Stochastic Game based Microgrid Clusters Energy Management Modeling and Strategy

The advanced energy management model and strategy is a powerful guarantee the economic benefit of microgrid (MG) system. In this paper, a perfect microgrid cluster model with two agents, which enables the energy management and trading, is constructed based on stochastic game theory. In order to solve this game problem, an improved deep Q network (DQN), which adopts upper confidence bound (UCB) to implicitly balance exploration and exploitation and the modified Q-value calculation, is proposed to leverages the novel exploration of state to accelerate convergence and achieve the optimal energy management strategy. Simulation results indicate that our proposed game model with constant and dynamic pricing has advantages in different microgrid operation situations.