Peer-to-Peer Energy Trading with Mid-Market Rate among Virtue Power Plants by Multi-Agent Deep Reinforcement Learning

Efficient energy allocation and utilization are critical challenges in the evolving landscape of decentralized power systems integrated with renewable energy sources. Peer-toPeer (P2P) energy trading offers a novel framework for direct energy transactions among prosumers, enhancing system flexibility and market responsiveness. This paper addresses the complex decision-making processes inherent in P2P energy markets by proposing a Multi-Agent Twin Delayed Deep Deterministic Policy Gradient (MATD3) algorithm. The proposed model leverages Virtual Power Plants (VPPs) to aggregate distributed energy resources, enabling autonomous and optimized energy trading within the P2P framework. VPPs act as intermediaries, enhancing market liquidity, optimizing energy distribution, and ensuring supply reliability. The MATD3-based approach effectively manages high-dimensional continuous action spaces and adapts to the dynamic nature of P2P energy trading. Simulation results indicate that the proposed MADRL model reduces overall energy costs and enhances the performance of VPPs in decentralized energy markets. This research contributes to the advancement of intelligent P2P energy trading mechanisms, emphasizing the pivotal role of reinforcement learning in developing sustainable and resilient energy infrastructures