Aiming at the problems faced in DC charging pile rectifier fault diagnosis, such as high fault data redundancy, low efficiency of temporal feature extraction, and difficulty of model tuning, a DC charging pile fault diagnosis system based on Bidirectional Long-Short Term Memory network optimised by time convolution network and improved sparrow search algorithm is proposed. Firstly, the simulation model of the DC charging pile rectifier is established to obtain different open-circuit fault signals of the rectifier; secondly, the variational modal decomposition is used to obtain the information of different frequency bands; subsequently, the time-frequency domain features of the signals are extracted by the wavelet transform, and combined with the quadratic feature extraction and the normalisation process, the sparse self-encoder is used to sparsely represent the fault features and retain the main features of the signals; then, the time convolutional network is used to extract local features from the sparse data. Sin chaotic mapping, Cauchy variational operator and inverse learning strategy are introduced to improve the sparrow search algorithm, which is used to optimise the hyper-parameters of the bidirectional long and short-term memory network; finally, the method proposed in this paper is used to carry out fault diagnosis of DC charging pile rectifier. The experimental results show that the fault diagnosis accuracy of the proposed method is as high as 97.5%, which is better than that of the traditional Bidirectional Long-Short Term Memory network.