Fault Diagnosis of Phase Shifted Full Bridge Converter Based on Wavelet Packet and Extreme Learning Machine Optimized by Firefly Algorithm

Abstract

As auxiliary power supply of locomotive, phase shifted full bridge converter has complex fault type and large amount of fault information. A novel fault diagnosis method for PSFB convert based on wavelet packet transform and extreme learning machine as classifier which is optimized firefly algorithm is proposed. The WPT is used to extract the feature vector of the output voltage of PSFB converter which is taken as the research object. Then ELM is employed to classify the feature vector of the output voltage, and the FA optimizes the input weights and the hidden layer of the ELM, which improves the classification performance of ELM. The simulation results show that the method can effectively diagnose the PSFB converter.