Electric Vehicle Charging Load Prediction based on Graph Attention Networks and Autoformer

With the widespread popularity of electric vehicles (EVs), the problem of energy structure could be alleviated, but it also increases the pressure on the power supply side, so the charging load prediction has a wide range of application scenarios and a huge commercial value. The most of existing EV charging load forecasting methods are modeled from the perspective of charging stations, ignoring the travel habits and charging needs from the perspective of users. In this paper, a temporal spatial neural network model based on graph attention and Autoformer is proposed to predict EV charging load, and a spatiotemporal graph data set based on user travel trajectory is constructed. The experimental results show that the proposed method can fully tap the distribution of user clusters in time and geographical space, to effectively improve the accuracy of charging load prediction.