Coordinated Optimal Dispatch of Distribution Grids and P2P Energy Trading Markets

With the increasing integration of distributed renewable energy, traditional power users are evolving into prosumers capable of both generation and consumption. However, their decentralized nature poses challenges in resource coordination. This study proposes a bi-level optimization framework for distribution networks integrating peer-to-peer (P2P) energy trading and shared energy storage. The upper-level model minimizes distribution system operator (DSO) operational costs, including network losses and storage management, while ensuring voltage stability. The lower-level model enables prosumers to maximize P2P market profits through adaptive load adjustments and shared storage utilization. To address the nonlinear, high-dimensional optimization challenges, an improved Convex-Soft Actor-Critic (C-SAC) algorithm is developed, combining deep reinforcement learning with convex optimization to achieve privacy-preserving distributed coordination. Case studies on an IEEE 33-node system demonstrate that the framework increases prosumer profits by 56.9%, reduces DSO costs by 23.6%, and lowers network losses by 21.5% compared to non-cooperative scenarios. The shared storage system reduces capacity and power requirements by 20% and 14.1%, respectively. The C-SAC algorithm outperforms traditional methods (DDPG, SAC) in convergence speed and economic metrics, showing scalability across larger systems (IEEE 69/118 nodes). This work provides a model-free solution for renewable-rich distribution networks, balancing efficiency and operational security.