Charging modules are the most critical component of electric vehicle DC charging piles. Considering the open circuits fault characteristics of core devices such as power switches and electrolytic capacitors, a fault diagnosis method based on wavelet packet transform（WPT） and sparrow search algorithm-back propagation（SSA-BP）neural networks is proposed. The method takes the output voltage of the charging module as the original signal. Firstly its DC component is rejected through pre-processing, and the processed signal is decomposed into wavelet packet. Then the energy of each sub-band signal is calculated, and the initial feature vector is obtained through normalization. Finally the DC component and the normalized feature vector as the final fault feature quantity are put into the SSA-BP neural network, and then the classification results are output to achieve fault diagnosis. In order to verify the feasibility and superiority of this method, a two-stage simulation model with an output of 15 kW is built under different operating conditions. Experiment results show that this method could effectively improve fault diagnosis accuracy with diagnosis rate of 93.85%. And it has practical guiding significance for fault diagnosis of electric vehicle DC charging piles.