Data-Driven Short-Term Electric Vehicle Charging Station Demand Forecasting

Electric vehicles (EVs) could effectively reduce the dependence of vehicle industry on fossil fuels; however, the complex charging behavior and large power consumption by EVs induce challenges to the operation of power grid. Accurate forecasting of power demand of EV charging station can provide important information for the operation planning of power grid. In this paper, a data-driven framework is proposed for short-term EV charging station demand forecasting. To fully mine the potential of machine learning algorithms in EV charging station demand forecasting, forecasting models based on different ma-chine learning algorithms have been developed including artificial neural network (ANN), support vector regression (SVR), and boosted regression trees (BRT). The performance of a machine learning algorithm highly depends on its hyperpa-rameters. To improve the forecasting performance, a hyperparameter opti-mization method is further developed based on Rao-l algorithm. To verify the effectiveness of the proposed data-driven framework for EV charging station demand forecasting, extensive computational experiments based on real data from different EV charging stations have been conducted. Experimental results illustrate that the proposed data-driven forecasting models perform much better than the widely considered persistence model while the ANN model and the BRT model outperforms the SVR model.