Electric Vehicle Charging/Discharging Control Method based on Deep Reinforcement Learning for Photovoltaic Energy Storage Stations

With the implementation of Internet of Things technology and the proliferation of electric vehicles (EVs), real-time control of EV charging/discharging is one of the keys to ensuring the safe, stable and efficient operation of smart grids. Nevertheless, existing EV charging/discharging control methods fail to comprehensively consider photovoltaic power generation, dynamic energy price, and the randomness and uncertainty of EV loads. This article proposes an EV charging/discharging control method for photovoltaic storage stations based on deep reinforcement learning (DRL). Considering the EV charging demand, photovoltaic energy output, and time-of-use electricity price, under the conditions of meeting the EV charging demand and photovoltaic energy utilization efficiency, the Markov decision process (MDP) of EV charging/discharging control based on DRL is established with the goal of minimizing user charging costs and enhancing user satisfaction. Considering the continuity of EV charging/ discharging action, an improved deep deterministic policy gradient (DDPG) algorithm is applied for the above scenario. Finally, simulation results are conducted on the proposed method, and comparable with disordered EV charging. The results verify that the recommend approach and model can significantly absorb photovoltaic energy and enhance user satisfaction while saving user costs by 27. 94%.