Reward Based Charging and Discharging of Electric Vehicles Using Deep Q-learning and Stackelberg Game

The growing adoption of electric vehicles (EVs) poses significant challenges for grid operators, especially in managing load fluctuations during peak charging times. This paper introduces a strategic optimization framework that leverages Stackelberg game theory combined with deep Q-learning (DQL) to efficiently manage grid load and optimize EV energy consumption. Within this framework, the grid operator acts as the leader, aiming to minimize peak grid load by incentivizing EVs to discharge energy during peak hours. This is achieved through a reward mechanism that encourages EV owners, the followers in this game, to maximize their returns from energy transactions while considering the costs associated with battery degradation. Participants decide on their involvement in the vehicle-to-grid (V2G) energy exchange by weighing the rewards against degradation costs. The integration of a deep Q-network (DQN) with the Stackelberg model allows EVs to autonomously learn and adapt their charging and discharging strategies over time. The study compares the performance of a conventional Stackelberg game against our DQN-enhanced Stackelberg game, revealing that the latter markedly improves the optimization of charging and discharging strategies. The results demonstrate the effectiveness of the proposed approach in optimizing grid load, minimizing peak demand, and maximizing EV owner benefits. This framework paves the way for a more resilient and cooperative grid management system in the face of EV proliferation.