文献类型：期刊文献

英文题名：Data reconciliation method for nuclear power steam turbine unit based on combined robust function and generalized regression neural network

作者：Wang, Huazhong[1];Yao, Xudong[1];Jiang, Qingchao[1]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, POB 293,Meilong Rd 130, Shanghai 200237, Peoples R China

年份：2026

卷号：446

外文期刊名：NUCLEAR ENGINEERING AND DESIGN

收录：;EI(收录号：20254619488419);WOS:【SCI-EXPANDED(收录号:WOS:001616665000002)】；

基金：This work was supported by the National Natural Science Foundation of China under Grant 62322309.

语种：英文

外文关键词：Robust function; Gross error identification; Data reconciliation; Generalized regression neural network

摘要：Nuclear power generation has been widely applied due to its advantages. To ensure the safe operation of nuclear power plants, data-driven monitoring methods for operational status have gained significant attention. This study addresses the issue that random and gross errors in the process data of nuclear power steam turbine unit can affect control quality and degrade the performance of the status monitoring system. A robust function is proposed and applied to the gross error detection and data reconciliation for the nuclear power plant's steam turbine unit. This robust function reduces the impact of standard error, thereby suppressing the propagation of gross error to normal data sources. Additionally, traditional data reconciliation methods, such as Interior Point Method and Sequential Quadratic Programming, often face difficulties in reconciling large datasets in a short time. Therefore, this study proposes a fast data reconciliation method. The proposed method initially employs Sequential Quadratic Programming to reconcile a subset of the measurement data and subsequently utilizes Generalized Regression Neural Network to learn from the results for further reconciliation. Experimental results show that the proposed robust function significantly reduces the false detection rate of gross error, improving the accuracy of gross error identification. This measurement-based detection method allows for non-intrusive diagnostics, effectively minimizing the impact of fault detection on nuclear power production. Furthermore, the method proposed in this study reduces the data reconciliation time for 30,000 sample points from 1330 s to under 9.58 s, demonstrating its strong potential for online reconciliation in practical nuclear power systems.