基于改进类均值核主元分析的控制系统传感器故障检测  

Sensor Fault Detection of Control System Based on Improved Class Mean Kernel Principle Component Analysis

作　　者：王印松[1] 蔡博 焦阳 朱向伟 WANG Yinsong CAI Bo JIAO Yang ZHU Xiangwei(School of Electrical and Electronic Engineering, North China Electric Power University, Baoding 071003, China)

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003

出　　处：《电力科学与工程》2017年第9期51-55,共5页Electric Power Science and Engineering

基　　金：中央高校基本科研业务费专项资金(9161715008)

摘　　要：针对复杂控制系统数据维度大、变量之间的耦合性高的特点,采用了一种基于粒子群优化的类均值核主元分析的故障检测方法。首先利用粒子群优化高斯径向基核函数的参数,避免其设置的盲目性,然后利用优化后的类均值核主元分析法将输入数据样本映射到高维特征空间中,构建类均值矢量进行主元分析,完成对控制系统传感器的故障检测。类均值矢量包含了原数据的全部信息,且维数低于故障类别,能够实现数据的无损失降维。实验结果表明,与传统核主元分析相比,该方法能有效提高控制系统传感器故障检测的准确性。In view of the large dimension of the data and the high coupling among variables in complex control sys- tems, a fault detection method based on particle swarm optimization using class mean kernel component analysisis pro- posed in this paper. The parameters of the Gauss radial basis function kernel function are optimized by particle swarm first to avoid the setting blindness, and then the input samples are mapped to high dimensional feature space by using the class mean kernel component analysis method to construct the class mean vectors by principal component analysis, fault detection and control system of sensor. The class average vector contains all the information of the original data, and the dimension is lower than the fault category, so it can realize the lossless reduction of the data. The experimental results show that compared with the traditional kernel principle component analysis, this method can effectively improve the accuracy of sensor fault detection in control system.

关 键 词：控制系统 传感器 粒子群优化 高斯径向基核函数 类均值核主元分析 故障检测 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]