Deep learning modeling and regulation method for spatio-temporal distribution characteristics of charging piles for power grid stability

This study addresses the challenges to power grid stability posed by the increasing integration of electric vehicles (EVs) through a novel deep learning approach. We propose a comprehensive framework featuring a Graph Convolutional Network and Long Short–Term Memory (GCN–LSTM) model to accurately forecast EV charging loads, capturing both spatial and temporal dependencies. Additionally, a bidding decision model for charging operators and a hierarchical power control strategy are developed to optimize participation in ancillary service markets and effectively respond to grid dispatch commands. Experimental results using real-world data from the North China ancillary service market demonstrate the superior prediction accuracy of the GCN–LSTM model and validate the effectiveness of the proposed methods in enhancing grid stability and operator profitability. The framework presents a significant advancement in large-scale vehicle-grid interaction, offering a robust solution for the future integration of EVs into power grids.