A Hybrid Multi-model Ensemble Feature Selection and SVR Prediction Approach for Accurate Electric Vehicle Demand Prediction: A US Case Study

Electric vehicles (EVs) have become an essential element of smart transportation systems within smart grids, providing significant environmental and economic benefits by reducing fossil fuel consumption and featuring zero-pollution technology. However, the behavior of EV drivers and their energy demands are unpredictable and intermittent, posing challenges for power system operators when integrating a large number of EV s. To address this issue, an efficient prediction of EV s' charging demand is crucial. In this paper, a novel hybrid ensemble feature selection SVR-based model is proposed to predict the energy demand of EV's charging stations. The proposed method consists of two main steps. Firstly, an ensemble multi-model feature selection approach using four different algorithms is developed to identify the most relevant subset of features. Subsequently, the chosen attributes are employed as input for SVR to predict EV s electricity load. The effectiveness of the proposed method is assessed by evaluating its performance on the energy load data of EV s at a charging station in Palo Alto, USA. The results demonstrate its superiority in predicting EV s' charging demand when compared to individual feature selection techniques.