Multi-Agent Based Optimization of Hybrid Solar PV, Hydrogen, and Battery Storage Systems to Maximize Energy Production

The increasing demand for efficient and sustainable energy resources has emphasized the need for hybrid energy systems. The integration of intermittent solar PV systems and battery storage systems presents a significant challenge in achieving controlled and maximized energy production. This study optimizes energy production in a DC microgrid system, integrating solar PV, a electrolyzer, and a controllable battery storage subsystem. Additionally, it proposes a Multi Agent Reinforcement Learning (MARL) framework that enhances coordination control of the storage system and operational efficiency within the microgrid. By allowing discrete agents as a PV availability, to make decisions and interact with the energy storage system, the framework adapts dynamically to changes in energy demand and available supply, maximizing energy and hydrogen production. The results demonstrate that the multi-agent approach achieves higher energy output compared to conventional optimization techniques, like Mixed-Integer Linear Programming (MIP), using yearly PV datasets in Indian scenarios. This study is performed using the Python PuLP tool, and the simulation findings include improved battery storage utilization and real-time energy dispatch. The proposed approach contributes to the development of MARL based solutions for hybrid energy systems, with implications for enhancing sustainability in distributed microgrids.