Fault Arc Detection Model Based on LSTM-Transformer

In order to quickly and accurately detect electric vehicle series arc faults, this paper builds an electric vehicle fault arc experimental platform and obtains data samples of power supply terminal voltage and trunk current under different working conditions. At present, most detection models only perform fault analysis through current signal characteristics, ignoring voltage and other signal characteristics. Based on the traditional single-layer LSTM model, a series arc fault detection model combined with LSTM-Transformer is proposed, and the signal characteristics of voltage and current are used to identify faults. Experimental results show that the fault arc recognition rate of the model reaches about 97% under all working conditions. It provides a feasible scheme for arc fault detection in electric vehicle electrical system.