基于主元分析的小型压水堆故障检测和辨识方法研究

Research on Fault Detection and Identification Method of Small PWR Based on Principal Component Analysis

故障检测和辨识对于小型压水堆的安全经济运行具有重要意义。反应堆中通常采用基于信号和专家知识经验的故障检测和辨识方法,操纵员往往不能从海量的故障数据信息中及时准确甄别故障类型,追溯故障原因。本文提出了采用主元分析进行小型压水堆故障检测和辨识的方法。首先利用RELAP5程序对小型压水堆建模,获得典型故障的样本数据。其次,基于主元分析理论对样本降维,并计算T^(2)和Q两个统计量,通过判断是否超出阈值来检测反应堆运行状态。然后,利用贡献率图方法分析了过程变量对于统计量的贡献率,从而确定了对故障特征变化起主要作用的变量,实现对不同故障的辨识。最终和实际物理过程分析结果进行对比,验证了该方法的有效性。

Fault detection and identification are important for the safety and economy of small PWRs. The fault detection and identification method based on signal and expert knowledge and experience is usually applied in nuclear reactors. However, operators are often unable to identify the fault type and trace the fault cause in time and accurately from the massive fault data information. A method of fault detection and identification of small PWR based on principal component analysis is presented in this paper. First, the model of a small PWR is established by RELAP5 code, and the sample data of typical faults is obtained. Second, the dimension of sample data is reduced by using principle component analysis method. T^(2) and Q statistics are calculated to detect the reactor operation condition by judging whether the thresholds are exceeded. Then, the contribution rate of process variables to statistics is analyzed by using the contribution rate graph method, so as to determine the variables that play a major role in the change of fault characteristics and realize the identification of different faults. Finally, the effectiveness of the method is verified by comparing with the actual physical process analysis results.