Soft Fault Diagnosis for DC-DC Converters with Wavelet Transform and Fuzzy Cerebellar Model Neural Networks

Identifying the soft faults of converters in power electronic converter is a significant problem for the stable and efficient operation of power systems. This paper proposed a novel soft fault diagnosis method based on wavelet transform and fuzzy cerebellar model neural networks (WT-FCMNN) for DC-DC Converters.

First, the multiscale feature extraction is achieved by multilevel signal decomposition to extract the feature information of different frequency ranges signal. Meanwhile, optimal wavelet decomposition scale and feature dimension reduction are used to reduce computational quantity and eliminate redundant information.

Then, in order to effectively diagnose soft faults in DC-DC converters, a classifier based on FCMNN is proposed to identify different operating states of the capacitor and power MOSFETs in push-pull circuits. Finally, two common fault diagnosis methods and the proposed FCMNN are performed for circuit fault diagnosis.

Compared with the BPNN and SVM, simulation results show that the proposed method has a better generalization, fast diagnosis speed and higher diagnostic accuracy that proves its effectiveness and feasibility in soft fault diagnosis.