ML20127A814
Text
-
r==.7)gy
- w m = ' -
. g r-7-
o m,.
s
~
{
s
.g DISTRIBUTION CENTRAL FILE AEB R/F AUG -0 3 1963 JMitchell.
. Plant File
" MEMORANDUM FOR:
L. G. Hulman, Chief Accident Evaluation, Branch
(
Division of Systems Integration FROH:
docelyn'Mitchell Accident Evaluation Branch D vision of Systems Integration
SUBJECT:
SOURCE TERMS FOR SEVERE ACCIDENTS LIMERICK GENERATING j
STATION, " REQUEST FOR COMMENTS' Enclosed is a report on source terms for severe accidents due to internal events for the Limerick Generating Station. The effort was undertaken to better characterize two aspects of the source terms reported by Brookhaven National Laboratory in NUREG/CR-3028, which is a review of the applicant's Probabilistic Risk Assessment. These two areas are separation of lower probability, more severe release accidents from " bins" of higher probability, less severe release accidents chosen by the applicant; and better characterization of the decontamination factor (DF) of the suppression pool foi selected sequences using a model recently developed by a consultant to the Accident Evaluation Branch, Dr. Arlin Oostma. Dr. Postma's report on developing the model is contained in its entirety in the enclosure as Appendix A.
The effort was undertaken to assess source terms using some aspects of the evolving technology; with the intention that further revisions and improvements could be made prior to study of prevention - mitigation i
features. Since the inclusion of improvements over WASH-1400 methodology must await review and approval by a broad based scientific
. panel, the results presented here should be considered as providing direction only. The improvements include:
investigation of the
'(sequence and time dependent) aerosol particle size distribution for calculation of pool DF; determination of the effectiveness of the standby oas treatment system and the main steam isolation valves and leakage c.ontrol system; an additional review of Classes I and III loss
~
of coolant accidents; and consideration of external event sequences.
- ?
i S'$O$ 156.O v2
......... ~. ~. '-N,....... - ~..... ~~.... ~J l
~
. ~......
..~.--.~..-
..~ ~ ~ ~ ~ ~ ;
............ ~......
-. ~... ~.. - ~
- ~... - - -..
.~..... - -.
~ ~.... ~. -
- - ~ ~ ~ ~
CATEk
.... - ~. - ~. -
.. - ~ ~ ~ ~ ~
- ~ ~ ~ ~ -
e roxw aisno,.o.uncu o 4o OFFICIAL RECORD COPY
- u.s.am nes-4oo-an ;
~ ~ ~ ~.
o....
3
~ dy
.gr L. G. Hulman Coments on the approach used are solicited by August 15, 1983.
Original signed trra Jocelyn Mitchell Accident Evaluation Branch Division of Systems Integration
Enclosure:
As stated cc:
R. Mattson R. Bernaro D. Huller M. Silberberg W. Butler A. Thadani E. Chelliah J. Rosenthal J. Meyer A. Postma U. Pasedag T. Pratt S. Acharya J. Read W. Gammill
.....S..I...:.AE..B. 4.... MA D3hAE D
OFF6CE)
.JM i t..c..h e l l..y...
- ye n
suamuc >
.... s/ //. 83 8
..... /.8 3.......
8 j....
om>
wne ronu ais po.eomscu oao OFFICIAL RECORD COPY usceo.ini-m. m
c s,
c s
~
c Source Terms for Severe Accidents Limerick Generating Station Introduction
~
- A revision of' source: terms for severe accidents caused by internal events at the Limerick Generating Station has been completed. The revision was undertaken to better characterize two aspects of the source terms given in NUREG/CR-3028 (a Brookhaven National Laboratory (BNL) review of the applicant's Probabilistic Risk Assessment (PRA)). The
~
effort began in late February 1983. The two aspects are: separation of lower probability, more severe release accidents from " bins" of higher
. probability, lower release accidents chosen by the applicant; and L
improvement of the suppression pool decontamination factor (DF) for selected sequences using a model developed by Arlin Postma, a consultant to the Accident Evaluation Branch. The salient features of this model are discussed below, and the development of the model.is given in Appendix A.
Since: improvements over WASH-1400 methodology must await review and approval by a broid based scientific panel, the results presented here 1.
should be considered as providing direction only. When the source term methodology is developed, it should be= incorporated into revised source terms for use in future studies.
4-Because of binning and because of changes in nomenclature at BNL and in i
~this review, no attempt should be made to equate the designations in what follows with the nomenclature for sequences developed in WASH-1400.
Binning-is the process of adding. probabilities of occurrence of "like" releases and chosing a set of plume characteristics and releases-to the environment to represent the bin and has been used since WASH-1400.
Its advantage is in cost reduction, since the cost of consequence analyses l
- are roughly proportional to the number of bins or " release categories"
. considered. The disadvantage is that if the plume characteristics and i-release fractions to the environment are not sufficiently alike (that is, sequences are misbinned), incorrect evaluations of risk can result.
The effect can be to artifically raise or lower risk.
h Even though unbinning has been part of this effort, numerous sequences remain binned. For the efforts in source tenn determination to be undertaken later, I recomend that all the individual sequences be taken to the determination of plume characteristics and release fractions to the environment. - These sequences should be, binned only prior to consequence analyses,-and then in a manner appropriate to the use of the analysis because consequence calculations are presently planned for four
- different uses:
to reach conclusions relative to undue risk to the public from individual sequences; to capability of risks with other l
- plants; to capability of risks.with risks of other types of accidents; and to the' efficacy of suggested prevention / mitigation features. That is, binning suitable for DES use may not be appropirate for either the SER or a prevention / mitigation study. Much more care should.be taken
'~
,., sr, n,,---.-
c
e,
~
for prevention / mitigation studies in the bins chosen because one analysis of risk may not be sufficient. A pair of analyses is required to assess the impact of'a potential change on risk.
There are sequences for which no release categories have been determined, y st notably Class I and Class III LOCAs. Together these represent 10 in probability, but with no estimate of importance to risk.
I recommend that for later efforts, an assessment of releases be made to determine if this is a significant omission.
There are two other areas of potential importance in fission product release determination as yet not fully evaluated. They are the leakage path through the main steam isolation valves and the effacacy of the standby gas treatment system in the face of the potential for large aerosol loadings.
In the present effort, credit has been given for SGTS and MSIV leakage control system effectiveness.
Steam Explosion and Hydrogen Detonation Release Categories The probabilities of the steam explosion accident categories were reassessed b BNL raise eventto10~gBNL.comparedto10gtheprobabilityofthesteamexplosion in the PRA. Brookhaven recomended that the classes (R2A, R2B, R2C and R2D) be separated, because of large differences in plume characteristics.or source terms. Only one source term category was changed, R2D, the hydrogen detonation category. BNL judged that the entire c' ore would not be involved in the release; one fifth of the core was chosen as reasonable and the Te and Ru classes were reduced by this factor. No other' changes were made as a result of this evaluation.
Probabilities are taken from the BNL evaluation, Table 8.6, and plume characteristics and release factors from Table 7.5, modified as noted.
Pool Scrubbing Model The pool scrubbing "models" used in WASH-1400 and in the PRA consisted of assignment of a decontamination factor (DF), the mass of material input to the pool divided by the mass of material output from the pool, based on whether the pool is or is not saturated. The values used are given in Table I.
In all models, the DF for noble gases and organic iodine is unity.
Dr. Arlin Postma, a consultant to the staff, developed a model of pool scrubbing which takes into account the size distribution of entering particles, the makeup of_ fluid carrying the particles (condensible and non-condensible gas fractions), the height of pool above the entrance location, as well as the temperature of the pool (which was the only variable in the WASH-1400 and PRA studies). The development of the model is given in Appendix A.
Other pool scrubbing models have been developed, for instance a General Electric model (DECON) used in the GESSAR probabilistic risk assessment, and an Electric Power Research Institute model (SUPRA) which has not been released and will not be further discussed. The GE model is, in all but one major area, comparable to the Postma model. The Postma
model includes an additional. tenn for evaporation into the' bubbles from pools as the bubbles rise through the water; this factor is only of importance for hot pools.
In addition.to evaporation, the effects considered.were condensation of the condensible fraction of the carrier fluid as the bubble comes to thermal equalibrium with the pool; gravitational settling; centrifugal-deposition due to water flow around
. the bubble which induces circulation flow within the bubble; and diffusion of particles within the bubble. Although the GE model has been documented in part in several places (the GESSAR probabilistic risk assessment and answers to staff questions thereon), the. exact definition of certain constants and input.is unclear and a detailed comparison with the present'model is not fruitful prior to GE's response to GESSAR second round questions. Postma's model has provided the basis for DF used by Battelle Columbus Laboratory in the recent assessment of the Peach Bottom reactor using the SPARC code. The code (SPARC) is an extension of the model to remove a limitation on particle size distribution present in this model and to allow the assumption of eliptical bubble shapes as a user option.
In. contrast to the PRA and WASH-1400, a sequence dependent DF as a function of pool and bubble conditions and as a function of time has been included for most sequences.
In order to apply the model in the present effort, ranges of conditions were chosen in consultation with BNL to represent the types of conditions that were calculated to be applicable to important sequences. A range of DFs were then provided to BNL, and the DF for conditions that most closely matched the calculated conditions was applied.
A crucial parameter is the particle size distribution. The present model assumes a log normal distribution defined by a number median diameter and standard deviation.
It cannot be overemphasized that the choice here and in other studies is based on " engineering. judgment." As shown in Table 4 of Appendix A, preliminary evaluations by Battelle gave particle size ranges; the chosen values are at the lower end of the
. ranges. Relative to the core-concrete interaction release, the value chosen, 1.3 pm aerodynamic mass median diameter, is given in a
?
memorandum from J. Brockmann to D. A. Powers, both of Sandia, dated May 13, 1982. The importance of the particle size distribution has been recognized and more guidance is expected as a result of the severe j-accident program in RES. Applications of SPARC have included l
utilization of size distributions calculated by TRAP-MELT.
l The parameters chosen to represent the pool and the bubbles for various T-calculations are summarized in Table II. The resulting DFs are given in Table III.
In all cases but one, one significant digit is given that reflects the potential inaccuracies in the choice of input.
For sequences that bypass the pool, the DF is obviously unity.
i Binning There.werit two major bins 'in 'the PRA; OPREL and OXRE. OXRE represented the steam explosions and hydrogen detonation and have been discussed previously. OPREL has been separated into three releases. The fraction 4
ofATWSleadingtoClassIIIcontainmentfailugehasbeenseparated, based on a probability of Class III (3.4 x 10- from Table 8.3) times the fraction that had been in OPREL (.775 from table 8.4, Class III in RI). The plume characteristics and release fractions were taken from Table 7.18, the BNL evaluation for Class III failure. Note that this still uses "old" DFs (Table I).
ATWS sequences also lead to Class IV failure modes y,y'and y"; these sequences are designated C4 y,C4 y' and C4 y " in the summary Table IV.
The probabilities are taken f~om the BNL evaluation, Table 8.6, but with r
the probability reduced for the LOCA Class IV which was unbinned, as discussed below. The plume characteristics were taken from Table 7.18
~and the release fractions reflect new DFs. These values were calculated by BNL, as part of a new test case on the removal of the restriction of 20 MARCH steps of pressure and temperature input to CORRAL. These sequences are the only ones that have this last improvement.
The other two release categories separated from OPREL.are the transient cases, which are bins of many sequences. They were divided by BNL into a high pressure bin, where the operator does not manually depressurize, and a low pressure bin where the operator does depressurize.
On the judgment that the operator would fail to depressurize 2/3 of the time, the probabilities of the two bins, C1HP and C11P, are determined from the probability of OPREL in Table 8.6, less the ATWS Class III probability, already discussed. The' release fractions reflect the new DFs.
The LOCA Class IV, with a total probability of 5 x 10-9, is a bin of large, medium and small break LOCAs, with the large break LOCA the dominant sequence. The probability of 4 LOCA is the sum of BNL C4 probabilities in Tables 5.21, 5.22, and 5.23.
The release fractions were taken from Table 7.28, Class IV LOCA with the plume characteristic provided by BNL. This sequence bypasses the pool.
The AJC2y category is again a bin of large, medium and small LOCAs. The probability is again the sum of BNL probabilities in Tables 5.21, 5.2 and 5.23, but in the C category. This category represents a release 7
into a failed containment, and from a failed pressure boundary, and hence bypasses the pool.
All of the characteristics of the release categories are given in Table IV.
For those sequences where new DFs were applied, the only t
documentation for the release fractions is this table.
Engineered Safety Feature Performance Main steam isolation valve leakage has not.been considered. This has been designated as a generic safety issue and may be important in the Limerick design.
The standby gas treatment system has been credited with full capability to reduce fission product releases.
No specific evaluation has been
~
made of the potential loadings of aerosols and what impact this might i
t
i 4
1 l
have.
In sequences such as AJC2y, for which the melt and vaporization releases are into a failed containment, this factor may be important.
f e
h 4
i i
i 1
e l
v I
. I I
i I
~
i.
i I
d i
I h
i-l' 4
(
l 4
s i
i 9
4 i
1 :-
4 e,wwerww--r*,--s
--e,w v -mm-
.<w
,---,-,-w---
,,w-e,ww,,-e-ww,-n,,,--,o.--m,--nes~yn--
,p-=----w-wn-,->-,ww ww--ww~mmw,--r--,
- - - - - - ~ -
Table I Pool DF for Limerick PRA and WASH-1400 Subcooled Pool
- Study Saturated Pool PRA 10 100
- WASH-1400 1
100 DF for noble gases and organic iodine equals I for all cases.
Table II Parameters of Pool and Bubble Conditions for Contemporary DF Calculations-0 Cool pool T = 90 7 0
Hot-pool T 4 204 p s
Saturated pool T = 212 F Melt Release NMD = 0.3 um (AMMD = 5.5 pm) c$*=2
~
Concrete Release NMD = 0.15 pm (AMMD = 1.3 pm o
=2 H, carrier 100 % H 0 % steam 2
Steam carrier 25 % H 75 % steam 2
~
Scrubbing Height 12 ft.
number median diameter and corresponding aerodynamic mass median diameter
- ** standard deviation 5
m y-
<-r
Table III Decontamination Facters for Limerick Conditions DF-Pool
- Carrier Stage coof -
H melt 60 2
. concrete 4
Steam melt 200 concrete 10 hot H2 melt 40 concrete 3
Steam melt 30 concrete 2
i saturated NA melt 10 concrete 1.5 e
M
.e
. < ~.
4
TABLE IV LIMERICK SOURCE TERMS Time To Warning Release Nogle Name Probability Release Duration Time Cal /Sec Ht Gas.
Organic-I Cs.
Te Ba Ru La
-6 6
ATWS 2.6x10 4.5 2.
3.7 8.4x10 25 l
.007
.12
.054
.19.0036
.017
.0024 4
AJC2Y 6.5x10-8 31.
2.
3.6 7x10 25 1
.008
.82
.69
.83
.078
.14
.01 6
R2A 9.8x10-8 2.
5 1.
9x10 25 1
.40
.40
.50 -.05
.50
. 003 6
R2B 2.1x10-8 39.
.5 8.
9x10 25 1
.096
.10
.40
.01
.40
.002 6
R2C 3.2x10-9 2.
.5 1.5 9x10 25 1
.096
.10
.40
.01
.40
.002 0
R2D 4.3x10-7 4.
.5 3.
8.4x10
'25 1
.20
.06
.10
.007
.08.00001 4
C4y 1.4x10-7 1.5 2.
1.
7x10 25 1
.007
.17
.22
.72
.016
.12
.0089 C4y' 7x10-8 1.5 2.
1.
7x10 25 1
.007
.14
.18
.55
.014
.09
.007 4
C4y" 7x10-8 1.5 2.
1.
7x10 0
1 0
.89
.80
.58
.095
.14
.008 4
4LOCA 5x10-9
- 2. 5 2.
.5 7x10 25 1
.007
.82
.75
.75
.086
.11
.010 6
CIHP 4.7x10-5 5.2 2.
3.5 8.4x10 25
.94
.007.0031
.0097. 0028.00071
.0021.00037 6
CILP 2.7x10-5
- 5. 2 2.
3.5 8.4x10 25
.94
.007.00097
.0080.0032.00044.0025.00041 I