Abstract: Focusing on the issue of power converter fault diagnosis, a method of power converter faults diagnosis based on wavelet packet energy spectrum and M-ary support vector machine(SVM) is proposed. Firstly, the wavelet packet decomposition is adopted to extract the output voltage energy values, and the fast Fourier transform (FFT) is adopted to analyze the frequency feature points, perform the dimension reduction and obtain the fault feature vectors. Then， various fault modes are isolated based on the M-ary SVM diagnostic model． Experiment results show that compared with traditional BP neural network and one-against-one SVM fault diagnosis methods， the proposed method has high diagnostic accuracy and needs far less number of sub-classifiers, and the diagnosis speed is improved a lot. The proposed method is also suitable for online diagnosis.


Key words: M-ary support vector machine, wavelet packet decomposition, feature extraction, power converter, fault diagnosis