Nash-MADDPG Improves V2V Energy Trading by 61.6%

The newly developed Nash-MADDPG framework employs multi-agent reinforcement learning by incorporating the Nash Bargaining Solution into the Multi-Agent Deep Deterministic Policy Gradient, facilitating vehicle-to-vehicle (V2V) energy trading among electric vehicles (EVs). This method promotes decentralized peer-to-peer energy transactions, thereby decreasing reliance on the grid and allowing for the monetization of excess capacity. Efficient bilateral pricing is established through Nash bargaining, and rewards based on Nash-guided price proximity steer agents towards optimal bargaining strategies. A 30-day continuous operation evaluation reveals a 61.6% enhancement in social welfare and a 62.9% increase in trading volume compared to Double Auction, while also ensuring greater fairness. The study tackles the complexities of coordinating self-interested EV agents with varying charging requirements and unpredictable schedules, addressing the shortcomings of centralized optimization and fairness in current methods.