ML22341A080
| ML22341A080 | |
| Person / Time | |
|---|---|
| Site: | Electric Power Research Institute |
| Issue date: | 12/08/2022 |
| From: | Licensing Processes Branch |
| To: | Electric Power Research Institute |
| References | |
| EPID L-2021-TOP-0006 | |
| Download: ML22341A080 (6) | |
Text
Enclosure EPRI REPORT 3002018337, "USE OF DATA VALIDATION AND RECONCILIATION METHODS FOR MEASUREMENT UNCERTAINTY RECAPTURE: TOPICAL REPORT REQUESTS FOR ADDITIONAL INFORMATION (RAIS)
SET 2 Regulatory Basis On January 27, 2021, the Electric Power Research Institute (EPRI) submitted Technical Report 3002018337, Use of Data Validation and Reconciliation [DVR] Methods for Measurement Uncertainty Recapture [MUR]: Topical Report" to the U.S. Nuclear Regulatory Commission (NRC or staff) for review and approval (Agencywide Documents Access and Management System (ADAMS) Accession No. ML21053A027). This topical report describes a process for using a mathematical DVR method for monitoring core thermal power and use of the methods for MUR power uprates.
Nuclear power plants are licensed to operate at a specified maximum core thermal power (CTP). Appendix K, "[Emergency Core Cooling System] ECCS Evaluation Models," of Title 10 of the Code of Federal Regulations (10 CFR), Part 50, formerly required licensees to assume that the reactor has been operating continuously at a power level at least 1.02 times the licensed power level when performing loss-of-coolant accident (LOCA) and ECCS analyses. This requirement was included to ensure that instrumentation uncertainties were adequately accounted for in the safety analyses. In practice, many of the design bases analyses assumed a 2 percent power uncertainty, consistent with 10 CFR Part 50, Appendix K.
A change to the Commission's regulations in 10 CFR Part 50, Appendix K, was published in the Federal Register on June 1, 2000 (65 FR 34913), which became effective July 31, 2000. This change allows licensees to use a power level less than 1.02 times the CTP for the LOCA and ECCS analyses, but not a power level less than the licensed power level, provided the proposed alternative value has been demonstrated to account for uncertainties due to power level instrumentation error. Licensees can use a lower uncertainty in the LOCA and ECCS analyses provided that the licensee has demonstrated that the proposed value adequately accounts for instrumentation uncertainties. As there continues to be substantial conservatism in other Appendix K requirements, sufficient margin to ECCS performance in the event of a LOCA is preserved.
In the requests for additional information (RAI) below, the staff is seeking information to define the proposed DVR methodology.
RAI-13
EPRI provided a significant amount of information to the NRC during this review. However, as demonstrated in the response to RAI-01, much of the information provided in the topical report was provided as information about the concepts of DVR and was not intended to define a specific DVR methodology. Therefore, the NRC staff is requesting that EPRI provide the complete list of requirements which would be necessary for appropriate application of a DVR methodology at a nuclear power plant. Specifically, identify those situations or conditions in which the DVR value would be considered out of service or not reliably accurate enough to base judgements about the current value of core thermal power.
RAI-14
Background The NRC staffs current understanding of the process used to calculate the reconciled uncertainty is that this calculation does not result in the variance of the reconciled measurement, but rather the variance in the mean of the reconciled measurement. Suppose n samples of the reconciled measurement were made, and then the mean of those n samples was generated. The reconciled value represents that mean, and the reconciled variance is the variance in that mean. However, the quantity of interest is not the variance in the reconciled mean, but the variance in the reconciled measurement itself. To get the variance in the reconciled measurement we need to multiply the variance in the reconciled mean (the outcome of the DVR process) by n.
For example, consider two flowmeters on the same pipe where one of the flowmeters is downstream from the other. While the flowmeters may be independent from each other they are measuring what should be the same variable (i.e., the flow rate in the pipe should be the same at both locations - assuming no leakage). Thus, we could take advantage of this redundant measurement and use DVR to determine a better estimate of the true flow rate in the pipe which has a lower uncertainty than either of the individual measurements. To demonstrate this, assume that we have a pipe with flow meters A and B as displayed in Figure 1 with the values given in Table 1.
Figure 1: Pipe with two flow meters A and B
Table 1: Example Values Flow Meter mean flow rate (kg/sec)
+/- 1.96 95% tolerance Interval (kg/sec) standard deviation (kg/sec) variance (kg2/sec2) variance in the mean (i.e., the standard error squared)
(kg2/sec2)
(number of samples to obtain the mean)
Flow Meter A
245.00 245 +/- 12.25 12.25 1.96
= 6.25 39.06 39.06 50
= 0.7812 50 Flow Meter B
250.00 250 +/- 12.50 12.50 1.96
= 6.38 40.67 40.67 100
= 0.4067 100 For this example, we are interested in combining the measurements from flow meter A and B into a single value using the DVR process. The function which defines the reconciled mean is given as the following.
(,,, =
=
1
+ A +
+ B
+ A + 1
+ B
Eq. 1 Thus, our reconciled mean is 247.45. To calculate the reconciled variance, we use the Taylor Series Method on Equation 1 which results in the following. (Note, we will consider the distinction between and later).
=
+
B
B
+
+
Eq. 2 Where:
is the variance in the new mean value (i.e., this is the standard error squared).
is the first derivative of the function with respect to mean
is the variance in the mean
is the first derivative of the function with respect to mean
is the variance in the mean
is the first derivative of the function with respect to the variance is the variance in the variance
is the first derivative of the function with respect to the variance
is the variance in the variance It is important to note that the output from the Taylor Series Method is not reconciled variance, but the reconciled standard error squared. However, we do not want the standard error, therefore we will have to convert this to a reconciled standard deviation. But before we do that, we can simplify this equation. In general, we assume that the variances ( and ) are constants and not random variables, therefore the variance of these variances ( and )
are zero. This allows us to simplify the equation to the following.
=
1
+
B
B Eq. 3 We can now calculate the partial derivatives
and
. Even though we have two different equations ( and ), each equation results in the same partial derivative.
= 1
+ =
+ = 0.5101 Eq. 4
=
+ = 1
+ = 0.4899 Eq. 5 Next, we need to calculate the variance in the mean values (
and
). Note, because our DVR function is a function of the mean value, we need to use the variance in the mean value (i.e., the standard error squared) and not just the sample variance.
=
= 39.06 50
= 0.7812 Eq. 6
=
= 40.67 100 = 0.4067 Eq. 7 Finally, we can calculate the new variance in the reconciled mean value ().
=
+
B
B
= (0.51010.7812 + (0.48990.4067
= 0.3009 Eq. 8 Notice that this variance is much lower than the variance in the either measurement A
(= 39.06) or measurement B (= 40.67). This is because it is the variance in the mean of
the reconciled measurement () and not the variance in the reconciled measurement itself
(). However, we can use the standard formular to turn the variance in the mean into the variance from the population.
=
Eq. 9 However, we do not know how to appropriately determine, and the choice of is directly proportional to the reconciled variance of the population. The following table provides the reasonable values for. Also provided is the special case where = for comparison.
Table 2: Possible values of
= = 50 15.045
=
+
2
=75 22.5675
= = 100 30.09
19.93 Question For such a situation in which the mean of all measurements is not calculated using the same number of samples, how is the reconciled variance calculated? Conversely, what is the correct reconciled variance that should be used from Table 1, and why?
RAI-15
In response to the first set of RAIs, EPRI provided many documents and some responses to the same RAI were spread across multiple documents. The NRC staff is requesting that EPRI provide a table containing each RAI and a list of which documents provide responses to those RAIs.
RAI-16
Is there a maximum limit to the uncertainty reduction using the DVR process? For example, given enough process measurement data points and computed additional redundancies, is it possible to arrive at a near-zero core thermal power or total feedwater flow uncertainty?
Describe the controls that are in place (e.g., data validation checks) and/or the backstops used to prevent this situation from occurring.
RAI-17
Chapter 3 of the EPRI Topical Report states that the accuracy (uncertainty) values for the measurement and auxiliary (unmeasured) variables for input to the DVR algorithm are provided as an input to the overall calculation. These accuracy estimates may include approximations of the systematic and random errors of the measurement based upon the user's knowledge of the measurement variables, or plausible assumptions using engineering judgement. An empirical covariance matrix is developed in the DVR algorithm that relates the errors of the measurement uncertainties and the auxiliary conditions.
Describe how the estimates of process measurement and sensing line instrument channel random and systematic error components are to be determined, what instrument channel modeling process and engineering standards should be used to determine them, and how these estimates of systematic and random uncertainties are used in the determination of reconciled results. The report describes the consideration of calibration uncertainties but does not seem to mention how estimates of the process-related (e.g., thermal stratification, flow bias effects, thermowell lag and heat transfer effects) or instrument sensing line-related uncertainties were considered in the estimates of channel uncertainty that are to be part of the reconciliation process, nor how the uncertainties (e.g., A/D conversion, anti-aliasing, steam table look-up modeling errors, algebraic and numerical methods, digital signal processing within the process computer and the data historian, and algebraic effects of the data historian) introduced by the plant process computer and the data historian are considered and are to be accounted for.