DRL-Based charging power and price optimization framework for EV charging stations

Abstract

With the rapid advancement of electric vehicles (EVs) in recent years, the role of charging stations has become increasingly important. One critical challenge of this is the optimal scheduling of charging power among EVs at charging stations while maximizing the profit of the electric vehicle charging station (EVCS). Existing research has not adequately focussed on predicting both charging price and charging power considering a lot of factors simultaneously. To address this gap, deep reinforcement learning (DRL) is leveraged along with the Proximal Policy Optimization (PPO) algorithm to predict optimal charging power and charging price at charging stations, considering user preferences, battery data, grid prices, and forecasts, EV energy demand forecasts, as well as the maximum available grid demand. The proposed method is an online and modelfree approach. Our simulations demonstrate that the proposed method can achieve higher profit compared to the independent operation of the charging station.