基于强化学习的综合能源系统管理综述

In order to meet the growing energy demand and reduce the damage to the environment, energy conservation has become a long-term strategic policy for global economic and social development. The enhancement of energy management can improve energy efficiency, as well as promote energy conservation and emission reduction. However, the integration of renewable energy and flexible load makes the integrated energy system (IES) become a complex dynamic system with high uncertainty, which brings great challenges to modern energy management. Reinforcement learning (RL), as a typical interactive trial-and-error learning method, is suitable for solving optimization problems of complex dynamic systems with uncertainty, and therefore it has been widely considered in integrated energy system management. This paper systematically reviews the existing works of using reinforcement learning to solve integrated energy system management problems from the perspective of models and algorithms, and puts forward prospects from four aspects: Multi-time scale, interpretability, transferability, and information security. Copyright © 2021 Acta Automatica Sinica. All rights reserved.