Power Scheduling and Cost Optimization of a Grid Integrated PV and BESS Fast Charging using SARSA Reinforcement Learning

The global surge in electric vehicle (EV) adoption demands a corresponding expansion of smart EV charging stations, leveraging intelligent management for efficient EV charging. Intelligent EV charging management can efficiently control EV charging scheduling and power flow among the electrical power grid, renewable energy sources, and battery energy storage systems (BESS) to achieve minimized charging costs. This paper proposes the use of the State-Action-Reward-State-Action (SARSA) reinforcement learning (RL) algorithm for power scheduling in a grid-tied model integrating photovoltaics (PV) and BESS of an EV charging station, specifically for fast charging. SARSA RL is selected for its significant potential to effectively manage the complexities associated with power scheduling, thus facilitating the scheduling and optimization of power flow in this scenario. The system employs a day-ahead planning approach, adopting forecasts for solar power generation, grid tariff, and EV charging demand as EV load consumption predictions. In addition, a Markov decision process is used to formulate the optimization problem. The simulation results conclude that the SARSA RL algorithm achieves good rewards by minimizing the cost associated with power scheduling.