An Early Warning Protection Method for Electric Vehicle Charging Based on the Hybrid Neural Network Model

Abstract

During the charging process of the electric vehicle (EV), a spontaneous combustion accident may occur due to overheating of the battery, causing personal danger and property damage. To address the charging safety of EVs, this paper proposes a new hybrid EV charging process early warning protection method by combining Convolutional Long-Short Term Memory (ConvLSTM), the sliding window method, and the residual analysis method. The method is fully trained by extracting the deep features of EV charging data through ConvLSTM, eliminating the influence of erroneous transmission data through the sliding window method, and setting a reasonable warning threshold through the residual analysis method. The cross-validation results showed that among the four training sets, the ConvLSTM model of training, set three, had the highest prediction accuracy compared with the CNN, LSTM, BiLSTM and CNN-LSTM models, with RMSE reaching 0.029, MAPE reaching 11.37, and r2 reaching 0.89. Training set one had the worst prediction in the four training sets, and after using it to set the warning threshold, the alarm task was completed five sampling points earlier. Therefore, the hybrid model can quickly complete the safety warning task, thereby ensuring the safety of EV charging.