Optimal charging schedules for EV charging stations considering hybrid smart and uncontrolled charging: A scalable framework

The randomness, temporal variability, and extended idle connection time of electric vehicle (EV) charging behavior impose significant load pressure and regulatory challenges on the grid and charging facility operations. Most studies have focused exclusively on smart charging, often overlooking the impact of uncontrolled charging. This singular focus has created a discrepancy between charging scheduling strategies and real-world conditions. To address these issues, this study investigated the factors influencing smart charging choices through a survey conducted in Jiangsu, China, and developed a smart charging choice model. Based on this model, a charging schedule method that integrates both smart and uncontrolled charging modes at stations was proposed. An energy boundary model and a relaxation mechanism for hybrid charging were employed to ensure alignment with charging demands. The charging process was modeled as a markov decision process, and a decentralized framework was proposed to provide charging power to each EV, using deep deterministic policy gradient reinforcement learning algorithms to determine charging strategies for multiple heterogeneous EVs. Numerical experiments confirm that the proposed method effectively reduces charging costs and peak loads at charging stations, and manages both homogeneous and heterogeneous charging demands. Additionally, centralized training of the decentralized framework demonstrates performance consistency across multiple charging units while consuming fewer training resources, thereby enhancing scalability.