Multi-agent reinforcement learning for microgrids

This paper presents a general framework for Microgrids control based on Multi Agent System Technology. The proposed architecture is capable to integrate several functionalities, adaptable to the complexity and the size of the Microgrid. To achieve this, the idea of layered learning is used, where the various controls and actions of the agents are grouped depending on their effect on the environment. Moreover this paper, focus on how the agent will cooperate in order to achieve their goals. The core of the cooperation is a Multi Agent Reinforcement Learning Algorithm that allows the system to operate autonomously in island mode.