ML18100B077

Forwards Response to Request for Addl Info Re Reactor Physics Methods Topical Rept NFU-0039,Rev 1

ML18100B077
Person / Time
Site:	Salem
Issue date:	05/12/1994
From:	Labruna S Public Service Enterprise Group
To:	NRC OFFICE OF INFORMATION RESOURCES MANAGEMENT (IRM)
References
NLR-N94092, NUDOCS 9405200293
Download: ML18100B077 (25)

Text

..

Public Service Electric and Gas Company I

Stanley LaBruna Public Service Electric and Gas 1Company P.O. Box 236, Hancocks Bridge, NJ 08038 609-339-1700 Vice President - Nuclear Engineering MAY 12 1994 NLR-N94092 United States Nuclear Regulatory Commission Document Control Desk Washington, DC 20555 Gentlemen:

RESPONSE TO REQUEST FOR ADDITIONAL INFORMATION REACTOR PHYSICS METHOD TOPICAL REPORT SALEM GENERATING STATION UNIT NOS. 1 AND 2 FACILITY OPERATING LICENSES DPR-70 AND DPR-75 DOCKET NOS. 50-272 AND 50-311 Attached is PSE&G's response to NRC's request for additional information on Salem's Reactor Physics Methods report, NFU-0039, Revision 2.

The questions were transmitted via NRC letter dated February 24, 1994.

As a result of our ongoing model validation process, additional benchmarking of the extended burnup model has been completed.

Since the submittal of NFU-0039, Revision 2, three cycles of operation have been completed, each of which utilized multiple burnable poison loadings.

We have included the results of these three recent cycles together with the five cycles presented in NFU-0039, Rev. 2, to evaluate the biases and reliability factors previously reported.

The results of this current benchmarking show consistent model performance with increased burnup, thus indicating that the NFU-0039, Revision 2 reliability factors remain applicable for the multiple burnable absorber loading and extended burnup designs, which are part of the present Salem fuel management strategy.

However, based on the recent benchmarking results and recent EPRI information, biases for Doppler defect and temperature coefficient have been revised, as discussed in the attachment.

200060 9405200293 940512

~DR ADOCK 05000272 PDR.

~\\

Document Control Desk NLR-N94092 MAY 12 1994 Should you have any questions regarding this information, please contact us.

Sincerely, Attachment C

Mr. T. T. Martin, Administrator - Region I U. s. Nuclear Regulatory Commission 475 Allendale Road King of Prussia, PA 19406 Mr. J. c. Stone, Licensing Project Manager -

Salem

u. s. Nuclear Regulatory Commission One White Flint North 11555 Rockville Pike Rockville, MD 20852 Mr. c. Marschall (S09)

USNRC Senior Resident Inspector Mr~ K. Tosch, Manager, IV NJ Department of Environmental Protection Division of Environmental Quality Bureau of Nuclear Engineering CN 415 Trenton, NJ 08625

Responses to the NRC Request for Additional Information PSE&G Salem Reactor Physics Methods Topical Report, NFU-0039, ~~vision 2

April29, 1994 TABLE OF CONTENTS BACKGROUND................................................,..................... 1 QUESTIONS and RESPONSES....................................................... 3 Question #1........................................................................ 3 Question #2........................................................................ 6 Question #3........................................................................ 7 Question #4........................................................................ 7 Question #5........................................................................ 8 Question #6........................................................................ 8 Question #7........................................................................ 9 Question #8........................................................................ 9 Question #9........................................................................ 12 Question #10.................................................................... 12 Question # 11 Question #12

...................................................................... 12

...................................................................... 13

SUMMARY

andCONCLUSIONS.................................................... 16 ATIACHMENT 1.................................................................... 17

April29, 1994 BACKGROUND A brief background discussion of the PSE&G Salem core physics design and analysis process is provided, prior to responding to the specific requests, to support the explanation of how changes in fuel designs or changes in computer codes are addressed at PSE&G.

The paragraphs that follow describe the fuel designs in use at Salem, swnmarize the methodology change that occurred to support the new designs, describe the PSE&G plans for future Salem cores, and swnmarize the overall process employed at PSE&G to deal with design and methodology changes. This discussion provides background information relevant to PSE&G's specific responses that follow.

The PSE&G design and analysis process deals with the reality that fuel designs change over time and the computer codes or methodology also change to support the new designs. Model benchmarking or validation is an on-going l-11ocess.

• Each cycle, reactivity and power distribution comparisons are made to verify that the model accuracy is within expectations and to determine if any trends in accqracy are d~veloping which may impact future predictions.

PSE&G is cognizant of and participates in industry methodology development programs.

Changes in fuel design for Salem are gradual. The table below shows the Salem cycles and the trends in burnable poison usage and the discharge burnup.

Unit/ Begin Feed BOC Burnup Cycle Burnup Lead Assembly

1. BPRs

1. IFBA Cycle Date U23s w% (MWD/MTU) (MWD/MTU) Discharge Burnup Loaded Pins Loaded 1/7 lC.'87 3.8 10,620 16,805 40,000 1576 0

1/8 2/88 3.8 11,210 13,766 40,000 1664 0

2/5 11/88 3.6/4.0 11,700 14,614 42,000 1728 0

1/9 6/89 3.6/4.0 11,420 16,521 45,000 1344 0

2/6 6/90 3.8/4.0 12,040 15,994 47,000 1632 0

1/10 4/91 4.0 16,750 11,818 44,000 512 3072 217 4/92 3.8/4.0 16,580 9,540 46,000 544 2560 1111 8/92 4.0/4.4 14,670 13,035 46,000 368 6400 2/8

• 6/93 4.0/4.4 14,610 16,900*

48,000*

418 5088 1/12 1/94 4.0/4.4 16,055 16,000*

46,000*

512 3072 (1.SX)

• denotes projected values for current cycles.

1

April 29, 1994 The change from the EPRI CELL based model to the PSCPM based model was necessitated by the need to utilize multiple burnable poisons for the extended bumup cores. The new fuel design and modeling technique were evaluated by performing sensitivity studies.

This established the impact of design decisions or modeling assumptions. Next, the methodology to be used was validated by performing comparative calculations to other codes. This establishes the confidence in the ability to predict the behavior of new designs before they are loaded into the reactor for the first time. Next, benchmarking was performed to prior operating data to compare the accuracy of the old model (CELL based) vs the new (PSCPM based). As changes to the fuel designs were utilized in Salem core reloads, on-going benchmarking was performed.

Salem 2 Cycle 9 will begin operation in the fall of 1994 and contingent upon approval of the licensing amendment request, the fuel will utilize several high bumup design features.

Included among these features are the use of Zirlo cladding, an increase in the plenum volume for fission gases, intermediate flow mixers, and a 50% higher ZrB2 linear loading (1.SX design). The l.SX IFBA design was loaded for the first time in Salem 1 Cycle 12. at the beginning of 1994. The LOX IFBA design was loaded starting with Salem 1 Cycle 10in1991.

In summary, the change to new designs is a methodical decision, and the changes in computer codes and the performance of all engineering analysis is controlled and documented per the PSE&G QA procedures.

As illustrated in the background discussion above, the PSE&G process is on-going and consists of; 1) sensitivity studies to evaluate the design or methodology change, 2) calculations to validate the performance of the analysis system, and 3) benchmarking to operating data.

In step one of the process, sensitivity calculations are performed to determine the impact of either design variables (i.e.; # of IFBA/assembly) or modeling assumptions (i.e.; IFBA coating smeared in clad region or treated explicitly). In step two, predictive calculations are performed and compared to other independent methods (i.e.,

another code or the fuel vendor results). In step three, measured versus predicted comparisons are made and the model biases and reliability factors are revised as appropriate.

As additional improvements are made to the EPRI-CPM code (current EPRI CPM-3 project),

these improvements will be evaluated for incorporation into PSCPM using the above process.

Any future methodology changes will be reviewed to determine if the Salem Reactor Physics Methods topical report needs to be revised.

April 29, 1994 QUESTIONS and RESPONSES Note that iii the questions and responses that follow, all reference numbers refer to the Salem topical report page 5-1 and 5-2 except for questions 1, 8 and 12. For questions 1 and 12, the list of references are the Doppler referenc~s found in the response to question 12. The response to Question 8 contains additional new references which are listed at the end of that response.

1 Provide a detailed discussion of the basis/or the PSCPM bias in the Doppler coefficient.

Have the Reference 12 and 13 benchmarking comparisons been made with the latest version of PSCP M? Discuss the applicability of the PSCP M Doppler bias and reliability factors and the calculations of References 12 and 13 to the intended Salem fuel designs.

The basis for the Doppler bias and uncertainty is the EPRI research documented in References 1 and 2 (see Question 12 response for list of Doppler references).

The Reference 12 and 13 (of Salem topical report) benchmarking comparisons were made with CPM2 which calculates the exact same Doppler as PSCPM. The impact on the Doppler uncertainty due to different H20/U02 ratios is shown in Reference 2 and based on this information it is concluded that the bias an i uncertainty derived in Reference 2 are applicable to the current and the planned Salem fuel designs. A detailed explanation of these responses is provided in the text that follows.

The original basis for the bias and uncertainty in the PSCPM Doppler coefficient and defect was taken from the investigation of a EPRI project as reported in Reference 1 (see Question 12 response for list of Doppler references). Reference 1 shows that CPM2 consistently gives a higher value for Doppler worth than EPRI-CELL2/Standard Library (CELL2/STD). Pertinent data in Reference 1 shows:

1. The difference between CPM2 and CELL2/STD Doppler worth values for the 3.9% enrichment is approximately 30% at E=O, approximately 20% at E=14 GWD/MT and approximately 18% at E=32 GWD/MT.

2. There is some dependence on enrichment; as the enrichment increases the difference between CPM2 and CELL2/SID decreases. -

At the time Revision 2 of the Salem Reactor Physics Methods was completed, the available information incorporated in Reference 1 was the Hellstrand experiments (Reference 4), and some in reactor measurements (References 3, 5, 6). Based on this, it was concluded that CELL2/SID results were the most accurate, and CELL2/STD was taken as the standard. As explained on page 4-10 of the Salem topical, the CPM2 Doppler worths are bounding on the high side as calculated by the code (CPM2 over

- i I

April 29, 1994 predicts Doppler). It was postulated that when the CPM2 values are multiplied by.80 they are bounding on the low side. For application to Salem, the typical average core exposure at BOC was taken to be approximately 14 GWD/MT and at EOC, approximately 32 GWD/MT (Reference 1 shows the difference between CPM2 and CELL2/STD is nearly constant at higher burnup). The data in the table on page 4-10 of the l:'SE&G Salem topical is from Reference 1, and the factor of.8 makes CPM2 values smaller than CELL2/STD for the exposures of interest. Since the Doppler results from CPM2 and PSCPM are exactly the same, the conclusions drawn for CPM2 are also true for PSCPM. Thus the derivation of the 10% bias and 10% uncertainty on page 4-10.

In July 1993, Reference 2 was published and presented new information. Reference 2 contains detailed comparisons of the various cross section generators with Hellstrand's measurements, Monte Carlo analyses and reactor trends.

Reference 2 showed that agreeing with Hellstrand results. is not, by itself, a sufficient determination of the accuracy of the Doppler calculation in a reactor lattice. It also showed that References 7 and 8 herein, and Reference 13 in the Salem Reactor nhysics Methods report have data that are erroneous. These cases were recalculated a.id reported in Reference 2. Reference 2 concludes that all the results show CELL2/STD to be giving the best estimate of the Doppler worth.

Quantitatively, CELL2/STD was within 2 to 3% of the Hellstrand nominal values, and within 3% of the Monte Carlo nominal values for the enrichments of interest. This supports the earlier Reference 1 approach of taking CELL2/STD as the standard. The later results in Reference 2 showed CPM2 to be 10% higher than the Hellstrand nominal value, and approximately 14% higher than the Monte Carlo nominal value for the mixed oxide case (i.e., the case emulating extended exposure). Reference 2 concludes that for CPM2 a bias of -25% with an uncertainty of +/- 10% is appropriate for fresh fuel, and a bias of -15% with an uncertainty of +/- 10% is appropriate for fuel at 32 GWD/MTU.

Based on the more recent results, the Doppler bias and reliability factor shown in the Salem topical report Table 4.0. l will be replaced with the following:

bias:

CAE ~ 32 bias= -(25% -.3125

• CAE)

CAE > 32 bias= -15%

where CAE =Core Average Exposure in GWD/MTU reliability factor (RFDC) = 10%

reliability factor (RFDD) = 10%

April29, 1994 For example; if PSCPMffRINODE calculated a Doppler coefficient of -2 pcm/F at a BOC core average exposure of 14 GWD/MTU, then the kast negative Doppler coefficient would be :

=nominal value* (1 + bias/100) * (1 - RFDC/100)

= -2.0 * [1 - (25 - 0.3125

• 14)/100] * (1 - 10/100)

= -2.0 *. 794 *.9 = -1.43 pcm/F If PSCPMffRINODE calculated a Doppler coefficient of -4 pcm/F at a EOC core average exposure of 32 GWD/MTU, then the mQfil negative Doppler coefficient would be:

=nominal value * (1 + bias/100) * (1 + RFDC/100)

= -4.0 * [1 - (25 - 0.3125

• 32)/100] * (1+10/100)

= -4.0

• 0.85

• 1.1 = -3.74 pcm/F In the evaluation of Doppler accuracy, the most important physical parameters are the moderator/fuel ratio and the pellet size. The biases and uncertainties determined based on the Reference 2 material are applicable to the current and planned designs as illustrated by the following table:

Typical Monte Carlo Pin Cell Salem Comparisons Comparisons Values (Ref2)

(Ref 1)

Pellet O.R.(cm)

.4186

.39306

.41169 Clad O.R. (cm)

.4759

.45802

.47587 Pitch 1.2649 1.26209 1,25984 H20/U02 1.614 1.924 1.645

- Table 4-12 in Reference 2 showed that a change in the H20/U02 ratio from 1.65 to 1.92 changes the difference between CPM2 and CELL2/STD U-238 absorption fraction and hence Doppler results by approximately 1 %. The difference in the fuel pin size and pitch for Salem compared to the Reference 1 and 2 data will have little affect on the results and hence the results of Reference 1 and Reference 2 are applicable to Salem fuel designs.

5

April29, 1994 2 Discuss in detail the changes introduced into the more recent EPRJ-CPM2 and PSCPM versions of CPM What changes have been introduced in PSCPM to calculate the integral fuel burnable absorber fuel designs and mixed core loadings? Are these changes included in the extended burnup model benchmark comparisons and, if not, how will these changes be validated?

Revision 1 of the Salem topical utilized a CELL based model in which the origillal EPRI-CPM code was part of the EPRI ARMP methodology (Reference 1). For Revision 2 of the topical, the starting point for the extended bumup model was the EPRI-CPM2 (Reference 11) code and library. Since the EPRI release of CPM2, no modifications to the PSE&G copy of the EPRI-CPM2 code or library were made to incorporate EPRI sponsored modifications. (Note: Changes have been made to CPM2 by the EPRI maintenance group mainly for BWR applications, but these have not been incorporated into the PSE&G version of CPM2).

The CPM2 code as released by EPRI was modifie<l by PSE&G to write punch files for the downstream linking codes and for the explicit treatment of IFBAs. The modified code was named PSCPM. The linking code modifications did not change any of the CPM2 computations; the original Punch option was extended to include data needed in the PDQ and TRINODE input. The PLINK code was written to process the CPM2 calculated items such as; reaction rates, number densities, macro cross sections, fluxes, and geometry data to create the PDQ input files. The BLINK code was written to process the CPM2 calculated items such as; K"', M2, K, Lr, VLr, and axe to create the TRINODE input files. Explicit treatment of the IFBA design was achieved by modifying CPM2 to permit an additional annulus in the fuel pin. The annulus added is outside of the pellet and is the ZrB2 coating. As outlined in the background discussion, the new methodology was validated by performing sensitivity studies to evaluate modeling assumptions and by performing comparisons to other computer codes. For the IFBA design, the assumption of an explicit ZrB2 region vs smearing the ZrB2 in the clad region was evaluated and pincell results were compared to EPRI-CPM (which permits four annuli in the fuel pin) and EPRI-CELL (both codes described in Reference 1). The new IFBA capability combined with the existing CPM2 capability for burnable absorber rods provides capability for modeling mixed assembly designs containing both IFBAs and burnable absorber rod inserts. The new code, PSCPM, was then used to generate the PDQ and TRINODE inputs which were then used to perform the extended burnup comparisons presented in Revision 2 of the Salem topical report.

6

April29, 1994 For future applications, PSCPM is being converted to execute on RISC based processors and will be designated CPM-2/WSE. This version is functionally equivalent to PSCPM for Salem reactor physics analysis and produces identical results.

• 3 Describe in detail the changes made in TRINODE including the treatment of burnable absorber reactivity input.

A comprehensive list of the changes made to TRINODE since Revision 1 of the Salem topical is contained in Table 1. A summary of the key changes for the extended burnup model is presented below.

The TRINODE code was modified to accept table input instead of curve fits and to explicitly account for the multiple burnable poison designs.

Prior to modifying TRINODE, PSCPM calculations were perfonned for various multiple burnable poison designs to evaluate each of the reactivity effect-; modeled in TRINODE. Based on this examination the traditional "B-constants" of EPRT-NODE-P (Reference 1) were replaced with explicit tables and the functional dependence of the independent variables was extended to multiple dependent parameters.

For example,

~Pexposure vs burnup was replaced by tables of K"' as a function of burnup, instantaneous moderator temperature, and moderator temperature history. The tables are input to TRINODE by fuel type. A fuel type in TRINODE is a unique plane of an assembly based on its fuel dimensions, U235 enrichment, number of BPR rodlets, and number of IFBA pins. Thus the reactivity

.)f the newer fuel designs is explicitly modeled as a unique fuel type.

4 Describe the fuel designs (e.g., multiple burnable absorber and integral fuel burnable absorber) to which the PSE&G methodology will be applied. Discuss the adequacy of the Chapter 4 benchmarking for qualifying the methodology for these applications.

Satem fuel is a l 7xl 7 lattice which may contain 0 to 24 Pyrex Burnable Poison rodlets inserted into the guide tube locations. The number of fuel pins which may contain the ZrB2 coating ranges from 0 to 164. The typical designs currently utilized are 64 or 104 IFBA with 0 to 12 BPRs. Occasionally some cycles may contain assemblies with inert rods, steel or zirc, which are used as filler rods to replace damaged fuel pins following fuel reconstitution. Salem 1 cycle 11 and Salem 2 cycle 8 are recent examples of this.

As explained in the introductory background text, the latest design utilizes a l.SX boron loading in the IFBA pins. Benchmarking is an on-going process at PSE&G. When Revision 2 of the Salem topical was submitted, the report contained the most recently 7

April 29, 1994 completed cycles for each reactor unit. Since that time, Salem 1 cycles 10 and 11 and Salem 2 cycle 7 have completed operations. All three of these cycles contain multiple burnable poisons and have been benchmarked as part of the on-going model validation. to this response presents the latest results.

PSE&G has no current plans to utilize* fuel designs other than those that are uranium fuel with boron as the burnable poison. Any future design changes would be evaluated and the methodology revised if warranted following the process outlined in the background discussion.

5 How do the Revision 2 rod worth standard deviation and reliability factor based on calculation to measurement comparisons compare with the Revision 1 values?

Revision 2 values compared to the Revision 1 values (reported on page 3 - 4 of Physics Methods Report) are 60 pcm vs. 91 pcm reliability factor for rods worth less than 600 pcm, a 9% vs. a 12% reliability factor for rods worth more than 600 pcm, and a 8% vs. a 16% reliability factor for total rod worth. In addition, each of the Revision 2 factors is less than the the extended bumup model reliability factors tabulated in Table 4.0.1, page 4 -2.

6 In Table 4.2.1, what is causing the large difference between the Revision 1 and Revision 2 predictions of the Cycle 8 Salem 1 isothermal temperature coefficient?

The extended burnup model contains explicit tables of K"' as a function of burnup, instantaneous moderator temperature, and moderator temperature history. This treatment of moderator temperature and history is a significant change from the original CELL based model. The original CELL based model used a BOLK"' vs moderator temperature and a ~Pexposure vs bumup at the average moderator temperature.

Also, the Doppler component of the isothermal temperature coefficient is larger in the PSCPM based model compared to the CELL based model (approximate difference is.4 pcm/F). The resulting earlier TRINODE calculations included analytically the use of a burnable absorber dependent adjustment. This analytical adjustment was used as part of the TRINODE methodology to produce the results presented in Table 3.2. l. The Salem 1 Cycle 8 CELL model results presented in Table 4.2.1 did not use the prior adjustment. The extended burnup model does not have a similar adjustment built into the methodology; instead a constant bias is applied after the TRINODE calculation (Table 3.2.l shows the mean=

0.0 vs Table 4.2.1 shows a mean of 1.31 ).

8

April29, 1994 7

What is causing the Revision I/Revision 2 bias in the isothermal temperature coefficient?

See the response to item 6 above.

8 Is the calculation of the delayed neutron.fraction Pe.ff based on the ENDFIB-V data? If not, provide justification for the value used.

PSE&G does not use the p values contained in the output edits of PSCPM nor the P values based on ENDF/B-V. The PSE&G methodology uses a PSE&G code called BETA to calculate core Peff values and the input p values to this calculation are from Tuttle (see references at the end of this response). The justification for the delayed neutron fractions used at Salem is based on an examination of the BET A code process and also on the component information that goes into the calculation of a core Peff* The following paragraphs first explain the PSE&G methodology and then discuss the component parts and the resulting impact of differences in the components on the core Peff*

The core Peff used in point kinetics calculations is calculated by the PSE&G BET A code.

The BET A methodology is shown below:

The three dimensional core average delayed neutron fractions are calculated as:

Im km (PWF) 7 L L (P1,k)

X L (Pn,m X Vn X F n,1,k)

A

_ 1=1 k=l n=l pm -

where; Im km (PWF) 7 L: L: (P1,k) x L: (vn x Fn1k) l=lk=l n=l m = delayed neutron group, 1 to 6

!=assembly, 1to193

-k =axial node, 1 to 12 P1,k = nodal power from TRINODE for assembly l, node k PWF = power weighting factor n = nuclide number, 1 to 7 Pn,m = delayed neutron fraction for nuclide n, group m vn = average number of neutrons per fission for nuclide n F n,l,k = fission fraction for nuclide n, assembly l, node k

April 29, 1994 The total delayed neutron fraction is calculated as:

6 Ptot = L Pm m=l The effective delayed neutron fraction is calculated using the importance factor, I, as:

Petr = I X Ptot In the BETA methodology the P1,k comes from the TRINODE case which simulates the reactor condition of interest and the F n,l,k comes from the cycle specific full core PDQ case. Thus these P and F values reflect the nuclear data library used in the lattice code, PSCPM. The Bn,m and the vn are input and may come from any external source. To determine the impact of using the current PSE&G p values (Tuttle) versus the ENDF/B-V P values, two BETA calculations were performed for a typical Salem 1 Cycle 9 BOC condition. The table below summarizes the

~ input data and the BET A code output results:

Nuclide Current PSE&G Values ENDF/B-V Values U-235 0.00683 0.00685 U-238 0.01639 0.01570 Pu-239 C.00224 0.00223 Pu-241 0.00532 0.00549 BET A results peff 0.00618 0.00627 As discussed in Section 3.5 of the Salem topical report, the uncertainties m the calculation of Pet1 are composed of several components, the most important being:

a. experimental values of p and A. by nuclide

b. calculation of the spatial nuclide inventory

c. calculation of core average Pelf as a flux weighted average over the spatial nuclide inventory

d. calculation of Pelf from the core average as Pelf = p

• I, where I = importance factor.

10

April29, 1994 The first three items are relevant to the justification of the Petr values to be used by PSE&G. For item a), the previous table showed the p by nuclide. This data contains the uncertainties of the experimental data and of the processing of the experimental data. The major significant difference is the U-238 values.

When the BETA calculations are performed in which the only difference is the values of p by nuclide, the impact of the current PSE&G values versus ENDF/B-V results in approximately 1.5% smaller Petr values.

For item b), the most important nuclide concentrations with respect to P are U-238, U-235 and Pu-239. An inspection of the EPRI-CELL versus the EPRI-CPM ability to predict nuclide concentrations (Table 3.4. l and Figures 3.4.1-3 vs Table 4.4.1 and Figures 4.4.1-3) shows that EPRl-CPM agrees better with experiment and for the three nuclides of interest the difference between measured and predicted is less than 1 %.

For item c), the uncertainty associated with the calculation of a core average p is dependent on the relative flux weighting of the individual assemblies in the core. This uncertainty does not change due to the use of different values of p by nuclide. The item d) uncertainty also does not change due to the Uje of different values of p by nuclide.

Thus, using the same approach as pages 3-21 and 3-22 of the Salem topical report, the combined 'uncertainty for the extended cycle burnup model using the PSE&G BET A methodology is:

l.5%(a) + l.0%(b) + 0.8%(c) + 0.5%(d) = 3.8%

Based on this review of the PSE&G BET A methodology and the items that influence the calculation of the core Petr, it is concluded that the 4% reliability factor for P stated in Tables 3.0.l and 4.0.l is appropriate for the PSE&G BETA methodology.

BETA Data Source

References:

Tuttle, R.J., "Delayed - Neutron Yields in Nuclear Fission," INDC (NDS)-107/G+

Special, from "Proceedings of the Consultant's Meeting on Delayed Neutron Properties,"

Vienna, (March 1979).

Tuttle, R.J., "Delayed - Neutron Data for Reactor Physics Analysis," NSE-56 37-71 (1971).

11

April29, 1994 9 Provide the fuel exposure dependence of the confidence limits for X(l, K, M) and X(L M) for the extended burnup model (Figures 3. 6.15 and 3. 6.18 for the Revision 1 model).

Figure 1 shows the exposure dependence for both models. The confidence limits for both the normal and the non-parametric statistics for all sub-populations fall below the bo;.:nding value reliability factors for FQ and F MI*

10 What is the maximum fuel exposure for which the extended burnup will be applied?

Discuss the adequacy of the benchmarking for these fuel exposures.

The lead assembly maximum burnup to which the extended bumup model will be applied is 65,000 MWD/MTU. This burnup bounds current fuel utilization plans.

For current fuel designs used at Salem, the maximum burnup is about 55,000 MWD/MTU for the lead assembly and 52/'00 for the batch average. Future cycles are being planned for which the lead assembl) burnup will be approximately 60,000 and 55,000 for the batch average. As shown in the table in the background discussion, the lead assembly discharge bumup has gradually increased with time. The on-going model validation process ensures that the model adequately predicts the behavior of future reloads. This is illustrated in Attachment 1 of this document, which shows the use of a scatter plot for the difference between measured and predicted integral power as a function of assembly burnup. Scatter plots such as this are used to determine if model biases or trends exist. In addition, the reliability factors are updated if necessary based on the on-going benchmarks. If the on-going benchmarking shows that smaller reliability factors are warranted, then the Salem Reactor Physics topical report will be revised.

11 How has the change from EP Rl-CELLIP DQ to PSCP M affected the determination of the detector fission rate-to-assembly power ratios? Is this change included in the extended burnup model benchmark comparisons?

There has been no change in the method for determining this ratio. Both the CELL based (Figure 2.0.1) and the PSCPM based (Figure 2.0.2) methods utilize PDQ for generation of the instrument factors (detector fission rate/assembly-power).

The box labeled "Normalization Reaction Rates Pin to Box" in both figures is the step which generates the instrument factors to be fed to the SIGMA code which performs the measured vs predicted reaction rate comparisons. Any change in the numerical values between the models is derived from PDQ's cross sections coming from EPRl-CELL or from PSCPM.

12

April29, 1994 The extended bumup model benchmark comparisons include the use of the PSCPM/PDQ instrument factors.

12 Provide all references used in determination of the Doppler bias and reliability factors.

1 J. Fisher, R. Grow, D. Hodges, J. Rapp, K. Smolinske, Utility Resource Associates.

"Evaluation of Discrepancies in Assembly Cross Section Generator Codes", Volumes 1 & 2, 1989, Volume 3, 1990. EPRI-NP-6147.

2 J. Fisher, R. Grow, Utility Resource Associates, 11Evaluation of Discrepancies in Assembly Cross Section Generator Codes", Volume 4 Doppler Evaluations,July 1993, EPRI NP-6147.

3 "Power Coefficient Validation of the Kewaunee ARMP Model." ANS Paper Circa 1981.

4 E. Hellstrand, P. Blomberg, S. Horner, "The Temperature Coefficient of the Resonance Integral for Uranium Metal '"1.d Oxide." Nuclear Science & Engineering, July 12, 1990, 8-497-506.

5 D. W. Dean. "Analysis of Axial Xenon Oscillations at Prairie Island." International Reactor Physics Conference, Jackson Hole, WY. September 18-22, 1988.

6 "Results of PI2 Cycle 13 Xenon Test." March 19, 1991. NSP Internal Report, K. L.

Wright 7

R. D. Mosteller, L. Eisenhart, S. Levy Inc. "ENDF/B-V Doppler Evaluations."

Presented at Eight Semi Annual EPRI Reactor Physics Users Group. November 1989.

8 R. D. Mosteller, L. Eisenhart, S. Levy Inc.; R. C. Little, Los Alamos National Lab; W. J. Eich, J. Chao, Electrical Power Research Institute. "Benchmark Calculations for Doppler Coefficient of Reactivity."

13

April 29, 1994 Table 1 Response to Question 3, List ofTRINODE Changes 1

Table look up of nodal K"' and Ap values for Doppler, boron, xenon, and control rods; M2, crxe* ~F* and u/K as functions of number of burnable poison fuel pins, number of BPR rodlets, exposure, moderator temperature history, instantaneous moderator temperature, fuel temperature, and boron concentration.

2 Add moderator temperature history input, calculation, and edits.

3 Add target ~ff and convergence criterion to boron search logic.

4 Change the definition of the control rod search target band from A I to axial offset.

5 Add a Ap poison core map edit for burnable poison fuel pins.

6 Calculate the Doppler fuel temperature for table look up as a function of nodal power.

7 Allow the axial variations in assembly desi;.1s to be modeled by assigning fuel types by axial level in each assembly.

8 Allow optional fuel temperature, TFUEL, input. This will set all nodal fuel temperatures equal to TFUEL.

9 Model the finite lattice effect on control rod worth. The finite lattice effect accounts for the reduction in rod worth in a core geometry compared to the PSCPM single assembly rod worth in an infinite lattice geometry.

14

Figure l Response to Question 9 Confidence Limits for X(l,M} vs Cycle Exposure Rev 1 Model vs Rev 2 Model vs RFFDH 0.12 ~--------------------------.

0.10.._

'E 0.08 !-------------------------!

J g 0.06 CD

'C =

c: 8 0.04 -

0.02 ~

0 0

0

\\ _______________________ -*-**-**- -

0.00 '----"'--.l...-

1 -"'--.l...-'-"'------..__---"'------'

6 8

10 12 0

2 4

Cycle Ex;;osurn ( Gw-j/t)

Normal Non-Parametric PSE&G RFFDH Normal Non-Parametric Rev 1 Rev 1 Rev 2 Rev 2 Confidence Limits for X(l,K,M} vs Cycle Exposure Rev 1 Model vs Rev 2 Model vs RFFQ 0.12 ----------------------------.

0.10 1-----------

• .)

'E 0.08 ~

0

J g 0.06 ~

CD ;g c: 8 0.04 -

0.02 ~

<)

o.ooL--..___.1....-*-..___.__._..___,__._..___.__,_..___.__._..____.

0 2

4 6

8 10 12 Cyde Exposure ( Gwd/t)

Normal Non-Parametric PSE&G RFFDH Normal Non-Parametric Rev 1 Rev 1 Rev 2 Rev 2 0

15 April29, 1994

April29, 1994

SUMMARY

AND CONCLUSiuNS This document provides PSE&G's response to the NRC request for additional informati"on for the PSE&G Salem Reactor Physics Methods Topical Report, NFU-0039, Revision 2. As the result of our on-going model benchmarking and validation process additional benchmarking work has been completed (Attachment 1) since the topical report submittal. Based on this work and the recent Doppler work described in response # 1 which was sponsored by EPRI, PSE&G will utilize the extended bumup model with the biases and reliability factors shown in the table below. Note that the temperature coefficient and Doppler defect biases have been revised since NFU-0039, Revision 2 was submitted. These changes are based on both the new PSE&G benchmarking and the EPRI results.

Revised Reliability Factors and Biases for PSE&G Extended Bumup Model Applied to Salem Parameter Rod Worth Meas ~ 600 pcm Meas < 600 pcm Totals Temperature Coefficient Moderator (MTC)

Isothermal (ITC)

Doppler CAE~32 CAE>32 Doppler Defect CAE~32 CAE>32 Power Distribution FQ P~.50 P<.50 Fm re p ~.30 P<.30 Reliability Factor RFROD= 15%

RFROD = 100 pcm RFROD= 10%

RFMTC = 2.1 pcm/°F RFITC = 2.1 pcm/°F RFDC= 10%

RFDC= 10%

RFDD= 10%

RFDD= 10%

RFFQ=0.10 RFFQ = 0.16- (0.12

• P)

RFFDH=0.08 RFFDH = 0.09 - (0.033

• P)

Bias 0

0 0

1.3 1.7

-(25% -.3125 *CAE)

-15%

-(25% -.3125 *CAE)

-15%

0 0

0 0

where: CAE=Core Average Exposure in Gwd/t,and P= Core rated thermal power 16

April29, 1994 Most Recent Benchmark Results Since the submittal of Revision 2 of NFU-0039, Salem Reactor Physics Methods, three additional cycles have completed operation at Salem. As shown in the BACKGROUND section discussion, these cycles utilized the multiple burnable poison design.

Comparisons of measured versus predicted rod worths, temperature coefficients and reaction rates are made as part of the on-going model validation. In response to question 4 the data from these recent cycles has been compiled in summary format in Tables A.I and A.2. Table A. l lists the three recent unit/cycles and also the flux maps analyzed.

These three most recent cycles are compared to the results for eight cycles (the initial five plus the recent three) in Table A.2.

Rod Worth Benchmarkin" The rod worth results (mean and standard deviation) are shown in Table A.2 for each of the three groups of cycles for the categories of measured worth equal to or greater than 600 pcm, measured less than 600 pcm and the total worth of all rods. The eight cycle results in Table A.2 can be compared to the extended burnup model reliability factors (Table 4.0.1 topical report) ofRFROD = 15%.for individual rods Meas~ 600, RFROD =

100 pcm for individual rods Meas < 600 pcm, and RFROD = 10% for total rod worth.

Table A.2 shows that the results are consistent between the different groupings of cycles.

Isothermal Temperature Coefficient Benchmarkin" Only one ITC measurement per cycle was made for the recent cycles in contrast to the many measurements for the earlier Salem cycles. For the extended burnup model the mean difference (M - P) for the eight cycles is + 1. 7 pcm/F. A meaningful reliability factor can not be determined for the recent cycles alone due to the small sample size, thus the alternate approach used for small populations w~ to show on a bounding value approach that all new data points lie inside the ITC bias plus reliability factor. Based on these results the ITC bias was changed to 1. 7 pcm/F, the MIC bias to 1.3 pcm/F (accounts for subtracting out the Doppler contribution to the bias, see response to #6), and the reliability factors remained at 2.1 pcm/F for both the ITC and MIC.

17

April29, 1994 Power Distribution Bencbmarkin~

Table A.2 shows the integral and local power distribution results. The results in Table A.2 can be compared to the extended bumup model reliability factors (Table 4.0.1 topical report) of RFFDH = 0.08 for power ~ 30% and RFFDH = 0.09 - P/30 for power < 30% ;

and RFFQ = 0.10 for power~ 50% and RFFQ = 0.16 - (0.12*P) for power< 50%. Table A.2 shows that the results are reasonably consistent between the different groupings of cycles.

Figure A.1 shows the scatter plot of the M-C integral residual for each assembly monitored versus assembly bumup at the time of the measurement. This data shows consistent model performance with increasing bumup.

\\.8

April29, 1994 Table A.1 - Recent Cycle Flux Map Database Salem 2 Cycle 7 Unit/Cycle/Map Cycle Bumup

%Power D Bank Position 2-701 17 21.8 173.00 2-702 66 46.3 186.00 2-705 2152 99.l 228.00 2-708 5134 99.2 228.00 2-712 9324 99.8 228.00 Salem 1 Cycle 10

• 1-1001 8

23.8 177.00 1-1002 18 46.5 180.00 1-1004 1269 100.0 228.00 1-1010 7863 99.9 228.00 1-1013 9955 100.2 228.00 Salem 1 Cycle 11 1-1101 8

20.l 156.00 1-1102 47 46.0 177.00 1-1105 2033 99.9 228.00 1-1108 4652 99.8 228.00 1-1112 8109 81.3 186.00 I

1-1116 11692.00 99.9 228.00 19

April 29, 1994 Table A.2 Summary of Statistical Results Recent 3 Cycles All 8 Cycles 8 Cycle Reliability Factors I) Rod Worths (µ,cr)

M ~ 600 (M-C)/C 2.6%+/-5.6%

2.8%+/-3.9%

8.6%

M <600 (M-C) 6.lpcm +/- 44.6 pcm 4.2pcm +/- 29.9 pcm 66pcm Total Worth (M-C)/C 1.9% +/- 1.8%

2.1% +/- 1.7%

5.7%

3) F MI (cr) (M-C)

High Power Maps 0.025 0.023

.040 Mid Power Maps 0.030 0.026

.046 Low Power Maps 0.036 0.029

.052

4) FQ (cr) (M:C)

High Power Maps 0.041 0.045

.075 Mid Power Maps 0.063 0.062

.104 Low Power Maps 0.065 0.077

.129 20

Figure A.1 Integral Power Distribution Accuracy vs Assembly Bumup Assembly Integral Residual vs Exposure Meas - Pred 0.20...----*-----------------------.

0.18 0.16 0.14 0.12 a:- 0.10

~ 0.08 i 0.06

.g 0.04

• m 0.02

~ 0.00 iii -0.02

~ -0.04

8 -0.06 0 gJ -0.08 0::: -0.10

-0.12

-0.14

-0.16

-0.18 0

0

• *0 0

Topical Recent Reliability 5 Cycles 3 Cycles Factor FDH 0

-0.20._._.....,_._,__.._.....J...J....._____. _

_........;....J.~............ ~-----'-'----............ J.......J.---.-'---------'

50 0

5 10 15 20 25 30 35 40 45 Assembly Exposure (GWD/MTU) 2l April29, 1994