Optimal Control Strategy for Plug-in Electric Vehicles Based on Reinforcement Learning in Distribution Networks

Electric vehicles (EVs) as distributed storage devices have the potential to provide frequency regulation services due to the fast adjustment of charging/discharging power. Along with the policy incentives, it is practical for EVs to take part in the regulation market through the aggregator. An optimal control strategy based on reinforcement learning (RL) for electric vehicles (EVs) in distributed networks is proposed in this paper. The overall goal is to follow the regulation signals sent by the system operator in the real time regulation market by controlling the EVs in the parking lot. To achieve this, the reinforcement learning algorithm is employed to optimize the charge and discharge strategy of the EVs, so that the aggregator optimally allocates the regulation power and the baseline charging power to EVs to respond to the regulation signals for the best regulation performance. Comprehensive simulation studies have been carried out based on the data of PJM electricity market and the results show that the regulation performance based on the control strategy is excellent in both cases of traditional and dynamic regulation signals.