Multi-Timescale Electric Vehicle Charging Load Forecasting Using RCEL-Net: A Residual CNN-EA-LSTM Network

The rapid growth of electric vehicle (EV) charging loads presents significant challenges to grid stability and load management. Therefore, establishing a high-precision multi-timescale charging load forecasting model has become a key requirement for grid regulation and resource optimization. The primary objective of load forecasting is to provide a reliable baseline for the day-ahead declared power curves at electric vehicle charging stations, supporting subsequent load control strategies to ensure the stable operation of the system. This study introduces a multi-timescale EV charging load forecasting approach using the RCEL-Net (Residual CNN-EA-LSTM) model. RCEL-Net combines Convolutional Neural Networks (CNN) with an encoder-attention mechanism in the LSTM structure, enabling parallel processing of spatial and temporal information. This allows the model to extract key features and enhance its response to crucial information in historical data, effectively capturing the nonlinear and multi-scale characteristics of EV charging loads. Experimental validation across different timescales shows that the proposed RCEL-Net model significantly outperforms traditional methods in prediction accuracy, providing solid data support and theoretical foundation for charging station load regulation and grid stability maintenance.