1.2+1 9.3+4 3.3+6 3 .10 1.8-3 3.4-2 2.8-5 1.4 +2 8.1+5 1.6+7 4 .11 3.7-3 1.1 1 5.75 9.7 + 1 5.3+5 7.4+6 5 .11 1.2-2 3.9-1 1.1-4 4.3 +2 3.2+6 4.8+7 l 6 .07 2.8-1 1.0 2.6-4 1.0+ 3 1.9+6 4.9 +8 7 .08 6.3-1 2.0+ 1 3.8-4 1.5 +3 3.0+6 9.3 +8 8 .06 6.9-1 2.7 + 1 4.6-4 2.9 + 3 1.2+7 9.1+8 I 9 .06 1.9 4.1 +2 3.8-4 2.3 +3 3.9 +6 1.2+9 10 .003 0.0 0.0 0.0 7.4-1 5.2+3 1.6+6 11 0.0 0.0 0.0 0.0 1.4 + 1 1.0+5 1.9+6 12 .02 1.9-4 1.0-1 2.9-6 7.5 7.3+4 2.6+6 ) 13 .0009 0.0 0.0 0.0 3.1 +1 2.4 +5 2.7+6 14 .0005 1.2-2 4.0-1 1.1-4 8.6+ 1 8.1 +5 8.4 +6 15 .003 1.4-2 3.9-1 2.2-4 2.1+2 1.6+6 3.4+7 16 .0009 8.4-1 2.4 2.6-3 8.2+ 1 7.4+5 8.8 +6 17 .0005 0.0 0.0 0.0 4.2 4.0+4 1.9+6 18 .003 4.7-4 1.3-1 7.3-6 3.2+ 1 2.9+ 5 3.1 +6 19 .007 3.7-4 5.6-1 5.7-6 2.0+ 1 1.8 +5 2.6+6 20 .02 0.0 3.7-4 0.0 4.0 + 1 2.5+5 2.6+6 21 .006 3.3-3 6.0-2 5.1-5 3.3 +2 1.8+6 6.0+7 22 .03 0.0 6.1-3 0.0 1.1 +2 6.5+5 9.8 +6 23 .006 3.9-3 1.4- 1 5.8-5 3.8 +2 1.8 + 6 5.6+7 24 .04 4.0-3 1.3 1 5.9 5 1.4 +2 9.5+5 1.4 +7 25 .0001 5.3-2 6.71 2.1-4 1.1+3 3.6+6 3.148 26 .0006 2.4-1 9.9-1 2.4-4 1.1+3 2.4 +6 5.4+8 27 .0008 4.1 1.1+ 1 3.3-3 2.8+3 1.2+7 1.1 + 9 28 .02 1.3-1 7.8-1 2.1-4 4.5+2 1.5+6 9.9+7 29 .008 4.6-1 2.4 2.1-3 1.2 + 2 1.1 +6 1.0+7 30 .001 9.9-1 7.0 5.2-4 2.2+3 9.5 +6 1.1 +9 5-35 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-12 (Continued) CRAC2 ESTIMATES OF MEAN CONSEQUENCES FOR PEACH BOTTOM LLH CLUSTERS Cluster Usage l Early Early Individual Latent Population Offsite l Factor Fatalities Injuries Risk Cancer Dos- 50 Miles Costs (01 Miles) Fatalities (Person-Rem) ($) ~ l 31 .005 1.2-1 8.8 1 2.4-4 4.0+2 2.5 +6 4.5 +7 l 32 .0002 4.8 1.2+ 1 3.5-3 2.5+3 1.4+7 5.9+8 33 .003 2.4-2 5.4-1 2.0-4 3.3 +2 1.6+6 3.5 +7 34 .003 4.3-1 2.5 3.4-4 1.8+3 7.6+6 3.3+8 ) 35 0.0 2.9 5.4+2 5.3-4 5.1 +3 1.5+7 3.3 +9 36 .002 4.3 1 2.0 3.6-4 1.7+3 9.1 +6 4.6+8 37 .003 2.1 1 1.1 2.8-4 4.9+2 3.6+6 7.6+7 38 .0002 2.4 3.0+2 5.1-4 3.4 +3 1.6+7 1.0+9 39 .0008 3.6 9.2 6.6-3 2.0+3 9.1 +6 6.2+8 40 0.0 3.2 1.0+3 5.3-4 5.8+3 1.9 +7 4.5 + 9 41 .002 4.1-1 1.5 3.8-4 1.7 +3 7.1 + 6 5.0+8 42 .003 6.2-1 4.2 3.8-4 2.9+2 1.1 +6 5.5+7 43 .0004 2.4 6.7 6.5-3 1.2 +3 7.1 +6 2.6+8 44 .0001 5.6 2.3 +2 1.1-3 3.3 +3 1.8 +7 7.5 +8 45 .0008 2.3 6.0 2.93 6.2+2 4.4 + 6 1.2+ 8 46 .0003 3.5 5.4+2 8.7-4 4.0+3 1.4 +7 3.4 +9 47 .001 7.5-1 7.6 5.1-4 1.5 +3 8.0 +6 1.6+8 48 .0009 6.2 1.4 + 1 3.5-2 6.2+2 4.7 +6 1.0+8 49 0.0 2.6 3.9 +2 5.1-4 4.3+3 1.4 +7 3.4 +9 50 .0009 2.1 1.1 +2 4.4-4 1.1 +3 2.3 +6 2.5+8 51 .0001 9.2 2.5 + 1 4.72 1.4 +3 8.2+6 3.5+8 52 .0001 2.1 3.3 +2 3.8-4 1.6+3 2.5 +6 6.0+ 8 i 53 0.0 4.6 1.2 + 3 3.8-4 2.2+3 3.5 +6 1.4+9 54 .15 0.0 0.0 0.0 0.0 0.0 0.0 1 This is the frequency of usage for each cluster, given core damage. 2Some values in this table are written in shorthand notation: 4.5 4.5 x 10-3 5-36 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l 5.5 RISK RESULTS . The results of the integration of the core-damage frequency, containment response, source-term and consequence analyses to estimate risk are presented in this section. Six risk measures are included: earl; f9talities, early injuries, latent-cancer fatalities, individual risk of early death, offsite costs (in 611ars), and population dose within 50 miles (in person-rems). All results were calculated r. sing both the MACCS and CRAC2 consequence codes. Latent-cancer fatalities and early f Atalities are discussed in detail first, followed by the results of the other risk measures. Mo e detailed listings of the results are included in Appendix C. Although the scope and goals of the Peach Bottom analysis are the same as for the other SARRP plants [5.,ft,2], the actual methodology employed for Peach Bottom included more detailed consideration of uncertainties and a more thorough integration of the uncertainty issues and their impact on the source term. Section 2 of this report describes the actual methodology. Section 5.5.3 includes observations concerning the risk results for the Peach Bottom plant. ! The LLH risk results reported in this draft are known to contain some inaccuracies. In particular, the treatment of drywell sprays did not properly credit the operation of sprays for l pressure reduction and radionuclides scrubbing and the containment analysis for the ATWS J scenarios is known to underpredict the consequences for some sample members. Overall these ] errors would not be expected to change the LLH uncertainty ranges, although the risk of individual sample members within that range might be different. These aspects will be corrected in the final report. 5.5.1 Results for Risk of Latent-Cancer Fatality and Eariv Fatality Tne risk results are displayed and discussed in this section. The results are presented ' first, including several breakdowns of risk to examine the relative importance of core-damage sequences, containment outcomes, and source terms. The important contributors to the uncertainty range represented by the LLH are then discussed. Risk Results The results for the risk of fatality due to latent cancer are presented in Figure 5-7. The results of the LLH indicate that the risk ranges from approximately 7 x 10-4 to 1 x 10-1 for the MACCS calculauon and from 2 x 10-4 to 2 x 10-2 for the CRAC2 estimate of consequences. The central estimate is approximately 2 x 10-2 latent-cancer fatalities per year as calculated with the MACCS code and approximately 9 x 10-3 fatalities per year as estimated using the CRAC2 code. As illustrated, the two consequence code pedict similar results, although for a given 5-37 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 0 LLH LLH (MACCS) (CRAC2) 8: ' M - RSS ! (Peach Bottom) D Central e a E (MACCS) h' O ($h"A 2) X il l 1 (Baseline) I lDCOR l g* g i l (Committed) x o io: 2 3 3
- I ;
i 1h Figure 5-7. Risk of Latent-Cancer Fatalities source term the MACCS code does pmdict consequences about a factor of thme higher than the CRAC2 model (see Section 5.4). The range of uncertainty illustrated by the LLH result corresponds specifically to the uncertainty issues and their levels and weighting factors assigned by this review group (as described in Section 4). The central estimates fall well within the LLH uncertainty range, although they are higher than the average LLH result. Since it was the goal of the program to address the change in estimated risk since previous risk assessments, the RSS point estimate of latent-cancer fatality risk for Peach Bottom [2] is also indicated on the risk profiles shown in Figure 5-7*. The RSS value for latent-cancer fatalities is at the top of the LLH range as calculated by CRAC2 (the RSS result was derived with CRAC, and some assumptions in the code input are different from those used here). The IDCOR result [4] is also illustrated in Figure 5-7. The IDCOR results are shown for two sets of boundary conditions, baseline (as the plant existed at the time of analysis), and committed, which accounts principally for changes associated with the ATWS rule and the ADS actuation logic. The boundary conditions in the ASEP evaluation of Peach Bottom are comparable to a plant status similar to the IDCOR committed profile. In either case, the *The RSS value was multiplied by a factor of 30 to convert from the yearly rate ofincidence of latent cancers over tb 30 years following the accident to total fatalities. 5-38 i NUREG/CR 4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) IDCOR result is in the middle of the LLH range, and somewhat below the SARRP central estimates. The estimated consequences of accidents are generally lower in IDCOR, but the total risk is comparable because the core-damage frequency (which was not assessed in as great of detail in IDCOR as compared to ASEP) is a factor of four higher in IDCOR than in the ASEP results. Although detailed comparisons ofIDCOR and SARRP are yet to be completed, it is believed that the difference in consequences is due to generally larger radionuclides releases from the primary containment, particularly as a result of core-concrete interactions, and lower retention in the reactor building in the SARRP analysis. In particular, the issue involving drywell failure by meltthmugh due to direct contact of the molten core is included in one-half of the LLH sample members, but this same failure mode is considered negligible in IDCOR. However, even considering the differences in core-damage frequencies, the IDCOR results appear to be within the uncertainty range depicted by the LLH calculation. , l The risk results for early fatadties are displayed in Figure 5-8. As illustrated, the LLH d analysis for MACCS ranges from 8 x10-9 to 8 x 10-6. The CRAC2 result has nearly the same range, although the distribution of individual sample member results is somewhat different. The central estimate results are about 2 x 10-5 and 1 x 10-5 early fatalities per year using MACCS and CRAC2, respectively. 1b Central & , (MACCS) ' NII a to ,
- O (Central CRAC2) h t,e I
>. i. 2 ! p i ! 3 I t .. w i . i.. 1 = . LLH LLH ,, (MACCS) (CRAC2) 4[, Figure 5-8. Risk of Early Fatalities 5-39 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Although the CRAC2 and MACCS results are nearly identical, there is a fairly significant difference between the LLH and central estimate results, the central estimates predicting higher risk than the LLH range. Once again, it was the intent of the central estimate to represent the mean ASEP core-damage frequency result, a best estimate containment response (as assessed by the SARRP analysts) alor g with a source-term analysis designed to mimic the STCP. For this particular analysis, the uncertainties as assessed by the the expert review group tended to have an overall effect oflowering the risk relative to the SARRP central estimate. A detailed examination of the results suggests that the principal reason for the difference between the LLH and the central stems from the weighting factor assignments for the issue involving the magnitude of the CCI releases. As illustrated in Section 4, there were four levels used to represent the uncertainty in the CCI releases. Each level was defined to I have a set of specific release fractions for each release group. For this issue, the reviewers weighted levels 2 and 3 the most significantly. The STCP results predict releases between levels two and three, except for one radionuclides gio :p, lanthanum. Sensitivity studies with CORCON-VENESA suggested that the nominal STCP predictions of La releases are near the upper range of the uncertainty, and the definition of the levels for the uncertainty analysis reflected this fact by defining the La releases to be similar to the STC"in level 4 alone, and this level was not we.ighted significantly by the review group. Thus, on the average, the LLH has substantially lower lanthanum releases the STCP-derived central estimates. In addition to the assessment of the STCP as overpredicting releases of La, an additional feature of the SARRP methodology served to magnify this difference between the LLH and the central estimate. The clustering methodology (discussed in Section 2 and Appendix B) was based on use of MACCS importance weightings to allow the clustering to take place on a manageable set of parameters. This process results in a translation of the La releases to cesium equivalents, potentially understating the consequences associated with the La releases. By design of the clustering algorithm, more emphasis was placed on characterizing the late effects than the early effects, thus the difference between the centrals and the LLH is most visible in the early risk measures. Although it is believed that the principal reason for the LLH-central estimate difference is the weighting factor assignment, the amount that this feature is magnified by the methodology is not fully known at this t me.- (Investigations of the role of the clustering with regards to this issue are currently being performed.) The results are also compared to the RSS result for Pesch Bottom and the IDCOR result for Peach Bottom in Figure 5-8. The SARRP LLH resuitt indicate that current estimates of risk are below the RSS value, although there are a few sample members with risk estimates 5 40 L l' u NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) higher than the RSS result. The SARRP central estimate of risk is quite similar to the RSS value for Peach Bottom: approximately a factor of three lower in risk. The IDCOR result for Peach Bottom predicted no early fatalities, a result that is consistent with the lower range of the SARRP analysis when the difference in evacuation assumptions is considered (IDCOR accounts for 100% evacuation while SARRP assumed 5% non-participation on the part of the affected population). In addition, the lower end of the SARRP LLH range is near the limits of resolution of the analysis, resulting in the conclusion that the IDCOR result is not too different from the SARRP result. Several methods of decomposing the risk results were utilized to aid in understanding the various contributors to risk. The contribution of each plant-damage state to risk is listed in Table 5-13 to provide a perspective on the important accident sequences. The results are illustrated for early fatalities and for latent-cancer fatalities for both consequence-code evaluations. The TBUX plant-damage state dominates all of the LLH risk calculations, while the TB damage state accounts for most of the remaining risk. These damage states are the two most frequent damage states as estimated by the ASEP analysis; in addition, the accidents in these damage states tend to be severe in terms of containment response since they all involve a loss of electric power. The relative contribution to risk is very similar for both early and latent cancer fatalities, for the calculations as performed with the two consequence codes and for the central estimate as compared to the LLH. The results presented in Table 5-13 must be used with some caution due to an error in the calculation which results in an underprediction of the consequences for the ATWS damage states (TC sequences). This enor is not of great importance to the range of the LLH results or the conclusions of the study, but the error does preclude any conclusions regarding the relative contribution of all of the ATWS damage states. This results from an error in the binning of the 1 containment event tree results which caused the containment failure location to be always assigned to the wetwell, regardless of the actuallocation predicted by the analysis. This error l could result in significant (orders of magnitude) underprediction of consequences for some failure modes which would otherwise bypass the suppression pool. The impact of this underprediction was assessed through analysis of individual sample members, with the conclusion that the LLH band would not change significantly, that the mean of the LLH could increase slightly and that individual sample members could have some significant adjustments. l As a result, the relative contributions of all the damage states would probably change slightly if , the error was corrected, and use of individual sample members results must be carefully. examined to determine the impact of this error. 5-41 1 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-13 COMPARISON OF LLH AND CENTRAL RISK ESTIMATES BY FRACTIONAL CONTRIBUTION OF PLANT-DAMAGE STATES Contribution Contribution to - to Early Fatalities Entent enneer Fatalitien Plant- CRAC2 MACCS CRAC2 MACCS Damage State Central LLH Mean* Central LLHMean Central LLHMean Central LLHMean TB 8% 15 % 26% 19 % 24 % 14 % 24 % 15 % TBUX 90 % 79 % 72 % 76% 72 % 82 % 72 % 82 % TBU 2% 1% 2% 1.2% 3% .7% 3%- 0.8% TCSX*
- 0.01 % 0.8% - 0.03 % 'O.8% 0.3% .8% 0.4% 0.7%
TCUX*
- 0.0% 0.01 % 0% 0.08 % 0.2% .2% 0.3% 0.2%
TCSRX2** 0.02 % 2.7% 0.06 % 2.0% 0.6% 1.4% 0.9% 1.3% TCSRV23*
- 0.8% 1.3% 0.01% 1.1% 0.1% .4% 0.1% 0.4%
*The fractional contribution is difficult to illustrate in the LLH, since each of the 150 sample members has a - different outcome. The mean of the LLH sample members was calculated, and it is illustrated in the table, but it must be emphasized that the sample cannot be interpreted as a probability distribution, in light of the manner in which the input information was developed. **As described in text, the results in this table are somewhat in error with regards to the relative contributions of ATWS (TC) damage states. A similar bmakdown is provided in Table 5-14 for the accident progression bins (only contributions greater than 1% are listed). Because of the similarity of MACCS and CRAC2 results only the former are listed in the table. The accident progression bins used here include ten dimensions which describe the principal attributes of the containment response.* The definitions of the bins are provided in Table 5-15. As illustrated by the table, a few bins tend to dominate the calculated risk results. Most of the important contributors are similar. The first four bins in the table account for about 70% of the risk for all the central and the LLH, for both risk measures, and for either consequence code. These bins allinvolve accidents in which the sequence type is best defined. as a fast-evolving station blackout (TBU or TBUX) in which the vessel does breach and core-concrete interactions occur in a dry cavity. All four bins also include an early containment failure at vessel breach: a drywell rupture to the reactor building. This latter failure mode is the *The results were actually calculated with 22 dimension bins designed to maximize the efficient calculation of source terms in RELTRAC. For presentation purposes the 22 bins were reduced to 10. The detailed listing of code results provided in Appendix C are listed in terms of 22 dimension bins. 5-42 i i 1 i i I NUREG/CR-4551 VOL.3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) ] I Table 5-14 COMPARISON OF LLH AND CENTRAL RISK ESTIMATES BY FRACTIONAL CONTRIBUTION OF ACCIDENT-PROGRESSION BINS: MACCS RESULTS Accident- Contribution to Early Fatalities Contribution to T ntent.rancer Fafnlitiet Progression Bin Central LLH Mean Central LLH Mean AAAAEAADAB - 29 % - 10 % AAAAEAADBB 30 % 8% 20 % 3% AAAAEAEDAB 4% 16 % 5% 36% AAAAEAEDBB 39 % 15 % 47 % 15 % AAAAKBADAB - - - 1% AAAAKBBDAB -- 2% - 2% AAAAKBBDBB - - - 1% AAAAKBEDAB - - - 2% AAAAKBEDBB - 1% - 2% AAAAKBFDAB - - - 1% j AAAAKBFDBB - - - 2% i AAAAKBGDAC -- 1% - 1% AABBEAABAB - - - 1% ' AABBEAEBAB -- 1% - 1% AABBEAEBBB - - 1% - AABBEAEBAB - - - 1% AABCEAEBAB - 1% - 1% AABCEAEBBB -- - 1% -- BAAAEAADAB - 4% - 2% BAAAEAADBB 14 % 4% 6% 2% BAAAEAEDAB 1% 4% 1% 4% BAAAEAEDBB 11 % 4% 14 % 4% , i BAAAFABDBB - 4% 2% ~ j CADBEAEDBB 15 13 = - Approximate Totals 100% 94 % 97 % 92 % 5-43 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-15 DEFINITION OF THE TEN DIMENSION ACCIDENT PROGRESSION BINS Bin Parameter Outcomes 1 SequenceType A TB, fast C TC, slow B TB, slow D TC, fast 2 VesselBrexh A Yes B No 3 Ac Power Availability A Not available B Recoveredlate l C Recoveredearly D Always available 4 Core <oncrete Interactions A Yes, dry drywell floor B Yes, wet drywell floor C Initially, but arrested late in the accident D None 5 Early Containment Failure A Drywell rupture to the environment B Drywellleak to the environment C Drywellrupture to the refueling bay D Drywellleak to the refueling bay E Drywell rupture to the reactor building F Drywellleak to the reactor building G Wetwell rupture or vent to the environment H Wetwellleak to the environment 1 Wetwell rupture or vent to the reactor building J Wetwellleak to the reactor building K No early containment failure 6 Containment Failure at A Yes VesselBrexh B No 7 Late Containment Failure A Drywellrupture to the environment B Drywellleak to the environment C Drywell rupture to the refueling bay D Drywellleak to the refueling bay E Drywell mpture to the reactor building F Drywellleak to the reactor building G Wetwell rupture or vent to the environment H Wetwell leak to the environment I Wetwell rupture or vent to the reactor building J Wetwellleak to the reactor building K No late containment failure 8 DrywellSpray Availability A Sprays available from the start B Late slaay availability only C Early sprays only D No sprays 9 Direct Heating A Yes B No 10 Suppression Pool Bypass A Pool bypass both early andlate B Late pool bypass only C None I 5-44 l I NUREG/CR.4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) .) l I principal failure mode associated with the drywell meltthrough. In the first first two bins the release becomes a release direct to the environment later in the accident, as a result of-hydrogen-burn induced failure of the building. For the third and fourth bins the leak continues , i to be to the reactor building. For all four bins the drywell sprays are not available (most likely j due to power unavailability) and the suppn:ssion pool is bypassed late in the accident. Finally, j two of the bins include direct heating (the first and third) while the other two do not. Thus the l bulk of the risk is associated with releases that are not effectively scrubbed due to lack of sprays (as a result of power loss) and bypass of the suppression pool. Destruction of the reactor building (leading to its loss as a retention mechanism) as a result of hydrogen burns is also a substantial contributor to risk. Two other bins appear to be most important to the risk, particularly the central estimate. These bins, BAAAEAADBB and BAAAEAEDBB, are similar to those described above but are associated with the station-blackout accident that occurs on a longer time scale. These two bins are otherwise similar to the second and fourth bins in the list which were discussed above. The contribution of each source-term cluster to LLH mean risk is provided in Table 5-16 as an aid to further understanding of the results. The clusten are useful for understanding the role of various radionuclides species. Four clusters account for about 90% of the risk for all the risk calculations considered in the table (clusters 6-9). Examination of the actual source-term magnitudes and release characteristics in Section 5.3 for the clusters with high fractional contributions in Table 5-16 yields additional insights. l The important clusters tend to have rather substantial releases for most of the radionuclides ) groups, including La as discussed earlier. In the MACCS calculations these clusters were also characterized by a two puff release, one very short, high energy release at the time of containment failure, and another prolonged release. The important clusters have very short , durations for the first puff release, this results from a stage in the RELTRAC source term estimation in which radionuclides are relocated instantaneously at containment failure. The ; consequence codes treated these releases as of one-time step duration, one and three minutes for MACCS and CRAC2 respectively. In reality, the time of the puff release will be larger and the energy associated with the releases would be spread over a longer time period. There is some possibility (currently being investigated) that the assignment of such short durations I could lead to underprediction of early effects. This could occur because the releases are being treated as having more energy than than they should, leading to a lofting of the release to higher atmospheric levels, reducing the exposures closer to the plant. 5-45 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-16 FRACTIONAL CONTRIBUTION OF LLH CLUSTERS TO RISK 1,2 Contribution Contribution to Source-Team to Early Fatalities Latent <ancer Fatalities Cluster CRAC2 MACCS CRAC2 MACCS 3 .08% 0.02 % 2% 2.1% 4 .2% 0.1% 1% 2.1% 5 .5% 0.9% 6% 5.0% , 6 11 % 18 % 14 % 13 % 7 24 % 4% 20% 27 % 8 27 % 13 % 40% 32 % 9 32 % 60 % 14 % 17 % 16 .08% .1% 0% 0% 22 0% .01% .2% .1% 24 .02% .2% .3% .2% 27 .3% .2% .08% .06% 28 .2% .1% .3% .3% 29 .7% .1% .06% .09% 30 .2% .2% .1% .1% 31 .09% .1% .1% .1% 32 .2% .1% .03% .02% 34 .1% .06% .2% .2% 36 .2% .3% .2% .2% 37 .08% .2% .06% .04% 38 .09% .1% .05% .03% 39 .3% .07% .06% .08% 41 .2% .07% .2% .2% 42 .3% .4% .04% .06% 43 .1% .N% .02% .03% 45 .3% .2% .03% .N% 46 .1% .2% .05% .04% 47 .1% .07% .07% .06% 48 .7% .4% .02% .03% 50 .2% .2% .03% .04% 51 .1% .06% .01% .01% I Clusters with contributions less than .1% for all four columns are not included. 2 The mean of the LLH sunple members is illustrated in the table, but it must be emphasized that the sample cannot be interpreted as a probability distribution, in light of the manner in which the input information was developed. 5-46 l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Complementary CumnIntive Distribution Functions Complementary cumulative distribution functions (CCDFs) corresponding to the risk i results of Figures 5-7 and 5-8 are shown in Figures 5-9 and 5-10 for the MACCS results (only the results obtained using MACCS are provided because CRAC2 does not produce CCDFs for the evacuation assumptions used in SARRP). As discussed in Section 2.2.2, these represent the pro'oability per year (on the ordinate) that a given number of fatalities (on the abscissa) will be exceeded. The LLH risk displays shown in the previous two figums represent the integral of the data shown in Figures 5-7 and 5-8, evaluated separately for each sample member. Symbolically, this may be repmsented as follows: R= ,o P(C>Co) dCo where R is the mean risk, P is the probability of exceedence, and C is the magnitude of the consequences. Thus, the CCDFs separate risk into its two constituents, probability and consequences. As with the displays of mean risk, the results for each sample member are plotted, as are curves representing the 5th and 95th percentile of the sample. Also shown are the CCDFs for Peach Bottom obtained in the RSS.* The CCDFs for early fatalities indicate that most sample members predict less than ten early fatalities per year, and that very few sample points in the CCDF have values for early fatalities greater than 100. The RSS result is generally above the SARRP result (higher risk), and the SARRP CCDFs generally have a sharper slope--thus indicating a lower incidence of large numbers of fatalities than the RSS. The RSS CCDF for latent cancer fatalities lies in the range of the SARRP CCDFs. It should be noted that the SARRP result is for the MACCS consequence code while the RSS curve is based on CRAC. As seen in Section 3, the MACCS code predicts higher values of latent cancer fatalities than CRAC2, suggesting that CCDFr Msed on CRAC2 for SARRP would generally be lower than the RSS CCDF. i J l *As noted earlier, the RSS value was multiplied by a factor of 30 to convert from the yearly rate of incidence to total fatalities. 5-47 l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) e !i itiii iilii itiii iiiii ee e e i 's'.. 8 # I i % g n . . ... . . .. . . . .. s . . .gg ll I .. , ye r . .~us '8
- l W *' .;,
.h #1
- I RSS
- al f t!
..y . .,. .. unm cam { l l Figure 5-9. CCDF for Latent-Cancer Fatalities (MACCS) ' #[ !! $ $ j * . . , , . g!Ili*. n a ll ( ' , o y o { \l J ** 836 j *ad IV N T u a @c" se H Og . l E .. *.'. *. ul u ll Ig o o. as !' i B g Igl 8 es aa e .le . l o88s ale 19 a . :.8 .eg,. l :18e: Is .. . . 1 . .* l 8* * !g . g ..l a . [g_ao .8 { g " g. 'y .- g sy "y Figure 5-10. CCDF for Early Fatalities (MACCS) 5-48 1 NUREG/CR-455 t, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Results and Insiehts from the Uncertainty Analysis f The LLH uncertainty ranges have been illustrated in the pmvious sections. The reasons for the ranges and the importance of the various issues are discussed in this section. The first method used to examine the relative importance of the LLH issues was a rank-regression j I analysis of the results. Multivariate rank regression was employed to estimate the importance of each issue in determining the size of the risk uncertainty. This method differs from the methods used for the other SARRP plants where the regression was done for each variable .(issue) individually. The differences between these two methods are briefly discussed in Section 2. The multivariate method is believed to be more robust. The rank-regression msults are reported in terms of the fraction of variance in risk (which can be represented by the rank-regression model) that can be attributed to a given issue. There are two limitations to this interpretation of the results. First, the variance in risk was determined by the rank of the sample members (e.g.,3 in 150) in the range of outcomes for a particular risk measure, rather than the values of the risk measure actually calculated (e.g.,0.012 latent cancers per year). As a consequence, the fraction of variance explained cannot be directly translated into the displays of the range of risk for the consequence in question. Second, the rank-regression model cannot represent all of the variance for any risk measure seen in the LLH sample. Thus the fraction of variance explained by an issue is the fraction of the total variance that can be , explained by the rank-regression model. In the current results, the variance which can be f explained by the multivariate rank-regression models is between 80 to 90% of the total variance. However, the results are reported in terms of the unique contributions of the j individual issues and these generally only account for about one-half of the modeled variance-- I the rest of the variance is due to combinations ofissues. The resalts reported here are based on the more the accurate multivariate rank regression, although a comparison to the results using the single variable methods of rank regression is also provided. The complete results of i i regression analyses using both methods are provided in Appendix D. l The multivariate rank-regression results for the LLH study of latent-cancer fatalities is shown in Figure 5-11 (only the results obtained using MACCS are shown, but the CRAC2 i results were nearly identical). The plot displays, for each issue which had statistical significance to the risk uncenainty, the fraction of the total explainable variance in risk which is uniquely attributable to that issue. In other words, if all of the uncertainty in a given issue were eliminated, the total variance in risk would decrease by at least the fraction indicated. (Because of correlations and synergisms between issues, the variance may actually decrease more.) The fractional decrease is measured by the rank of the sample members and the change in the 5-49 NUREG/CR-4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) 0.30 2 Rank Regression: R = 0.90 I $ N 6a $e $ . .sr . g $ z u . 8 0.20 - E E eI E' *p. f k h 5 .oE b 5 I 9 f -@ s3 o E g . e E 3 g 8 $ g o $ ,3 0.15 - 73 8 e & 3 5 5 % $ g 8 37 as 3 3 a
- a E
3 3 u $He -15 g, a e 3 - 8 3 e 34010' .B u1 S m h o h*s 4 17; >
- a. c. m& c!5o 3 3 e a@
O.05 - 0.00 Issues Figure 5-11. Multivariate Rank-Regression Results for Risk of Latent-Cancer Fatalities (MACCS Consequence Calculations) variance of risk may therefore be slightly smaller or larger than the fraction indicated. In any event, the rank-regression plot in Figure 5-11 points to a few issues as important and significant to the uncertainty in latent cancer risk: the failure rate assessed for the common-mode failure of the de power system (about 20% of the variance), the probability of ac power recovery following a loss of offsite power (about 7% of the variance), the probability of drywell meltthrough (also about 7% of the variance), and the reactor building decontamination factor (about 4c5 of the variance).* The de power issue is important to the uncertainty since it directly affect . .A most mquent accident sequence, and because the LLH range for the issue was very large--over four orders of magnitude. The second issue, ac power recovery, is important because it directly affects the probability of core damage for all of the important station blackout scenarios. In addition, this issue can have a significant impact on the containment response since the availability of water, the operability of the sprays and the ability to vent are all affected by the power recovery values. One containment issue appears to be important: the issue dealing with the probability of drywell meltthrough. The importance of this issue is expected since it involves a direct failure mode of the drywell and has a range from a probability of zero to a probability of unity that is sampled only over those two values. Thus *As noted previously, the relative role of the ATWS sequences is not accurately characterized in these results and the rank regressions may not indicate the correct importance for the first issue, SLC initiation probability. 5-50 NUREG/CR.4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) this issue results directly in the stratification of the sample members into two groups defined by the outcome for this issue. One source term issue was found to be significant. The amount of decontamination associated with the passage of the release through the reactor building is quite uncertain and very important to risk since it is the last barrier to radionuclides release and any uncertainty in this parameter translates nearly directly to uncertainty in risk. It should be noted that the results of the rank-regression analysis using the single variable method included in the RISQUE code (and used for the other SARRP plants) had very similar results in terms of the relative importance of the uncertainty issues. Overall the four issues discussed above account individually for about 40% of the 90% of the total variance that can be explained with the rank-regression model. Since the variance explained uniquely by ; individualissues is less than one-half the total variance, the interactions between issues are an important feature of the uncertainty. This suggests that the reasons for the uncertainty are fairly complex and interrelated, and that a change in uncertainty associated with the change in the uncertainty for a given issue is not easily derived without additional calculation. In order to better understand the containment and source-term issues with a significant uncertainty in terms of risk, the risk was recalculated conditional on core damage. Specifically, the risk for each sample member was divided by the core-damage fmquency for that sample member to obtain an LLH variation of the mean consequence given core damage. The risk results indicated that the LLH range of uncertainty was reduced by about a factor of six when the effects of the sequence issues on core damage fmquency are removed. The rank-regmssion ) results for this case are illustrated in Figure 5-12. The regression for this case is based on the single variable rank-regression models incorporated directly in the RISQUE code. (See the discussion below concerning early fatality risk rank-regression analysis for a comparison of the ] results of the two rank-regression methods). These results further illustrate the importance of 1 the drywell meltthrough issue and the reactor building DF uncertainty. Two other issue become more important: the probability of reactor building bypass and the CCI release magnitudes. In addition, the front-end issues still indicate some importance because they determine the relative contribution of each plant-damage state to the core damage frequency. Since some damage states intrinsically lead to higher consequences (because of the containment , response), these relative contributions of plant-damage states impact risk. The issue goveming the non-recovery probability fo; ac power also impacts containment response directly for the station blackout damage states 'TB, TBU, and TBUX) and was thus expected to retain its risk importance for the case of risk conditional on core damage. 5-51 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) C Statistically Significant Not Statistically Significant i 0.5 - l i 5 E 0 .4 - I 8hN *
- fI
.$ m $e g .o ke E $ [* S m A a g R @ !:M j h8 [IE2 $9 js$$
- e I. e e en E .5 ti o 8 5 1 l' s >g $ E s %.6 89 ou ii si a s d d,3 )5, ug g Fo.3 35
-h0'3 -E E" E 8 e J.h >li ?E g .!. R g ma - 1 0.2 - $ b !! h fi! N e e b b $jg d!sek 9 98 R !IS l .5 3 e B S B " vSSai$$$3$.2 = [ u e 3 $ 3 h
- g. 6 k3ek$L ie 8 4 ?3 1 e 3g -,.,-$- E 3 h, sy I !!D ! M 8 .c 0.1 - 3 [ h3 g .& 3: d i .seIE ed wh 44 =E2 m o d d d d >8 d" E m a g sommemo r ma !
0 Rm : mn_ _ r,Rmn C m ...U:a Issues Figure 5-12. Single Variable Rank Regression-Results for Risk of Latent-Cancer Fatalities (MACCS Consequences, Conditional on Core Damage) The results of the rank-regression analysis for the risk of early fatalities are illustrated in Figure 5-13 (again, only the MACCS results are illustrated, since the CRAC2 results were ; similar). The same issues as discussed for the latent cancer fatality results were important here 1 as with the latent-cancer risk measure, although their relative importance is different. The most important issue is the uncertainty in the decontamination associated with the reactor building (about 10% of the variance). A number of the other issues important to latent cancer uncertainty contribute approximately equally to the early fatality risk measure (~5% of the. variance each): the de power failure probability, the ac power recovery probability, the probability of drywell meltthrough and the reactor building decontamination factor. Some other issues also appear in this rank regression including the uncertainty in the magnitude of the radiological releases associated with the core-concrete interactions, the decontamination factor for volatile iodine in the suppression pool, the probability of reactor building bypass due to damaging hydrogen burns. Other containment issues were calculated to be significant: the vessel failure mode and the extent to which the suppression pool is bypassed by a stuck-open safety / relief valve vacuum breaker. As with the latent cancer results presented above, the total variance explained uniquely by individual issues is approximately 40 % 5-52 Q. NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,1987) 0.30 2 Rank Regression: R = 0.87 m ? 0.25 - o * , E m > e o b
- o $ e
" $ 0.20 - ! c 6 0 t 5 E $ 5 E $ h Q- u @ g [ @ g 2 i e g g i 43 % E g . .g E 3 $ 8 E 5 o f B r 0.1s ; 8 e e a g a - 18 8 j 8 37 /! .s 3 3 a
- B 3 3 o u
$He k$.s 3 e 3 8 T> i 0.10 ' O ja w hho o <c o 15 > h o c. o
- c. w h g c:
o O 4gcc a u I y 0.05 ' A 0.00 Issue Figure 5-13. Multivariate Rank-Regmssion Analysis for Early-Fatality Risk (MACCS Consequence Calculations) For the early fatality risk measure. the results of the rank regression using the single variable methods of analysis am presented in Figure 5-14. The rank-regression model for this method is generally able to explain only about 40% of the variance and the relative importance rankings am not exact. A comparison of the results in Figure 5-14 with those in Figure 5-13 illustrate some differences in the relative importance of issues. In particular, the issue concerning the probability of drywell meltthrough was seen as most important in the simpler rank regression while the more accurate representation in Figure 5-13 indicated that this issue was less important that the reactor building DF issue. In addition, there are some differences in the lower level contributors, with some that are significant in one set of results being not significant in the other. (The results for the latent cancer risk measure were very similar for the two different methods.) The multivariate methods allow for a more accurate picture of the rank ofimportance of the uncertainties, and the values of fraction of variance explained are more accurate. The general conclusions for either method are similar, but detailed consideration of the effects of removal of one uncertainty could be somewhat different as assessed by the two methods. The regression analysis was also performed on the individual levels of the issues found to be important. This regression is performed relative to the first level for each issue, providing a listing of the relative importance of an issue level to risk compared to the first level 5-53 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) for the issue. The results of this analysis for the latent cancer risk measure are indicated in Figure 5-15. Levels 3 and 4 of the de power cormnon mode issue are important relative to the first level, a result that is expected since this represents the range of uncertainty in this issue whic) was found to be the single most important issue to risk uncertainty (Figure 5-11). The resalts in Figure 5-15 generally mimic the results in Figure 5-11, indicating the importance of the levels at the other end of the range from level 1 as being important. A complete listing of the rank regression analyses by level are included in Appendix D. 0.6 Statistically Significant Not Statistically Significant , 0.5 - .c { Im 3 gg ggohlN m 0.4 - > gg .o e e g ;;; e 3 h t: a e m g -;g e E hTh r g e .o i c e m$ e e-Bee e " .o Nb 4 e .ls e = u. e .e 3.0.3- g g ,C g 3 gogg , 4 7 y i g g g ,,y g gg c u.3 E aE1>E ."i1E .u **e.5 E j 16 e e 3 . $ 5 dde{5QE@Q0 oEcg3 3cc - e G E .o eg oc $x 0'2 - 2 0 h U atSB > v S S = $ $ $ $5 NI* e h m aemg3eeo%%%%s*s --- 2 a JT Y s E '5 ..=2[ k m %e e {j @ .h s at N L y man e sicin d y - :e 5&& e as 0.1 - d812l kel58888$8&m E E .g bd && selc& U $2 ' ~ g _ n - _nn_ ___ ___ n [ n : - n - n n_ lssues Figure 5-14. Single Variable Rank-Regression Analysis for Early-Fatality Risk With the RISQUE Code Regression Models The regression analysis was also performed on the individuallevels of the issues found ; to be important. This regression is performed relative to the first level for each issue, i providing a listing of the relative importance of an issue level to risk compared to the first level for the issue. The results of this analysis for the latent cancer risk measure are indicated in Figure 5-15. Levels 3 and 4 of the de power common mode issue are important relative to the j first level, a result that is expected since this represents the range of uncertainty in this issue which was found to be the single most important issue to risk uncertainty (Figure 5-11). The results in Figure 5-15 generally mimic the results in figure 5-11, indicating the importance of i the levels at the other end of the range from level 1 as being important. A complete listing of the rank regression analyses by level are included in Appendix D. 5-54 l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 0.70 Rank Regression: R = 0.81 2 0,60 - N $~ n 4 N g 28 s s s a N e v Q & **** . e *1 . 3 N n
- 3 i g i i n v So g 3 n J J J i i li'u>
Ta l c 1 c 4& lY l ! ! ! b & .c c *c *c ii*c 3 Q 3 Q 7, E 0'3 ' @ @ s i$ $ _j' E F s _g . 3? e e --'[F 8 .8 $ g ' oc o.20 8 8 e e Q - =i a 8 5 8 8 3- y $ a a a a a g 8 2 - k k $ $ 5 " E E 8 0 0 0'10-a O b & 3 1 3-1 cO cO cO m
- 8 C
o.00 i 1ssue Figure 5-15. Rank-Regression Results for Risk of Latent-Cancer Fatalities by Level As described in Section 2, a chi . squared test for goodness of fit was also performed on the LLH results. This test was used to examine the number of samples of a given level of a given issue above and below the median, and to compare these values with what would be expected due to chance alone. Thus, if any level of an issue is present predominantly in the sample memben at either the high or the low end of the sample, this test should highlight that effect. As with the rank regression by level, this approach allows examination of the effects of ] individual levels of a particular issue. The parameter calculated, the significance, is the probability that the effect of the issue is due to random causes. For example,if the significance calculated in the test is 0.12 for a given level of a given issue, then there is a 12% chance that the stratification in risk space of results with this issue level would occur in a random distribution, conversely, an 88% chance that the result is actually the consequence of the impact l of that issue level on risk. Because this test is performed for each level of every issue, even if 1 one level of an issue is particularly significant in determining risk uncertainty it effect can be seen. This is particularly true if the risk (rank) is not monotonically dependent on issue level. The results of the chi-squared test are listed in Table 5-17. The results of the LLH are listed, as are the results for the sensitivity case involving only the uncertainty in mean consequences given core damage. The chi-squared test points to the drywell shell meltthrough uncertainties as being the most significant with regards to both risk measures, once again this l 5-55 NUREG/CR 4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) issue divides the 150 sample members equally between assured drywell meltthrough and no meltthrough when vessel breach occurs The de power common-mode failure probability is the second most imponant issue. As described earlier, this issue has a major impact on the frequency of the dominant accident sequence. One other sequence frequency issue, the recoverability of offsite power was also found to have a major effect. This issue effects directly the containment response, as well as the accident frequencies. One source-term issue appears to be significant: the uncertainty associated with the reactor building decontamination factor. As described previously, this issue is important i because it is the last barrier for most of the releases. Two other source-term issues indicated significance for some of their levels. The magnitude of the CCI releases indicated some significant contribution. The CCI releases are the most important risk determinants for Peach Bottom so this is not surprising. The suppression pool scrubbing of volatiles was also somewhat important for early health consequences. l l The role of the other issues can be seen from the individual entries in the table. Any j I conclusions based on the chi-squared results must be qualified by an understanding of the simplicity of the test employed here: the risk space is partitioned only into ranges above and below the raedian and the statistical results are therefore not rigorous. The insights derived from the chi-squared analysis merely support the rank regression results presented previously. j 5-56 l NUREG/CR-4551, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) 3 Table 5-17 j i RESULTS OF CHI-SQUARED TEST (MACCS CONSEQUENCES) x 2Significance Early Fatalities Latent-Cancer Fatalities Comments issuel and 1.1R Mean Cons. IDI Mean Cons. Note: In this study, values of the test greater q Outcome Risk Given Core Damage Risk Given Core Damage than 0.1 are considered significant. j a Issue 1 (1-lh Failure to Actuate SLC Level 1 1.0 0.6 0.3 0.6 This issue is not a significant contributor Level 2 0.4 0.3 0.3 0.3 to uncertainty in risk, by this test. Level 3 1.0 1.0 0.6 0.4 Level 4 - 0.1 0.1 1.0 1.0 Issue 2 (1 2h De Power Common Mode Level 1 0.1 1.0 0.04 0.6 This issue is a very significant contributor Level 2 0.8 0.1 0.02 0.2 to uncertainty in risk because it has such Level 3 0.8 0.03 0.1 0.8 a large impact of the frequency of the l Level 4 0.01 0.3 0.0 0.1 most frequent damage state. The issue is j most important to latent cancer fatalities d since that risk measure is most determined l by the total release which is directly I affected by the core-damage frequency. Issue 3 (1-3h Failure to Vent--ATWS Level 1 0.3 0.6 0.03 0.03 One level of this issue appears to be l Level 2 1.0 0.8 0.6 0.8 significant to the latent risk measure. Level 3 0.8 1.0 0.4 0.4 However, conclusions regarding the ATWS Level 4 0.1 1.0 0.1 0.1 scenarios cannot be made based on these results (see text). l 1 Issue 4 (15h Ac Power Recoverv ' l Level 1 0.04 0.04 0.002 0.002 This issue is a very significant contributor ' Level 2 0.3 0.4 0.07 1.0 to uncertainty in both risk measures. This ! Level 3 0.07 0.07 0.02 0.07 issue affects the containment response as Level 4 0.6 1.0 0.04 0.6 well as the frequency of the station l blackout accidents. l Issue 5 (2-lh Tai!rine Vacuum Breaker Level 1 1.0 1.0 0.6 0.6 This issue is not a significant contributor Level 2 0.8 0.8 0.4 0.8 to uncertainty in risk, by this test. Level 3 1.0 0.6 0.4 1.0 Level 4 0.6 0.6 0.6 0.3 Issue 6 (2-2h HPSW Drvwe!! Serav Recovery Level 1 0.6 0.6 0.6 0.6 This issue is not probably not important Level 2 0.8 0.8 0.1 1.0 to uncertainty in risk. Although the fourth Level 3 1.0 1.0 0.2 0.4 level is significant for some cases, the Level 4 0.01 0.3 1.0 0.04 trend is not consistent. Issue 7 (2 3h Failure to Vent--No Ac Power Level 1 0.3 0.6 0.1 0.1 This issue is not indicated as being Level 2 0.3 0.07 0.4 0.07 important because the LLH levels and Level 3 0.4 0.1 1.0 0.4 weighting factors did not indicate any Level 4 0.6 1.0 1.0 0.6 uncertainty (less than a factor of 2). Issue 8 (2-4P Failure to Vent..Ac Recovered Level 1 0.3 0.6 0.6 0.6 This issue was not significant because the Level 2 0.4 0.2 1.0 0.6 levels assigned all indicated a fairly Level 3 0.6 0.8 0.6 0.8 reliable action to vent in these l Level 4 0.1 0.1 0.1 0.3 circumstances. 5-57 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-17 (continued) RESULTS OF THE CHI-SQUARED TEST (MACCS CONSEQUENCES) x 2Significance Early Fatalities Latent-Cancer Fatalities Comments issuel and Lili Mean Cons. LUI Mean Cons. Note: In this study, values of the test greater Outcome Risk Given Core Damage Risk Given Core Damage than 0.1 are considered significant Issue 9 (2-5h Size of Sueoression Pool Bvoass Level 1 0.9 0.6 0.7 0.7 This issue is not significant because issue Level 2 1.0 0.4 0.7 0.7 5 was net significant, and this is an another factor in the uncertainty in that issue. Issue 10 (24)* Containment Faiture Pressure (T<5n0) Level 1 Na Na Na Na This issue is generally not significant, Level 2 1.0 1.0 0.5 0.6 except for a few levels. Level 3 refers Level 3 0.03 0.007 0.07 0.3 to the lowest pressure capacity for the Level 4 1.0 1.0 1.0 1.0 drywell knuckle location. levels 8 and 9 Level 5 0.7 0.2 1.0 0.7 are the highest pressure capacities for the Level 6 0.5 0.3 0.7 0.7 drywell. The statistics of this issue are not Level 7 0.2 0.2 0.4 0.7 particularly good due to the number of Level 8 1.0 1.0 0.4 0.01 levels. Level 9 1.0 0.6 1.0 0.08 Issue 11 (2 7h Containment Failure f_ocation (T > En0) Level 1 Na Na Na Na This issue is not significant, by this Level 2 0.7 0.7 0.8 0.7 tes t. I.evel 3 0.6 0.6 0.8 0.6 issue 12 (2-8h Containment Failure Sire (Racid Pressurintion) 1.evel 1 1.0 0.01 1.0 1.0 This issue is not significant except to Level 2 1.0 0.4 1.0 1.0 early fatalities after the sequence frequency effects are removed. Issue 13 (2 oh Containment Fai!ure Size (Slow Pressurization) 1.evel 1 1.0 1.0 1.0 0.6 This issue is not significant by this test. Level 2 1.0 1.0 1.0 0.6 Issue 14 (2-10h Vessel Failure Mode Level 1 0.5 0.2 0.8 0.2 This issue is not significant by this test. Level 2 0.5 0.2 0.8 0.2 issue 15 (2-11k Containment Pressure Before Vessel Breach--I_nne-term Blackout level 1 0.1 0.04 0/)5 0.1 This issue appears to have some level of level 2 1.0 1.0 0.8 0.4 significance, mostly after the impact of the level 3 0.2 0.09 0.2 0.02 frequency issues are deleted. This issue 1.evel4 Na Na Na Na helps determine the probability of ; failure at vessel breach. ? Issue 16 (2-12h Pressure Rise at Vessel Breach level 1 0.5 0.3 0.7 0.7 This issue was generally not significant. Level 2 0.07 0.03 0.4 0.6 The levels assigned to this issue only l level 3 1.0 0.7 0.2 0.7 spanned a factor of two in uncertainty. Level 4 0.4 0.7 0.2 0.4 1.evel5 0.3 0.3 0.1 0.3 Issue 17 (2-13h Pressure Due to Sumo Water Interactions Level 1 0.4 0.8 0.4 0.4 This issue is not significant to risk level 2 0.5 0.6 0.2 0.2 uncertainty, by this test. g Level 3 1.0 0.6 0.1 0.1 5-58 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-17 (continued) RESULTS OF THE CHI-SQUARED TEST (MACCS CONSEQUENCES) x 2Significance Early Fatalities I.atent-Cancer Fatalities Comments issuet and IDI Mean Cons. IDI Mean Cons. Note: In this study, values of the test greater Outcome Risk Given Core Damage Risk Given Core Damage than 0.1 are considered significant Issue 18 (2-14h ' Probability of Drvwell Meltthmuch Level 1 0.001 0.006 0.001 0.0 This issue is the most significant overall. Level 2 0.001 0.006 0.001 0.0 The sample members for nearly all risk measures are nearly completely stratified by the outcome for this issue. Fifty percent of the sample members have containment failure by this mode. Issue 19 (2-17h Reactor Buildine Bveats Level 1 0.003 0.8 0.6 0.02 One level of this issue appears to be Level 2 0.2 1.0 0.8 0.4 somewhat important, but the trend was not stable between the cases with and without the effects of the sequence frequency issues. Issue 20 (3-lh In-Vessel release from the Fuel Level 1 0.3 0.4 0.7 0.5 This issue does not appear to be Level 2 0.3 0.2 0.5 0.8 significance, by this test. Level 3 0.3 0.3 0.5 0.3 Level 4 Na Na Na Na Issue 20 (3-2h Csl Decomposition Level 1 0.4 0.1 0.5 1.0 This issue does not indicate any Level 2 0.8 0.1 0.6 0.6 significance, by this test. Level 3 0.6 0.8 1.0 0.8 Level 4 0.3 0.3 0.6 1.0 Issue 21 (3-3h Retention in the RPV Level 1 0.1 0.1 0.3 0.3 This issue does not appear to be Level 2 1.0 0.4 0.6 0.3 significant, by this test. Level 3 0.5 1.0 1.0 0.5 j Level 4 0.7 0.7 0.09 0.3 i Level 5 1.0 1.0 1.0 1.0 Issue 22 (3 4h Sureression Pool DF--aemsols Level 1 1.0 0.3 1.0 1.0 This issue does not appear to be Level 2 0.6 0.4 0.8 0.8 significant, by this test. Level 3 0. s7 0.7 1.0 0.3 Level 4 0.4 0.7 0.7 1.0 Level 5 Na Na Na Na Issue 23 (3-Sh Sureression Pool DF--volatiles Level 1 Na Na Na Na This issue does not appear to be Level 2 0.04 0.3 0.3 0.5 significant, by this test. Level 3 0.3 0.02 0.8 1.0 Level 4 0.2 0.2 0.8 0.8 Level 5 0.002 0.0 0.6 0.6 Issue 24 (3-6h Revolatili7ation fmm the RPV Level 1 Na Na Na Na This issue does not appear to be Level 2 0.2 0.6 0.8 1.0 significant, by this test. Level 3 0.8 0.5 0.5 0.5 Level 4 0.1 0.1 0.1 0.1 j l 5-59 - NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) Table 5-17 (continued) RESULTS OF THE CHI-SQUARED TEST (MACCS CONSEQUENCES) - y,2 Significance Early Fatalities Latent-Cancer Fatalities Commetits i Issuel and IDI Mean Cons. IDI Mean Cons. Note: In this study, values of the test greater l Outcome Risk Given Core Damage Risk Given Core Damage - than 0.1 are considered significant j issue 25 (3-7h Maanitude of CCI Relensee level 1 0.2 0.01 0.3 0.06 level 1 of this issue is sign!ficant Level 2 0.6 0.6 0.8 1.0 when the sequence frequency issue effects Level 3 0.3 0.1 1.0 0.6 are removed. The CCI releases are very level 4 0.6 0.6 0.6 0.3 important to the risk, but they do not appear to be very important to risk uncertainty - Issue 26 (3-8h Reactor Buildine DF Level 1 0.02 0.02 0.02 0.005 This issue appears to be very significant 12 vel 2 0.1 0.03 0.8 0.2 to both risk measures. The reactor Level 3 0.03 0.01 0.05 0.006 building is the last barrier for the release 12 vel 4 0.04 0.01 1.0 0.1 and the DF as.igned here translates directly to the risk results for any sequences that have failures which vent through the reactor building. Issue 27 (3-9h Refueline Bay DF level 1 0.6 1.0 0.6 1.0 Although this issue is similar to that above level 2 0.3 0.7 0.7 0.7 in nature, it is not that important because Level 3 0.2 0.2 0.8 0.8 it has limited uncertainty and it applies level 4 0.8 0.3 0.8 1.0 only to a limited number of failure modes. Issue 28 (3-10h late Release of Iodine Level 1 1.0 1.0 1.0 1.0 The iodine releases are generally not Level 2 1.0 0.8 0.8 0.8 important to the risk at Peach Bottom. Level 3 0.4 0.8 0.8 1.0 Level 4 0.2 0.7 0.7 0.7 lissue numbers in parentheses refer to their denotation in the risk-code printout provided in Appendix C. 2Na implies that there were too few sample members with this outcome to support a chi-squared test. 4 5-60 ! i j l NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 5.5.2 Results for Other Risk Measures The n:sults for the other risk measures (early injuries, individual risk of death, offsite costs, and population dose) are presented in this section. The results are presented with less discussion, since the insights in the previous section apply to these results as well, and no new insights were discovered through the calculation of these additional measures. Further detailis provided in Appendix C for the interested reader. Figure 5-16 illustrates the results for the risk of early injury. The central estimates and the LLH ranges are indicated for both the CRAC2 and MACCS consequence calculations. As expected, the general pattern of results is similar to the early-fatality risk measure, with the exception that the CRAC2 results span a wider range for this risk measure, and the top of the CRAC2 range is higher than the MACCS range. As discussed in Section 5.4, calculational differences between the two consequence codes account for the differences in the results. The rank-regression analyses are not illustrated, since they are very similar to those for early fatalities. Note also that the differences between the LLH and the central estimate seen in the early fatality results are also seen here, for the same reasons. The results for individual risk of fatality are presented in Figure 5-17. This is essentially the risk to an individual in the first populated interval (from the grid used in the consequence calculations); it is calculated as the number of fatalities within the first interval l divided by the total population in that interval at the start of the accident. The rank regression for this risk measure is illustrated in Figure 5-18. This case is slightly different from the other early effects. The most important uncertainty is the de power common-mode probability, ) followed by the reactor building DF uncertainty which was the most important issue to the other early effects risk measures. The magnitude of CCI releases is not significant to the uncertainty in individual risk, nor are the uncertainties in vessel failure mode or the size of suppression pool bypass which were important to early fatalities. One issue is seen to be somewhat important to individual risk uncertainty which was not important to early fatalities or l illness: the uncertainty in the probability of initiation of the standby liquid control system. This issue might have also been important to other risk measures if the ATWS treatment had I been more accurate. The population dose results are illustrated in Figure 5-19. These results parallel the trends seen in the results for the risk of latent-cancer fatality discussed in the previous section. The MACCS code predicts higher population dose than CRAC2 by factor of two to five. The rank regression analysis for this case is plotted in Figure 5-20. The important issues for this 1 5-61 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) so' LLH LLH (MACCS) (CRAC2) . .-3, Central :* Central : 6 (CRAC2) . g ifj (MACCS) 4 $ e a
- a. .I Ih .
s ' ! t l ! a 1 10 ! I I, 7 . 10 A e ,e - Figure 5-16. Annual Risk of Early Injury s o I Central ,,, E (M ACCS) ! ; O (bYAC2) e i . < i i o ,o , 3 5 e I i ,. . go. i ~ b -11 io : : LLH LLH (MACCS) (CRAC2) . ,\ 1C Figure 5-17. Individual Risk of Fatality 5-62 1 4 _ - _ _ _ - _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _ _ - - - _ _ _ _ _ _ - _ _ _ _ _ _ _ _ _ _ _ . I NUREG/CR 4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 0.30 . 2 Rank Regression: R = 0.86 I 0.25 - g .7 e .a .c ' m 3 > s8 e a e m Q o R .O o m a 1 - e e i 2 o S 3 u. 8 .e e 0.20 - c y .c e E e a 8's m a 5 9 .* 8 @ > A 0.15 - _d ~ o ~ o . .e c .)e - y a7 8 -- g 3- i 3 8 N.g ' n -- t B y' 3 8 {3 m o 0.10- 3 { B g e[ e 3 e E- .E}W O 2 E' 3 o h $. o + $ R id - .S 4 o17; > a. a. m e o a a g 0.05 - 0.00 Issue Figure 5-18. Multivariate Rank-Regression Analysis forIndividual Risk of Fatality , LLH
- (CRAC2) e : 8 e i+ . : .
- Central !
E E (MACCS) l. g A g (Central CRAC2) . .1 8 ' 5 e . c 8 10 ) e:. s E l' i ..- + I 3 LLH i ., (MACCS) i i ' Figure 5-19. Results for Risk Measure of Population Dose per Year-i i i t S-63' NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l l i 1 0.30 Rank Regression: R = 0.91 $ 0.25 - g 8 [p.E 0.20g - 5 5 5 b$ E e E$ d5 0.15 - a 8 E 8 EE .) h fs$ e e E M 3 e E j S 8 ] A u. 3 E 8 = s w a $a .5 3 45 .5 a a S 3 {, 15 g, g g 8 j a 0.10 - 3 o u1 o { S E o j m u $ d a -)- E $ E e o $ $ 8 2 $ $ ? ? $ E o E E a 0.05 - 0.00 Issue Figure 5-20. Rank Regression Analysis for Population Dose risk measure are essentially the same as for latent cancer fatalities (one relatively minor issue, the probability of standby liquid control initiation, is not seen in the population dose rank regression), but their relative importances are somewhat different. The de power common-mode probability uncertainty is the single most important issue, accounting for about 25% of the total variance. The ac power recovery is the second most important uncertainty. These two accident-frequency issues are important because they have a significant effect on the probability of release which is one of the principal determinants of the population dose. Drywell meltthrough is the next most important uncertainty to this risk measure. l The risk results in terms of offsite costs are indicated in Figure 5-21. As described in Section 5.4, the MACCS results are generally higher because MACCS considers the inhalation dose due to resuspension in calculating the need for interdiction but CRAC2 does not; this leads to higher estimated costs for decontamination. Once again, the insights into uncertainty derived from the rank-regression and chi-squared analyses are essentially the same as those discussed for latent cancer fatalities. 5.5.3 Observations Concernine the Risk Results l There is a significant quantity ofinformation presented in this section; the key insights regarding the risk n:sults for the Peach Bottom plant are summarized below. J 5-64 l l { NUREG/CR-4551 VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) io' q I LLH i o', (CRAC2) ? E(Nc'$s) , iot i ! i. O(!$$E2$ 5 'I l , . i s ' : l I_ j LLH j (MACCS) , i o' Figure 5-21. Results for Risk in Terms of Offsite Costs The early fatality risk as calculated by the LLH method is lower than the previous study of Peach Bottom, the RSS. The IDCOR result for this facility does not predict any early fatalities but this is a product of the evacuation assumption more than any other aspect of the analysis. The central estimate of early fatalities was higher than the LLH range, and similar to that estimated in the RSS. The principal difference between the central estimate and the LLH is the amount of lanthanum predicted to be released. The central estimate was based directly on the STCP while the LLH included estimated releases that reduced the lanthanum releases by an order of magnitude or more. This reduction was based on the results of sensitivity studies, as well as on the expert opinion of the reviewers who provided the LLH input. The effect of the lanthanum release on early health , I effects was also decreased by the source-term clustering process used in the I LLH, possibly magnifying the difference between the LLH and the central estimate. The importance of this latter effect is currently being assessed. The risk of latent-cancer fatality is also low compared with the RSS value ) for Peach Bottom (when calculated with similar codes, CRAC and I CRAC2). The MACCS result covers a range that is generally a factor of l two to five higher than the range' predicted by CRAC2. As described in Section 5.4, the MACCS code models for population dose and latent cancer ris3: ' ve been updated, resulting in higher consequence pn: dictions. The central estimates of latent cancer risk are well within the LLH range, although they are in the upper one-thini of the uncertainty range. This result is due to the weights assigned by the experts to two key source term issues: 5-65 i e NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRU ARY,19b7) reactor building decontamination and core-conente interaction releases. The experts weighted reactor building decontamination factors at least as large as the STCP nine times more heavily than factors lower than the STCP. Ninety eight percent of the weight was assigned to CCI releases no greater than those predicted by the STCP; almost three-quarters of the weight was given to lower releases. . The latent cancer results for IDCOR compare well with the SARRP results. Two IDCOR results are illustrated because that study provided results for the plants as they existed at the time of the analysis and after consideration 3 of committed changes. The ASEP plant analysis would fall somewhere in between the two IDCOR results in terms of plant boundary conditions. This comparison is at least somewhat fortuitous however, smce the features { of the analysis which drive the results for the two studies are totally different. Overall, the most important uncertainty in the Peach Bottom analysis is the probability of common-mode failure of the de power system. The importance of the de power uncertainty derived from three factors: 1)It is the leading cause of severe core damage; 2) Recovery of ac power is assumed to be far mom difficult when de power is unavailable for resetting - circuit breakers and switches; 3) The safety / relief valves are not available for ! reducing pressure in the reactor pressure vessel and thus the containment loading which results from vessel breach is more severe. The second most important uncertainty in the study of Peach Bottom involves the occurmnce of a drywell meltthrough after the core has left the vessel. The impact of this uncertainty on risk is somewhat overstated in these results by the assumption that meltthrough occurs at vessel breach. However, the SARRP expert reviewers generally supported this treatment as being appropriate given the slump-type (coherent) in vessel melt progression. This failure mode is significant since it leads to suppression pool bypass, even if the containment was previously vented through the suppression pool. Relatively early on in this analysis, it was recognized that the reactor building decontamination factor is a key uncertainty. The results obtained clearly support this hypothesis. While several calculations were commissioned as an adjunct to this study, the uncertainty in this issue has not been substantially reduced as of this writing. The probability that hydrogen deflagrations will result in catastrophic failure of the reactor building (leading to negligible retention of radionuclides) is a related uncertainty which is also important to the results. The type of in-vessel melt progression (slump versus flow--cohemnt versus incoherent) may impact risk uncenainty to an even larger extent than is reflected here. Some experts believe strongly that a drywell meltthrough is certain if a slump-type melt occurs. The other experts' lower weighting for this particular case of the meltthrough issue probably reflects their lack of belief in this model for melt progression. In addition, a slump-type core , melt progression may lead to direct containment heating with a probability ! greater that the 0.1 value which was assumed. Neither the probability of direct heating not the containment pressure rise attributable to direct heating was considered as an uncertainty is this analysis. 5-66 NUREG/CR.4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) . The uncertamty in the probability of ac power recovery affects nearly all the dominant accident sequences since they principally involve a loss of electric power. In addition, the containment response can be very significamly affected by the availability of electric power. - Recovery of core and/or containment cooling after initiation of core damage affects the results, but the effect is limited. About 20% of the ASEP core-damage frequency results in conditions in which the core is damaged but the vessel is not breached. The uncertainties associated with power recovery are probably understated in these results. In particular, the probability of arresting the accident in-vessel and the probability of preventing core-concrete interactions subsequent to vessel meltthrough were not treated as being uncertain. It was a particular goal of this study to provide perspective on the uncertainties that had the greatest effect on risk. The LLH method was selected as the primary approach to characterizing the uncertainties. It should be emphasized that the objective was to obtain " reasonable" ranges of uncertainty in the results, not rigorous statistical bounds. A number of those providing input to the LLH analysis were adamant in their desire that no probabilistic interpretation be attached to the LLH risk outcomes, but rather that they serve only as a range of uncertainty. As discussed in the following section,it must be recognized that because of the l limitations in the LLH approach, some of the observations noted here could change when this plant is analyzed in more detail, and other issues could play a more important role in the characterization of uncenainty. l The Peach Bottom analysis was completed with a more advanced methodology for source-term estimation, embodied in the RELTRAC code, than were the other SARRP ! analyses. The analysts are confident that this treatment provides a more thorough treatment of the source term uncertainties and a more consistent interface between the source term and containment analyses. However, since there is no direct comparison to the methodology employed for the other SARRP plants, the actual impact of the methodology change is difficult ) to establish, other than in analysts confidence. It is felt that the true value of the work which is ) summarized here lies not only in the results obtained but also in what can be learned from the I i model for risk which has been developed. There are some surprises in the results. For j instance, the containment failure location, the containment pressure rise at vessel breach, and ] the probability of containment venting wem not shown te be important sources of uncertainty. However, there are clearly some instances where these inputs to the model are significant. The current results are thus indicating that those instances are not significant to risk within the ranges assigned to these issues and what has has been assumed relative to other uncertain issues. Much more can be learned from the model than has been obtained from this analysis j 5-67 ) 1 , t. '3 t i ( NUFEG/CR-4551, VOt. 3f DRAFT REPORT FOR COMMENT (FEBRUARY,1987) with a single limited L tit $ hypercube sample. Certainly there are regimes of behavior in which the uncertainties listed above (and perhaps others) are inportar,t but have not been investigated. l 5.6 LIMITATIONS l r ' 1 The limitations associated with probabilistic risk assessment have been discussed in many forums, a53 a reiteratiotiof the entite subject is not needed here. There are some specific limitations that should be recognized ~when interpreting the results of this study. The individual i discussions of the principal tasks in the main report and the appendices include a description of the known limitations. Some of theincee important of these include the following: t - The kope of analysis was limited due to' schedule constraints, and this ! resulted in^some potentially important limitations for this study: In L, particular, the drywell liner meltthrough was the most significant failure o mode but it was treated rather rimply in the model. The actual releases associated with this failure could be smaller than indicated by the current results. The results apply to the Peach Bottom plant only and must not be inferred as being generally applicable to other plants. The accident sequences reflect the plant systems and procedures at Peach Bottom. The issues found to be important to uncertainty here are based on specific elements of the plant response. The results and conclusions are based on internal initiating events only, since extemal events were not included. Extemal events are scheduled to be included at a later date. The study was also limited in that not all areas of uncertainty are considered in the results. The uncertainties due to data are not systematically included in all of the results. Uncertainties in consequence calculations were not addressed, except that some indication of modeling uncertainty is available from the CRAC2 and MACCS comparisons. + The consideration of operator actions was limited by the groundrules to actions associated with written precedures. In particular, neither errors of commission nor innovative actions the operator could take to cope with the accident are considered. The techniques for human reliability assessment suffer from the same limitations that affect all risk assessments. The appropriateness and interpretation of the LLli results remains the subject of considerable discussion among analysts familiar with the process, and even among most of those who participated (see Appendix E). The la* gest area of concent i: relative to interpretation of the results; a number of l participants provided input with the understanding that it be used to define ! ranges, but that no probabilistic interpretations should apply. Therefore, although the mean values of the LLH samples are discussed in this report, extreme caution must be exercised in their use, smce they do not correspond to the mean values of probability distributions. The rank-regression study of the results is known to be less than ideal. With the methodology used here, the rank-regression analyses do not l l 5-68 i NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) provide many insights if one or two issues dominate the uncertainty. The results were calculated with removal of the effects of the sequence issues (which wem very significant) to help identify the importance ranking of the remaining issues. Another limitation of all of the tests used to analyze the results is that they i do not recognize the importance of combinations of issues that could be very significant to the overall uncenainty. Plant modifications, changes in the dominant accident sequences, or reanalysis with increased scope could possibly change the results such that other issues or combinations ofissues not identified hem could be important. . The ASEP study of core-damage frequency was not intended to be as comprehensive as some risk assessments, and it was completed with less resources than a full-scope PRA. Although there was interaction with the plam staff, there was no direct utility involvement. Some of the containment analysis relied on other studies or simple calculations rather than direct calculations for Peach Bottom. For example, pressures associated with some phenomena were scaled from other studies or from input of reviewers based on crude calculations. The containment event tree, while very detailed, still includes a great many simplifications, and there are numerous branch points for which the split fractions are highly uncertain. The source-term analysis is based on the STCP for the central estimates. The STCP has limitations, including lack of treatment of some phenomena such as revaporization after vessel failure, Csl decomposition, in-vessel flow recirculation, buoyancy-driven containment flow, direct heating and late iodine release from the pool. The treatment of these issues is discussed in the detailed sections conceming the SARRP source-term analysis. 4 The RELTRAC code used to calculate the source terms offers a substantial I improvement over the methods used in the other SARRP analyses, although j the impact of limitations in RELTRAC are not yet fully known. < + The LLH results are limited to the views of the participants in the process, and are further limited by the number ofissues considered. The results are also applicable only to a single sample oflimited size that does not allow for firm conclusions regarding most mdividual contributors. Therefore, the i uncertainty representation that results is not a comprehensive, statistical one, and use of these results must be properly qualifiedL While many of these limitations appear to be severe in the abstract, most are not too different from the limitations associated with any risk assessment. The methods selected were based on the input of many in the reactor safety community. Some reflect the bounds of the state of the art, while others reflect restrictions in schedule or budget. The limitations in the uncertainty analysis reflect the current state of PRA, in that there are no generally accepted methods for fully characte.rizing many sources of uncertainty. The value of the study is manifested in the engineering and scientific insights which are derived through the process and examination of the results. 1 5-69 l \ l r l NUREG/CR4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) ! l i References for Section 5
- 1. Kolaczkowski, A. M., et al. Analysis of Core Damage Frequencyfrom Internal Events:
Peach Bottom Unit 2. U.S. Nuclear Regulatory Commission Report NUREG/CR-4550, Volume 4, Sandia National Laboratories, Albuquerque, NM: February,1987. ' 2. Reactor Safety Study--An Assessment of Accident Risks in U.S. Commercial Nuclear Power Plants. U.S. Nuclear Regulatory Commission Report WASH-1400 (NUREG-l 75/014), Washington, DC: 1975. l 3. Amos, C. N. et al. Containment Event Analysisfor Pos:ulated Severe Accidents: Peach Bottom, Atomic Power Station U.S. Nuclear Regulatory Commission Report l NUREG/CR-4700, Volume 3 (Draft), Sandia National Laboratories, Albuquerque, NM: February,1987. l I 4. IDCOR Task 23.1: Integrated Containment Analysis. Indust 1y Degraded Core Program, Technology for Energy Corporation, Knoxville, TN: 1984.
- 5. Amos, C. N., et al. Evaluation of Severe Accident Risks and the Potentialfor Risk Reduction Grand Gulf, Unit 1 U.S. Nuclear Regulatory Commission Report NUREG/CR-4551, Volume 3 (Draft Report for Comment), Sandia National Laboratories, Albuquerque,NM: February,1987.
- 6. Benjamin, A. S., et al. Evaluation ofSevere Accident Risks and the Potentialfor Risk Reduction: Surry Power Station Unit 1 U.S. Nuclear Regulatory Commission Report NUREG/CR-4551, Volume 1 (Draft Report for Comment), Sandia National Laboratories, Albuquerque, NM: February,1987.
- 7. Benjamin, A. S., et al. Evaluation of Severe Accident Risks and the Potentialfor Risk Reduction: Sequoyah Unit 1 U.S. Nuclear Regulatory Commission Report NUREG/CR-4551, Volume 2 (Draft Report for Comment), Sandia National ,
Laboratories, Albuquerque,NM: February,1987. i 5-70 1 NUREG/CR-455t, VOL. 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) 3 Section 6 RESULTS OF RISK-REDUCTION ANALYSES In addition to updating the estimate of risk for the Peach Bottom plant, some areas in which reductbns in risk might be achieved were also investigated. As outlined in Section 3, this was accomplished by identifying a number of modifications to plant procedures and/or systems aimed at particular aspects that were assessed to contribute most to risk. The risk measures were then recalculated to identify the effectiveness of each of the modifications. In order to provide perspective on the relative merit of the modifications, an estimate of the costs { i and other impacts of each option was also developed, It is important to note that within the time and resources available for this assessment, it was not possible to fully address the possible ways in which these modifications might actually increase the risk of sequences other than those that they were primarily intended to address. For the safety options aimed at reducing the frequency of core damage, it was assumed that the modifications would be carefully evaluated and designed to minimize the potential for inadvertent operation that could increase the likelihood of demands on other systems. Where such potential was identified in performing these analyses, it was noted in assessing the impacts. For the mitigative options, the results of the containment event tree analysis were 3 examined, in terms of the conditional probability of accident progression bins, to identify the l impacts of the options. This review was limited, of course, to the plant-damage states ( identified by ASEP as leading to severe core damage. No assessment of the impact of the safety options, either positive or negative, was made on the accidents not currently in the ASEP profile. Because of uncertainties regarding many of the containment phenomena, there was a distinct possibility that, for some options, the likelihood of containment release modes implying higher consequences could be increased, resulting in a net increase in the estimated risk. These values are reflected in the results summarized in this section. l This section provides a synopsis of the results of the risk-reduction calculations. Additional detail may be found in Appendix D. Finally, it must be emphasized that any conclusions concerning the risk-mduction analyses presented in this section must be considered in the light of the limitations of the current study, many of which were summarized in Section 5.6. These risk-reduction analyses are useful for engineering insights and comparison of relative value, but any decisions concerning a need to implement changes must be supported by more detailed study. . 6-1 I NUREG/CR4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 6.1 EFFECTS OF PREVENTIVE OPTIONS ! l The identification of options intended to reduce the frequencies of sequences j contributing most to con:-damage frequency and to risk is described in Section 3. The options considered which reduce the core-damage frequency were: PI: A reduction in the probability of a common-mode de power failure by improving the procedures to limit maintenance and test activities that could affect both trains of power. P2: The use of a fifth, low-capacity diesel generator, which is already on-site, for providing power to a control rod drive pump as a means of injecting water to the core. P3: The use of the diesel-driven fire pump, already available at the plant, to provide an additional method of low-pressure injection of water for the core. Depressurization of the reactor vessel would be maintained or accomplished by manually charging one or more SRV actuators with nitmgen from bottles. ; i The fint option is intended to eliminate the sequence assessed as being dominant in the ASEP analysis, TBUX--a loss of all electrical power due to a loss of offsite ac power and onsite de power (which precludes operation of the emergency diesel generators). The suggested modification is procedural only, thus offering a relatively low-cost method of addressing the dominant accident. It should be noted however, that there is substantial uncertainty in the methodology for assessing the probability of such a common-mode error and any conclusions concerning either the base case study or the proposed modification for this scenario must consider the uncertainty in the results presented here. The next two options are also intended to address station-blackout scenarios. Option P2 provides another mechanism for injecting water into the core when all power is lost. While this extra diesel could not make up for the loss of one of the four main diesels, it could be used i to power a control rod drive pump to provide for water injection into the core. However, because a single control rod drive pump has a relatively low flow rate, this option is only useful for long-term accidents in which the flow required to cool the core is reduced because of previous cooling and the reduction in decay heating. As described in Section 3, this option would also entail some changes to the venting procedure to establish a cooling path for the rejection of decay heat. I Option P3 addresses the power loss accidents by establishing procedures for the use of l the diesel-driven fire pump in a station blackout accident. The fire pump could provide water ! from the fire protection system to the core through a special piping arrangement, although the output would be limited in pressure (125 psi). Because the core-damage frequency was 6-2 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) dominated by high pressure scenarios, this option was always coupled with a modification of the ADS such that it could operate in the absence of de power. This allows the reduction of the vessel pressum to the point where the fire pump could have a successful impact. , The effect of these options on the accident-sequence frequencies is illustrated in Figure 6-1. As illustrated, the options tend to have a relatively small impact on the overall core-damage frequency. There are enough limitat ions associated with the options proposed that they do not apply to all of the important scenarios. Other options were initially considered, but most that could be effective on a large enough percentage of the accidents to result in substantial reduction in core-damage frequency would also be prohibitively expensive (they would involve new systems.) Because the core-damage frequency assessed in the ASEP study of Peach Bottom was fairly low (in comparison to the results of other PRAs), it was recognized that the addition of entirely new systems would not likely prove cost effective. 10 ' s y j Base P1 P2 P3 tu10 y } , LLH 95 o Y $ 10% . l . S ;f 10b i 2 - \ = E j
- Oentral Estimate KLLHS 1
10 l l l l Figure 6-1. Illustration of the Impact of Preventive Options on Core-Damage Frequency I 1 6.2 EFFECTS OF MITIGATIVE OPTIONS ON CONTAINMENT RESPONSE After reviewing the phenomena contributing most to the conditional likelihood of the i more severe containment release modes, a number of modifications were proposed that could possibly reduce the consequences of core-damage accidents. The options that were explicitly considered in this analysis were the following: M1: Modification of the automatic depressurization system (ADS) to allow its operation in the absence of de power; 1 l 6-3 ! 1 _ - _ _ _ _ _ _ _ _ . i NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) M2: Addition of a number of featuret--a drywell curb, option M1, and improved venting--to address key accident phenomena. i M3: Addition of a drywell curb to delay or prevent drywell meltthrough, , one of the key phenomena and uncertainties in the study; j M4: Installation of another completely independent train of drywell spray [the results for this option were not available for this draft, will be ; included when the new runs are available]; M5: Improved hardware and procedures for wetwell venting station blackout scenarios; and M6: Installation of area-wide fire sprays in the reactor building. The first option directly addresses the dominant station-blackout scenarios by providing . a means to depressurize the reactor vessel. This option can have a mitigative effect by allowing I depressurization to the suppression pool either before or during core damage, thus reducing the severity of containment loadings associated with vessel breach. This option is also part of the P3 safety option discussed above. The combination of P3 and this option better characterizes the maximum benefit to be derived from modifications of this type. The second option includes all of the first option and other features designed to pmvent risk-significant phenomena. The option includes a drywell curb which is intended to reduce to probability of direct contact of molten core debris on the drywell. It is assumed that depressurization of the vessel will eliminate this containment failure mode. This option is combined with the first option because high pressure ejection of com debris could negate the effects of a curb. Finally,if the probability of drywell meltthrough is reduced, containment venting becomes relatively more effective in reducing radionuclides releases. The fm' al part of option 2 therefore improves the probability of successful venting by changing some valve power supplies and procedures to allow venting in station-blackout scenarios. The third option provides for the drywell curb alone to measure the impact of that option with no other changes, while the fifth option addresses the improved venting aspect by itself. The founh option includes the addition of hardware to improve the effectiveness of drywell sprays. As described previously, the core-damage profile is dominated by accidents involving power loss, thus precluding drywell spray operation. This option is designed to eliminate dependency on plant ac and de power as well as to provide an independent source of water to allow drywell spray operation in a bmader spectrum of accidents. The drywell sprays have three potential benefits: 1) they mduce pressure loads in the drywell,2) the sprays act as a very effective radionuclides scrubbing mechanism, and 3)if activated before vessel breach they may arrest com-concrete interactions. 6-4 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) The final option only addresses the radionuclides release aspect of the accidents. The reactor building provides a final scrubbing mechanism for any release paths that open to the reactor building rather than the environment. Because analysis of other plants had pointed out the potential for large scrubbing due to the fire protection system (sprays), one of the options selected hem was the modification of the fire protection system at Peach Bottom to ensure more ; complete coverage of the reactor building (the plant currently has spray headers, but they are limited to a curtain between units). 6.3 EFFECTS OF SAFETY OPTIONS ON RISK The risk was recalculated for each of the safety options by modifying the appropriate portions of the analysis to reflect the reductions in core-damage frequency or changes in containment response. The reduced risk oflatent cancer and early fatalities for each option is illustrated in Figures 6-2 through 6-5. The figures illustrate the range of the LLH which was selected as the 5th to 95th percentiles of the sample member outcomes. These bounds should i not be interpreted as the bounds of uncenainty on risk, since as has been noted several times in this report, the LLH does not entail a statistical weighting that allows the creation of a distribution of risk uncertainty. As stated in Section 5, the risk results in this report am known to include inaccurate treatment of some phenomena; in particular, the effects of the drywell sprays were underestimated and the containment response results for some of the ATWS scenarios were inaccurate. For the risk-reduction options, the former error could result in the l I overstatement of benefits for some of the proposed modifications, including the improved vent j and the area-wide fire sprays, but the basic trends seen in the analysis here would be expected to hold. These errors will be corrected in the final version of this report. As shown in the figures, most options due not have a major effect on the risk measures. Option P3 combined with M1 does appear to result in a downward shift in risk for both risk measures, but the j impact is not major. Option M6, the area-wide fire sprays in the reactor building, tends to have a noticeable impact also, primarily on the latent cancer risk measure. Option P1 also reduces the risk because of a reduction in core-d. mage frequency, but that effect is also quite limited (the top of the uncertainty is lowered by a factor of two). Overall, it can be generally l concluded that the risk-reduction options selected for this study had limited impact, particularly 1 when uncertainties and the limitations of the analysis are considered. Them is a potential that some of the mitigation options could increase risk. This facet of the results, which is not clearly seen in the displays in Figures 6-1 and 6-2, is explored further in Section 6.5. While all of the options (except one) result in a reduction in risk for 6-5 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 0 10 , i j Bace P1 P2 P3 & M1 M1. M2 M3. M4* MS M6 LLH 95 10' r - - g -
- a. -
$ 30 2,'. e . , g ! . e x .i
- d. 10 1 -
g - LLHS - = 4 W 10 1- e CentralEstimate 5 - . .Not compieted: To be . Included in the Final Report l Figure 6-2. Comparison of Safety Option Risk of Latent Cancer Fatalities: MACCS Calculation 10 f h P1 P2 P3 & M1 M1 M2 M3 M4' MS M6 > -1 . Base LLH 9gh
- a. -
g _ g - - _ ~ ~ - E # 10' "! * .
- e e
- 4
- u. .
10' l' 6 ~ 10 1" - ' . th _ e CentralEstimate 'Not Completed: To be neluded in the Final Report -5 _ 0 Figum 6-3. Comparison of Safety Option Risk of Latent Cancer Fatalities: CRAC2 Calculation 6-6 '1 l i i NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l Base P1 P2 - P3 & M1 M1 M2 M3 M4' MS M6
- e
~ e e e
- c. 10 5] -
LLH 95th _ - - l 9 : m e e g .104] " 3 m - W . h 10' ]- ! g . o
- u. .
> 8 I d 10 l' - _ { d . LLH yh - l 10 l e CentralEstimate --- - j 'Not Completed: To be included in the Final Report 10 t 10 l Figure 6-4, Comparison of Safety Option Risk of Early Fatalities: MACCS Calculation 4 10 ' , Base P1 P2 P3 & M1 M1 M2 M3 M4* M5 M6 10" I"- e - '_ t. - c . 610' g ! z . $ 10' 1-E : H 8 - ""~ 610 1 - - l 3 _ d . 9 LLHS g10 ] e' Central Estimate *Not Completed: To be ,3 g included in the Final Report . 0 l Figure 6-5, Comparison of Safety Option Risk of Early Fatalities: CRAC2 Calculation 6-7 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) most of the sample members, all of the mitigation options have the show the potential to increase risk for the conditions associated with certain sample members. The implication of this is that before further consideration is given to these options, the uncertainty must be reduced such that there is high assurance that the proposed option does not, in fact, lead to risk increase. 6.4 COSTS OF RISK-REDUCTION OPTIONS The investigation of the relative merit of the proposed risk-reduction measures mquires a thorough evaluation of the impact to the plant owners that would result from their implementation. As discussed in Section 3, three types of costs were considered for each l safety option: (1) The total cost associated with installation of the proposed modification; (2) The recurring costs to the utility (maintenance, increased plant downtime, etc.) that arise from the modification; and (3) The radiation exposure of personnel responsible for installation and maintenance of the option. The costs for the modifications considered for Peach Bottom are presented in detail in Appendix D. All of the estimates were derived from the data base provided in Appendix E. At this time, the costs for replacement power in the event that a shutdown of the plant would be necessary for implementing the modification were not included, under the assumption that all modifications could be accomplished within scheduled outages. This may well be an unrealistic assumption for a number of the modifications; however, it is not likely that the overall conclusions regarding the relative merit of the options would change, particularly in light of the substantial uncertainties from other parts of the analysis. Appendix D provides a perspective on the potential impact of replacement-power costs. The results of the cost evaluation are listed in Table 6-1. The values in the table represent the present value of total cost for each option, including all maintenance and operating costs over the life of the plant. Some of the cost ranges are quite large because,in addition to , all the uncertainties in the estimation of a cost for a designed option, some of the options include a range of costs that reflects the uncertainty in what will be required to gain the defined improvement. Thus the improvement of the drywell spray operability in a station blackout could involve some relatively minor modifications or it could involve the installation of a new spray system. 6-8 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) . Table 6-1 ESTIMATED COSTS FOR RISK-REDUCTION MEASURES Total Cost (Present Value) Option Low Central High P1 De common-mode reduction $140,000 $280,000 $740,000 - P2 Use of fifth diesel generator $240,000 $720,000 $4,000,000 P3 ' Fire system use as injection source and . & M1 de indanandmt ADS - $4,300,000 .$6,500,000 $17,600,000 M1 De i:vianandant ADS $3,100,000 $4,800,000 $14,240,000 M2 , Drywell curb, de independent ADS and improved venting capability . $8,300,000 $13,000,000 $35,000,000 - M3 Drywellcuit . $5,000,000 $8,000,000 $20 000,000 M4. Improved drywell sprays $240,000 $720,000 $10,000,000 MS Improved venting capability $190,000 $560,000 $6,100,000 M6 Reactor building globalfire sprays $4,900,000 $9,400,000 $21,200,000 l l 6.5 COMPARISON OF COSTS AND BENEFITS In order to provide a direct comparison.for the averted risk and the costs associated with each of the modifications, it was necessary to develop a measure of benefit in terms of a dollar value. Because a number of different approaches to the characterization of the benefits associated with reduced risk have been used in the past, the benefits for the safety options were - calculated in four ways: (1) Averted offsite costs, consisting of the differences in costs of property - damage (as calculated by MACCS or CRAC2) plus an assigned value for reduction in health effects ($1 million for each averted early fatality and $100 thousand per averted latent-cancer fatality and early injury); (2) Averted offsite costs calculated as $1000 per person-rem averted within a 50 mile radius of the plant; (3) Averted offsite costs, as in (1), plus averted onsite costs (calculated as described in Section 3); and (4) Averted offsite costs at $1000 per person-rem plus averted onsite costs. These measures of benefits, as calculated from the ranges of LLH outcomes, are presented in Figures 6-6 (for MACCS) and 6-7 (for CRAC2). The range of costs for each option from Table 6-1 is also shown to provide a comparison between the costs and benefits. In developing the ranges for the nessures of averted risk, the risk estimate for each LLH 6-9 _ - L__ _ _ NUREG/CR-4551. VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) 1 l ?
- A F,educ
,,_,ed PM,,,,,, ,,, ,o,f .__ i / . _-_ .__ < 1 o--. v ;i% W//////A hout .2 O Aos oc + o / M2 , DryweB Cue, ADS Withoul4- . , , oc. a imoroved vent 1 O cum *~ - - -- - o M4 Not Corroemd;To be improved Drywed trukond in the Final Report . Spreys * + improved Vert _ _ _ _ _ _ i Re.acio udding ep,.,s / n 1 a e i a sa a a i a a a e a e n ia a a n i 10 10 1[ 10 10 10 COSTS (DOLLARS) Y.aL (1) Avened offsas costs Cost of the safety option i (2) Avened of'sso costs @ $1000$erson rem www (3) Sum of (1) and onsas costs 4. ,,,,,gnifcant negative tenettt, and the fuu averwd nsk range le 1 i (4) Sum of (2) and onsne costs not shown en tNs plot. The b o CentralEstimme(Not Avalable i in. .arsinrepresents oe ih. usi ,esui the 6th ,o, optons) Figure 6-6. Cost-Benefit Compadson for the Safety Options as Calculated by MACCS 6-10 NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) I i l P1 , ' ' ' ' 'f 1 Reduced Probabitty of ie Common-Mode Battery Falure ~"<// ] ] 7.. .,j-----..y DneeWnermor P3& M1 ; Wra" - 1 ll j E I i ADS WR DC + m...,.ma~,4....... m. M2 Drywot Curb, ADS Withoul+ DC, & improved Vent ' " "^ ' " ' " ' ' " ' ' * ' " " ' " ' " ' " * " " ) ] M3 4 0 /g - - C. - j O ! smvoNorvwa Sorays 5%L'*L V//////////// j j l ,m-.ed vem - + __ W/#/HH/ Re,,,o acto Sweding+ S,a,s ] ti 10 10 id 10 10 10 COSTS (DOU.ARS) Eat m (1) Averted offatto costs x Cost of the salary option i (2) Averted offsite costs @ $1000@erson rem N (3) Sum of (1) and onsas cmsis + for s6pnamara negative beneta, I i (4) Sum of (2) and onste costs and the full averted nsk range is not shown on this plot The bar represents the 5th to the B5th of the LLH resus o CentralEstwnste(Not Availabw for AlOpoons) 1 1 Figure 6-7. Cost-Benefit Comparison for the Safety Options as Calculated by CRAC2 6-11 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) sample member for an option was subtracted from the estimate for the same sample members in the base case. The differences were then re-ortiered, and the 5th and 95th of the 150 sample members serve as the ranges depicted in the figures. It is therefore possible, and in fact the case, that for some sample members the options implied an increase in risk rather than a decrease. The options that would be most cost-effective are generally those that indicate the largest overlap between the costs and the averted risk. As indicated in Figures 6-6 and 7, only the first prevention option, the improved msistance to a de power common mode, has any real overlap between the costs and benefits. Even for this option however, the ratio of costs and ! benefits is only favorable for the upper regime of the LLH uncertainty. Another featum of the cost-benefit comparisons is that some of the mitigation options showed some negative benefit for some sample members (the area-wide fire sprays only showed very slight negative benefit and only for the CRAC2 result). For these cases, the cost / benefit comparison is best illustrated by a linear display, as presented in Figure 6-8. (Only MACCS results am provided, the CRAC2 results are very similar, except the global fire sprays l show a very small negative range.) This display suggests that the uncertainty in the containment response is large enough that conclusions regarding the impact of the options are difficult to establish. Some of the options have as big a potential for negative benefit (inemased risk) as they do positive. The reasons for the negative outcomes in this display are not always immediately obvious. Investigation of the physical masons giving rise to these results is time consuming since it involves the examination of the individual sample members, and this investigation has not been completed for Peach Bottom at the time of this report. The analysts do have some hypotheses as to phenomena resulting in increased risk. For options that include j provisions to actuate containment venting in the event of a loss of ac power, there are several ; factors which can lead to negative benefit: The venting path, which is initiated from the wetwell airspace, is assumed to bypass the reactor building. Hence, the potential benefit of i deposition in the reactor building is not realized. In some of the LLH sample members, the reactor building decontamination factor is larger than the suppression pool DF. Therefore the releases to the environment would have been lower if venting had not been initiated. Many of the accident progressions involving containment venting also involve recovery of ac power at some later time. In some of these { cases, recovery of power would have enabled resumption of l containment heat removal possibly averting containment failure. For j this type of scenario, containment venting could lead to release of radionuclides that would have otherwise been retained in containment. 6-12 1 I l NUREG/CR-4551, VOL. 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) l l M1 - d. ADS Without DC 4:::in """ " 838 M2 ~ * ~ ~ - - d I Drywell Curb ADS Without swim saw l DC. & Improved Vent ! IM ) M3 I" d ~ l Drywell Curb d'"" 6" ~ improved Drywell ~~ inmus.8 m e.F** R.p.t V((///////// E FA+ spreys j mE tmprov Vent en (((((((( (( M6 , Reactor Building Fire Spreys -+set Y..sw I I f 1 I .$1M -$ 8M 4 GM -84M 4 2M $0.0 8.2M S4M S6M S.0M $1.0M $12M $1.4M $1.6M COSTS s. , $ C.~. io.e b **"" *""""'" _ . , . _ _ l l Y orine. M = Figure 6-8. Cost-Benefit Comparison for the Mitigative Options, Linear Scale , I Containment venting allows some releases to occur earlier in time than , might have been the case if venting had not occurred. This has an j impact on the warning time for emergency response, the potential for . radionuclides decay and the effectiveness of natural deposition processes. Certain containment failure modes are not appreciably affected by. I venting and can occur whether venting is successful or not. These include meltthrough of the drywell shell, overpressurization due to rapid i pressurization in the drywell (e.g., due to direct heating), and drywell failure due to high temperatures. The possible negative benefits of an ac-independent containment vent for station-blackout accidents should not be construed as negating the benefits of the current venting procedures. The current procedures are principally responsible for lowering the probability of cme damage from accident sequences involving loss oflong-term heat removal (e.g. the TW sequence). l 6-13 1 1 l j l .1 i 5 9 l .. 1 1 I NUREG/CR-4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) Section 7 INSIGHTS AND CONCLUSIONS The results of the SARRP study of risk and risk-reduction for Peach Bottom have been - outlined in the previous two sections, along with a description of the insights derived from , those results. Several of the more important insights are summarized in this section relative to j the objectives of the program: (1) to reassess the risk of the Peach Bottom facility with current technology and to understand the changes in the perception of risk, (2) to provide insights as to issues most important to the uncertainty in the estimates of risk, and (3) to examine the i potential for reduction in risk in a cost / benefit framework. These objectives have been satisfied in the program with the generation of a significant quantity of information regarding reactor safety with a special emphasis on the uncertainties in current understanding. Many of the . insights that were developed are applicable only to certain areas of risk-assessment technology j and are too detailed for inclusion in this summary, but such insights have been included in the appropriate sections of the main report and appendices. 7.1 INSIGHTS AND CONCLUSIONS FROM THE RISK REBASELINING Several of the more important insights and results of the Peach Bottom risk rebaselining are listed briefly below. These are given first with respect to the core-damage frequency, then, . in order, for the containment, source term, consequence, and finally the risk analyses. The core-damage frequency for Peach Bottom as calculated by ASEP has a ] mean value of 8.2 x 10-6per year. Uncertainties associated with reliability j data could cause the core-damage frequency to be a factor of three higher or six lower. In addition, the uncertainty associated with models or assumptions in the systems analyses could lead to higher or lower total core-damage frequency by about a factor of five higher or ten lower. Recent information from the utility leads to the expectation that one of the ASEP sensitivity studies more accurately represents the plant. In that study the mean core-damage frequency was assessed to be 6.9 x'10-6 per year. The mean core-damage frequency is about a factor of four lower than the previous assessment of Peach Bottom performed as part of the RSS. The analyses point to substantially different aspects of the plant as important contributors. The reasons for the change in assessed frequency and relative ; contributors include the following: (1) plant modifications, including a number of changes'directly in response to the results of the RSS which helped to remove some previously important contributors; (2) changes in understanding of plant response, for example, the new' understanding concerning accidents involving containment failure prior to core damage and (3) additional evidence from industry experience that has been incorporated 1 l 7-1; NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) into PRA data and methods, for example, in the identification of new failum modes. The principal contributor to core damage was found to be station blackout. This is in substantial agreement with the findings of recent industry assessments (e.g., GESSAR PRA). The large diversity of water injection methods in a BWR is the principal reason why the station blackout accidents dominate The evaluation of containment response in this study indicates that the probability of early containment failure, excluding intentional venting, (i.e., l the probability of failure occurring before or at the time of vessel breach) , given core-damage ranges from approximately 0.02 to 0.8. More than one-half of the sample members had probabilities greater than 0.7 for early containment failure. One specific failure mode dominated the containment response: meltthrough of the drywell caused by direct contact of molten core debris. This failure mode was assigned to the early failure category as a simplification of the methodology. In reality this failure would occur somewhat later in time, depending on the specific accident progression. Without inclusion of the drywell meltthrough failure mode, the probability of early containment failuru ranges from approximately 0.01 to 0.4. The bulk of the sample members have an early containment failure probability less than 0.3 for this case. The probability of early intentional venting ranged from approximately 0.01 to 0.4 The probability of no containment failure or vent was estimated to range from 0.08 to 0.2 for this LLH sample. The containment event tree considered termination of the accident after core damage had been initiated. About 20% of the core-damage frequency was estimated to follow an accident progression resulting in core damage but no vessel failure. The source term results for the central estimate are based on the Source Term Code Package (STCP). The LLH calculation was based on the input of an expert review group concernmg specific aspects of the release (uncertainty issues). The reviewer input was used to alter the STCP releases to mflect both uncertainties and to supplement the STCP output where the code is currently lacking,in the eyes of the reviewers. For this study, the uncertainty issues were treated through manipulation of a parametric code, RELTRAC, designed to allow consistent and fairly detailed consideration of specific issues that would affect source term methodology. For the other SARRP plants the LLH source terms were developed in a much simpler manner. From a comparison of central estimate and LLH results, one can make the general observation that the STCP was assessed by the experts to be somewhat conservative in its predictions of the releases for the specific core-damage profile examined here. This was specifically due to the input of the reviewers that lowered the predicted lanthanum releases associated , with core-concrete interactions: generally an order of magnitude lower than i the STCP predictions. For this facility, the core-concrete interaction 7-2 i NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMEM (FEBRUARY,1987) releases were very important to risk. In addition, the reviewers generally attributed a larger aerosol retention capability to the reactor building than i was calculated by the STCP. The results reported hemin have been calculated with both the CRAC2 and MACCS consequence codes. There were some differences between the MACCS results and those calculated using CRAC2. In general, the MACCS code predicts higher consequences for latent effects and population dose by factors of two to eight, with an even larger diffemnces (factor of ten or more) for offsite costs. The consequrmce uncertainties were not evaluated in this study, except for the insihhts from the CRAC2 and MACCS comparison provided in Section 5. Risk was mealculated in SARRP with methods that reflected improvements in each area of analysis. The resulting SARRP central estimate of public health risks were found to be somewhat lower than the values cited in the RSS for Peach Bottom. The LLH representation of risk suggested that the SARRP central estimates (based on the STCP) and the RSS results am near the upperlimit of uncertainty. As described above, the STCP predicted higher releases oflanthanum than those predicted as a result of the LLH process. This difference lead to a difference in risk for early effects, with the SARRP central estimate being above the LLH range. Studies are currently undenvay to fully examine the differences and to verify the magnitude of these risk differences after consideration of the uncertainties and simplifications of the SARRP methodology. The uncertainty was represented in this study by the LLH method, described in Section 2. This method attempts to characterize the uncertainty associated with modeling of phenomena that affect systems, containment and source term analyses. (Uncertainties in consequence modeling wem not included due to time and budget constraints.) The LLH method was based on the input of a group of experts. The SARRP analysts and experts selected a limited number of the uncertainties thought to be most significant and provided their views on the uncertainty in each. It is this treatment of uncertainty which allows additionalinsight through the analyses performed in SARRP compared to previous risk assessments. The results of the LLH study indicate about two orders of magnitude uncertainty for the risk of latent cancer fatalities and three orders of magnitude for early fatality risk. The uncertainty analyses indicated a number of individual issues were significant to the uncertainty in risk: The sequence-frequency uncertainty of the probability of a common-mode de power failure was found to be significant, particularly to latent effects. This uncertainty had a large range and it directly affected the probability of the most frequent plant-damage state. The uncertainty in the recovery of offsite power was also found to be significant, because of the importance of station blackout I scenarios in the core damage frequency assessment. In addition 'to its frequency impact, this issue also had direct impact in the accident progression in containment, particularly in the ability to 1 7-3 NUREGCR 4551, VOL 3: DRAIT REPORT FOR COMMENT (FEBRUARY,1987) vent containment through the suppression pool, to supply water to the core or to use the spray systems. The probability of drywell failure resulting from meltthrough of the drywell shell by the molten code debris was found to be a significant uncertainty. It was assumed in one-half of the sample members that drywell meltthrough would occur with a probability of 1.0, while the rest of the sample members precluded this failure mode. -- The magnitude of the CCI release was also an important uncertainty. The SARRP reviewers assigned weighting factors that caused the central estimate to differ substantially from the LLH in this area. -- The uncertainty in the decontamination effectiveness of the reactor building as releases pass through it was significant to all risk measures, particularly early effects. For many types of releases this is the last retention mechanism and its effective decontamination factor is very important. . The Peach Bottom analysis included a more sophisticated method of analysis of the LLH results that removed some of the weaknesses of the rank regression analyses performed previously within the RISQUE code. These mort detailed regression analyses lead to three observations: The uncertainty in the LLH is due to a fairly large (approximately
- 10) number of issues, none of which have an overwhelming effect on the total uncertainty in risk. Specifically, no issue uniquely accounts for more than about 10% of the total variance. Thus in l order to significantly reduce the uncertainty in risk it is probably l necessary to remove the uncertainty in more than one issue.
The identifiable uncertainty due to unique issues is less than one-half of the total uncertainty--the balance of the uncertainty is probably due to combinations of issues. The statistics of the current sample do not allow for definition of all specific , combinations which may be very significant to risk. In some cases, only one or two levels of an issue are important. Improved understanding of an important level of an issue could reduce the total significance of an issue. The methodology developed for this study represents a significant advancement in PRA technology. A fully integrated treatment of uncertainties has been developed. The entire process has also been automated to allow efficient development of risk estimates for sensitivity studies and other analyses of the plant. The current calculation of risk for Peach Bottom is subject to substantial uncertainties in l most analysis areas. There have been improvements in understanding of possible accidents and their consequences, some of which point to lower risk than previously estimated. In other areas, further analysis has raised concerns which cause the rather wide ranges of uncertainties in the risk results. Finally, the results presented here are plant-specific, and changes in the 7-4 l l NUREGCR-4551, VOI. 3: DRATT EPORT1DR COMMENT (FEBRUARY,1987) j accident sequence profile, the specifics of the containment safety systems designs or in the j plant-specific procedures could change these results substantially. 7.2 RISK REDUCTION INSIGHTS AND CONCLUSIONS The overall conclusion fmm the risk-reduction analysis is that it is not likely that a cost effective change could be suggested based on this study. With regards to modifications intended to reduce the core-damage frequency, there were few options identified that could have a major impact and still be within a range of reasonable costs, because of the relatively low core-damage frequency. While one prevention option, which involved changes of a procedural nature to limit the probability of a common-mode de power failure, did show cost effectiveness for some part of the uncertainty range, the result was not conclusive. The mitigation options were intended to address specific aspect of the containment and , source-term methodology known to be important to risk. While these options did have a ) desired effect on specific phenomenon,it also indicated the importance of the integrated view ) ) of the effect of the issues--all of the issues had negative impact for some sample members. In several cases the negative impact of proposed modifications is not intuitive: investigation as to the specific cause of negative impacts is continuing. The overall conclusion however, is that the containment response is a complex process with many subtle interactions that cannot be predicted without detailed examination. Even with detailed analysis, the uncertainties in current l understanding make it difficult to establish a clear-cut benefit to many potential modifications. l 7.3 LIMITATIONS There are severallimitations noted in each discussion of the task activities associated with SARRP. It should first be recognized that this study is subject to the same uncertainties as any risk assessment conceming the potential for inaccuracies or incompleteness. The results and conclusions are also Peach Bottom, and no extensions of the insights reported herein to other facilities is appropriate without further analysis. External initiating events are not considered, and the consideration of operator action is limited, primarily to those actions called for in procedures. Thus, both innovative actions outside procedures and errors of commission are not generally included. The budget and schedule limitations led to some simplifications which are described in the appropriate task descriptions and summarized in Section 5.6. As described in Section 5, the uncertainty approaches used yielded a number of observations regarding the importance of individual issues, but the complex interrelationships could not be fully evaluated. Finally, after the results reported in this draft were obtained and reviewed, some errors in execution wem detected. These errors and their potential effects am discussed 7-5 NUREG/CR-4551, VOL 3: DRAFT REPORT FOR COMMENT (FEBRUARY,1987) in' Sections 5 and 6. The errors could change some of the risk results within the LLH uncertainty bands, but the uncertainty ranges and the conclusions of the study are not expected to change. All results will be corrected prior to final publication of this report. i 7-6 4 ] l DISTRIBUTION: l U. S. Government Printing Office I Receiving Branch (Attn: NRC Stock) l 8610 Cherry Lane Laurel, MD 20707 100 copies for AN 3141 S. A. Landenberger (5) 3151 W. L. Garner 6400 D. J. McCloskey 6410 N. R. Ortiz 6414 A. S. Benjamin (25) 6412 A. L. Camp 6412 F. T. Harper 6412 D. M. Kunsman 6512 D. M. Ericson, Jr. 6415 F. E. Haskin 6415 J. L. Sprung 6513 D. D. Carlson 6420 J. V. Walker 6422 D. A. Powers 6427 M. Berman 6440 D. A. Dahlgren 6442 W. A. von Riesemann i 6449 K. D. Bergeron 6449 D. C. Williams 8024 P. W. Dean l , DIST-1 l l Frank Abbey U.K. Atomic Energy Authority Wigshaw Lane, Culcheth Warrington, Chesire United Kingdom. WA3 4NE . Steven Aftergood Committee to Bridge the Gap 1637 Butler Avenue, #203 Los Angeles CA 90025 C. N. Amos j Technadyne, Inc. ) 300 Virginia SE Albuquerque, NM 87108 George Apostolakis University of California, Los Angeles Boelter Hall, Room 5532 Los Angeles, CA 90024 James W. Ashkar Boston Edison Company 800 Boylston Street Boston, MA 02199 Donald H. Ashton Bechtel Power Corporation 15740 Shady Grove Road Gaithersburg, MD 20877 Mark Averett Florida Power Corporation P.O. Box 14042 St. Petersburg, FL 33733 Michael B. Aycock TERA Corporation 7101 Wisconsin Avenue Bethesda, MD 20814 Raymond O. Bagley Northeast Utilities P.O. Box 270 Hartford, CT 06101 J. Bagues Consejo de Seguridad Nucleare Sorangela de la Cruz 3 Madrid 38046 Spain DIST-2 H.'Bairiot Belgonucleaire SEA Rue-de Champ de Mars 25-B-1050 Brussels -Belgium. Fred J. Balkovetz ;EG&G Idaho, Inc. P.O.. Box 1625 Idaho' Falls, ID- 83415 Patrick Baranowsky U.S Nuclear Regulatory Commission . Office of Nuclear Regulatory Research Washington, DC 20555 Robert A. Bari Brookhaven National Laboratories Building 130 Upton, NY 11973 Richard Barrett Baskin U.S. Nuclear Regulatory Commission Office of Nuclear Reactor Regulation Washington, DC 20555~ S. Kenneth S. California Edison Company P.O. Box 800 Rosemead, CA ( 91770 Werner Bastl l Gesellschaft Fur Reaktorsicherheit Forschungsgelande 8046 Garching Federal Republic of Germany Paul Baybutt Battelle Columbus Division j 505 King Avenue 'l Columbus OH 43201 Bruce B. Beckley l Public Service Company j P.O. Box 330 : Manchester, NH 03105 Myer Bender Querytech. Inc. 408 Cedar Bluff Road Knoxville,'TN 37923
- i DIST-3 j i
R. C. Bertucio Energy Incorporated 1851 So. Central Place, Suite 201 Kent, WA 98031 Peter Bieniarz Risk Management Association 2309 Dietz Farm Road, NW Albuquerque, NM 87107 Adolf Birkhofer , Gesellschaft Fur Reaktorsicherheit 1 Forschungsgelande 8046 Garching 1 Federal Republic of Germany James Blackburn Illinois Dept. of Nuclear Safety 1035 Outer Park Drive Springfield, IL 62704 Dennis C. Bley Pickard, Lowe & Garrick, Inc. 2260 University Drive Newport Beach, CA 92660 Roger M. Blond Science Applications Int. Corp. 1710 Goodridge Drive McLean, VA 22102 A. John Bourne U.K. Atomic Energy Authority Wigshaw Lane, Culcheth Warrington, Chesire United Kingdom, WA3 4NE G. J. Boyd Safety & Reliability Optimization Services 9724 Kingston Pike Knoxville, TN 37922 K. J. Brinkmann Netherlands Energy Res. Fdtn. P.O. Box 1 1755ZG Petten NH Netherlands DIST-4 R. H. Buchholz. Nutech 6835 Via Del Oro ' San Jose, CA' 95119 {' Fred Buckman Delian Corporation 296 Duff Road Monroeville, PA 15146 " Robert J. Budnitz Future Resources Associates b 734 Alameda I ' Berkeley, CA 94707 A. R. Buhl International Technology Corp. 575 Oak Ridge Turnpike Oak Ridge TN 37830 Gary R. Burdick U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research - Washington, DC 20555 Kenneth S. Canady Duke Power Company 422 S. Church Street . Charlotte,'NC 28217 Lennart Carlsson Swedish-Nuclear Power Inspectorate Box 27106 Stockholm, Sweden 08-635560 Sweden Annick Carnino i Electricity de France ; 32 Rue de Monceau 8EME Paris, France F5008 l France ! N. G. Cathey q Idaho National Engineering Lab. ] P. O. Box 1625 ! Idaho Falls, ID 83415 j Ivan Catton 22436. Galilee Street Woodland Hills, CA 91364 DIST-5 mlui unssh s' f'uIu A u' sa g r ' S. Chakraborty Radiation Protection Section Div. De La Securite Des Inst. Nuc. 5303 Wurenlingen. Switzerland Switzerland Sen-I Chang Institute of Nuclear Energy Research P.O. Box 3 Lungtan, 325 Taiwan Robert F. Christie Tennessee Valley Authority 400 W. Summit Hill Avenue, W10D1 90 Knoxville, TN 37902 Frank Coffman U.S. Nuclear Regulatory Commission Office of Nuclear Reactor Regulation Washington, DC 20555 Larry L. Conradi Energy Incorporated 1851 South Central Place, Suite 201 Kent, WA 98031 Peter Cooper U.K. Atomic Energy Authority Wigshaw Lane, Culcheth Warrington, Chesire United Kingdom, WA3 4NE Michael Corradini University of Wisconsin 1500 Johnson Drive Madison, WI 53706 Michael Cullingford IAEA P.O. Box 200 A-1400 Vienna, Austria Garth Oummic:gs Lawrence Livermore Laboratory L-91, Box 808 Livermore, CA 94526
- Mark A. Cunningham U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Washington, DC 20555 W
DIST-6 James J. Curry 7135 Salem Park Circle Mechanicsburg, PA 17055 Peter Cybulskis Battelle Columbus Division 505 King Avenue Columbus, OH 43201 Peter R. D;vis Intermountain Technologies. Inc. P.O. Box 1604 Idaho Falls. ID 83401 Jose DeCarlos Sor Angela De la Cruz, No. 3, Planta 8 Madrid 28016, Spain Richard Deem New York Power Authority 123 Main Street White Plains, NY 10601 Richard S. Denning Battelle Columbus Division 505 King Avenue Columbus, OH 43201 Raymond DiSalvo Battelle Columbus Division 505 King Avenue Columbus, OH 43201 Adel A. El-Bassioni U.S. Nuclear Regulatory Commission Office of Nuclear Reactor Regulation l Washington, DC 20555 J. Mark Elliott International Energy Associates, Ltd. 000 New Hampshire Ave, NW, Suite 600 Washington, DC 20037 Energy Issues Rep. Sierra Club 1228 North Street #31 Sacramento, CA 95814 Malcolm L. Ernst U.S. Nuclear Regulatory Commission 101 Marietta Street, suite 3100 Atlanta, GA 30323 DIST-7 ) 1 Executive Director Natural Resources Defense Council 122 East 42nd Street New YorkNY 10168 P. Fehrenback Atomic Energy Canada, Ltd. Chalk River Ontario, Canada A. Fiege Kernforschungszentrum Postfach 3640 7500 Karlsruhe Federal Republic of Germany George F. Flanagan Oak Ridge National Laboratories P.O. Box Y Oak Ridge, TN 37831 Mario Fontana . International Technology. Corp. 575 Oak Ridge Turnpike l Oak Ridge. TN 37830 1 Ray R. Fray Pacific Gas and Electric Co. 215 Market Street, Room 206 San Francisco, CA 94106 James Fulford NUS Corporation 910 Clopper Road Gaithersburg, MD 20878 , J. B. Fussell JBF Associates, Inc. 1630 Downtown West Boulevard Knoxville, TN 37919 Raymond H. V. Gallucci Battelle Pacific Northwest Labs. P.O. Box 999 Richland, WA 99352 o John Garrick Pickard, Lowe & Garrick, Inc. ' 2260 University Drive Newport Beach, CA 92660 DIST-8 I John Gaunt British Embassy 3100 Massachusetts Avenue, NW Washington, DC 20008 J. Gieseke' Battelle Columbus Division 505 King Avenue Columbus, OH 43201 ) James C. Glynn U.S. Nuclear. Regulatory Commission i Office of Nuclear Regulatory Research l Washington, DC 20555 j i L. Hammar l Statens Karnkraftinspektion P.O. Box 27106 S-10252-Stockholm, Sweden M. R. Hayns l U.K. Atomic Energy Authority Wigshaw Lane, Culcheth Warrington, Chesire United Kingdom WA3 4NE Robert E. Henry i Fauske and Associates, Inc. 16WO70 West 83rd Street f Burr Ridge, IL 60521 Daniel Hirsch University of California A. Stevenson Prog. on Nuclear Policy Santa Cruz, CA 95064 Steven Hodge , Oak Ridge National Laboratories P.O. Box Y Oak Ridge, TN 37831 Thomas C. Houghton KMC, Inc. 1747 Pennsylvania Avenue, NW Washington, DC 20006 Dean Houston U.S.. Nuclear Regulatory Commission Advisory Comm. on Reactor Safeguards Washington, DC 20555 DIST-9 Der Yu Hsia Taiwan Atomic Energy Council 67 Lane 144, Keelung Rd., Sec. 4 Taipei, Taiwan R. E. Jaquith Combustion Engineering, Inc. , 1000 Prospect Hill Road M/C 9490-2405 Windsor, CT 06095 Kjell Johannson Studsvik Energiteknik AB S-611 82, Nykoping, Sweden D. H. Johnson Pickard, Lowe & Garrick, Inc. 2260 University Drive Newport Beach, CA 92660 Vojin Joksimovich NUS Corporation g 16885 W. Bernardo Drive ( San Diego, CA 92127 H. R. Jun Korea Adv. Energy Research Inst. P.O. Box 7, Daeduk Danju Chungnam 300-31, Korea Peter Kafka Gesellschaft Fur Reaktorsicherheit Forschungsgelande 8046 Garching Federal Republic of Germany M. Kahatib-Rahbar Brookhaven National Laboratory Associated Universities, Inc. Upton, NY 11973 Geoffrey D. Kaiser NUS Corporation 910 Clopper Road Gaithersburg, MD 20878 W. Kastenburg , _ ( UCLA 5532 Boelter Hall Los Angeles, CA 90024 DIST-10 W. Kato Brookhaven National Laboratory Associated Universities, Inc. Upton, NY 11973 William Kerr University of Michigan, North Campus Department of Nuclear Engineering Ann Arbor, MI 48109 George Klopp , Commonwealth Edison Company P.O. Box 767. Room 35W Chicago, IL 60690 Klaus Koberlein Gesellschaft Fur Reaktorsicherheit Forschungsgelande 8046 Garching Federal Republic of Germany E. Kohn Atomic Energy Canada Ltd. Candu Operations, Mississauga Ontario, Canada Herbert J. C. Kouts Brookhaven National Laboratories Building 179C Upton, L.I. NY 11973 Thomas Kress Oak Ridge National Laboratories P.O. Box Y Oak Ridge, TN 37831 Jeanne-Marie Lanore CEA B.P. No. 6 92660 Fonenay-Au-Roses, France John C. Lee University of Michigan, North Campus, Dept. of Nuclear Engineering Ann Arbor, MI 48109 Harold W. Lewis University of California Physics Department Santa Barbara, CA 93106 DIST-ll Librarian Atomic Industrial Forum 7101 Wisconsin Avenue Bethesda, MD 20814 Eng Lin Taiwan Power Company 242, Roosevelt Rd., Sec. 3 Taipei, Taiwan J. P. Longworth Central Electric Gen' Ting Brd. Berkeley Gloucester GL13 9PB United Kingdom William J. Luckas Brookhaven National Laboratories Building 130 Upton, L.I., NY 11973 Hans Ludewig Brookhaven National Laboratories Building 130 Upton, L.I., NY 11973 P. MacDonald EG&G Idaho. Inc. P.O. Box 1625 Idaho Falls, ID 83415 A. P. Malinauskas Oak Ridge National Laboratories P.O. Box Y Oak Ridge, TN 37831 Giuseppe Mancini Commission European Comm. CEC-JRC Eraton Ispra Varese, Italy Alan McGowan Scientists Institute for Public Info. 355 Lexington Avenue New York, NY 10017 Ralph Meyer U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Washington, DC 20555 DIST-12 Joselyn Mitchell i U.S. Nuclear Regulatory Commission l Office of Nuclear Regulatory Research j ~ Washington, DC 20555 Joseph A. Murphy U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Washington, DC 20555 Sue Niemczyk 1545 18th St. NW, #112 Washington, DC 20036 Edward P. O'Donnell Ebasco Services. Inc. 2 World Trade' Center, 89th Floor New York,,NY 10048 David Okrent University of California, Los Angeles Boelter Hall, Room'5532 Los Angeles, CA 90024 i William Parkinson i' Delian Corporation 1340 Saratoga-Sunnyvale Road Suite 206 San Jose, CA 95129 G. Petrangeli ENEA Nucl. Energ. ALT Disp Via V Brancati 48 00144 Roma, Italy I Robert D. Pollard Union of Concerned Scientists 1616 P Street, NW, Suite 310 Washington, DC 20036 R. Potter UK Atomic Energy Authority Winfrith Dorchester Dorset DT2 BDH United Kingdom William T. Pratt Brookhaven National Laboratories Building 130 Upton, NY 11973 l DIST-13 l David Pyatt U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Washington, DC 20555 Ingvard Rasmussen Riso National Laboratory Postbox 49 DK-4000 Roskilde. Denmark Norman C. Rasmussen Massachusetts Institute of Technology 77 Massachusetts Avenue Cambridge, MA 02139 Dennis Richardson Westinghouse Electric Corporation P. O. Box 355 Pittsburg, PA 15230 Robert Ritzman Electric Power Research Institute 3412 Hillview Avenue Palo Alto, CA 94304 Denwood F. Ross U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Washington, DC 20555 Frank Rowsome U.S. Nuclear Regulatory Commission Office of Nuclear Reactor Regulation Washington, DC 20555 John Siegal Atomic Industrial Forum 7101 Wisconsin Avenue Bethesda, MD 20814 Chester P. Siess Advisory Comm. on Reactor Safeguards 805 Hamilton Drive Champaign, IL 61820 Melvin Silberberg U.S. Nuclear Regulatory Commission Office of Nuclear Regulatory Research Washington, DC 20555 DIST-14 I K. Soda Japan Atomic Energy Res. Inst. Tokai-Mura Naka-Gun IBARAKI-KEN 319-11, Japan Leonard Soffer U.S. Nuclear Regulatory Commission Office of Nuclear Reactor Regulation Washington, DC 20555 Themis P. Speis U.S. Nuclear Regulatory Commission Office of Nuclear Reactor Regulation Washington, DC 20555 Klaus B. Stadie OECD-NEA, 38 Bld. Suchet . . 75016 Paris, France I s l Wayne L. Stiede Commonwealth Edison Company P.O. Box 767 P_ Chicago, IL 60690 . I: John Taylor Electric Power Research Institute 3412 Hillview Avenue Palo Alto, CA 94303 Merrill A. Taylor . U.S. Nuclear Regulatory Commission Office of the EDO Washington, DC 20555 Technical Library Electric Power Research Institute P.O. Box 10412 Palo AltoCA 94304 Mark I. Temme General Electric, Inc. P.O. Box 3508 Sunnyvale, CA 94088 T. G. Theofanous University of California, S.B. Dept. of Chemical and Nuclear Eng. Santa Barbara, CA 93106 DIST-15 Gordon Thompson Inst. for Resource & Security Studies 27 Elsworth Avenue Cambridge, MA 02139 Long Sun Tong 9733 Lookout Place Gaithersburg, MD 20879 A. F. Torri Pickard, Lowe & Garrick, Inc. 191 Calle Magdalena, Suite 290 Encinitas, CA 92024 Chao-Chin Tung Atomic Energy Council 5 67 Lane 144, Sec 4 Keelung Rd. Taipei, Taiwan Ian B. Wall Electric Power Research Institute P.O. Box 10412 Palo Alto, CA 94303 David Ward Savannah River Laboratories Building 773-A, A-219 Aiken, SC 29808 Edward Warman Stone & Webster Engineering Corp. P.O. Box 2325 Boston, MA 02107 Norman Weber Sargent & Lundy Co. 55 E. Monroe Street Chicago, IL 60603 Lois Webster American Nuclear Society 555 N. Kensington Avenue La Grange Park, IL 60525 Paul Wood Delian Corporation 296 Duff Road Monroeville, PA 15146 DIST-16 John Wreathall Battelle Columbus Division 505, King Avenue Columbus, OH 43201 i Roger'Wyrick Inst. for Nuclear Power Operations-1100 Circle 75 Pa'rkway, Suite 1500 .i Atlanta, GA 30339 Jonathan Young i Energy, Inc. 1851. South Central Place Suite 201 Kent, WA- 98031 l Steven Sholly MAB Technical Associates. 1723 Hamilton Avenue, Suite K San Jose .CA 95125 Sam Hobbs Mississippi Power & Light Post Office box 1640 Jackson, Mississippi 39215-1640 Gary Smith . Mississippi Power & Light j Post Office box 1640 ) Jackson, Mississippi 39215-1640 j Mr. A. R. Diederich Supervising Engineer Environmental Branch Philadelphia Electric Company 2301 Market Street , Philadelphia, Pennsylvania 19101 Greg Krueger Philadelphia Electric Company 2301 Market Street- t Philadelphia, Pennsylvania 19101 , Margaret Federline U.S. Nuclear Regulatory Commission Office of the Commission Washington, D.C. 20555 Walter Lowenstein Electric Power Research Institute Post Office Box 10412 Palo Alto, California 94303 DIST-17 _. __ __ a William Vesely Science Applications Int. Corp. 2929 Kenny Road, Suite 245 Columbus, Ohio 43221 Gareth Parry NUS Corporation 910 Clopper Road Gaithersburg, Maryland 20878 Doug Rickeard Virginia Electric Power Company James River Plaza Post Office Box 26666 Richmond, Virginia 23261 R. J. Hardwick, Jr. ) Innsbrook Technical Center 5000 Dominion Boulevard Glen Allen, Virginia 23060 Roberta Gallante Tennessee Valley Authority 400 West Summit Hill Drive W4A24C-k Knoxville, Tennessee 37902 DIST-18}}