ML20023B201

Forwards Summary of Loss of Offsite Power Comments Presented at 821215 Meeting,Compilation of Requests for Addl Info for BNL Re PRA Review & Summary of Comments on Remaining Issues

ML20023B201
Person / Time
Site:	Limerick
Issue date:	12/22/1982
From:	Boyer V PECO ENERGY CO., (FORMERLY PHILADELPHIA ELECTRIC
To:	Eisenhut D Office of Nuclear Reactor Regulation
References
NUDOCS 8212270141
Download: ML20023B201 (23)

Text

-

.a PHILADELPHIA ELECTRIC COMPANY 2301 M ARKET S1 REET P.O'. BOX 8699 PHILADELPHI A. PA.19101

v. S. DO Y E R sa. vicr antsiocar December 22, 1982 NUC LE A R POWE R e r> d 5 z-353 Mr. Darrell G. Eisenhut Division of Licensing Office of Nuclear Reactor Regulation U, S. Nuclear Regulatory Commission Washington, DC 20555

Subject:

Limerick Generating Station - PRA Documentation of Comments and Additional Information Requests Based on 12-15-82 Meeting on BNL Draft Review of the LGS-PRA

Dear Mr. Eisenhut:

At subject meeting between Philadelphia Electric, Brookhaven National Laboratory, and the NRC staff, it was agreed that PE would submit comments and requests for additional information relative to BNL's review of the LGS-PRA. It was also agreed that PECO's comments would be provided in two phases in order to expedite information flow. This letter transmits Phase I, which consists of three enclosures:

1) A summary of Loss of Offsite Power comments presented at subject meeting and supportiAg information.

2) A compilation of requests for additional information needed by PECO in order to properly address the BNL draft.

3) A summary of comments on certain issues based on cur present understanding.

Phase II will provide the results of our page by page review of the BNL report. In addition, it will include any information generated as a result of the NRC staff's response to the Phase I question.

We thank you for your cooperation in this matter and the opportunities you have afforded us to interact with the staff and BNL. We believe that this kind of active technical exchange is essential to the successful use of PRA techniques.

,1 8212270141 ~ 821222~ ~ ~ ~ '

Sincerely, l PDR A

ADDCK 05000352 PDR cc: See Attached List . . 7 3 g/

7

[ *p, UNIT 13 STATis NyCLEAR REGULA7 CRY CCMMISSICN i }%1.f't) usmuaien. c. c. acssa 1.

%, kal%/yj

. .?

'E40RA.TJG1 FOR: TI.M CCRPCRA!!CN TRCM: US NRC/ TOC /00CDEC MANAGDEC 3RA.NCE SU3.!ICO : Sp ecial "Docu:::en Handling Requir<aents

1. Please use the folicwing special distri':ution *ist for .he .

a: ached docu:::ent.

!/C $/ W 2 h

• ggg $&k = h gg4 m t Zf JAd/$ - / Hnf ~/

4 pis - t MM/D/A-l 6e fle , /

og/A/&l //ff 5 -/6 a A ff$ ~ f WM OA -I pg . ,

2. The attached dccu=ent requires -he f:ll: wing special consider 2; ions:

Do no send eversi e enclosures to the $3C PDR.

bCnlyoneoversizeenclosurewasreceived-pleasere:r.

for Regulato.~/ File storage.

b Proprietary info: :ation - send Only non-proprie:a_~f portions to the NRC PDR.

Other: (specify)

=c: c.va Files rpt,:xs au: hor :e_ 3:.gna-=,

l-p .

i l

9

--~:, e --,.-o ,. _ , . - .,,,e,, ,_,. , _ ,_ _ _ _ _ _

cc: Judge Lawrence Brenner Judge Richard F. Cole Judge Peter A. Morris Troy B. Conner, Jr. , Esq.

Ann P. Ilodgdon Mr. Frank R. Romano Mr. Robert L. Anthony Mr. Marvin I. Levis Judith A. Dorsey, Esq.

Charles W. Eiliott, Esq.

Mr. Alan J. Nogee Robert W. Adler, Esq.

Mr. Thomas Gerusky Director, Pennsylvania Emergency Management Agency Mr. Steven P. liershey James M. Neill, Esq.

Donald S. Bronstein, Esq.

Mr. Joseph II. White, III Dr. Judith II. Johnsrud Walter W. Cohen, Esq.

Robert J. Sugarman, Esq.

Rodney D. Johnson Atomic Safety and Licensing Appeal Bosrd Atomic Safety and Licensing Board Panel Docket and Service Section

!, i V. S,. BOYER TO D. EISENHUT, 12-22-82 ENCLOSURE 1 l l

LOSS OF OFF-SITE POWER

1. LOSS OF OFF-SITE POWER (DATA)

The data base used to calculate the frequency of loss of off-site power for the Limerick PBA is site specific. These data are '

based on the NAAC rollability region nuclear plants and the PECo fossil plants. The rationale for selecting this data base was to ,

provide site specific intornation comparable to the population and meteorology data bases.

We believe th6 UAAC reliebility region, which includes PECo, l has a network reliability higher than the national average due to  !

a number of reasons. The PJM Interconnection, which is identical l to the MAAC reliability region, is the oldest major i interconnection in the United States dating back to 1927 and is  !

the most experienced. Figure 1 illustrat'es the projected MAAC l

system in 1992, a few years af ter the planned service date of t Limerick #2. The only circuits shown are transmission lines at j 230 Ky or higher voltages. The figure illustrates the ,

compactness of the MAAC system which is only about 330 miles i vide. This compactness with high load density leads to shorter  ;

lines and more circuits per substation. Table 1, which is based  ;

on data in the NEHC 1981 Annual Report, details the compactness  !

of the MAAC system. MAAC is the smallest system with less than '

half the area of the second smallest system. Yet the NAAC Load  ;

is greater than that of three of the eight other systems. Most  ;

unique is the ratio of the NAAC transmission mileage to its area. -

It is two and a half times the national average.

Another important network reliability aspect is the mode of l operation by the PJM Interconnection. The PJM Interconection has a central control office which economically dispatches the PJM l generation without regard to company boundaries but with full regard to network reliability. Semi-annual joint studies by the PJM nenber companies determine all operating restrictions for the

forthcoming period. These studies are up-dated on a weekly basis by the PJM Interconnection office to conform with the currently available equipment. These off-line studies minimize the chances i of operating problems, but the major prevention of troubles is l

l l

7

l 1 .

the real-time control system. Power flows on all 500 KV lines, critical 230 Ky lines, and tie lines are telemetered every fourteen seconds to the PJM control center. If any line reaches eighty percent of its normal rating the PJM operator is alerted by a CRT message. At a one hundred percent of normal rating, an

audible alarm is sounded which must be acknowledged. The chances of flows actually reaching their normal rating are unlikely

, because the PJM Interconnection operates to first contingency limits. That is, no single contingency shall cause a circuit to excede its emergency rating. The real-time power flows are computer analyzed to calculate the effect of about 500 different outages of circuits and generators. If any circuit is projected to reach its normal rating for any outage the operator is e.lerted by a CRT message. At one hundred percent of emergency rating an

! audible alarm is sounded. Trends of flows are calculated so that corrective action vill be initiated before limits are reached.

Corrective action usually entails redispatch of generation. The PJM Interconnection has over 7000 MW of rapid-start combustion turbines and over 2000 Mw of rapid start hydro units. Relief of potential problems can be rapidly accomplished with this generation. The PJM control computer is backed up by an off-line machine which automatically as?unes control for an outage of the on-line nachine. The power supply for the control center is backed-up by both diesel and battery supplies. In addition to the PJM Cottrol Center, the member PJM Cogpanies have their separate control computers.

The transmission and substation f acilities of Limerick station, itself, will provido a minimum risk of loss of off-site power. Figure 2 shows the conn actions of Li. arick to the

.network . There are two 16.5 mile Limerick to Whitpain end a 55.8 mile Limerick to reach Botton 500 KV lines. There are two high capacity 1300 MVA 230 Kv lines, each approximately 23 miles long:

one Limerick to North Wales and the other Limerick to Plymouth Heeting. All circuits are on separate towers. The Limerick substations will use the breaker and a half design. This design allows required periodic breaker maintenance to be performed without removing a circrit and its protection from service. The two start up (S .U . ) supplies, one on the tertiary of #4 500/230 Ky bus-tie transformer and the other #10 transformer on the 230 KY bus, are isolated by an extra breaker. A failure of either transformer with a coincident breaker failure vill not remove the other S.U. from service. The reliability of a station's off-site supply is a function of the number of supply lines. The reliability of a supply line is a function of its length. The large number, five, of supplies to Limerick and the relatively short line lengths will give Limerick Station an above average reliability of supply of off-a te power.

j l

l*

l The use of historic HAAC nuclear and PECo fossil loss of off-l site power experience to determine the expected frequency at l

Limerick is site specific. Since the PECo fossil plants use the same network, the same central control system, and the same pool of operating personnel; we believe their experience is indeed appropriate for use in determining the expected Linerick risk.

Therefore, in estinating the frequency and restoration times of loss of off-site power initiators, the LGS-PRA used a data base consisting of PECo fossil plant (with three or more transmission lines) experience from January 1973 through July 1980 plus MAAC nuclear plant experience from initial critically through July 1980. As shown in Table 2, in 94.7 plant-years of exposure there were only 4 occurrences of loss of off-site power.

The date of each occurrence along with the number of circuits affected and the duration of the event are given in Table 3. We believe it is important to note that the affected plants were connected to this systen by only 2 or 3 circuits where as Limerick will be connected by 5 circuits.

This data base was recently updated through May 1982 for the BAAC nuclear plant data and through September 1982 for the PECo fossil plant data, 26.6 plan'-years of exposure were added with no additional occurrences.

Since BNL based its analysis of MAAC nuclear plant experience on Scholl's data (reference 8 in Section 4 of the BNL Draf t) , we investigated the differences between Scholl's reported occurrences and those of the LGS-P3A data base. Through contacts with the respective ut.ilities, we obtained background on each of Scholl's reported incidents at Calvert Cliffs and Oyster Creek.

Based on this information, which is summarized in Attachment 1, we believe that the LGS-PRA data base is a correct statement of the MAAC nuclear plant total loss of off-site power experience.

We would, of course, recommend that contact be made with the appropriate companies for clarification of these incidents. The apparent inaccuracies in the HAAC portion of Scholl's data lowers the credibility of the Scholl's national data base and leads us to question the appropriateness of its use in BNL's overall analysis.

In the event of 1)ss of off-site power to Limerick station, PECo has an cbove average ability to rapidly restore service.

Figure 3 shows how PEco could rapidly restore service in the event of an entire network shutdown. The conovingo station has eleven hydro units which can rapidly be restored without any off-site power supply. Once the conowingo units are restored, breakers would be reclosed to bring power to Plymouth Meeting and

to the Buddy Run pumping hydro station where an additional eight l hydro units can rapidly be restarted. PECo has had a continuing concern for rapid system restoration following a shutdown. PECo has a 200 page manual detailing step by step procedures for restoration. These procedures are periodically updated and analyzed in simulation studies for correctness. With the l operation of Limerick station these procedures will be revised to emphasize the restoration of the Limerick supply. Demonstrations by PECo have shown that ConcWingo and Muddy Run units can be started with service restored to Plymouth Heeting in less than twenty five minutes. Limerick'will be only one breaker away from Plymouth Meeting. Theref ore, rstoration of of f-site power at Limerick should occur in less than thirty minutes. Figure 3 illustrates only two primary paths to get service to Limerick.

There are oth9r redundant paths in the event that sone of these circuits would be unavailable. Alao, the PECo combustion turbines which can be started without off-site power provide further assurance that service could be rapidly restored. There are three 15 n= combustion turbines near Limerick which could feed into Limerick through the Cromby 230/69 Ev transformers.

1 B. LOSS OF OFF-SITE POWER fSTATISTICAL METH_0DOLOGY)

The analyses conducted by the LGS-PR A provif e the most reasonable estimate of loss of off-site power (LOO P) because of:

. the method and data base used

. the agreement of the LGS-PRA estisates with the data collected since the analysis

. the poor (f ailing the " reasonableness" test) fit of the BNL estimates to the data collected since the analyses.

Each of these points is dealt with below.

1. LGS-PRA Method and Data Base Loss of offsite power, by definition, depends only upon the power grid, transmission lines, and switchgear involved. These are not unique to nuclear plants but depend upon the nature of the grid, local conditions, company management, and design philosophy, etc. The LGS-PRA based its calculations on the most appropriate data available, i.e., BAAC nuclear plant and PECo fossil plant data. Table 2 shows these data. A total of four l LOOP events occurred in the 94.7 plant years of data available at j the time the LGS-PRA was written. No plant had more than one I LOOP incident. Consequently, it was not appropriate to consider l

_== _ _ = = _ . --- -

~ -

that each plant had its own LOOP parameter. In f a ct , a more I homogeneous body of data would be hard to imagine. Consequently, l

t a single estimate of the LOOP parameter was made by conservativley adding one incident to the observed four and l

dividing it by 94.7 plant years of exposure. The result, .0528 l failures per year provides a more reasonable estimate than would  !

be obtained if plant variability were considered. As will be I shown below, this result is quite consistant with the f act that in the subsequence 26.6 plant-years of exposure there were no additional incidents.

In reviewing the approach above one ought to keep in mind that what is desired is a specific forecast for the Limerick plant. All pertinent information should be used in that estimate. Estimates based upon national average data are at f ault because they ignore the unusual reliability associated with the MAAC and PECo experience.

2. Acreement of LGS-PRA Estimates With Subsequent Data Both the LGS-PRA and BNL agree that LOOP incidents follow a poisson distribution, i.e., the probability of x incidents, p (x) ,

equals

-n1 X C fD

>( ! Ob where "e" is the base of the natural logarithm systen and "a" is the expected number of occurrences. In the 26.6 plant years since the LGS-PRA there have been no incidents. According to the LGS-PRA estimate of 0.0528 events per plant year 1.4 incidents could have been expected during 26.6 plant years. Using Formula 1 above we find the probability of observing 0 events when 1.4 are expected, is 0.2455 a very reasonable result. If the rate is reestimated based upon the entire 121.3 plant years of data, one obtains 5/121.3 = .0412 or an expected 1.097 events in 26.6 years. The probability (from Formula 1) of observing 0 events when 1.097 are expected, is 0.3341, again a very reasonable result.

3. Unreasonableness of BNL Estimates When the national Lverage value of 0.1221 (EPRI) is used, the probability falls to 0.0013, a very unlikely and unreaconable result. If the value of 0.29, which BNL obtains from Scholl's data is used, the result is even more unreasonable, i.e., the probability of 0 occurrences in 26.6 years is 0.0004. The BNL

estimates do not pass a " reasonableness" test while the LGS-PRA estimat'es do. Incidently, the maximum likelihood estimatn of incidents per plant year given the 26.6 years with no incident is about 0.025 incident per year. Thus the LGS-PRA estimate of 0.0528 is clearly conservative.

The causes of the BNL unreasonable estimates are difficult to determine without the computer codes, data, and preprocessing procedures used by BNL. The use of a single prior for all initiators, the use of unimodal distributions where bimodal may be appropriate, truncation problems, or the attempt to provide a general procedure for any plant rather than a specific estimate for Limerick may be causes. Each of these possibilities needs investigation.

l l

l

e

• C. CONCLU SION S For the reasons detailed above, both the data base and methodology of the LGS-PRA are a sound basis for estimates of loss of offcite power initiator frequencies. In summary, the appropriateness of the LGS-PRA approach is clear for the following basic reasons:

1. Site or regional characteristics (similar to population and meteorology) are recognized.

2. The LGS-PRA estimate of loss of offsite power initiator frequency is consistent with subsequent data collection.

3. The results are conservative since the Limerick transmission system is more redundant than the transmission systems of the other plants in the data base.

I

..g -

N'

!. i E

li l-y

[!

d 4 g, M.i lll!!l ll-

.g . ei ;;

c-II s

s.

?

-u-n resp

! E !! .

3 !

• ji*E=d 4 /-,

F wi! il,b( W l i e

'M. m i

ir.i.s>.a t c3 i .

= x 3m i

.' 1  !!!! i I'k j il 7

.=  !! -

l i lI

!.\'lflesifl.sjT.i,1N1(%.

' ~

-i p, -

~. ;y

[ [

1; .

. ' , ~

,/ I:

I I

! 4$ ,l!!sg[)j'j.'&

il*

\- 'I

. " ' ~~~ ,

,,, ;b, I .-

g [n ;; 7l ,ri, liss g 9,,.t -

'N j'A' r,, 's  :-  ? -

s i

i. ir, - '/ j /l tr'"}g 1:y? ,.

it -

I!

I8

'3 . '* i '~l g ,

1 ..

~

, igit.t-

!<ii,d!i

' W E

p- ,/,' l  :: l ) s t

.x ff

[

r' ,gegl jils j '~

,lli \t) v

.,, il ,>} '~ )-- y l u$

<! J , , x ri <

l I,; I

/

ki ,

.sI Ja ' \.ys ,ll' i I i

f I s _;,%=- 4l; !!! l, , ((s , ;ll --

l y

N s .

I,n g[ r_ s

1. .

, p d'.  ; \d'f , .. I!!

ll s

-: , ... /1 r 'g, H.p I / T- il,1  : ,! - . 6 t -- i

. s c' y',/ Jp  !! l '7 % ..,j l i ,g, l i. i i i - .,

{Is 'e l =

/  ; . 3 I  !

d\ ,n I

Ix _ _

/ g

- =

'.e-- we -

i il TABLE 1 1981 NERC SYSTEMS A) (B)

POPULATION AREA PECK LOAD TRANSMISSION -

6 (10 ) (MILES ) (Mw) (MILES) B/A 1

MAAC 21.8 48,700 34,575 6,073 .125 ERCOT 10 195,000 32,866 5,089 .026 *

, MAIN 18 170,000 33,730 5,121 .030 MAPP (US) 13.6 420,000 18,200 12,000 .029 SPP 23 500,000 45,007 NA (-)

WSCC 44 1,800,000 83,966 93,553 .052 NPCC (US) 27.3 112,527 39,493 5,447 .048 SERC 37.7 345,636 95,065 22,305 .065 ECAR 36 192,000 64,585 14,474 .075 (less SPP) 3,283,863 164,062 .05

[ NA = NOT READILY AVAILABLE

2-500 KV LINES NORTH ELROY WALES .

!L".!!!ECE u

II  :

- I

[] [3 W HIT-% VN V% Y g C3 .

$ 43-230 KV LINES -

O #2 unit

[] [3 E.3 "

i g

1. 4 y g- V 4-230 KV LINES

1. 1 UNIT ] [] [] 8M's 500 KV O >[ ,_

0 E CROMBY - 230 KV

>E l El 9 ----

E I LIMERICK CONNECTIONS g W #10 S.U. TO SYSTEM NETWORK PEACH BOTTOM 3-500 KV LINES Fze,vax Z .

LOSS OF OFF-SITE POWER EXPERIENCE EXPOSURE (PLANT-YEARS?

PJM NUCLEAR OCCURRENCES LGS /PRA UPDATED CALVERT CLIFFS 1 5.8 7.6 OYSTER CREEK 1 11.3 13.1 -

PEACH BOTTOM O 6.9 87 '

SALEM O 3.7 5.5 THREE MILEI O 6.2 8.0 -

PECO FOSSIL CHESTER O 7.6 9.8 i CROMBY O 7.6 9.8 I CROYDON 1 7.6 9.8 DELAWARE O 7.6 9.8 EDDYSTONE O 7.6 9.8 SCHUYLKILL 0 7.6 9.8 SOUTHWARK 1 7.6 9.8~

RICHMOND 0 7.6 9.8 TOTAL 4 94.7 121.3

. Kn 7 AS LE. 2..

LIMERICK SERVICE RESTORATION

1. 2 UNIT g/ .

[] [ ] WHITPAIN

- LIMERICK i i1 l1 l #4 w []

q#1 UNIT V

[] [] , h( ) PLYMOUTH -

gj g -i MEETING l

~3 3 l

@ #10 S.u.

! PEACH BOTTOM CC Il BRADFORD

= :m IN I3

[3 "'~ ~

l yy CONOWINGO

! g3 [] HYDRO mem SOO KV O MUDDY RUN [] STjTION

- 230 KV "

r g j PUMPED HYDRO STATION I g

I A l g

-3 4-110 ~MW [ l 4-110 MW UNI S 4-65 MW UNITS UNITS UNITS

! NOTE: ONLY CIRCUITS AND BREAKERS NECESSARY TO

RESTORE SERVICE ARE SHOWN Freurs 3.

e

• TABLE 3 LOSS OF OFF-SITE POWER OCCURRENCES LGS /PRA DATA BASE I

DURATION PLANT DATE CIRCUITS AFFECTED (MINUTES)

CROYDON 3-3-77 3 out of 3 2 SOUTHWARK 10-27-73 3 out of 3 48 CALVERT CLIFFS 4-13-78 2 out of 2 330 f

OYSTER CREEK 9-8-73 2 out of 2 90 J

l i

l l '

J L

D ATTACHMENT,1 l REVI2W OF SCHOLL *S El AC NUCLEAR PLANT LOSS OF OFF-SITE POWER DATA BASE - BEVISIONS 3 AND 4 The LGS-PRA and EPkI data bases are in agreement with the following two incidents of total loss of offsite power reported by Scholl:

Calvert, Cliffs 4-13-78 Both 500 KV buses were deenergized as a result of abnormal operations of various 500 KV circuit breakers. The abnormal operations were caused by the presence of AC voltage in the DC breaker control circuit. -

Oyster Creek 9-8-73 While attempting to transfer station loads to start up transformers during a plant shutdown, both start up transformers were deenergized as a result of inadvertent opening of associated breakers. These incorrect operations were due to incorrect tap settings on the current transformers of differential relays on both start up transformers.

In addition to the two events described above, Scholl includes four occurrences which in our estination should not be included for the reasons cited:

?

Culvert Cliffs 12-20-73 Although we were unable to determine whether or not this was an actual total loss of off-site Power event, it is inappropriate to include in either the LGS-PRA or the BNL assessment since it occurred prior to foel loading and thus is outside the exposure period as defined by both the LGS-PRA and BNL analyses.

4-11-78 Atter a tripping of Unit No. 1, erroneous operations

of various 500 KV circuit breakers resulted in the deenergization '

of one of the two 500 KV buses. This resulted in the loss of control power to and tripping of a unit No. 2 steam generator j feed pump. Unit No. 2 was then manually tripped because of rapidly decreasing steam generator water levels. At ac time i'

during this transient was there a complete loss of offsite power to either unit. This event should be reclassified as a partial loss of offsite power, i

j

8-19-82 During normal operation of both units, while conducting routine maintenance on a 500 KV circuit breaker, a malfunction of a disconnect occurred. Repair of the disconnect i required that one of the two 500 Ky buses be deenergized.

Throughout this planned bus outage one of the two offsite power sources remained operable. Both units remained in operation throughout the incident. This event should be deleted from the data base.

Oyster Creek 9-10-73 According to personnel at Oyster Creek there was no loss of off-site power event recorded on this date.

However, a planned loss of power test was logged by the operato rs . This event should be deleted from the data base.

G I

.r'

'R .

.a V. S. BOYER TO D. EISENHUT, 12-22-82 ENCLOSURE 2 BEOUESTS FOR ADDITIONAL INFORMATION A. INITIATOR FREQUENCY In order to perform a detailed review of BNL's estimates of initiator frequencies, we need the following additional information.

1. A copy of BNL's 2-stage Bayesian Computer Program. We request that delivery of the program not be held up by BNL's desire to provided additional documentation. We would like a computer readable form of transmittal but if this proves too time consuming to prepare, a listing of the program would suffice in the short ters.

2. The data used in estimating f requencies for each class of intiator. This should include copies of the original source plus a list of all modifications, if any, made by BNL in preparing the input to their program. A copy of actual input data is also requested.

3. A copy of Reference 4-4 of the BNL draft in its current form.

4. Copies of senstivity studies performed by BNL as outlined in item B.3 of Enclosure 1.

B. SYSTEM DEPENDENCIES AND SYSTEM UNAYAILABILITIES

1. HPCI reliability estimation is important in the modifications made by BNL to the event tree reguantification. In order to assess the reasonableness of the increase in HPCI unreliability used by BNL, y

please supply the data, model, or subjective assumptions used in the reguantification of the HPCI fault tree.

g This additional justification is necessary to substantiate the reasonableness of the BNL assumptions.

w..nn.--~-- - . - ____ _ _ _____ _

~< 0 7 . .

2. Manual depressurization quantification is another key parameter which was modified by BNL and which has a significant impact on the increase in core melt frequency as calculated by BNL. From the draft report the choice of the human error appears arbitrary. Please identify the data and/or human error model used in the quantification, to explicitly identify the basis for the subjective judgement.

3. ADS Inhibit includes a human error. BNL should supply similar information to that requested in item 2 above.

BNL should provide any insight or modeling considerations ir their evaluation which would bear upon the contribution of the conditional probability that ADS ,

inhibit would be required. l

- \

4. BNL should provide sufficient information to confira j that their calcualtion of TFQW is correct. j specifically, the cutset and/or cutsets used to make up the value of 4.5x10SX*A0.SI'B7,/Rx-yr.

5. Additional discuss 3on of the specific WSW dependencies h in class II accident sequencesis required. Failure I modes of specific equipment should be identifiedby accident sequence.

6. For each of the four ATWS events please identify the SLC initiating signal or signals assumed to be present.

7. For recovery of W (PCS) did the evaluation of human error consider that the operator was following normal plant shutdown procedures during a relatively low stress time period? Please provide explanation and basis for assumption made.

8. What was the basic for determining that human error in both the "Q" and "W" functions of the "Q-Wa se3 vnces nre highly correlated?

9. For the joint failure probability of the Q-W sequence please provide information on the methods used for evaluation of system failures. Please explain differences between event initiated and transient initiated syste= failures as well as the type of equipment for which repair was assumed in the analysis.

10. What basis was used for not evaluating the System Level Fault Trees provided in an explicit fashion for l

w-.--n.~-~....-...-.-.....-..-. . . . . . - . . . . . . . . . . . . . .

r j

. . l 1

l I

derivation of the time dependent total loss of D.C.

power? I C. SOURCE TERMS AND CONT AINMENT EVENT TRIES

1. Please provide the basis for the use of the unaudified OXRE source ters for the events classified by the BNL modified containment event trees as steam explosions.

2. Please provide the bases for the assumed probability of location of primary containment overpressure failure.

D. CONSEQUENCES ANALYSIS 1

! 1. Please provide the basis and rational for selection of evacuation parameters and models used in the draf t. It appears iron the draf t that the parameters and models used were for convenience only. See comments in Enclosure 3 iten D.1.

l t

1

..s .. , ..... ~ .... ~ ..- . w - - .. . . . _ . .- -

l V. S. BOYER TO D. EISENHUT, 12-22-82 ENCLOSURE 3

SUMMARY

COMMENTS j A. 1NITIATOR FREQUENCY

1. Other than the presentation material from the 12/15/82 meeting which focused on LOOP (See Enclosure 1) and the inappropriate use by BNL of first year data to characterize the expected performance over plant life, further understanding cannot be stated without the information requested in Enclosure 2 (See Iten A) .

B. SYSTEMS DEPENDENCY AND SYSTEM UNAVAILABILITIES

1. Differences between the LGS design and the D.C. systen l descibed in NUREG-0666 include the four division design without inter-bus ties and the performance of maintenance on batteries during unit outages.

2. Page A-100 of appendix A of the LGS-PRA should be corrected as follows: item 2., first sentence, delete

" trimmed values" insert "the column based upon the 36 1 LWRS". P (2/1) = 0.42 not 2.34E-1 and P (3/2) = 0.17 not 5.52E-1. Insert the sentence "This data is similar to that found for Plant I and Zion / Cook multiple diesel failure information.d at the end of ites 2. The value for common mode failure of all four diesels is correctly used in the PRA as 1E-3 C. SOURCE TERMS AND CONTAINMENT EVENT TREES

1. The discussion in Appendix D of the LGS-PRA should be clarified in that a D.F. of 10 was applied for particulates as well as iodine for saturated pools in the LGS-PRA calculation.

D. CONSEOUEECE_ ANAL'/ SIS

1. Following receipt of the NRC's May 6, 1980 letter requesting PECo to perform a risk assessment for LGS, considerable discussions and debate took place about the 1

,~...-a......c_.....___.._. . -.._ . _ - - -

e e

• j a . .

appropriate grcundrules for the assessment. These l

[ discussions included the assumptions and calculational  !

bases for the ris:. curves to be provided and culminated )

in a meeting with NEC on Bay 21, 1980.

The May 6 letter contains the following statements about the consequence modeling:

"The staff requests that you conduct a preliminary risk assessment of the Limerick facility utilizing the WASH-1 1400 methodology, but taking into account significant differnaces between the WASH-1400 reference plant and the Limerick facility."

" Meteorological, population, and hydrological data specific to the Limerick site should be used in evaluating the conseguences of selected accidents."

The specific guidance provided by NRC and documented in the May 23, 1980 summary of the May 21 meeting includes:

! "The NRC staff emphasized that the study must utili u

( the same option of the CRAC code as was used in WAiH-1400 for the base case (Limerick at Limerick site using WASH-1400 data) ."

"The staff directed PEco was to use 1970 population data in order to provide a valid comparison with WASH-1400, but also requested a similar analysis for the projected midlife population surrounding the facility."

Because of these statements, the consequence analysis, including evacuation modeling performed for the LGS-PRA

, was done in a manner consistent with that for WASH-1400.

This was done so that the results of the LGS-PRA could be validly compared with WASH-1400. It is recognized by PECo that the evacuation modeling of WASH-1400, while representing the state-of-the-art in 1973-1975, does not necessarily represent how evacuation would be modeled today. However, any change in the model would invalidate any comparison with WASH-1400 and would require a computer code revision. It should be noted that changes in evacuation modeling from that used for WASH-1400 to that which might today be considered appropriate for LGS would be due to improvements in our understanding of evacuation modeling rather than any specific aspect of the LGS site.

l

e o . .

Contrary to the above, the BNL review of the LGS PRA utilized an evacuation model different from that used in WASH-1400. This greatly and unnecessarily complicates the BNL review. We believe that the discussion of evacuation models and the appropriateness of the model used in the LGS-PRA is beyond the scope of the BNL review since the use of the WASH-1400 model was in effect specified by NBC. PECo has never aale any claims as to the validity of the WASH-1400 evacuation model for application to LGS.

2. Page E-24, Table E.6 of Appendix E should have values of 1.0 and .5 added for cloud and ground shielding f actors for moving populations. These values are those used in the LGS-PRA and were inadvertantly left off the table.

They should be included for corpleteness.

3. Attention is called to the changes made on pages E-11 and E-12 of Appendix E in revision 5 to the LGS-PRA which more clearly describe the methods used in the LGS-PRA.

4. Attention is called to our response to Question E.10 (c) on page Q-146K of Volume 1 of the PRA. This information was also presented at the September 3, 1982 neeting in Bethesda.

i

(

l