基于小波神经网络的逆变器开路故障诊断方法研究

Research on Open-circuit Fault Diagnosis Method of Inverter Based on Wavelet Neural Network

针对逆变器常见的开路故障,提出了一种基于小波神经网络的故障诊断方法。对逆变器功率管开路故障进行分类和编码,在MATLAB/Simulinlc仿真环境下构建三相电压型逆变器系统,以输出端的三相电流作为特征参数,提取经db4小波基函数5层分解后的信号低频系数能量值和计算得到的波形辅量值构成故障的六维特征向量。最后将得到的学习样本用于BP神经网络的学习训练,对测试样本加入噪声后输入到训练好的网络进行测试,能准确诊断出逆变器不同模式的开路故障。

Oriented to the open-circuit faults of inverters, a diagnosis method based on wavelet neural network is proposed. Firstly, the inverter transistor's open-circuit faults are classified and coded. The threephase voltage source inverter system is constructed in the MATLAB/Simulink simulation environment from which we can get the three-phase current signal used as the characteristic parameter, the signal low frequency coefficient energy value that extracted by db4 wavelet basis function as well as the calculated waveform auxiliary value form a six-dimensional feature vector. Finally, the BP neural network that trained by the obtained learning samples, is tested by the test samples with added noise, and the results show that various types of open-circuit faults of the inverter can be accurately diagnosed.