ML20127C469

From kanterella
Jump to navigation Jump to search
Responds to 841231 Memo Re Preliminary Review of PRA Study. Reactor Power Vs Reactor Water Level Studies in Progress. Rept on Results of Preliminary Review on Back End PRA Expected by 850331
ML20127C469
Person / Time
Site: Shoreham File:Long Island Lighting Company icon.png
Issue date: 02/06/1985
From: Sheron B
Office of Nuclear Reactor Regulation
To: Thadani A
Office of Nuclear Reactor Regulation
Shared Package
ML20127A367 List:
References
FOIA-85-199 NUDOCS 8502110559
Download: ML20127C469 (1)
v • d • e


Text

_ - .

na:g'o

,, UNITED STATES

! o NUCLEAR REGULATORY COMMISSION

( $ WASHINGTON, D. C. 20555

,,,,,/ FEB 6 1985 MEMORANDUM FOR: Ashok Thadani, Chief Reliability and Risk Assessment Branch, DST FROM: Brian W. Sheron, Chief Reactor Systems Branch, DSI

SUBJECT:

RSB COMMENTS ON PRELIMINARY REVIEW OF SHOREHAM PRA STUDY (FRONT END)

Reference:

Memorandum from Themis P. Speis, Director, DST to Darrell G.

Eisenhut, Director, DL, " Preliminary Review of Shoreham PRA Study," dated December 31, 1984 In response-to the RRAB review request, the Reactor Systems Branch has reviewed the subject study and our comments are attached. For ranvenience, we have identified pages with comments by paper clips. We will be glad to discuss the comments with you~at a mutually convenient time. .

Our inquiries with the licensee (LILCo) have indicated that the Shoreham ATWS procedures reviewed by BNL were prepared according to revision 18 of the Generic Procedure Guidelines. The latest, Revision 3 of the Generic Proce-dures Guidelines is an iniprovement over Revision 18. LILCo is committed to follow the latest revision of the Generic Procedure Guidelines in thei~r next

/

revision of the ATWS procedure.

We believe that when the reactor water level is at the top of active fuel (TAF), the reactor power is between 10% to 201 Reactor power vs. reactor water level studies are still in progress at EPRI, BNL and other industry sources. The issue is still not settled.

RSB is planning to issue-for review and co'mment the results of a preliminary review done by BNL on "Back End" Shoreham PRA by March 1985.

YW Brian W. Sheron, Chief Reactor Systems Branch, DSI

Enclosure:

As stated cc: w/o enclosure R. Bernero R. Houston A. Buslik 6

CONTACT: G. Thomas, RSB, x29445 I w

ge q l FALW6 M -

[

h-a-e* . ; w::*: , -

~

/gg'O LONG ISLAND LIGHTING COMPANY SU e,, SHOREHAM NUCLEAR POWER STATION P.O. BOX 618, NORTH COUNTRY ROAD . WADING RIVER. N.Y.11792 JOHN D. LEONARD.JR.

VICE PRESIDENT. NUCLEAR OPER ATONS February 25, 1985 SNRC-ll49 Mr. Harold R. Denton, Director Office of Nuclear Reactor Regulation U.S. Nuclear Regulatory Commission Washington, DC 20555 i LILCO Comments on Preliminary Review of Shoreham PRA Study Shoreham Nuclear Power Station Docket No. 50-322

Reference:

Letter from A. Schwencer (NRC) to J. D. Leonard, Jr. (LILCO) dated January 24, 1985

Dear Mr. Denton:

Attached please find LILCO's comments on NUREG/CR-4050, "A Review of the Shoreham Nuclear Power Station Probabilistic Risk Assess-ment", dated November 1984. At this time, additional effort by LILCO is deemed necessary in order to prepare detailed comments on the Brookhaven review. To expedite this process, a meeting between Brookhaven and LILCO at the working level is recommended to resolve differences in the analyses.

We trust this letter addresses in part Brookhaven's review of the PRA study, and we look forward to discussing the outstanding com-ments with their staff. If you require additional information, please contact this office.

Very truly yours, k-o . Leonard, Jr.

Vice President - Nuclear Operations NRL:ck Attachment cc: P. Eselgroth K. Shiu, Department of Nuclear Energy Brookhaven National Laboratory MW 1 O dIQ

,.+.n.-. - . ,. n . . - . - - . . , - . . - - - .- ~ ,

.
  • r ..  ;
  • I l

ATTACHMENT I SNRC-1149

SUMMARY

The BNL review indicates a core vulnerable frequency of 1.5E-4yr.

while the SNPS PRA calculated a value of 5.5E-5yr.- This differ-ence is not large, but it is judged that the BNL value is not a best estimate analysis and has incorporated the following conservative biases:

o The initiator. frequencies are increased by utilizing the " Bayesian Two Stage" analyzer.

o The initiator frequencies are biased by including the first year of data directly. It is a well known fact that the first year data is not characteristic of

~

mature plant operation. The unweighted inclusion of first year data results in conservative values.

o The initiators are not discriminated due to initial power-level for the ATWS evaluation. This is an important conservative bias introduced by BNL.

o The LOOP initiator frequency is conservatively assessed by ignoring LILCO grid-specific information, and by  !

conservatively assessing the data presented in NSAC/80.

o The present on-site power capability of LILCO is not included.

o The reactor water level instrumentation analysis neglects the changes made by LILCO which include:

analog trip system, and j additional level instrumentation.

o The postulated common-mode failure of the HPCI and RCIC batteries is included in the Shoreham analysis. BNL increased this probability without substantiation.

Removal of the conservative elements cited in this summary will result-in reducing the BNL calculated core and containment vulnerable frequency from 1.5E-4/yr. to approximately 8E-5/yr.

OVERVIEW

l. . Within the context of PRA evaluations and their inherent uncertainties and limitations, there does not appear to be any' substantial disagreement in the evaluation of what sequences are important.

' Attachment I SNRC-ll49 Page 2

2. It is judged that the BNL numerical estimates are conserva-tive and therefore, the use of the BNL study point estimates should.be used with caution.
3. The BNL point estimates are referred to as core damage fre-quencies.' However, the.SNPS PRA-frequencies that are cited are not core damage frequencies, rather they are end-states referred to as core or containment vulnerable conditions.

This distinction is important if the purpose of the BNL evaluation is to compare to the proposed secondary safety goal of core melt frequency. The SNPS PRA values cited by BNL do not correspond with core melt, and it can only be assumed that the BNL values are meant also to be consistent with the SNPS intent.and are therefore not core melt. BNL review needs to, clarify the end-states since they currently appear.to be inconsistent.

4.- The BNL review provides a good sensitivity study to demon-strate what types of variations in accident sequence fre-quency can be obtained-through changes in input parameters.

Most'of the BNL changes may be referred to as sensitivities since there is not aLways a technical justification which would support the BNL value as a best estimate value.

5. The differences in numerical values cited by BNL are

- generally due to differences in generic input data values.

6. For initiators, the use of EPRI NP-2230 would make a small increase in the turbine trip initiator frequency.
7. There appears to be agreement that the modeling of the plant has been performed accurately and that generally only the point estimate input values are in question.

, 8. The BNL document identifies some areas in which their

analyses are conservative. These conservatisms should be considered if the BNL quantification is to be used for anything other than sensitivity study. Examples include

, p3: MSIV initiator frequency is high due to LILCO '

committed plant-specific design modification not reflected in data from older plants. (Level 1 isolation).

p7: The LOOP sequences are conservative due to the lack of modeling of the actual plant electrical

configuration which includes

two independent switchyards black-start gas turbine onsite additional diesels onsite extensive black-start gas turbine capability offsite 4

e , e 1 --.+ -y,-m.w-e-e.w --vi-i-ry,~,m-wewy -yy w w- wee--4---*,.-,,,-w-e--- v-,, .-r-~-,.,--a- -- - - - - - - w- - , , ,--r- mm- , - , - - . ,r-,---w-

. .- . , - . - _ . - - , o_. _ _. - _. _ _

.-m. - . .

  • Attachm2nt I SNRC-ll49 Page 3 p50-5: The ATWS, initiator frequency is conservative'due

+

to the inclusion of'startup initiated events.

Approximately one-third of transient initiators are at low power.

9. The BNL assertion on page xi that there are large changes in
the ranking of dominant sequences as a result of the BNL quantitative estimates. This, however, is misleading since the rearrangement of sequences as a result of the BNL assump-tions does not change the basic set of sequences identified
in the SNPS PRA.

Detailed Comments

- 1. Initiator Frequency l o Use of the weighted average of years is appropriate for the evaluation of risk associated with Shoreham during i

mature plant operation. Calculation of risk associated with initial years of plant operation or end-of-life is a separate problem. This use of the data by BNL causes a conservative bias in the calculation of risk for mature plant operation.

p' o Bayesian two-stage analysis tends to increase calculated values.

o IORV frequency is biased high since it does not reflect the plant specific shoreham design (Target Rock two-stage SRVs).

The LOOP frequency is biased high,hecause: -

o NSAC/80 data including BNL modifications to data (pg 4-9) is not appropriate to the LILCO specific grid.

o The plant-specific shoreham design is not included in LOOP recovery; i.e., black-start gas turbines, backup diesels, independent switchyards.

2. The contributions to Class II and IV events are referred to as core damage sequences. In fact, these sequences are containment vulnerable conditions which may not necessarily lead to core damage, i
3. By placing the ATWS' Class I sequences in Class IV, BNL has not' fully considered the logical possibility of sequences which cause core melt prior to containment challenge. This c is judged to be a conservative bias in the public risk evaluation.

l 4. There are a number of additional failure probabilities inserted into the Shoreham analysis which seem to double count the failure probabilities already included in the

, Shoreham analysis. For example, time-phased functional l event trees on power conversion system appears to double I i count feedwater failure.

4

-- - - , ,-, , ~ , . ,,------, ,, ~ _ , -n-, $1,Ylr. N,5 w a --r.-- - - - -n,.n,-n-cn,-,, -----~ n -- ,, ,,-e-,-,

, "Attschmsnt.I SNRC-Il49 Page 4 COMMENTS BY SEQUENCE ATWS:

o The initiator frequencies are high as identified above.

o The failure to discriminate between high power and low power initiated transients leads to approximately a 30% increase in the ATWS sequence frequencies, o p5D the table is misleading by including only high power initiators from Shoreham while including all power level initiators from BNL evaluation.

LOOP:

o The BNL initiator frequency is conservative and is not characteristic of the LILCO grid, o The Shoreham dual switchyard configuration is not included in the quantification.

o . Table SB the evalu'ation of the reactor water level indi-cation during LOOP does not recognize the scenario which is postulated (i.e., DC buses and the third diesel available) and does not reflect the emergency procedure actions which would occur. It does not reflect the likelihood of continued operation of a working water level indication system.

o Table SB the justification for a Class II accident scenario during station blackout is not correct (T I IV W).

The data referenced for recovery of offsite power Endicates that all observed failures have been recovered within ten hours. The relatively high probability for failure to recover that BNL assumes and the fact that BNL only allows 15 hours1.736111e-4 days <br />0.00417 hours <br />2.480159e-5 weeks <br />5.7075e-6 months <br /> to recover offsite power before calling it a Class II sequence is very conservative. AC power recovery from i

Table 4-7 indicates that there are no observed LOOP events longer than 10 hours1.157407e-4 days <br />0.00278 hours <br />1.653439e-5 weeks <br />3.805e-6 months <br />. The crucial time before containment vulnerable conditions would be on the order of 30 to 40 hours4.62963e-4 days <br />0.0111 hours <br />6.613757e-5 weeks <br />1.522e-5 months <br />, not 15 hours1.736111e-4 days <br />0.00417 hours <br />2.480159e-5 weeks <br />5.7075e-6 months <br /> as~ assumed by BNL.

i INTERFACING LOCA o The BNL review represents a good sensitivity study to identify potential variations.

o There appears to be a neglect of the valve interlocks which would be violated to cause such an event.

~

o There appears to be a lack of credit for the required inservice testing of the valves.

4 f

. - , . ._ _ .. _ ._ _ _ _ n _ , _ _ . _ _ . - . _ . _ _ _ _ . . _ _ . _ _ . _ _ _ .

. AAttachment I SNRC-ll49 Page 5

  • t o .The3 mathematics used to combine failure rates (.02/yr. and 10 /d) seems questionable.

o The 10 -3 d (MOV spurious. opening) seems arbitrary and is useful only as a sensitivity. Spurious Mov opening is not a

, failure mode generally seen in the data, especially for an interlocked valve.

o The neglect of the operator probability to reclose.the valves is an important omission in light of the suc'cessful performance by the operator in the LER data.

o When the above changes are included in the analysis, the

. point estimate should be reduced by several orders of magnitude.

Reactor Water Level Reference Leg o BNL conservatively neglected two major design changes:

Analog Trip System: This eliminates the need for on-line testing involving the reference legs.

additional HPCI level transmitters.

o BNL's reference to LERs where more than two reference legs have been affected is not supported. Such LERs have not occurred to our knowledge or been verified by S. Levy in reports SLI 8218 or SLI 8221.

P o The assertion that miscalibration is the same as the loss of a DC bus is a conservative assumption which introduces conservative bias into the quantification.

S l

9

. - - . - - . , . . _ . , -- .- ,,,----,, ,- --,,.,,- ,--..,.,, +, ,,--,- ,.. ,.. ., n. , - , . - -. . - - . - n.- ac.

  • *a. j

. e

  • 1

. ~.Attcchmsnt I l

SNRC-1149 '

Page 6 REVIEW OF FLOOD EVENTS Overview:

The increase in frequency of floods estimated by BNL is in large part due to conservatism within their modeling process. The BNL analysis description contained insufficient information to track in detail their quantification process, only summary tables were provided. The use of Markov models did not enhance the trace-ability of the results, and did not allow for straight-forward indication of plant states. In light of the conservatisms included in the BNL study, an appropriate reanalysis should reduce the estimated frequency of core vulnerable conditions developed by BNL.

Specific Comments:

The following describes some, but not all, of the problems encountered in the review:

1. Inclusion of the Brown's Ferry Event in the analysis is very conservative due to the fact that the Brown's Ferry flood resulted from failure *of aluminum pipe while Shoreham has stainless steel.
2. The initiator frequency for MSIV closure of 4.42/y used in the flooding analysis by BNL is very conservative with respect to the value used in the PRA review .67/Rx year.
3. Hourly failure rates derived for motor driven and turbine driven pumps, and major maintenance are conservative. All failures of these devices do not result in disassembly of these devices.
4. It appears that maximum allowable outage times were used.in the Markov models for maintenance induced floods. Mean time to repair should be used.
5. The BNL time-phased event tree calculations were not
traceable.
6. The conclusion that HPCI and RCIC are disabled in Phase II is

~

incorrect. Equipment between l'3" and l'10" is not necessary l for successful operation of those systems.

.

Retrieved from "https://kanterella.com/w/index.php?title=ML20127C469&oldid=2903835"