Prediction of Aggregated EV Representation Using XGBoost and LightGBM

Electrical vehicles (EVs) presence forecasting on a parking lot with charging points is critical for charging schedule optimisation with providing demand response to the power grid. The paper leans on our previous work where we proposed innovative aggregated representation of EV population. The historical dataset is transformed in the aggregated representation, and it is analysed and used for training of two gradient boosting models in order to forecast future EV population. The predictions are generated on 2 hours forecasting horizon. Achieved accuracy is compared to week-of-day average and persistence.