A Fault Warning Method Based on Modified ARMA-GRU Network for EV Charging Piles

As electric vehicles are gradually being promoted and popularized, the condition monitoring and maintenance problem of charging piles have garnered increasing attention from researchers and charging station operators. In this paper, a fault warning method based on modified ARMA-GRU network is proposed to address the fault warning problem for EV charging piles. In the proposed method, the multimodal signals from electric vehicle charging piles are firstly preprocessed using autoregressive moving average sifting mechanism, thus the anomaly data points are sifted to avoid the inaccurate results caused by measurement errors. Then the pre-processed data is used to train and test the modified gated recurrent unit (GRU) networks with modified cell mechanism. Thus, the working condition monitoring and fault warning can be achieved utilizing the memory characteristics for time-varying signals of GRU. The simulation results shows that the proposed method has higher fault warning accuracy for the condition monitoring problem of EV charging piles.