Optimal Operation of Integrated Energy System Based on Deep Reinforcement Learning

Abstract

Integrated energy system (IES) combining multi energies such as power, heat, cooling and gas, can be applied to improve the overall energy efficiency based on the energy cascade utilization. In order to enhance the operation efficiency of IES, an operation optimization method is proposed in this paper based on deep deterministic policy gradient (DDPG). Two categories of models of linear ones and nonlinear ones are established, then a nonlinear optimization model is constructed by taking the annual operating cost as the objective. With regard to the instability in its training process of DDPG, two improved strategies of scaled initial state spaces and dynamic noise compensation are presented. In the case study, the proposed method is employed to obtain the optimal operation strategies for a typical scenario of a practical IES. The experimental results demonstrate that the proposed one not only obtains the most economical strategies, but also exhibits the robustness of the algorithm.