Short-Term Load Forecasting of EV Charging Stations via Adaptive K-means Clustering and Deep Hybrid Networks

With the rapid increase in the number of electric vehicles, the strong time-variability and behavioral diversity of charging loads have posed significant challenges to the stable operation and planning of urban distribution networks. To address this issue, this paper proposes a short-term load forecasting method that combines a Newton–Raphson optimized K-means clustering algorithm with an integrated deep learning model. First, charging stations are adaptively clustered according to their load profiles, improving the segmentation accuracy under different charging behaviors. Then, for each cluster, personalized prediction models are developed by integrating multi-source data such as historical load, meteorological information, and electricity price, utilizing a combination of convolutional neural networks (CNN), long short-term memory networks (LSTM), and attention mechanisms (AM). Finally, the effectiveness of the proposed approach is validated with real-world regional data, and the results demonstrate significant improvements in both forecasting accuracy and generalization capability.