Abstract: The open-circuit faults of EV charging piles affect power quality of power grid and threaten charging safety. Researches on the open-circuit fault diagnosis are of great significance to ensuring the safe and stable of the power grid and reducing the maintenance cost of the charging pile. According to the multidimensional characteristics of open circuit fault signals of charging piles, a tensor reshape fusion diagnostic model is proposed in this paper. In this method, the multidimensional feature parallel extraction ability of residual network (ResNet) and the sequential feature extraction ability of gated recurrent unit (GRU) are used to extract the front and rear stage fault features of the charging pile circuit, the fusion diagnosis is performed for front and rear stage features, and the high accuracy fault diagnosis of front and rear stage of charging pile is realized. A pre-processing method of front-stage three-phase fault signals based on tensor reshape method is proposed, which avoids the graphical input or one-dimensional input used by traditional deep learning algorithms and gives full play to the diagnostic performance of deep neural network. Compared with the traditional fault diagnosis method, the proposed method uses deep learning technology, which does not require manual selection of fault characteristic parameters. The simulation experiments show that the average accuracy of the proposed method can reach above 96% under the influence of different intensity noise; especially in the case of high noise with signal-noise ratio (SNR) of 10dB, the accuracy is still above 90%, and the accuracy of this method reaches 94.54% in experimental, which further verifies the effectiveness of this method.