Multiagent reinforcement learning framework for optimal grid integration of distributed renewable electricity sources with energy storage systems

This study develops a topology-aware multiagent reinforcement learning framework that coordinates distributed renewables and storage for transmission-level control. Using a 24-month Saudi Eastern Province dataset, the framework reduces curtailment by up to 69.1% versus traditional economic dispatch and 10.3% versus MPC, cuts total annual operating costs by 27.9%, maintains frequency within ±0.1 Hz during 97.3% of periods, and adapts with 234 ms median latency. Emissions decrease by 0.85 to 1.46 Mt CO2-equivalent annually. Results demonstrate scalable, sub-second control that improves stability and economics while enabling higher renewable integration.