Fault Diagnosis Of Power Electronic Circuit Based On Wavelet Transform-Neural Network

With the rapid development of the power electronic technology, it has great significance to study power electronic circuit fault diagnosis technology. According to the characteristics of the output voltage waveform in power electronic circuits, this paper uses wavelet transform to extract faulty features, forming feature vector, and use neural network as a fault classifier, achieve diagnosis of open fault in a three-phase full-controlled thyristor bridge rectifier circuit.In this paper, at first, the significance and background of the power electronic circuit fault diagnosis are introduced, and the characteristics of power electronic circuit fault diagnosis fault diagnosis are explained, the basic theory of wavelet transform is summarized, and various properties of the wavelet base are analyzed in order to choice wavelet bases. Taking the Three-phase full-controlled rectifier circuit for example, study the simulation method of the circuit failure based on the Matlab / Simulink, the rectifier output waveform and the raw data of various fault status are obtained. Wavelet transform is used to deal with faculty signal, based on the good characteristics of wavelet analysis at time-frequency localization, the wavelet transform modulus maxima is used to detect of faculty the mutation point. The analysis of wavelet multi-resolution is also used to decompose the signal in different frequency bands and the energy statistic of wavelet coefficients is analyzed in the each frequency band. The feature vectors formed by feature signal extract the energy value of original data in different frequency bands as a feature vector of the faulty classification. BP neural network is used as a fault classifier with its good characteristics of nonlinear mapping. This paper gives the simulation results which based on the design methods and practical training - testing of BP neural network in Matlab, and the simulation results show the BP neural network fault diagnosis system is effective.