基于最小二乘支持向量机的电力电子电路故障诊断应用研究

Research on Fault Diagnosis of Power Electronic Circuits Based on LS-SVM

采用最小二乘支持向量机预测算法对电力电子电路进行故障预测.以基本降压斩波电路为例,选择电路输出电压作为监测信号,提取输出电压平均值及纹波值作为电路特征性能参数,并利用LS-SVM回归预测算法实现故障预测.仿真结果表明,利用LS-SVM对基本降压斩波电路输出平均电压与输出纹波电压的预测相对误差均低于2%,能够跟踪故障特征性能参数的变化趋势,有效实现电力电子电路故障预测.

The paper proposes the least square support vector machine forecasting algorithm for the power electronic circuit fault prediction. With basic buck-chopper circuit, choose circuit output voltage signal as monitoring signal, extract output voltage ripple and average value as circuit features performance parameters, then using LS-SVM regression algorithm to the fault prediction. The experimental result shows that the use of the output circuit LS-SVM average voltage and output voltage ripple of the relative prediction error less than 2%, it can follow the fault feature performance parameters change trend, realize the power electronic circuit fault prediction effectively.