An Estimation Method For Bias Error Of Measurements By Utilizing Process Data, An Incidence Matrix And A Reference Instrument For Data Validation And Reconciliation

With further applications of AI, IoT, and digital twin technology to plant operation and maintenance, it is becoming increasingly important to ensure data reliability. Data validation and reconciliation (DVR) represents one promising technique to ensure data reliability by minimizing the uncertainty of measurements based on statistics. DVR has been widely applied to nuclear power electrical generation plants in Europe and the United States in recent years. The most important input for DVR analysis is measurement uncertainty. In Japan, performance management of nuclear power plants is often done by measuring condensate flow rate. While the uncertainty of other flowmeters is handled by the JIS standard, the condensate flowmeter is specially calibrated every few cycles. This leads to reduction of effectiveness of DVR analysis due to variations in measurement uncertainty management. To overcome this issue, we propose an estimation method for measurement uncertainty by utilizing process data, an incidence matrix between sensors, and a reference instrument. The conventional method proposed in the previous study only treats the random error. The proposed method quantitatively estimates not only random error but also bias error by considering the uncertainty of the reference instrument. Using several benchmark problems, we found that the proposed method was applicable to various flow conditions, including physically fluctuating flow such as that observed in the feedwater flow in nuclear power plants. We anticipate that the proposed method will promote use of DVR analysis in nuclear power plants in Japan.