Fault Diagnosis of DC Charging Pile Based on Convolutional Neural Network and Wavelet Packet Feature Extraction

Electric vehicle DC charging stations have always been plagued by frequent malfunctions, difficult maintenance, and high repair costs, but traditional fault detection methods are inefficient. Therefore, a diagnostic method is proposed for the operational status of DC charging station charging modules based on wavelet packet decomposition and convolutional neural networks (CNN), capable of detecting abnormalities and pinpointing faults in the charging modules. Initially, the paper models the DC charging station charging modules. After obtaining current data, it uses wavelet packet decomposition to extract feature vectors to form a dataset, which is then used to train the constructed CNN network model. Simulation verification demonstrates the superiority of the CNN model in identifying faults in DC charging station charging modules.