Fault Diagnosis of Modular Multilevel Converter Based on RNN and Wavelet Analysis

Modular multilevel converter (MMC) is very common in DC transmission system. But the power electronic components in the MMC sub-module usually fail and affect the operation of the circuit. Consequently, a method of MMC fault diagnosis based on recurrent neural network (RNN) and wavelet analysis is proposed. This proposed method combines the wavelet transform and the energy spectrum entropy to reduce noise interference and extract characteristic signals more effectively. During the experiment, the wavelet transform is used for signal processing, and then the fault signal is extracted by the wavelet entropy. After the fault feature data are obtained, the fault diagnosis model is established by using the RNN. Finally verified by experiment, the validity and the high accuracy of the new method are showed by comparing with the method based on the RNN and the BP with wavelet analysis.