Energy Trading of Multiple Virtual Power Plants Using Deep Reinforcement Learning

This paper proposed a data-driven deep reinforcement learning (DRL) for the optimal energy trading structure between multiple virtual power plants (VPP) at the point of common coupling (PCC). Furthermore, constrained DRL is utilized by considering the operational limits. To coordinate the energy trading among multiple VPPs, the model-free approach is exploited by considering the uncertainties of load and intermittency of renewable energy resources to maximize every VPP revenue by subjecting it to different constraints and depends less on the utility by maximizing the energy sharing among VPPs and also the profit. After the training process, the optimal policy is executed to dispatch their local resources without solving a complex optimization problem, and P2P cooperative energy trading with distributed DQN based RL method is proposed. A case study has been performed for energy sharing among multiple VPPs, and results outperform the proposed method.