NUREG/CR-5496, Issues Final Rept,Evaluation of Loss of Offsite Power Events at Nuclear Power Plants 1980-1996,NUREG/CR-5496.Observations & Findings of Study Listed.Without Encl

From kanterella
(Redirected from NUREG/CR-5496)
Jump to navigation Jump to search
Issues Final Rept,Evaluation of Loss of Offsite Power Events at Nuclear Power Plants 1980-1996,NUREG/CR-5496.Observations & Findings of Study Listed.Without Encl
ML20195J517
Person / Time
Issue date: 11/16/1998
From: Rossi C
NRC OFFICE FOR ANALYSIS & EVALUATION OF OPERATIONAL DATA (AEOD)
To: Gillespie F, Holahan G, Lainas G
NRC (Affiliation Not Assigned)
References
RTR-NUREG-1032, RTR-NUREG-CR-5496 NUDOCS 9811250001
Download: ML20195J517 (18)
v • d • e


Text

{{#Wiki_filter:m. ,__m- -.-m- _. _ _ ~ _. 4 j - A RE*pqk-UNITED STATES ~. y s= { NUCLEAR REGULATORY COMMISSION e WASHINGTON D.C. 20066-4001 V %. jf ; November 16, 1998 i MEMORANDUM TO: Frank P. Gillespie, Director, DISP:NRR j Gary 'M. Holahan, Director, DSSA:NRR Gus C. Lainas, Acting Director, DE:NRR Jack W. Roe, Acting Director, DRPM:NRR F R. Lee Spessard, Director, DRCH:NRR Lawrence C. Shao, Director, DET:RES Thomas L. King, Director, DST:RES John W. Craig, Director, DRA:RES ' James T. Wiggins, Director, DRS:RGN-l Charles W. Hehl, Director, DRP:RGN-l Bruce S. Mallett, Director, DRS:RGN-ll Loren R. Plisco, Director, DRP:RGN-Il John A. Grobe, Director, DRS:RGN-ill Geoffrey E. Grant, Director, DRP:RGN-Ill Arthur T. Howell, Director, DRS:RGN-IV Thomas P. Gwynn, Director, DRP:RGN-IV 'b FROM:- Charles E. Rossi, Dir Safety Programs Division 1 Office for Analysis and Evaluation of Operational Data 4'

SUBJECT:

ISSUE OF FINAL REPORT-EVALUATION OF LOSS OF OFFSITE POWER EVENTS AT NUCLEAR POWER PLANTS: 1980-1996 (NUREG/CR-5496) The final report, " Evaluation c,f Offsite Power Events at Nuclear Power Plants: 1980-1996," [ NUREG/CR-5496, has been completed. The report will be distributed to you by NRC Publications. The objective of this study was to update the loss of offsite power (LOSP) estimates (frequency and duration) found in NUREG-1032, " Evaluation of Station Blackout 'j S Accidents at Nuclear Power Plants," which spanned the time period from 1969 to 1985. This Q study used the same data classification scheme as in NUREG-1032, i.e., plant-centered, grid, d[;* and severe weather-related events. In addition to estimating LOSP frequency and duration during power operation, the study also estimated the LOSP frequency and duration during Ci shutdown. Earlier drafts of this report were provided to the Office of Nuclear Reactor j Regulation (NRR), the Office of Nuclear Regulatory Research (RES), the regions, and industry bi. organizations for peer review and comment. The summary of resolutions to comments is i ii attached. s 9 A number of comments from intemal reviews of similar AEOD reports indicated a difficulty in e, y '; obtaining qualitative insights and related detailed information from the reports, especially for inspection-related activities. To help the staff to better identify and relate this detailed information to various risk-important regulatory applications, we have provided a foreword to the p, [b, in the report. The foreword also indicates the appropriate type of engineering re report that provides directions to the relevant quantitative and qualitative information contained M'd D. information needed for application on a plant-specific basis. However, it should be recognized 9811250001 981116 CR-5496 C eDR 1 &// h,.@//x4eej-{- eon wunse

Multiple Addressees that this report is not a handbook for inspection. It is a source of information that can be used to focus both generic and plant-specific risk-informed activities involving the LOSP. Cognizant AEOD staff will be available to consult with users of this report. Notable observations and findings of the study incluae the following: NUREG-1032 found that plant-centered events accounted for the majority of the losses of offsite power. This study of more recent operating experience supports that finding, with plant-centered events clearly dominating LOSP frequency during power operation (0.04 per unit critical year), as well as during non-power modes of operation (0.18 per unit shutdown year). Events induced by severe weather are much less frequent (0.01 per site calendar year), and grid-related events are still less frequent (0.004 per site calendar year). The LOSP frequency for plant-centered events is significantly higher during shutdown modes of operation than during power operation, by a factor of about four. For severe-weather events and grid-related events, too few events occurred to distinguish between power and shutdown operations. Plant-centered initiating events per year based on the 1980-1996 experience have become less frequent since the time period studied by NUREG-1032 (based on 1968-1985 experience). A clear downward trend can be seen in the frequency of combined data from 1969 through 1996. Analysis of the data did not reveal any apparent reason for this decrease that could be directly related to the Station Blackout Rule (10 CFR 50.63), which was published in June 1988. The LOSP frequency from grid-related events in the period covered by this report, 1980-1996, was very small. During this period, there were only five site events that could be classified as grid-related. The frequency of grid-related LOSP events is less than that found in the period studied in NUREG-1032 by a factor of about 10. No grid-related events occurred in the 1990s, in spite of the occurrence of several widespread losses of power to the public. Only one LOSP event occurred that resulted in total and sustained voltage loss to all safety buses due to a grid disturbance. A fire in 1985 near Turkey Point caused a grid failure that resulted in both units experiencing a LOSP event. The frequency of LOSP events due to severe weather exhibited statistically significant site-to-site variability. This is to be expected, as some power plants' geographic location will tend to have increased their exposure to severe weather. The majority of pMnt-centered LOSP initiating events at power were caused by equipment faults (58%), with a smaller portion being induced by human error (23%). During shutdown modes, the opposite holds, with human errors being the major contributor (58%). The same patterns occur when only sustained events or only momentary events (less that two minutes duration) are considered. The major contributors to equipment faults were failures of transformers, protective / sensing circuits, circuit breakers, and relays. The major contributors to human error were maintenance, testing, and switching errors.

Multiple Addressees As found by the earlier NUREG-1032 study, the recovery times of sustained LOSP events were significantly longer for severe-weather events than for plant-centered events. Too few grid-related events occurred during the period (1980-1996) of this more current report to permit comparison of their recovery times with plant-centered or severe-weather recovery times. Statistical analysis of the 1980-1996 experience did not detect a difference in recovery times between switchyard design classes as defined in NUREG-1032. Analysis of station blackout risk was outside the scope of this study. However,16 station blackout events were identified during the data review in which a power plant unit had no AC electrical power from any source for up to 1 hour. Only two of these events occurred during power operations, and the longest of these two events lasted 11 minutes, which is well below the minimum coping time specified in U.S. NRC Regulatory Guide 1.155. The duration of each event was small and the need for accident mitigation systems powered from emergency AC power was not present in the events. These findings are discussed in more detail in the report. Graphical and tabular displays, along with specific discussions, are included. Specific failures and failure mechanisms are identified and characterized. USE OF STUDY RESULTS TO IMPROVE RISK-INFORMED REGULATORY ACTIVITIES: Based on these findings and conclusions, the following can be used to improve risk-informed regulatory activities: inspections. Inspectors should be made aware of the higher occurrence of LOSP e events while the plant is shutdown so that plant activities (especially maintenance, testing and switching activities in the switchyard) that could adversely impact the availability of offsite power can receive relatively higher attention in plant inspections. The plant is more susceptible to offsite power losses due to human error caused by increased maintenance, testing and switching activities in *he switchyard coupled with the greater chance that the unit electrical system has be' aligned in a non-standard configuration. Grid Reliability Task Action Plan. In light of the extremely small contribution of grid-induced LOSP events found in the 1980-1996 experience and the availability of data on grid-induced LOSP events as a performance indicator for grid reliability, we recommend that no further regulatory action is needed with respect to Milestone 4 of the task action plan on grid-reliability. Operating experience has shown that grid-induced LOSP events have decreased by a factor of 10 from the frequency originally estimated in NUREG-1032; the last grid-laduced LOSP event occurred in 1989; and the reliability trend of mitigating systems, such as emergency diesel generators, auxiliary feedwater systems in pressurized water reactors, and high pressure coolant injection and reactor core isolation cooling systems in boiling water reactors, are either constant or improving. Furthermore LOSP events are reportable, and any increase in the trend of grid-induced LOSP events can be observed long before they become significant in terms of the potential for significant increase in core damage frequency.

Multiple Addressees l l' l L There have been only four grid-related LOSP events since 1980, with no events occurring since 1990. Four or five grid-related LOSP events would have to occur between the years 1997 and 2000 before the grid-related LOSP frequency would be - statistically different from the frequency estimated in the report. During this same . period, we would expect to see approximately 12 LOSP events from all causes. Thus, increases in grid-related LOSP frequency can be identified through routine monitoring and analysis of operating experience before they become a significant contributor to risk-from LOSP events. cc w/att.: A. Thadani, RES S. Collins, NRR R. Zimmerman, NRR B. Sheron, NRR . B. Boger, NRR E. Adensam, NRR l I Distribution w/att.: Ii File Center JLarkins, ACRS JStolz, NRR DEDE RRAB R/F MMarkley, ACRS GParry, NRR PWi! son, NRR SPD R/F ARubin, RES LMarsh, NRR DCoe, NRR Public MMayfield, RES JFlack, RES JTrapp, RI . TTMartin REmrit, RES - MRubin, NRR TShediosky, RI FCongel MCunningham, RES AEl-Bassioni, NRR - WRogers, Ril KRaglin NSiu, RES JCalvo, NRR RBernhard, Rll DHickman MDrouin, RES RJenkins, NRR MParker, Rlli JRosenthal HVanderMolen, RES CPoslusny, NRR SBurgess, Rlli RBarrett, NRR WJones, RIV JMitchell, OEDO JShackelford, RIV DOCUMENT NAME: H:\\DMR\\LOSP-LET.WPD

  • See previous concurrence To receive a copy of the document, Indicate in the box "C" wo/ attach /enci "E" wlattach/enci "N" no co py OFFICE RRAS/RRAB E

RRAS/RRAB E RRAB E SPD 6 NAME DRasmuson* SMays* PBaranowsky* CERospg4 DATE-10/15/98 10/16/98 10/16/98 // / //,/98 i OFFICIAL RECORD COPY l

\\ Multiple Addressees There have been only four grid-related LOSP events since 1980, with no events occurring since 1990. Four or five grid-related LOSP events would have to o r between the years 1997 and 2000 before the grid-related LOSP frequency uld be statistically different from the frequency estimated in the report. During i same period, we would expect to see approximately 12 LOSP events from al uses. Thus, increases in grid-related LOSP frequency can be identified through r tine monitoring and analysis of operating experience before they become a signifi nt contributor to risk from LOSP events. cc w/att.: j A. Thadani, RES ) S. Collins, NRR R. Zimmerman, NRR B. Sheron, NRR B. Boger, NRR E. Adensam, NRR Distribution w/att.: FUe Center MMarkley, ACRS GParry, NRR DCoe, NRR ~ RRAB R/F ARubin, RES LMarsh, NRR JTrapp, RI SPD R/F MMayfield, RE JFlack, RES TShedlosky, Rl TTMartin REmrit, RES MRubin, NRR WRogers, Rll FCongel MCunningh ,RES AEl-Bassioni, NPI. RBernhard, Rll KRaglin. NSlu, RES RBarrett, NRR MParker, Rlli DHickman MDrouin ES JMitchell, OEDO SBurgess, Rlli JRosenthal HVand Molen, RES WTravers, DEDE WJones, RIV JLarkins, ACRS JStol, NRR PWilson, NRR JShackelford, RIV DOCUMENT NAME: H:\\ MR\\LOSP-LET.WPD l

  • See pr lous concurrence To ive a copy of the document, Indicate in the box "C" wo/ attach /enci "E" w/ attach /enci "N" no copy kFICE RRAS/RRAB E

RRAS/RRAB C. RRAB g SPD l i4AME DRasmuson* SMays 8yn PBaranowsky/M CERossi DATE 10/15/98 la / /b /98 /[/// /98 / /98 l 7

Multiple Addressees, 1 'There have been only four grid-related LOSP events since 1980, with no ev nts j - occurring since 1990. Four or five grid events would have to occur befor he grid 1' frequency for the four year period (1997-2000) would be statistically di rent from the frequency estimated in the report, ignoring the previous LOSP expe

  • nee. During this' same period, we would expect to see approximately 12 LOSP eve s from. all causes.

Thus, increases in grid-related LOSP frequency can be identifie rough routine monitoring and analysis of operating experience before they me a significant contributor to risk from LOSP events. .r cc w/att.: A. Thadani, RES S. Collins, NRR R. Zimmerman, NRR ' 1 B. Sheron, NRR B. Boger, NRR E. Adensam, NRR Distribution w/att.: File Center MMarkley, ACRS GParry, NRR DCoe, NRR . RRAB R/F ARubin, RES LMarsh, NRR JTrapp, RI SPD R/F MMayfield, RES JFlack, RES TShediosky, RI TTMartin REmrit, RES MRubin, NRR WRogers, Ril FCongel - MCunningha, RES AEl-Bassioni,' NRR RBembard, Ril KRaglin NSiu, RES RBarrett, NRR MParker, Rlli DHickman

MDrouin, S.

JMitchell, OEDO SBurgess, Rill JRosenthal HVande olen, RES WTravers, DEDE WJones, RIV 'JLarkins, ACRS JStolz, RR PWilson, NRR JShackelford, RIV DOCUMENT NAME: H:\\D \\LOSP-LET.WPD I To receiv a copy of the document, indicate in the box "C" wolattach/enci"E" wlattach/enci"N" no copy OFFjdE RRAS/RRAB RRAS/RRAB RRAB SPD NhE DRasmuson SMays PBaranowsky CERossi dATE 10/15~ /98 / /98 / /98 / /98

ATTACHMENT RESOLUTION OF COMMENTS FOR DRAFT REPORT EVALUATION OF LOSS OF OFFSITE POWER EVENTS AT NUCLEAR POWER PLANTS: 1980-1996 Summaries of the comments received on the draft report titled, " Evaluation of Loss of Offsite Power (LOSP) Events at Nuclear Power Plants: 1980-1996," November 1997, along with their associated resolutions follow. The final version of the LOSP report incorporates these resolutions. 1. Comment (from Region IV): The major grid disturbance, which occurred on August 10, 1996, was not captured in the data due to the definitions used in determining what constitutes a LOSP event. This event resulted in a dual unit trip at the Diablo Canyon site even through the emergency diesel generators did not receive an auto-start signal due to specific circumstances surrounding the disturbance. Even though an event such as this may not meet the exact definition of an LOSP as defined in the study, such events have significant implications from an operational safety perspective and should somehow be captured in the Agency's concems with potential impacts of economic deregulation on the electric utility industry. Comment (from NRR): Page 22: The draft report discusses an August 10,1996, westem grid disturbance that, despite national media attention, did not result in any LOSP events. While there were no LOSP events, there were four plant trips as a result of the disturbance (both units at Diablo Canyon and two units at Palo Verde). Resolution: Section 4.1.2 was revised to include an expanded discussion about this event. The AEOD case study, Grid Performance Factors (C97-01), discusses grid disturbances during the 1985 - 1996 time period in detail. 2. Comment (from Region IV): The study goes into considerable detail regarding the statistical treatment and analysis methods, which were used to analyze the data, but very little information is provided as to the qualitative insights which might be available from the data reduction. For example, statements are made to the effect that most LOSP events at shutdown can be attributed to " human error " However, the study is silent as to the specific characterization and/or nature of the human error involved. Similarly, the study concludes that at power events are dominated by equipment failures but does not provide information as to the specific components or failure modes which are involved. These types of insights would be particularly valuable from an inspection perspective in that it would allow regional and resident inspectors to target their activities on the most susceptible human actions and components. Resolution: The human error and equipment failure events have been reviewed to determine the subcategory of cause. Section 4.2 was revised to include the results of this more detailed review. 3. Comment (from NRR): The executive summary is not entirely clear regarding the trend of LOSP frequency. On page xiii, the second bullet states that a decreasing trend in time was not quite statistically significant for plant-centered initiating events at power. But the fifth bullet on this page states that a clear downward trend can be seen in the frequency [of plant-centered initiating events) from 1969 through 1996. 1

.~ l Resolution: The difference in the two trends is the size of the data set used in the l. trend analysis. The second bullet in the conclusion used data from the current study (1980-1996) and the fifth bullet used data from the current study combined with data from NUREG-1032 (1969-1996). Sometimes a trend can be seen with a large data set, but not seen with a smaller portion of the data. The executive summary and Section 5, Conclusion, were revised to include the time period of the trends. In addition, a i discussion was added in Section 3.5.1 that discusses the difference between the two

trends, j

4. Comment (from NRR): The executive summary states that there has been an increasing trend in the recovery times for plant-centered LOSP events. This is an interesting finding and it may be helpful to identify some of the possible explanations in the executive summary such as the interpretive nature of some of the data and the fact that operator actions have become more deliberate in recent years. Comment (from Westinghouse Owners Group): Table C-6: The recovery time for a given loss of offsite power is more strongly related to the nature of the failure than it is to j the particular arbitrary grouping by switchyard characteristic used in NUREG-1032. As an example, Table C-6 assigns individual plants to switchyard categories from NUREG-1032 that are correlated to offsite power recovery times. The Comed plants have been correctly categorized according to the NUREG-1032 categories, however, it is questionable whether that categorization really means anything. This categorization puts Comed plants in the worst categories, with the longest average AC power recovery . times. Yet, the Comed switchyards have characteristics that promote rapid recovery of offsite power. All Comed switchyards have six or more transmission lines, leaving the site in at least two or more different directions. In the 13 categorization, the plants get penalized, in part, because these numerous lines come together in one switchyard. But the switchyard arrangement is such that independence and diversity is maintained. The NUREG-1032 categorization does not reflect these switchyard design strengths. Furthermore, the NUREG-1032 categorization assigns shortncovery times only when an automatic switch to an alternative source of offsite power is provided. This overestimates the benefits of r.;.) automatic circuit. Manual realignment is accomplished via simple control room actions. Those actions can be completed in an amount of time that is small compared to the differences claimed in recovery times for the second and third switchyard groupings, compared to the first. Therefore, it is recommended that this NUREG-1032 grouping for offsite power recovery time not be given emphasis in the INEEL report. I Comment (from Westinghouse Owners Group): Section 3.1.3, first paragraph, second sentence: This sentence states that all reported recovery times were used. The last paragraph on page 14 discusses changes to certain recovery times. These changes should be mentioned in the first paragraph of section 3.1.3. ( Comment (from Westinghouse Owners Group): Section 3.4.1, last paragraph: Are the i' five cases discussed the only cases in which the recovery times were changed? If not, it i is recommended that a discussion of all such cases be included (either in this section or j in one of the appendices). j Comment (from Westinghouse Owners Group): Section 3.4.4, third paragraph: Conceming "The precaution presented in section 3.4.1.2 applies here as well. The i recovery times have not been systematically reviewed by a second engineer." The 2

report does not contain section 3.4.1.2, and a similar precaution was not found in section 3.4.1. It is recommended that a second review be performed or the results not be published. It will be difficult for readers of this report to know how to use the analysis results as presented. Resolution: The recovery times for all LOSP events were reviewed. The recovery times for some events were revised to reflect the time that a licensee could have restored power to a safety bus (based on engineering judgement) instead of the actual restoration time as documented in the licensee event report (LER). The revised recovery time data results showed no increasing effects as seen in the data in the draft report. The reanalysis of recovery times showed no statistically significant relation to design group. 5. Comment (from NRR): Page 19, Figure 3-10: The figure depicts the log (recovery time) of all plant centered LOSP events. The discussion on this page states that, "The results displayed on Figure 3-10 are attributable entirely to the shutdown events." This statement may not be clear in that it may imply only shutdown events were used for data in the figure. This is probably not the case, and it may be more clear if there were figures for shutdown events, operating events, and then the totals. Resolution: Section 3.5.5 was revised to include separate log (recovery time) plots for plant-centered trip events and plant-centered shutdown events. 6. Comment (from NRR): In a few places in the report, notably Pages xiv and 29, a statement is made regarding the complete loss of AC power events which have occurred. This statement in effect says that "the duration of each event was small and the need for accident mitigation systems powered from emergency AC power was not present in the events." Since " accident mitigation" systems are not required to be functioning following station blackout (SBO), the wording should be changed to indicate that none of the SBO coping systems were needed to mitigate the event (or something similar). Resolution: The executive summary, and Sections 4.4 and 5 were revised to clarify station blackout events as suggested by this comment. 7. Comment (from NRR): (a) One LOSP event (LER 334-94-005) identified in AEOD report C97-01 is not identified in subject report. (b) Please clarify the roles and conclusions of the two reports with respect to the analysis of LOSP events. Resolution: (a) The LER event described a dual unit reactor trip caused by the failure on the Unit 1 main unit transformer. One Unit 1 emergency diesel generator (EDG) started but was not required to load, due to a fast bus transfer to an available offsite power source. The Unit 1 transformer failure initiated a voltage disturbance in the switchyard which caused an inadvertent protective relay actuation on the line supplying one Unit 2 station service transformer that supplies one 4kV safety bus. One Unit 2 EDG started and loaded onto its bus due to the loss of voltage to the safety bus following the loss of the station service transformer. Since the start and load of all EDGs in neither plant occurred, neither of these events was classified as a loss of offsite power. 3

1 (b) The objective of the AEOD report C97-01, titled Grid Performance Factors, was to examine the factors affecting the reliability of offsite power by collecting current operating experience and providing background information on grid performance and the impact of grid performance on nuclear power plants. The grid report concluded j (Section 8.3) that on the whole, the grid is stable and reliable but due to uncertainties introduced by restructuring of the electric industry indicates the need to monitor grid conditions on a regular basis. The LOSP report supports the conclusion that the grid is stable and reliable based on an observed decrease in grid-related LOSP events found in the 1980-1996 operating experience by a factor of about 10 as compared to the j frequency found in NUREG-1032, Evaluation of Station Blackout Accidents at Nuclear Power Plants. NUREG-1032 was based on 1969-1985 operating experience. 8. Comment (from NRR): It should be noted that the loss of offsite power events as defined in the subject report do not correspond to the definition of LOSP which is used for regulatory purposes. For example, those plants which provide power to their safety buses from the main turbine generator could lose both offsite sources in an event. This event would not be characterized as an LOSP in the subject report. Resolution: The definition of a LOSP event used in this report corresponds to the LOSP initiators typically used in probabilistic risk assessments (PRAs). The above example refers to those few plant alignments where all power from offsite sources could be lost, no reactor trip would be generated, and the main turbine generator would have carried all plant loads. These events, which are applicable only for a few plants that have the capability to bypass 100% of the steam load to the condenser and the capability for the main generator to supply house loads during operations, would not be classified as a LOSP event in this study because the EDGs would not have needed to start and load onto the safety buses. 9. Comment (from RES): The analysis of reported failures shows that equipment failures caused the majority (60%) of plant-centered LOSP events at power. From our point of view, the following information, if included, will be extremely useful: (a) types of equipment failed; (b) root-cause of the equipment failure (s); (c) time-trending of equipment failures with vintage of plants; (d) identification of affected systems; and data on potential for common-cause failures. Resolution: (a) Section 4.2.1 was revised to include a breakdown of the equipment failures, based on the type of equipment failed. (b) The source data (LERs) usually do not contain sufficient detail to determine root cause, so this additional analysis was not performed. (c) Equipment age is not available through the data source used for this study. Therefore, time trending as a function of equiprnent age was not performed. (d) The identification of affected systems and collection of data on common cause failures were beyond the original scope of this study. The NRC common-cause failure (CCF) database contains some information of common-cause failure of circuit breakers which may be more informative. The CCF database is described in NUREG/CR-6268, Common-Cause Failure Database and Analysis System, Volumes 1 through 4. 10. Comment (from INPO): There are several places in the study where the authors state that some nuclear power plants are designed to withstand a complete loss of offsite power, defined as the simultaneous loss of power to all safety buses, while continuing to operate using emergency power. We believe the study should point out that this condition may not be permitted by the plant's technical specifications. This would 4

remove any suggestion that a facility would actually continue operations under those conditions. Resolution: Section 2.2 was revised to clarify that in some cases plants actually are designed to (and do) continue operating under LOSP conditions as permitted by technical specifications. 11. Comment (from NRR): Page 7, Table 3.3: The median time to recovery for plant centered LOSP events is given as 24 minutes, while on page 8 (Figure 3.4) it is slightly different (22.12 minutes). Resolution: The number presented in the table is the median of the data; the value from the figure is the median of the fitted lognormal distribution. The reference to the median value in Figure 3.4 was deleted to eliminate any confusion. 12. Comment (from NRR): In Tables 3.8,3.9, and 3.10 the median recovery time for the present study is compared to the NUREG-1032 study. Since the mean recovery time is somewhat larger than the median recovery time, we recommend the inclusion of comparisons of mean recovery times in these tables. Resolution: The mean of a highly skewed data set is strongly influenced by a small number of events with large recovery times. NUREG-1032 did not include means, only medians. Therefore, a comparison was not made. 13. Comment (from NRR): (a) The study indicated that the number of grid events was sparse. We noted, however, that other studies had identified grid LOSP events during the time period in the study. We recommend a review of the following events for possible inclusion into the grid LOSP analysis: Calvert Cliffs (07/23/87); indian Point 3 (07/12/84); Peach Bottom (07/10/87); Pilgrim (11/19/86); Susquehanna (07/15/84). (b) Also, NUREG/CR-6538, " Evaluation of LOCA With Delayed LOOP and LOOP With Delayed LOCA Accident Scenarios," notes that one of the events listed on page 4-4 involved a loss of the grid. None of those events are listed in the AEOD updated LOSP study. Resolution: (a) Of the events listed in the comment, two were included in the LOSP study, although not as grid events (Calvert Cliffs,07/23/87 (plant-centered) and Pilgrim, i 11/19/96 (weather)). The LERs of the other three events reported that the safety buses fast-transferred to an alternate transformer, thus there was no loss of power to all safety buses. (b) Of the ten events listed in NUREG/CR-6538, six were included in the LOSP l study. The remaining four events were classified in the original data source (AEOD/E-93-02, Engineering Evaluation Report - Evaluation of Loss of Offsite Power Due to Plant-Centered Events) as partiallosses of offsite power. 14. Comment (from NRR): The AEOD study can be applied to our evaluation of events. l The updated frequencies and non-recovery curves can be modeled in the Accident l Sequence Precursor (ASP) models with the exception of the grid LOSP events. This is i due to the sparse number of grid LOSP events, which results in a discrete distribution and not in a best-fit curve. We have a need to model the current grid LOSP frequency and non-recovery curve in the ASP mode!s. Therefore, if within the scope of the study, we recommend further work to generate an updated best-fit grid LOSP non-recovery curve which can be used with the ASP models. l I 5

Resolution: The report contains a grid-related LOSP non-recovery curve. It should be pointed out that the number of events is very sparse and the uncertainty in this curve is large. 15. Comment (from NRR): (a) The use of the Empir; cal Bayes distribution of population variability of frequencies as a prior distribution may not be appropriate. First, the distribution already includes the data for the plant in question, so updating using plant-specific data double counts that data. Second, the posterior is being influenced by plant data that are irrelevant because conditions are totally different. This is particularly true for weather-related events. (b) When a difference between plants is known, the data that's most relevant should be used, for example, grouping the data from plants that have similar weather conditions. Resolution: (a) In the ideal theoretical situation, it is desirable to use the data only once. However, in practical situations both the Empirical Bayes and Hierarchical Bayesian methods use the data twice. The data are used to estimate the prior distribution and then in the update of the prior to obtain a plant-specific posterior distribution. The weight given to the data by the prior is usually very small. (b) Grouping plants with similar weather conditions was considered. It was not used because of the uncertainty in defining such groups. In addition, the extra complexity of statistically determining groups was not justified with so few events. 16. Comment (from Westinghouse Owners Group): Section 4.1.3, first paragraph, first sentence: The logic of the sentence appears to be inaccurate. As stated in the rest of the paragraph, the reason for plants being more vulnerable to severe-weather LOSP events at shutdown is because of the plant configurations at shutdown, not because the frequency of LOSP is higher. The LOSP frequency is higher because of the plant shutdown configurations. Another contributing factor to the severe-weather shutdown LOSP frequency may be the fact that some plants shut down at the threat of severe weather. Comment (from NRR): Section 4.1.3: one of the reasons why there are more severe weather events at shutdown might be related to the fact that some coastal plants shut down when there is a hurricane watch or warning. Resolution: We agree with the statements made in these two comments. In fact, the four hurricane-related LOSP events found in the 1980-1996 operating experience in which the plants shut down in response to approaching hurricanes were classified as reactor trip initiators, since these plants did not modify their switchyard configuration after plant shutdown. As the result of an re-evaluation of the analysis of the severe-weather LOSP data, the severe-weather shutdown events were combined with the at power events for the reasons discussed in the modified Section 4.1.3 in the final report. Furthermore, the first paragraph in the section was deleted. 17. Comment (from NRR): Section 5: A PRA is used to evaluate the frequency of core damage in an average sense, not as a function of which calendar year it is. Given that, it is not clear that there is a need for creating wider interval estimates as discussed in the fourth bullet. 6

t l l \\ Resolution: In the draft report, the fourth bullet stated that an " unexplained variance was seen between years, and this variance was used to obtain an interval estimate for the frequency, wider than a simple confidence interval." This variation associated with the trend was reevaluated in the final report. The annual event counts show a variation about the mean that was larger than expected. This variation was attributed, in part, to dependence between units at sites. It was used in estimating interval estimates. 18. Comment (from NRR): NUREG-1032, Evaluation of Station Blackout Accidents at Nuclear Power Plants, noted that "Most of the event descriptions in the LERs and in other documentation in the NRC files did not contain sufficient information to provide an accurate database for estimating frequencies and durations of losses of offsite power." in order to compensate for this information deficiency the authors of NUREG-1032 contacted utility engineers, the Institute of Electrical and Electronics Engineers and the Electric Power Research inctitute in order to get a better description of the causes and sequences of events, and the times and methods of restoring offsite pcwer. Given the need for accurate data regarding LOSP and recovery times as discussed in the draft report as well as the increased importance of LOSP events, should not the subject report make recommendations to change the licensee reporting requirements in order to properly characterize events from an engineering and risk perspective? Response: AEOD will work with NRR to identify potential improvements in reporting of these events in the upcoming demaking to revise LER (10 CFR 50.73) reporting requirements. 19. Comment (from Region 1): The new mnt frequencies for LOSP events from the final report should be used to update the SP/ K model basic event data. Response: It is the practice to update the ASP model data with results of completed AEOD reliability studies. 20. Comment (from Region 1): Section 5, the report conclusions, provides interesting insights, which may be useful to NRC resident inspectors in planning their routine in-plant obsentations. It should be promulgated outside of the final report. Response: AEOD agrees with this comment. AEOD will continue to work closely with NRR and the senior reactor analysts (SRAs) to integrate engineering insights from AEOD reliability studies in routine plant inspections. 21. Comment (from Region 1): The individual plant data from the report should be correlated against the Station Blackout Rule Site Characterizations to validate the licensee's assumptions. Response: This effort was beyond the original scope of this study, and due to resource limitations, this assessment could not be performed. 22. Comment (from NRR): The study appears to have gathered data on multi-unit site LOSP events. If within the scope of the study, a comparison of current multi-unit LOSP frequencies to those used in past PRA studies would provide further useful insights. 7

Resolution: This effort was beyond the original scope of this study, and due to resource limitations, this analysis could not be performed. The LOSP database can be used to obtain this information. 23. Comment (from Westinghouse Owncrs Group): The report states that the data covers the years from 1980-1996, inclusive. NUREG-1032 was published in 1988 and includes data from 1968 to 1985. An explanation should be given as to why 1980 was selected as the starting point in the report. Given the changes in the nuclear industry over the past 15 years, was consid6 ration given to using a shorter time interval (e.g., 5 to 10 years), then subsequent updates would be based on data from a rolling 5 to 10 year interval? Resolution: The 1980 start date of the study coincides with the beginning year of the NRC> Sequence Coding Search System (SCSS) database. One purpose of the update LOSP study was to identify any noticeable changes in LOSP frequencies after the implementation of the station blackout rule. Therefore,1980 was a logical start date. The report provides raw data so that data from any time interval of interost can be used ' in future analysis. 24. Comment (from Westinghouse Owners Group): The report distinguishes between LOSP events and LOSP initiating events, where the distinction is based on whether the plant trips or continues operating at power. However, as the report notes in Section 2.2 (first sentence), some LOSP events may not cause a plant trip, but he plant may be required to perform controlled shut down. (a) It is not clear how these everds c. ave been categorized. (b) In addition, further justification should be provided for excluding events which did not cause a trip at one plant but may have caused a trip at another and for eliminating events from the at-power frequency calculation which occurred at shutdown and would have caused a trip at power. Resolution: (a) A LOSP event that does not cause a reactor trip is not classified as an initiating event. (b) Causes of LOSP events are very plant-specific. Most LOSP events originate in the plant switchyard. Due to the plant-specific nature of switchyard designs, a failed component in the switchyard that results in an LOSP at one plant may not cause an LOSP in another plant. Therefore, this study used only actual operating experience in the analysis. Section 2.2 was revised to add clarification on which events were excluded in the analysis. 25. Comment (from Westinghouse Owners Group): Section 3.1.1 mentions that four of the plant-centered events were not initiating events because the trip preceded and caused the LOSP. It would be useful to have an estimate of the frequency of LOSP resulting from reactor trips. Resolution: This effort was beyond the original scope of this study, and due to resource limitations, this analysis could not be performed. Another AEOD study on the rates of initiating events that is nearing completion does include information on the frequency of LOSP events including those occurring after a reactor trip. 26. Comment (from Westinghouse Owners Group): Table 3.3 states that there are 117 events with reported recovery times (listed by site). Our review counted 119 events, after eliminating four same-date, multiple-unit events (Salem 6/2/84, Peach Bottom 7/29/88, Calvert Cliffs 7/23/87, and Sequoyah 12/31/92). We did not include the 8 L

} O; l Arkassas 6/24/80 event, since " unknown" was listed as the recovery time for the second j unit. It is possible that additional same-date, multiple-unit events could exist, but it is difficult te verify this since the events are not sorted by event date. Comment (from Westinghouse Owners Group): Figure 3.3 shows a total of 116 events i whereas Table 3.3 states there are 117 events. l-Resolution: The data was reviewed and corrected. Table 3.3 and Figure 3.3 were revised to reflect the actual number of site events with reported recovery times. l 27. Comment (from Westinghouse Owners Group): Section 3.3, top of page 10: Lightning i is listed as a plant-centered event rather than a severe-weather event. The weather i fronts which cause lightning can indeed be widespread making them candidates for i severe-weather events. Based on pages 3-5 and 3-6 of NUREG-1032, it seems that the i distinction between plant-centered and severe-weather events is the difference in j recovery duration. If that is part of the basis for the categorization of the lightning events,-it is recommended that it be mentioned in Section 3.3. i j Resolution: The distinction between plant-centered and severe-weather events in j NUREG-1032, " Evaluation of station Blackout Accidents at Nuclear Power Plants," was i not based on recovery duration. The categorization of lightning strikes as plant-1 centered events in the updated LOSP report was chosen to duplicate that used in NUREG-1032 in order to make the results compatible for comparisons. 4 Comment (from Westinghouse Owners Group): The total number of severe-weather I events in Tables 3.5 and 3.6 is 22. Table 3.7 covers 17 events. Where are the other 5 events, with regard to recovery? I ResoMion: Tables 3.5 and 3.6 were combined when the data for severe-weather LOSP emnts during shutdown was combined with power operations (now Table 3.5 in the final repmt). The second Table 3.7 in draft report (now Table 3.6) includes only sustained severe-weather events (greater than two minutes in duration). - Table 3.5 was j revised to provide a breakdown of severe-weather events. 3 i 29. Comment (from Westinghouse Owners Group): Section 3.4.5, first paragraph: Brief { definitions are provided for 11 and 13 category plants, but not 12. Recommend adding a ~ brief definition of 12 category plants. i Resolution: Section 3.4.5 was revised to include a definitbn of 12 category plants. j i 30. Comment (from Westinghouse Owners Group): Appendix B, Table B-2: The report j. derives plant-specific event occurrence rates for plant-centered events during shutdown, l-but not for operation. In the Executive Summary, it is stated that no statistically significant plant-to-plant variability in LOSP frequency was found for plant-centered initiating events at power. It is recommended that this conclusion be supported in Appendix B by showing plant-specific occurrence rates during operation. Resolution: As described in Appendix A, Methods of Data Analysis, a statistical test is performed to detect plant-to-plant variatAlity. Plant-specific occurrence rates are only provided in the report when a plant-to-plant variation was detected. I =

P Q i 31. Comment (from Westinghouse Owners Group): Table C-1: (a) For completeness, the meaning of " " should be specifically mentioned in the text above the table.. (b) It may I be more informative to sort the data in Tables C-1 through C-3 by the event date rather than the LER number. (c) As presented, some of the LER numbers are out of order. 4 Resolution: (a) Section C-2 was revised to include the meaning of " ". (b) The sorting of data by LER number (or docket number) is convenient for NRC staff to search LOSP events at a particular plant. (c) The entries in Tables C-1 through C-3 are not sorted only by LER number. They are sorted on status (P-power operations, S-shutdown, and T-reactor trip) and then LER number, ? i 32. Comment (from Westinghouse Owners Group): Appendix C: A comparison was made 1 of a few of the events documented in Appendix C with the events documented in EPRI TR-106306, " Losses of Off-site Power at U.S. Nuclear Power Plants - Through 1995," and the following differences were noted: a. Oyster Creek event on 5/18/89 was incbded as an initiating event in Table C-1, !~ however, the EPRI document indicated that the reactor tripped and then the ] LOSP occurred (LER 219/89-015). Quad Cities 2 event on 6/22/82 was included as an initiating event in Table C 1, however, the EPRI document indicated that ] the reactor tripped due to other causes (LER 254/82-012). i i b. The recovery time for the Millstone 2 event on 9/27/85 was documented as 1570 i minutes in Table C3, however, the EPRI document indicated that recovery was possible after 211 minutes. This information was not provided in the LER but was j provided in a letter to the NRC dated 12/31/85. However, the Millstone 1 recovery time documented in the EPRI report may have been overly optimistic because it credited the cross-tie. (LER 245/85-018) c. The Point Beach 1 event on 7/25/85 was attributed to equipment failure in Table C-1, however, the EPRI report attributes it to rain. (LER 266/85-004) d. There are events included in the INEEL report which are not documented in the EPRI report. For example, the Prairie Island i event on 7/18/84 (LER 282/84-004) or the Salem 2 event on 4/11/94 (LER 311/94-007). Resolution: The LER for each event discussed in the comment was reviewed. The results of the review are as follows: (a) The two events were reviewed. They were coded consistently with the rest of the events in the database. Each event was due to the same initiator that resulted in the reactor trip. The EPRI document referenced for the 6/22/82 event said the trip was caused by a feedwater pump trip. Our review revealed that the root cause of the feedwater pump trip was the transient that led to the LOSP. The definitions of initiating event and TRIP

  • were clarified in the report to distinguish between plant events that cause both a reactor trip and LOSP and those events in which the reactor trip and LOSP are independent of each other. (b) The recovery times were revised for Millstone Unit 1 and Unit 2 to 211 and 330 minutes respectively, based on information in the subject letter and engineering judgement. (c) A failed relay in the switchyard caused the Point Beach Unit 1 LOSP. (d) The Prairie Island Unit 1 event was deleted, and the Salem 2 event was classified as a LOSP event with no reactor trip since all EDGs started and loaded onto the safety buses.

10

O 33. Comment (from Westinghouse Ouners Group): Table 3.9: It appears that the dramatic difference between the number of grid-related initiating events in NUREG-1032 and the INEEL document may be due to the difference in time periods covered by the two documents. Additional discussion and justification should be provided to explain the differences. l l Resolution: The time periods covered by the two reports for grid-related LOSP events l are 18 years (1968-1985) in NUREG-1032 and 17 years (1980-1996) in the updated LOSP report. Only two events at Rancho Seco occurred during the 1980-1983 overlap in the studies time period which NUREG-1032 did not include in its database, There l appears to be no significant differences in the data between the two studies that would bias the comparison of the two results. 34. Comment (from NRR): Section 3.3.1, first sentence: This, as written, is ambiguous, since the units of frequency are not specified. Maybe if it were written as the frequency of p' ant-centered initiating events was "... per unit time during power operation rather l than the LOSP frequency per unit time during shutdown," it might be a little clearer. Comment (from NRR): While it is clearer, later on in the document, how the at power frequency has been evaluated (# trips /(critical hours)), it should be noted that this is not i the frequency that is used in PRAs/IPEs, which is # trips /(calendar years). This is becsuse the core damage frequency (CDF) is measured per reactor year, and the j' frequency is reduced by the fact that, for part of the time, the plant may be shut down. Perhaps this should be stated clearly somewhere. Resolution: The definition and the units were reworded for clarification in the final report. ) i j i l l 1 ( l 11 - _,...., _ _ _,. ~. ~

Editorial Comments: The editorial comments listed below were received from reviewers. They were incorporated in the final version of the report, if applicable, because the calculations were redone for the final report. 35. Comment (from NRR): Page xiv, first bullet: In the second sentence "...during operation and during power operation," the first " operation" should be " shutdown" or " shutdown operation," Apparently the word shutdown was inadvertently omitted. 36. Comment (from NRR): Throughout the report references are made to other sections of the report, such as section 3. These " sections" should be capitalized when used in conjunction with a specific section such as Section 3, the same as Table 3, or Appendix 3 are capitalized. 37. Comment (from NRR): Please verify whether Table C9 total numbers are correctly stated in the subject report. 38. Comment (from NRR): Reference (p.30) regarding the AEOD Grid Performance Report (i.e., AEOD C97-01) needs to be updated to reflect the final issuance date of March 20,1997. 39. Comment (from NRR): Table C-6 lists Rancho Seco and Washington Nuclear as sites which were categorized by NUREG-1032. Please verify whether the subject plants were categorized in the subject report in which case they should be designated by a

  • symbol.

40. Comment (from Westinghouse Owners Group): Page xiv, second bullet: Second sentence should state "...no statistically significant difference could be seen between recovery times during operation and during shutdown." 41. Comment (from Westinghouse Owners Group): Section 3.1.1, second paragraph: In the second sentence the word "non-vital" is misspelled. 42. Comment (from Westinghouse Owners Group): Section 3.4.5, third paragraph, fourth sentence: The sentence structure of "The results qualitative of the recovery time.. " should be revised. 43. Comment (from Westinghouse Owners Group): Figure 3.4: Title should read "...during both power operation and shutdown." The current text implies that the survival curves in Figure 3.4 are for power operation only, when they are for both power operation and shutdown. 12 L. ~ _}}

Retrieved from "https://kanterella.com/w/index.php?title=NUREG/CR-5496,_Issues_Final_Rept,Evaluation_of_Loss_of_Offsite_Power_Events_at_Nuclear_Power_Plants_1980-1996,NUREG/CR-5496.Observations_%26_Findings_of_Study_Listed.Without_Encl&oldid=4117830"