基于新小波神经网络的电力电子电路故障诊断  

Fault Diagnosis of Power Electronic Circuits Based on New Wavelet Neural Networks

作　　者：龙伯华[1] 谭阳红[1] 许慧[1] 文娟[1] 

机构地区：[1]湖南大学电气与信息工程学院,湖南长沙410012

出　　处：《计算机仿真》2009年第5期266-270,共5页Computer Simulation

基　　金：国家自然科学基金(50277010);湖南省自然科学基金(03JJY1010;04FJ2003;05GK2005);教育部博士基金(20020532016);高等学校新世纪优秀人才支持计划(NCET-04-0767)

摘　　要：电力电子电路模型具有非线性,所以对它进行故障诊断比较困难。针对这种情况,将小波变换的时频局部化特性和神经网络的非线性映射与学习推理的优点结合起来,建立一种新的小波神经网络。并以双桥12相脉波整流电路为例,首先列出电路的故障类型,并用软件模拟电路的各种故障下的故障信号,然后使用建立的小波神经网络学习和存储双桥12相脉波整流电路的故障信号和故障类型之间的映射关系,实现电路的故障诊断。仿真结果证明了这种故障诊断方法的正确性和准确性。A new wavelet neural network is presented by combining the properties of time - frequency localization of wavelet transform with the advantages of non - linear mapping and syntheszing of neural network to deal with the difficulty in diagnosing the power electronic circuit models due to their non - linearity . Taking the Twinbridge 12 phase pulse wave rectifier circuit as an example, frist,the fault features are listd, and the fault signals relating to the fault features are obtained through softwave, and then the fault diagnosis is realized by using the wavelet network to learn and save the mapping relations between fault models and fault features of the circuit. The simulation results show that this method is correct and accurate.

关 键 词：电力电子电路 故障诊断 脉波整流电路 小波神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]