Optimizing Charging Control for Medium and Heavy-Duty Electric Vehicles Using Deep Reinforcement Learning

The rapid electrification of medium and heavy-duty vehicles (MHDVs) poses considerable challenges for power systems due to their high energy demands. The simultaneous charging of multiple MHDVs can place significant stress on the electrical grid, potentially causing transformer overloads and driving up operational costs. To preserve grid stability and ensure a reliable power supply, effective management of the charging process is essential. This paper introduces a reinforcement learning (RL) strategy for centrally managing the charging and discharging of MHDVs at a dedicated station. The RL agent dynamically handles vehicles entering and exiting the station throughout a one-day simulation, to minimize energy costs and avoid transformer overloads while ensuring that all vehicles are fully charged before their scheduled departures.