Distributed Energy Scheduling for Microgrid Clusters: A Multi-Agent Deep Reinforcement Learning Approach

The fresh strategy for organizing energy within a cluster of an Integrated Energy System (IES) featuring renewable energy generation is introduced, leveraging multi-agent deep reinforcement learning. In this method, each IES can be viewed as an agent. The primary goal is to minimize overall economic costs, including carbon trading expenses, thereby encouraging agents to prioritize low-carbon choices. Through experimental simulations, it is shown that the agents can adapt and derive optimal strategies in real time by learning from environmental data, effectively managing the variability of clean energy (CE).