Research on Optimization Operation of Multi-entity Microgrid Based on Heterogeneous Multi-agent Reinforcement Learning

The micro-grid multi-agent optimization operation including smart power users, EV charging systems and solar energy storage systems is currently an effective way to reduce fossil energy dependence. For the traditional multi-agent reinforcement learning algorithm, there is only one total objective function, which cannot determine the realization of each agent’s own goal. In this paper, a heterogeneous multi-agent reinforcement learning method is proposed to solve the problem of optimal operation involving different stakeholders. While achieving the maximum of the total objective function, it clearly shows the realization degree of each agent’s own goal. The heterogeneous multi-agent reinforcement learning algorithm is applied to solve the micro-grid optimization operation model of different investment entities including EV charging system, PV/wind/storage system and smart power users. The study shows that the heterogeneous multi-agent reinforcement learning algorithm can provide effective strategies for the multi-agent optimization of micro-grid, and effectively display the realization effect of the objective function of each micro-grid agent.