Optimal energy management of multi-microgrids connected to distribution system based on deep reinforcement learning

As an effective way to integrate renewable energy, more and more microgrids (MGs) are connected to distribution system. However, the model-based energy management approach is confronted with challenges as the MGs data scale increases rapidly. The data-driven analysis and decision approach is widely utilized to maintain the secure and stable operation of MG. Hence, this paper firstly proposes a bi-level coordinated optimal energy management (OEM) framework for the distribution system with Multi-MGs. In this framework, the distribution system operator (DSO) makes decisions at the upper level, and the MGs make their own decision at the lower level. Secondly, an interactive mechanism based on a-leader-multi-followers Stackelberg game is provided to improve the utility of both sides by dynamic game, where the DSO is the leader, and the MGs are followers. Furthermore, a data-driven multi-agent deep reinforcement learning (DRL) approach is investigated to calculate the Stackelberg equilibrium for the OEM problem. Finally, the case study in modified IEEE-33 test systems with multi-MGs demonstrates the performance of the proposed approach. The computation efficiency and accuracy are verified by the dispatch result.

Automated summary

This paper proposes a bi-level coordinated optimal energy management framework for distribution system with Multi-MGs, with an interactive mechanism based on a-leader-multi-followers Stackelberg game to improve utility, and investigates a data-driven multi-agent deep reinforcement learning approach to calculate the Stackelberg equilibrium for the OEM problem. The proposed approach is validated through a case study in modified IEEE-33 test systems with multi-MGs.