1College of Electrical Engineering and New Energy, China Three Gorges University, Yichang 443002, China;2CSG Electric Power Research Institute, Guangzhou 510000, China

The large-scale integration of EVs poses substantial challenges to the stable operation of the power grid. Therefore, accurate load forecasting for EV charging stations is critical for ensuring power grid stability. Nevertheless, charging station loads are influenced by various factors, making accurate forcasting highly challenging. To address the insufficient utilization of spatial factors in existing load forecasting methods for EV charging stations, this paper proposes a forecasting method based on the improved spatio-temporal graph convolutional network (ISTGCN) model. Firstly, a spatio-temporal graph convolutional network (STGCN) is constructed by integrating graph convolutional networks (GCNs) and temporal convolutional networks (TCNs). This architecture extracts spatial features between charging stations from both distance and virtual matrices while effectively overcoming the gradient explosion problem commonly encountered in recurrent models. Secondly, to enhance forecasting accuracy and efficiency, a convolutional block attention module (CBAM) is introduced to further facilitate information flow within the network. Furthermore, to address the difficulty of hyperparameter tuning, the crested porcupine optimizer (CPO) algorithm, known for its superior global search capability, is incorporated into the model to improve forecasting performance. Finally, short-term load forecasting is conducted based on real-world data from charging stations in a specific city. The experimental results demonstrate that the proposed method has significant advantages in forecasting performance.