Research on Fault Identification Method of Series Arc in Electric Vehicle Charging System Based on Machine Learning

In this paper, the fault detection method of DC series arc in electric vehicle charging system based on deep neural network is studied. As the arc current of series fault is easily influenced by high-frequency noise in the charging system of electric vehicle, it is difficult for traditional fault identification methods in time-frequency domain to detect accurately. Machine learning-based fault detection approaches can be trained by models to obtain better recognition accuracy. By comparing and analyzing the characteristics of Goo-gleNet, AlexNet, VGG16 and other common network models, the VGG16 network with relatively simple structure and high recognition accuracy is selected to design the arc fault detection algorithm, and the arc current spectrum data is used to train and verify the arc fault identification algorithm. The outcomes indicate that the model achieves a recognition accuracy of over 98%.