Foresight Decision-Making Method for Dynamic Energy Management of Microgrid

The dynamic energy management (DEM) problem of microgrid (MG) is difficult to address as the computational complexity grows with the increase in the number of optimal periods. In this paper, a foresight decision-making method for DEM of MG is proposed. The optimization period of DEM problem is described by Markov Decision Processes (MDP) framework, wherein the value function represents the sum of the cost of subsequent periods. However, it is only required to obtain the costs of the initial few periods by the solver, while the costs of the rest periods can be obtained by deep Long Short-Term Memory (LSTM) natural network directly. In this way, it can reduce computation time significantly and maintain foresight performance, as it instantly calculates the value function. Numerical simulations demonstrate that the proposed method is favorable in terms of accuracy and computational efficiency.