基于小波变换和神经网络的逆变电路故障诊断方法

An Fault Diagnosis Method for Inverter Based on Wavelet Analysis and Neural Network

提出一种基于小波分析和神经网络的逆变器开路故障诊断方法;采用小波变换方法将逆变器的三相输出电压分解为高频系数和低频系数,以三相低频系数的平方和作为该相输出电压的特征向量,将逆变器开路故障进行分类和编码,建立一个三个输入、五个中间节点、一个输出的神经网络模型,实现逆变器故障桥臂定位,最后利用逆变三相电压同一桥臂故障电压的对称性的特点,用一种简单的判断逻辑实现故障元件的分离;仿真结果表明,该模型的诊断准确率达到98.6%以上,表明方法的有效性。

A novel algorithm based on wavelet analysis and neural network is presented to diagnose opening faults in inverter. The output three--phase voltages of inverter are decomposed into high frequency coefficients and low frequency coefficients by wavelet analysis method. The quadratic sums of low frequency coefficients of three-- phase voltages are calculated to be acted as the feature vectors. The break faults of inverter are classified and coded. Then the three quadratic sums are acted as inputs, the code of opening fault of inverter is acted as outputs, and a neural--network based model including five middle nodes is constructed to locate the break faults of inverter bridges. At last, a simple judge strategy for locating the fault switches based on the symmetry of three--phase fault voltage is presented. The simulation results show the accuracy of the model reaches 98. 6% and it is practical to detect the fault of the inverter.