A Novel Statistical Model for the Assessment of Feedwater Flow Calibration Factor Based on Ultrasonic and Deaerator Mass and Energy Balance Measurements

Abstract

The most accurate method of determining reactor thermal power in CANDU reactors is the Heat Balance around the secondary side. The Heat Balance reactor calculation is based mainly on secondary side process measurements. The feedwater flow and temperature are the two most important contributions. Therefore, it is an established practice to verify station feedwater flow and temperature readings against high precision Ultrasonic Flow and Temperature measurements and to implement the flow calibration factors and temperature corrections in the Heat Balance Program. Generator output of one of Ontario CANDU units has been significantly lower than its full power rating, and a secondary side performance test did not identify any deterioration in the turbine cycle performance. Therefore, possible mis-calibration of the feedwater flow was suspected as the cause of the lower unit output. DeAerator Mass Energy Balance (DAMEB) test was performed concurrently with the secondary side performance test. The total feedwater flow measured based on DAMEB test was found almost 1.8% lower than the factor based on ultrasonic feedwater flow measurements. Since there was no root cause identified of the high ultrasonic flow readings and considering safety implications of raising reactor power by over 1.5%, it was prudent to view the two measurements as two independent methods and evaluating how much of the difference is statistically significant, considering the combined uncertainties of each method. A novel measurement error model was developed, which consists of systematic measurement errors, repeatability and reproducibility random components. It includes the true calibration factor as an unknown parameter which is estimated by solving Maximum Likelihood Equations (MLE). Formal analysis of the MLE yields expressions for determining the model uncertainty. This is first time that the MLE approach has been used to evaluate the uncertainty of the most important contribution to the reactor thermal power and thus to ensure reactor power compliance. The measurement error model was validated by a Monte Carlo simulation and was found to represent the measurements well. The Monte Carlo simulation also provided an assessment of the magnitude of the model uncertainty.