Load Forecast and Regulation Capacity Assessment for Charging Stations Based on Probability Statistics and LightGBM

This paper proposes a charging load prediction algorithm that combines clustering algorithms with machine learning models, along with an assessment method for its adjustable capacity. First, typical charging load curves for each charging station are derived using clustering algorithms, which are used as input features for the LightGBM prediction model. Additionally, features such as weather and date information are also considered as inputs to the LightGBM prediction model. After training the model, prediction models for the load of each charging station are obtained. Secondly, the prediction results of the load and historical load probability statistics are combined to assess the adjustable capacity of the charging stations' load during different time periods in the future.