Enhanced wind power prediction using feature selection: Implications for grid integration of renewable energy and electric vehicles

In recent years, wind power and electric vehicles (EVs) have received increasing attention for their environmental benefits. However, the intermittent nature of wind and the irregular charging behavior of EVs pose challenges to power system reliability and cost-effective operation. This study proposes a GTO-CNN-BiLSTM model with integrated feature selection for short-term wind power forecasting, addressing accuracy limitations arising from the complex and nonlinear characteristics of wind energy. Recursive feature elimination (RFE) is employed to remove redundant inputs and retain the most informative predictors, improving model generalization. Experimental results show that applying RFE increases the average R 2 from 0.9127 to 0.9213, and incorporating GTO further raises it to 0.9526. To illustrate practical applicability, a localized demand response scenario is simulated in which a wind farm is directly connected to an EV charging hub. This setup demonstrates that improved forecasting enables effective coordination between renewable generation and EV charging, supporting local energy management, reducing renewable curtailment, and enhancing utilization efficiency. The study emphasizes methodological advancement in short-term wind power forecasting, with EV charging coordination presented as a representative application.