ML063410177

Attachment a to Topical Report DOM-NAF-5, Application of Dominion Nuclear Core Design and Safety Analysis Methods to Kewaunee Power Station

ML063410177
Person / Time
Site:	Kewaunee
Issue date:	12/06/2006
From:	Gerald Bichof Dominion Energy Kewaunee
To:	Document Control Desk, Office of Nuclear Reactor Regulation
References
06-578A
Download: ML063410177 (43)

Text

Dominion Energy Kewaunee, Inc.

5000 Dominion Boulevard, Glen Allen, VA 23060 December 6, 2006 U. S. Nuclear Regulatory Commission Attention: Document Control Desk Washington, DC 20555 p Dominion" Serial No.

06-578A N L&OS/C DS: R-1 Docket No.

50-305 License No. DPR-43 DOMINION ENERGY KEWAUNEE. INC.

KEWAUNEE POWER STATION ATTACHMENT A TO TOPICAL REPORT DOM-NAF-5. "APPLICATION OF DOMINION NUCLEAR CORE DESIGN AND SAFETY ANALYSIS METHODS TO THE KEWAUNEE POWER STATION (KPS)"

On January 31, 2006 at a public meeting with the NRC staff, Dominion Energy Kewaunee, Inc. (DEK) discussed application of Dominion nuclear core design and safety analysis methods to Kewaunee Power Station (KPS) (reference 1). On June 13, 2006, DEK and NRC staff engaged in subsequent discussions that focused on the content of a new topical report.

By letter dated August 16, 2006 (reference 2), DEK submitted topical report DOM-NAF-5, "Application of Dominion Nuclear Core Design and Safety Analysis Methods to the Kewaunee Power Station (KPS)."

This letter stated that DEK would provide supplemental material to support the NRC review of DOM-NAF-5, in the form of attachments A and B to DOM-NAF-5.

Enclosed is attachment A of DOM-NAF-5. Attachment A documents application of the Core Management Systems (CMS) methods, as documented in topical report DOM-NAF-1, to Kewaunee Power Station.

Two additional submittals are planned to provide NRC staff with material for review of DOM-NAF-5. Attachment B to DOM-NAF-5, documenting application of the Dominion RETRAN methodology to KPS, is scheduled to be submitted by January 31, 2007. A final submittal will provide a license amendment request (LAR) to include DOM-NAF-5 in the list of approved analytical methods in the KPS technical specifications. DEK anticipates this LAR will be submitted in the third quarter 2007.

DEK intends to apply DOM-NAF-5 to KPS cycle 29, scheduled to begin in the spring of 2008. In order to support application of these methods to KPS cycle 29, DEK requests NRC staff review and approval of DOM-NAF-5 by July 15, 2007. The requested date for NRC staff approval of the forthcoming LAR is anticipated to be January 31, 2008.

Serial No. 06-578A Submittal of DOM-NAF-5, Attachment A Page 2 of 2 Should you have any questions, please contact Mr. Craig D. Sly at 804-273-2784.

Very truly yours, Gerald T. Bischof

(/

Vice President - Nuclear Engineering

References:

1. Summary of Meeting on January 31, 2006, To Discuss the Applicability of Dominion Safety and Core Design Methods to Kewaunee Power Station," (TAC No. MC 9566),

(ADAMS Accession Number ML060400098).

2. Letter from G. T. Bischof (DEK) to NRC, "Request for Approval of Topical Report DOM-NAF-5, 'Application of Dominion Nuclear Core Design and Safety Analysis Methods to the Kewaunee Power Station (KPS),"' dated August 16, 2006 (ADAMS Accession Number ML062370351).

Attachment:

1. DOM-NAF-5 Attachment A, "CMS Benchmarking Information," dated November 2006.

Commitments made in this letter: None cc:

Mr. J. L. Caldwell Administrator, Region U. S. Nuclear Regulatory Commission Region Ill 2443 Warrenville Road Suite 21 0 Lisle, Illinois 60532-4352 Mr. L. Ragavan Project Manager U.S. Nuclear Regulatory Commission Mail Stop 0-7-D-1 Washington, D. C. 20555 Mr. S. C. Burton NRC Senior Resident Inspector Kewaunee Power Station

Serial No. 06-578A ATTACHMENT A TO TOPICAL REPORT DOM-NAF-5, "CMS BENCHMARKING INFORMATION," DATED NOVEMBER 2006 KEWAUNEE POWER STATION DOMINION ENERGY KEWAUNEE, INC.

DOM-NAF-5 Attachment A CMS Benchmarking Information Prepared by:

Christopher J. Wells Robert A. Hall Walter A. Peterson I

Supervisor, fluclear Core Design

Table of Contents TABLE OF CONTENTS..................................................................................................... 2 LIST OF TABLES............................................................................................................... 3 LIST OF FIGURES............................................................................................................ 4 SECTION 1 - INTRODUCTION AND

SUMMARY

............................................................. 5 1. 1 INTRODUCTION 5

1.2

SUMMARY

6 SECTION 2 - STATISTICAL METHODS........................................................................... 8 2.1 NULL HYPOTHESIS TESTS FOR NORMALITY 8

2.2 DETERMINING TOLERANCE LIMITS ASSUMING NORMALITY 8

2.3 DETERMINING NON-PARAMETRIC TOLERANCE LIMITS..................................................... 9 SECTION 3 - HIGHER ORDER CODE BENCHMARKING........................................ 10 3.1 CASMO BENCHMARKING TO HIGHER ORDER CALCULATIONS

...................................... 10 3.2 SIMULATE BENCHMARKING TO HIGHER ORDER CALCULATIONS

................................. 12 SECTION 4 - SIMULATE BENCHMARKING TO MEASURED DATA............................ 15 4.1 CRITICAL BORON CONCENTRATION 1 6 4.2 INTEGRAL ROD WORTH..............................................................................................

19 4.3 PEAK DIFFERENTIAL ROD WORTH...........................................................................

21 4.4 ISOTHERMAL TEMPERATURE COEFFICIENT

.............................................................. 2 3 4.5 DIFFERENTIAL BORON WORTH...................................................................................

25 4.6 DOPPLER COEFFICIENTS AND DEFECTS 28 4.7 REACTION RATE COMPARISONS 29 4.8 DELAYED NEUTRON AND PROMPT NEUTRON LIFETIME DATA........................................

33 SECTION 5 - NORMAL OPERATION POWER TRANSIENTS....................................... 34 SECTION 6 - CONCLUSIONS......................................................................................... 38 SECTION 7 - REFERENCES........................................................................................... 40

Table List of Tables Title Page Kewaunee Nuclear Reliability Factors CASMO-4 Reactivity Benchmarking Versus Monte Carlo Codes CASMO-4 W-prime and Pin-to-box Ratio Comparisons Critical Boron Tolerance Limits and Nuclear Uncertainty Factors Integral Rod Worth Tolerance Limits and Nuclear Uncertainty Factors Peak Differential Rod Worth Tolerance Limits and NUFs Isothermal Temperature Coefficient Tolerance Limits and NUFs Differential Boron Worth Tolerance Limits and Nuclear Uncertainty Factors Integral and 32 Node Reaction Rate Tolerance Limits and NUFs Comparison of North Anna/Surry and Kewaunee Nuclear Reliability Factors

Figure List of Figures Title Page Critical Boron Comparison Integral Rod Worth Comparison Peak Differential Rod Worth Comparison Isothermal Temperature Coefficient Comparison Reactivity Computer Bias Influence Differential Boron Worth Comparison Partial Power & HFP Flux Map Integrated Reaction Rate Comparisons Partial Power & HFP Flux Map 32 Axial Node Reaction Rate Comparisons KPS C27 November 2005 Transient - Power and D-bank Changes KPS C27 February 2006 Transient - Power and D-bank Changes KPS C27 November 2005 Transient - Axial Flux Difference KPS C27 February 2006 Transient - Axial Flux Difference KPS C27 November 2005 Transient - Critical Boron Concentration KPS C27 February 2006 Transient - Critical Boron Concentration

SECTION 1 - INTRODUCTION and

SUMMARY

1.1 Introduction The NRC approved topical report DOM-NAF-1,Rev. 0.0-P-A, "Qualification of the Studsvik Core Management System Reactor Physics Methods for Application to North Anna and Surry Power Stations" (Reference I), details the Dominion methodology for validating the StudsvikJCMS core modeling code package for use in reactor physics calculations. The CMS system consists of CASMO-4, CMS-LINK, and SIMULATE-3 (described in detail in DOM-NAF-1). In DOM-NAF-1 the accuracy of the CMS system was demonstrated through comparisons with measurements from over 60 cycles of operation at the Surry and North Anna Nuclear Power Stations.

The benchmarking methodology, as described fully in DOM-NAF-1, consists of the following broad steps:

Higher order codes are used to identify any biases in CMS model key parameters (such as control rod worth, burnable poison worth, fuel temperature (Doppler) defect and soluble boron worth). Most bias corrections are applied prior to assembling the SIMULATE cross-section library (Reference 1). Control rod corrections are applied using a grey factor in SIMULATE. Higher order codes are also used to verify the accuracy of peak-to-average pin power calculations in CASMO and SIMULATE.

SIMULATE-3 predictions are compared with measured data for the key physics parameters and normal operation power transients are modeled to test the integrated CMS models in a dynamic manner.

Where applicable, Nuclear Uncertainty Factors (NU F) and Nuclear Reliability Factors (NRF) are derived for key physics parameters using statistical methods. For parameters that cannot be directly evaluated using statistical techniques the basis for arriving at conservative estimates for the NUF and NRF is presented.

1.2 Summary This report (DOM-NAF-5, Rev.O.0 Attachment A) documents the application of the DOM-NAF-1 methodology to Kewaunee Power Station. Ten cycles (cycles 18-27) of Kewaunee have been modeled with CMS for the benchmark effort. These ten cycles of operation span a wide range of design and operating history, including transitions in fuel enrichment, fuel density, spacer grid design and material, fuel vendor, core operating conditions (full-power average moderator temperature and rated thermal power) and burnable poison material and design.

The Kewaunee CMS models have been validated by a set of benchmarks to both higher order calculations and ten cycles of measured data. Based on these benchmarks, the following set of nuclear reliability factors (NRF) were determined to account for model predictive bias and uncertainty.

Table 1 Kewaunee Nuclear Reliability Factors I

I Upper I

Lower I

Parameter NRF I

Differential Control Rod Bank Worth 1

1.20 1

0.80 1

Integral Control Rod Bank Worth (Individual banks)

Integral Control Rod Bank Worth (Total of all banks) 1.10 1.10 Critical Boron Concentration Differential Boron Worth I

Prompt Neutron Lifetime 1

1.05 1

0.95 1

0.90 0.90 Isothermal and Moderator Temperature Coefficient Doppler Temperature Coefficient Doppler Power Coefficient Effective Delayed Neutron Fraction I

FAH 1

1.04 1

NIA I

+70 ppm 1.05 I

FQ 1

1.05 1

NIA 1

-70 ppm 0.95

+2 pcmI0F 1.10 1.10 1.05

-2 pcmPF 0.90 0.90 0.95

In addition to the Nuclear Reliability Factors, two Kewaunee normal operation power transients have also been modeled to demonstrate how well the CMS system can predict simultaneous changes in multiple parameters such as core power, moderator temperature, fuel temperature, and control rod bank position. For both modeled transients SIMULATE provided accurate predictions of the timing and magnitude of overall reactivity (critical boron) and core axial power (axial offset).

The following sections present the details of the CMS benchmarking for Kewaunee Power Station. Since the basic methodology is the same as described in DOM-NAF-1,

the results of the Kewaunee benchmark are presented directly while relying heavily on Reference 1 for background and supporting information.

SECTION 2 - STATISTICAL METHODS The following sub-sections review the statistical methods used to construct tolerance intervals for the predicted versus measured plant data. The full discussion of these methods can be found in Reference 1.

2.1 Null Hypothesis Tests for Normality Since no single test for normality is particularly strong, the standard for accepting the hypothesis of normality for any sample data set is such that at least two out of three tests must indicate that the hypothesis of normality is not rejected. The normality tests employed for the Kewaunee predicted versus measured data are the same as used in Reference 1 : Shapiro-Wilk, Kolmogorov-Smirnov, Kuiper, and D'Agostino (Dl) tests.

Note that not all tests apply to all sample sizes. All statistical normality tests are conducted at the 0.05 level of significance.

2.2 Determining Tolerance Limits Assuming Normality From DOM-NAF-1-Rev.O.0-P-A, the one-sided tolerance limit (TL) for data that is assumed to be normally distributed is given by:

where X, and o are the mean and standard deviation of the sample data set. The multiplier K is chosen such that 95% of the population is less than the value of TL, applied in a conservative direction, with a 95% confidence level. Several references (References 2, 3) describe how to obtain the multiplier K. The K values in this report were calculated using the following formula:

The values of z refer to the inverse of the cumulative normal distribution function. P refers to the population that is less than the TL (95% for our case), while yrefers to the desired confidence level (also 95% in this case). Both ~ ~ ( 0. 9 5 )

and ~~(0.95) evaluate to

-1.645.

2.3 Determining Non-parametric Tolerance Limits If the assumption of normality for a certain data set cannot be supported then non-parametric methods must be used to determine the tolerance limits. The non-parametric ranking method of Somerville (Reference 4) is used to determine the mth value of a sorted list that bounds 95% of the population with 95% confidence. This method of computing distribution-free tolerance limits is also referenced in USNRC Regulatory Guide 1.1 26 (Reference 3).

SECTION 3 - HIGHER ORDER CODE BENCHMARKING 3.1 CASMO Benchmarking To Higher Order Calculations As part of the development of the Kewaunee models, Dominion has performed a comparison of CASMO and Monte Carlo code (MCNP-4B and KENO-V.a) calculations of reactivity worth for soh ble boron, gadolinia (gad) loading, AIC (silver-indium-cadmium) control rods, temperature defect, and Doppler defect. These comparisons identify any significant biases so that corrections may be applied prior to assembling the SIMULATE cross-section library (Reference 1). Control rod corrections are applied using a grey factor in SIMULATE.

Table 2 shows the results of the CASMO reactivity benchmarking. Statistical uncertainty associated with each Monte Carlo calculation was limited to a range of 0.00004 to 0.00034 AK (one standard deviation). The data represents a range of fuel enrichments from 3.0 to 5.0 wlo U-235, soluble boron concentration from 0 to 2000 ppm, and temperature from 100 to 547 O F. Doppler comparisons are for enrichments of 3.0 and 5.0 wlo U-235 (burned and new fuel) over a fuel temperature range of 300 to 900 K. Statistically significant biases are apparent for Doppler defect, control rod worth, and soluble boron worth. These results are consistent with results for both the 15x1 5 and 17x1 7 fuel from Reference 1. No statistically significant bias is apparent for gadolinia worth.

Table 2 CASMO-4 Reactivity Benchmarking Versus Monte Carlo Codes AIC Control Soluble Boron 1

Gadolinia 1 Doppler Defect Mean

(% difference) 1.70 23 Std. Deviation

(% difference) 0.6 Note: % Difference is 100 x (CASMO WORTH - MONTE CARL0 WORTH) / (MONTE CARL0 WORTH)

3.2 SIMULATE Benchmarking To Higher Order Calculations The method of comparing CASMOISIMULATE versus Monte Carlo code calculations of pin-to-box ratios and flux thimble instrument reaction rate ratios used in combination with normalized flux map reaction rate comparisons to determine appropriate peaking factor (FAH and FQ) uncertainty factors is fully described in DOM-NAF-1.

In order to estimate the W-prime (normalized ratio of assembly power to flux thimble instrument reaction rate) and pin-to-box uncertainty for the Kewaunee models, five 2x2 assembly models and one 5x5 model have been constructed using both SIMULATE and MCNP. The 2x2 cases include combinations of:

1) Two fuel enrichments (4.0 & 5.0 wlo)

2) Two soluble boron concentrations (0 & 2000 ppm)

3) The maximum gadolinia loading used to date at KPS (20 pins @ 8 wlo Gd2O3)

4) Control Rod Insertion

5) An approximation of once-burned fuel (4wIo u~~~

at 20 GWDIMTU)

The 14x1 4 Kewaunee assemblies are inherently asymmetric, with an off-center instrument thimble and asymmetric gadolinia loadings. In light of this, the 5x5 model is intended to test the ability of CASMOISIMULATE to predict the effect of large flux gradients (approximately a factor of 2 diagonally across the target assemblies) on the W-prime for asymmetric assemblies on opposite sides of the core.

The set of cases modeled (5 2x2 cases and 1 5x5 case) produce large power gradients but do not encompass the full range of burnable poison loadings, fuel enrichments, and fuel burnup that can occur in face-neighbors. However, large power gradients across the fuel assembly faces produce large pin-to-box uncertainties, resulting in a more bounding analysis.

CASMOISIMULATE uses a gamma smoothing technique to account for redistribution of fission energy released as gamma radiation. This method redistributes approximately

7% of the assembly power, effectively flattening the intra-assembly pin power distribution. The MCNP flux tallies do not account for gamma smoothing directly, therefore the MCNP pin-to-box ratios have been adjusted to include this effect. Both the CASMOISIMULATE and Monte Carlo pin powers include the effects of gamma smoothing and are compared on an equal basis.

Results of the SIMULATE / MCNP model comparisons are presented in Table 3. All W-prime data was determined to be normal and all pin-to-box ratio data was determined to be non-normal using the methods described in Section 2. The one-sided tolerance limit for pin-to-box is -1.9% for all pins. These results are bounded by the 2% pin-to-box uncertainty determined in DOM-NAF-1. It is important to note that the KPS results could be improved by eliminating non-limiting low power pins (such as rodded assembly and gadolinia pins) as was done in DOM-NAF-1.

Including the MCNP uncertainty, the W-prime tolerance interval is 1.7%'. If the MCNP uncertainty is excluded via root sum square from the overall uncertainty then the W-prime tolerance interval is recalculated to be 1.5%, which is the same as the DOM-NAF-1 value. It is important to note that the results of the 5x5 case clearly demonstrate that the CMS models are able to accurately account for the effects of offset instrument thimble and asymmetric burnable poison loadings when constructing W-prime values.

For the 5x5 "core", MCNP predicted 4.4% difference in W-primes between cross core symmetric partners while SIMULATE predicted a difference of 4.6%.

The RSS (root sum square) combination of W-prime and pin-to-box uncertainty for use in determining measured peaking factors is 2.6% (1.026 multiplier) using the conservative W-prime tolerance and 2.5% (1.025 multiplier) using the W-prime tolerance with the MCNP uncertainty removed. These values are in line with the 3.0%

and 2.5% RSS combination of W-prime values obtained for the 15x1 5 / 17x1 7 fuel in Reference 1.

• A conservatively low estimate of the MCNP statistical uncertainty for W-prime is 0.4% based on MCNP case output.

Table 3 CASMO-4 W-prime and Pin-to-box Ratio Comparisons Parameter I ':::Ie I Mean (%)

( ~ x c l u d i n ~

1 44 1

MCNP 0.01 (Including MCNP Uncertainty)

Uncertainty) I Std. Dev.

(%)

1.26 0.82 44 0.01 Normal Note: Difference is ((SIMULATE - MCNP) 1 SIMULATE) x 100%

Tolerance Limit Yes

• Both the Monte Carlo and CASMOISIMULATE models are adjusted to include the effects of gamma smearing.

1.5%

SECTION 4 - SIMULATE BENCHMARKING TO MEASURED DATA The following sections present the results of comparisons of SIMULATE-3 predictions with measurements from Kewaunee Power Station. The calculations were performed using full core, 32 axial node, 2x2 X-Y mesh per assembly geometry.

All comparisons of SIMULATE predictions with measured data will by nature represent a combination of SIMULATE bias, SIMULATE uncertainty, measurement bias, and measurement uncertainty. These comparisons will be used to derive appropriate uncertainty factors for SIMULATE predictions. In cases where the comparison data lead to unrealistically high estimates for SIMULATE uncertainty, attempts to quantify and account for measurement bias and uncertainty will be made. Statistical methods are discussed in Section 2. Both the nuclear uncertainty factors and nuclear reliability factors for each parameter are presented in the following sections. In the statistics presented, the sign convention used is such that a positive value indicates over-prediction of the magnitude of a parameter by SIMULATE, and a negative value indicates under-prediction by SIMULATE. Percent differences are determined by the following: (SIM-Measured)/SIM x 100%.

The following sections are summaries of the benchmarking effort for each parameter, with the tolerance limits and Nuclear Uncertainty Factors presented directly, followed by the determination of Nuclear Reliability Factors. More discussion on the general effect of measurement bias and uncertainty on the total observed bias and uncertainty can be found in Reference 1.

4.1 Critical Boron Concentration Table 4 presents the statistical benchmark data for critical boron concentration. Figure 1 presents the difference between SIMULATE and measured boron concentrations in histogram format. Since Kewaunee currently recycles its soluble boron, only boron measurements at end-of-cycle (< 50 ppm) and measurements for which there is a corresponding isotopic analysis of the RCS boron are compared against SIMULATE.

Based on Figure 1, the critical boron differences do not seem to come from a normal distribution. In fact, the boron data appears to be grouped into several separate distributions. From examining the underlying data (not presented), several observations are made about the critical boron comparisons:

1) Before Cycle 27, there is a consistent, tightly distributed bias centered around -28 ppm. This bias persisted through changes in fuel design, transition from discrete burnable poison to gadolinia, and transition from annual to 18 month operating cycles.

This is the "middle" distribution seen in the Figure 1 histogram.

2) During Cycle 27 it appears that the negative bias disappeared. The Cycle 27 boron differences are centered close to 0 pprn and most observations fall within +I 0 ppm.

3) There are several data points that are very negative, grouped about the -50 pprn point. These samples are from cycles before 27. The number of samples in this subset is too small to determine if these points are outliers or are part of distinct third distribution.

From the NUF values alone, the NRF for critical boron could be set to some conservatively bounding value, say, +60 ppm. Although +60 pprn is supported by the benchmark data, there appears to be some unknown driver that is biasing the boron comparisons. This bias does not appear to originate from CMS because in Cycle 27 the bias driver appears to have disappeared. Until the cause of the bias can be positively

identified or enough data is collected to determine that the driver has disappeared, the NRF for critical boron comparisons is conservatively set to k70 ppm.

Table 4 Critical Boron Tolerance Limits and Nuclear Uncertainty Factors (Nonparametric Distribution Assumed)

Mean Std. Number Dev. of Obs.

( P P ~ )

( P P ~ )

-I-Figure 1 Critical Boron Comparison SIMULATE Minus Measured Max. NUF Value (PPm) 55

-56

-51

-46

-41

-36

-31

-26

-21

-16

-11

-6

-1 5

10 PPM Sorted List mth value 2

Lower Tolerance Limit (ppm)

-55 Upper Tolerance Limit (ppm) 11 Min. NUF Value (PPm)

-1 1

4.2 Integral Rod Worth Table 5 presents the statistical benchmark data for integral rod worth. Figure 2 presents the percent difference between SIMULATE and measured integral rod worths in histogram format.

A nuclear reliability factor of +I 0% (multiplier range 0.90 to 1.1 0) is chosen for integral rod worth calculations as it bounds the NUF and is therefore conservative for SIMULATE integral rod worth calculations. Since the uncertainty for the total rod worth cannot be larger than the uncertainty for individual banks the NRF for total bank worth is also set to +I 0%.

Table 5 Integral Rod Worth Tolerance Limits and Nuclear Uncertainty Factors (Normality Assumed)

Std. Dev. /

Upper 1

Lower Multiplier Tolerance Tolerance Number of Obs.

6 1 Mean

(%)

-0.3 Figure 2 Std.

Dev.

(%)

4.6 K

Integral Rod Worth Comparison SIMULATE Minus Measured

-7

-5

-3

-1 1

3 5

7 9

11 13 15

% Difference Limit (%) Limit (Oh)

4.3 Peak Differential Rod Worth Table 6 presents the statistical benchmark data for peak differential rod worth. Figure 3 presents the percent difference between SIMULATE and measured peak differential rod worths in histogram format.

The NRF multiplier range for Peak Differential Rod worth is set to bound the NUF at 0.80 to 1.20.

Table 6 Mean

(%)

Peak Differential Rod Worth Tolerance Limits and Nuclear Uncertainty Factors (Normality Assumed)

Std. Number Std. Dev.

Upper Lower Min. NUF Max. NUF Dev. of Obs. Multiplier Tolerance Tolerance Multiplier Multiplier

(%I K

Limit (Oh)

Limit (%)

Figure 3 Peak Differential Rod Worth Comparison

% Difference

4.4 Isothermal Temperature Coefficient Table 7 presents the statistical benchmark data for the isothermal temperature coefficient. Figure 4 presents the difference between SIMULATE and measured isothermal temperature coefficients in histogram format. From the histogram, it appears one point is not grouped with the others. If this data point (a Cycle 20 predicted-measured comparison) is removed from the sample pool and sample statistics are computed, the point lies over 6~ from the mean. Also, from documentation of the measurement, it appears that there may have been difficulty in obtaining the measured ITC. Therefore, the ITC difference data point for Cycle 20 is assumed to be an outlier and has been removed from the tolerance interval dataset.

A nuclear reliability factor of k2 pcrnl°F conservatively bounds the NUF of -0.96 / +0.84 pcd°F and is therefore considered appropriate.

Table 7 Figure 4 Isothermal Temperature Coefficient Comparison Isothermal Temperature Coefficient Tolerance Limits and Nuclear Uncertainty Factors (Normality Assumed, Cycle 20 Outlier Removed)

-1.75

-1.375

-1.I25

-0.875

-0.625

-0.375

-0.125 0.125 0.375 Difference Max. NUF Value (pcmPF) 0.84 Min. NUF Value (pcmPF)

-0.96 Lower Tolerance Limit (pcmPF)

-0.84 Upper Tolerance Limit (pcmI0F) 0.96 Mean (pcmPF) 0.06 Num.

of Obs.

9 Std.

Dev.

(pcmPF) 0.30 Std. Dev.

Multiplier K

2.99

4.5 Differential Boron Worth Use of measured data to determine differential boron worth (DBW) uncertainty can lead to unrealistically large uncertainty estimates. As discussed in DOM-NAF-1, reactivity computer bias can be a large portion of the DBW measurement uncertainty. Figure 5 clearly demonstrates the correlation between DBW error and RW error for 8 measurements at Kewaunee. The correlation coefficient of 81 % indicates a strong relationship between DBW error and RW error, which except for the reactivity computer are independent quantities. Also, based on the best estimate and standard error of the slope of the least squares fit line, the true slope of the fit line is most certainly greater than zero. Therefore, the data shows that reactivity computer bias is contributing significant uncertainty to the DBW measurements.

Figure 6 presents the difference between SIMULATE and measured differential boron worth in histogram format. Using the same methodology described in DOM-NAF-1, tolerance limits and NUF values for DBW have been calculated with the uncertainty due to reactivity computer measurements removed (Table 8). Even without reactivity computer bias, the derived NUF range is unreasonable.

Much like the North Anna and Surry DBW NUFs presented in DOM-NAF-1, the Kewaunee DBW NUFs do not seem consistent with other evidence we have about boron worth. The maximum HZP BOC critical boron concentration difference between SIMULATE and measured is -55 ppm. Based on the DBW NUF range however, it is expected we would see boron differences due to boron worth error alone in the range of 98 ppm (7% x 1400 ppm) to 276 ppm (1 2% x 2300 ppm). The fact that the worst observed HZP boron difference (-3%) does not fall within this range indicates that the true DBW uncertainty is significantly lower than the DBW statistics suggest.

Therefore, based primarily on the evidence of critical boron concentration difference data, a NUF and NRF of +5% (multiplier range of 1.05 to 0.95) is considered to be sufficiently conservative for SIMULATE differential boron worth predictions.

Figure 5 Reactivity Computer Bias Influence DBW % Difference Versus Rod Worth % Difference I

1

+ DBW vs RW Difference I

-Linear (DBW vs RW Difference) 1 5.0 t-7

-10.0

-8.0

-6.0

-4.0

-2.0 0.0 2.0 4.0 6.0 8.0 Control Rod Worth % Difference

Table 8 Differential Boron Worth Tolerance Limits and Nuclear Uncertainty Factors (Normality Assumed, Reactivity Computer Influence on Variance Removed)

Figure 6 Differential Boron Worth Comparison I

I I

I I

I

-8

-4 0

4 8

12

% Difference Min. NUF Multiplier Mean

(%)

Max. NUF Multiplier Std. Dev.

Multiplier K

Std.

Dev.

(yo)

Upper Tolerance Limit (Oh)

Number of Obs.

Lower Tolerance Limit (Oh)

4.6 Doppler Coefficients and Defects As discussed in DOM-NAF-1, direct determination of the NRF for Doppler feedback is very difficult. The value of 1.I 0 for the Doppler Temperature Coefficient and Doppler Power Coefficient that was proposed for North Anna and Surry models in DOM-NAF-1 is proposed for Kewaunee CMS predictions. There are three indications that support the use of a +I 0% NRF (multiplier range 1.10 to 0.90)

1) Benchmarking of CASMO Doppler Temperature Defects to Monte Carlo methods enables a best-estimate correction to be performed that effectively eliminates the theoretical CASMO Doppler bias (as determined using higher order calculations) from the SIMULATE model.

2) The Kewaunee fuel pellet diameter is almost identical to the Surry fuel pellet diameter. The Doppler feedback is therefore expected to be just as well predicted for Kewaunee as it currently is for Surry. DOM-NAF-1 presented data for Surry that supported the use of a +I 0% NRF.

3) The Kewaunee Cycle 27 February 2006 operational transient modeled in Section 5 includes an undamped xenon oscillation. The good agreement of the measured and predicted axial offset oscillation magnitude in the modeled operational transient demonstrates that the xenon-Doppler balance is reasonable.

4.7 Reaction Rate Comparisons Table 9 presents the statistical benchmark data for the reaction rate comparisons.

Figure 7 presents the percent difference between SIMULATE and measured integral reaction rates in histogram format. Figure 8 presents the percent difference between SIMULATE and measured 32 node reaction rates in histogram format.

Table 9 presents the combined tolerance limits (the pin-to-box tolerance limits root-sum-squared with the reaction rate tolerance limits) and NUFs for integral and nodal reaction rates. Since under-prediction of reaction rates is the only concern from a safety analysis standpoint, only the lower tolerance limit and corresponding NUF is presented.

Based on the Table 9 values, a NRF for FAH of 1.04 is conservative and an NRF for FQ of 1.05 is conservative.

Table 9 Integral and 32 Node Reaction Rate Tolerance Limits NUF Multiplier 1.036 1.047 and Nuclear Uncertainty Factors (Nonparametric Distributions Assumed)

Note: The observations in this table are from a combination of partial power and HFP maps.

Data Type Integral 32 Node Mean

(%)

0.02

-0.1 2 Std.

Dev.

(%)

2.08 2.72 Number of Obs.

838 23333 Combined Tolerance Limit

-3.63

-4.65 One Sided Reaction Rate Tolerance Limit

-3.03

-4.20 Pin-to-box Tolerance YO

-2.0

-2.0

Figure 7 Partial Power & HFP Flux Map Integrated Reaction Rate Comparison

-9

-8

-7

-6 4

-3

-2

-1 0

1 2

3 4

5 6

7 8

% Difference

Figure 8 Partial Power & HFP Flux Map 32 Axial Node Reaction Rate Comparisons

4.8 Delayed Neutron and Prompt Neutron Lifetime Data The same effective delayed neutron and prompt neutron fraction NRFs developed in DOM-NAF-1 are also assumed for Kewaunee. A NRF of +5% (multiplier range of 1.05 to 0.95) is considered conservative for both of these values. The Tuttle delayed neutron set was used for all CASMO modeling of Kewaunee fuel.

SECTION 5 - NORMAL OPERATION POWER TRANSIENTS Two Cycle 27 power maneuvers were modeled with SIMULATE. These maneuvers are useful for demonstrating the ability of the CMS model to accurately predict core behavior involving combinations of large power changes, control rod movements, temperature variations, xenon worth changes, and boron changes. Figures 9 and 10 show the time dependent power and D-bank positions for the 11/2005 and 212006 events, respectively.

Results of the modeling (Figures 11 and 12) show that SIMULATE axial offset changes closely match the pattern of the measured ex-core axial offsets. The magnitude of the changes are reasonably well predicted (generally within 2-3% delta-I). Some of the differences may be due to the dependence of the ex-core detectors on a few peripheral fuel assemblies that are less axially sensitive to D-bank insertions than average.

Critical boron predictions show a modest initial bias of 10-20 ppm and follow the measured changes within approximately 20 ppm. The total boron change is -600 ppm for the 11/05 event and -200 ppm for the 12/06 event. No adjustment for B-10 depletion effects have been made. Based on the consistency of the boron bias before and after each event, there does not appear to be a significant change in B-1 O/B-11 ratio during the event. Results for A10 and boron are shown in Figures 13 and 14.

Figure 9 KPS C27 November 2005 Transient Power and D-bank Changes Figure 10 KPS C27 February 2006 Transient Power and Pbank Changes

Figure 11 KPS C27 November 2005 Transient Axial Flux Difference Time (hours)

Figure 12 KPS C27 Febuary 2006 Transient Axial Flux Difference 0

10 20 30 40 50 60 70 80 90 100 Time (hours)

Figure 13 KPS C27 November 2005 Transient Critical Boron Concentration 0

20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 Time (hours)

Figure 14 KPS C27 Febuary 2006 Transient Critical Boron Concentration 0

10 20 30 40 50 60 70 80 90 100 Time (hours)

SECTION 6 - CONCLUSIONS Using the methods and processes delineated in DOM-NAF-1, the accuracy of the CMS models has been demonstrated through comparisons with reactor measurements and through comparisons with higher order Monte Carlo neutron transport calculations. As Table 10 demonstrates, the Kewaunee NRF values are consistent with those derived from the assessment performed for the Dominion Surry and North Anna units.

Table 10 Comparison of North Anna / Surry and Kewaunee Nuclear Reliability Factors I

North Anna1 Surry I Kewaunee Parameter Critical Boron Concentration 1

+50 ppm 1 -50 ppm

+70 ppm 1 -70 ppm NRF Upper I ~ower Differential Boron Worth 1

1.05 1

0.95 1.05 1

0.95 NRF Upper I Lower Doppler Temperature Coefficient 1 1.I 0 1

0.90 1 1.1 0 1

0.90 Doppler Power Coefficient 1

1.10 1

0.90 1.10 1

0.90 Effective Delayed Neutron Fraction 1.05 0.95 1.05 0.95 Prompt Neutron Lifetime 1

1-05 1

0.95 1 1.O5 1

0.95 FAH

The largest NRF difference between North Anna/Surry and Kewaunee is for the critical boron concentration parameter. Although the critical boron tolerance interval would support a NRF of +15/-60 ppm at Kewaunee, both the shape of the difference distribution and the limited number of samples (93 samples for Kewaunee versus over 1000 for North Anna and Surry) introduce uncertainty as to how well the Kewaunee CMS models predicted critical boron. A +70 ppm NRF is chosen because it is sufficiently wide for conservative safety analysis but not so wide as to mask potential plant operation issues. The critical boron NRF can be adjusted in the future as more boron samples are collected andlor a cause for the bias between measured and predicted values is determined.

The Dominion Kewaunee CMS models, including appropriately determined NRF values, have been demonstrated to be appropriate for use in reload applications such as core reload design, core follow, and calculation of key core parameters for reload safety analysis at Kewaunee Power Station. The robust model development process, in conjunction with code and model quality assurance practices, as described fully in DOM-NAF-1 and applied specifically to Kewaunee Power Station in this document, provide assurance that future changes to core, fuel and burnable poison designs will be modeled with accuracy and appropriate conservatism.

SECTION 7 - REFERENCES

1. R. Hall, R. Kepler, J. Miller, C. Wells, W. Peterson, "Qualification of the Studsvik Core Management System Reactor Physics Methods for Application to North Anna and Surry Power Stations," Dominion, DOM-NAF-1 -Rev.O.O-P-A, June 2003.

2. M. G. Natrella, "Experimental Statistics," National Bureau of Standards Handbook 91, August 1963.

3. "An Acceptance Model and Related Statistical Methods for the Analysis of Fuel Densification," U.S.N.R.C. Regulatory Guide 1.126, Rev. 1, March 1978.

4. P. N. Somerville, "Tables for Obtaining Non-Parametric Tolerance Limits," Ann.

Math. Stat., Vol 29, 1958.