Hydraulic Model Calibration of a Nuclear Plant Service Water System
AFT Fathom™ Technical Paper
Authors: Erin Onat, Purple Mountain Technology Group; Trey Walters, P.E., Applied Flow Technology; David M. Mobley, AREVA Inc.; and James J. Mead, Duke Energy
Presented at the ASME 2016 Power Conference, June 30, 2016
ABSTRACT
As pipe networks age, build-up [scaling] and corrosion decrease pipe diameter and increase pipe roughness, leading to significant pressure drops and lower flow rates. When modeling the hydraulics of these systems, calibrating the pipes to account for additional scaling and/or fouling can be vital to accurately predicting the hydraulic behavior of the system.
An automated, multi-variable goal-seeking software was used to calibrate the raw water system of the Duke McGuire Nuclear Station (MNS). This calibration process involved three phases. The first phase was the testing of the automated, multivariable goal-seeking software on a previously calibrated system.
The second phase was the calibration of a partial data set. The third phase was the calibration of a complete data set. The automated goal-seeking software (AFT Fathom GSC) was found to have varying degrees of success in each phase.
At the conclusion of the calibration process, the partial data calibration of two parallel systems at MNS yielded average overall calibration accuracies of 2.1% and 1% for flow rates, and 1.2 psig (8.4 kPa-g) and 1.7 psig (11.9 kPa-g) for pressures. The complete data calibration of one of these systems at MNS yielded an average overall calibration accuracy of 2.3% for flow rates, and 1.4 psig (9.5 kPa-g) for pressures.
Conclusions
The Beta method, a new, automated process, was used along with manual iteration during various phases of calibration to assist in calibrating the Duke McGuire Nuclear Station raw water system. The Beta method was performed in three phases. In the first phase, the Beta method was used to calibrate a previously calibrated model. In the second phase, it was used on two systems with partial data sets. In the third phase, it was used on one system with extensive data collection.
During the first phase, the Beta method consistently found roughness values that matched the test data much better in the best cases. In the worst cases, it matched the data equally well. During the second phase, the Beta method was used to calibrate flows in a partial data set to an average overall accuracy of 2.1% and 1%, respectively, and pressures to an average overall accuracy of 1.2 psig (8.4 kPa-g), and 1.7 psig (11.9 kPa-g), respectively. During the third phase, difficulties with the Beta method were experienced due to a variety of reasons that will be explored. In this phase, manual iteration was used in conjunction with the Beta method to calibrate flows in a complete data set to an average overall accuracy of 2.3% and pressures to an average overall accuracy of 1.4 psig (9.5 kPa-g).
Duke does not currently have any plans to investigate the discrepancies discussed because the minimum requirements of the EC were met even with these discrepancies, and they were met with adequate margin above the required flow. Therefore, Duke is confident that the minor load piping will receive (at least) the minimum required flow and that these discrepancies do not need to be examined at this time in order to ensure the safe operation of the plant.
Because the only requirement of the flow balance tests for the minor load piping was to record the flows through these pipes (i.e., the pressure measurements were performed for additional precision in support of the EC and were not a requirement in the flow balance tests), the minimum criteria were determined to have been met satisfactorily and, conservatively, with sufficient margin. Further, flow balances are routinely performed to monitor and trend flows through all flow paths to ensure that adequate flows are met.
Even without further investigation of these discrepancies, calibration of the raw water system at Duke MNS provided several benefits to the analysis of the EC. Within the limited time that the calibration was performed, errors in the hydraulic model input, as well as errors in field data, were discovered and corrected as a direct result of the calibration process. Additionally, results acquired from the calibrated model were more realistic and conservative (i.e. lower flows and pressures) than would be calculated from an uncalibrated model.
These more conservative results provide additional certainty in asserting the suitability of the proposed EC. In the long term, the calibration performed will likely help those at Duke MNS to quickly pinpoint specific areas in their system that require additional attention. Such areas are indicated in the calibrated model by pipes with large scaling factors (which could indicate highly scaled pipes), or areas in the model that were unable to achieve the flows or pressures (which could indicate faulty measurement devices or other problems with data collection).
INTRODUCTION
McGuire Nuclear Station is located in Huntersville, North Carolina, USA off Lake Norman, midway in the chain of lakes created when the flood-prone Catawba River was dammed. Unit 1 began commercial operation in 1981 followed by Unit 2 in 1984. Figure 1 shows a photo of the station.
The Nuclear Service Water System (RN) is a safety related, open loop cooling system that provides cooling water from Lake Norman or the Standby Nuclear Service Water Pond (SNSWP) to various station heat exchangers during all modes of operation. In addition, the system acts as an assured source of makeup water for several other safety-related systems, including the Auxiliary Feedwater System (CA).
The CA system is provided as a backup for the Main Feedwater System and is designed to dissipate heat from the Reactor Coolant System when normal non safety-related systems are unavailable. The RN system delivers water to each of the two power station Units (Units 1 and 2). Two trains (A and B) supply water to each Unit. Therefore, Unit 1 is supplied by the 1A and 1B trains. Unit 2 is supplied by the 2A and 2B trains. These systems share common intake and discharge piping, as well as one heat exchanger, but are otherwise independent from each other.
The A Train and B Train systems also share common supply and discharge piping, so in reality, all four systems have some level of interconnection. Figure 2 shows a high level schematic of the RN B Train system. Note that there are 47 heat exchangers in the RN B Train system (23 heat exchangers in each Unit, plus one shared heat exchanger) counting the three major load heat exchangers in each Unit.
The original design configuration of the RN to CA assured supply placed the flow path downstream of the Diesel Generator Engine Cooling Water System (KD) heat exchanger, near the RN return header. This configuration resulted in high supply temperatures, low Net Positive Suction Head (NPSH), and air entrainment concerns to the CA pumps. Engineering changes (EC’s) were developed to relocate the assured supply to upstream of the KD heat exchanger which alleviates these concerns.
Hydraulic models of the RN and CA systems were needed to evaluate the new flow and pressure conditions of these systems after implementation of the EC. As previously mentioned, the RN system utilizes raw water from Lake Norman or the SNSWP, neither of which are chemically controlled. Over time, this raw water causes buildup (scaling) and corrosion which decreases pipe diameters and increases pipe roughness, leading to significant pressure drops and lower flow rates (see Figure 3).
After the model was developed from piping drawings, a benchmark (the MNS term for model calibration) was therefore required to ensure the model accurately reflected the current conditions in the plant.