基于小波分析及网络的电力电子电路故障诊断方法

针对电力电子电路故障，结合小波多分辨分析，将信号的特征提取作为网络的第一层，用小波函数代替普通神经网络中的S函数，提出了函数型和权值型两种不同的4层小波神经网络方法，给出了相应的数学模型和学习算法。以三相整流桥电路为例，建立了小波神经网络的输出与故障元之间的对应关系，实现了电路的故障诊断，并与用普通BP网络诊断的结果进行了比较。仿真结果验证了两种方法的正确性。

Fault diagnosis methods of power electronic circuits based on wavelet analysis and network

In this paper, two different four - layer wavelet neural networks, in which the features of signal are extracted as first - layer of network and the S - function in general neural network are replaced by wavelet function, are constructed integrated with wavelet analysis for fault diagnosis of power electronic circuits, and the mathematical model and learning arithmetic are given. In three -phase rectifier circuit, based on wavelet neural networks constructed, the relationship between the output of wavelet neural network and the faults is founded, the fault diagnosis of circuits is achieved, and the diagnosis results by two networks are compared with its by general BP network. The two methods are verified by simulation.