A Novel Forecasting Approach to Schedule Electric Vehicle Charging Using Real-Time Data

Abstract

The rapid adoption of electric vehicle (EV) has increased the need for precise demand estimates to ensure grid stability, reduce operational costs, and strategically plan the expansion of charging stations. Existing forecasting approaches struggle to capture the complexity and change of the EV load patterns, especially over time. The effective development and optimization of the charging infrastructure are critically dependent on accurate EV load forecasting. This paper proposes a hybrid forecasting approach that combines long short-term memory models with advanced decomposition methods like empirical mode decomposition, ensemble empirical mode decomposition, and complete ensemble empirical mode decomposition using adaptive noise and seasonal-trend decomposition to address this challenge. The proposed framework is tested for 15, 30, 60, and 120 min to show its adaptability and robustness. Statistical evaluations show that decomposition approaches using long short-term memory increase predicting accuracy across all time intervals. STL-LSTM reduces the forecast error by 52.38% between hybrid methods. Kolmogorov–Smirnov, Shapiro–Wilk, and t-tests confirm the results, improving consistency and dependability. This paper shows that hybrid decomposition-based forecasting models can scale and accurately manage future EV charging demands, overcoming the limits of traditional techniques.