AI Manages Microgrid for Reliable, Renewable Power

Researchers at an unspecified institution developed an artificial intelligence system to manage energy flow in a community microgrid in Rockhampton, Australia. The system combines solar, wind, battery storage, hydrogen production and fuel cells, and even diesel backup to maximize renewable energy use and keep the lights on.

Using a technique called proximal policy optimization, the AI learns to dispatch energy from these sources efficiently. The AI achieved a 91.2% renewable energy fraction and reduced carbon intensity to 0.085 kg CO2/kWh in normal operation, generating A$195,690.67 in annual revenue.

Even when simulating grid outages, increasing the probability from 1% to 5%, the system maintained 98.79% demand fulfillment, though this required significantly more battery discharge and diesel use. The research indicates promise for AI-driven, coordinated energy dispatch, but notes performance is sensitive to renewable energy variability and the boundaries used for evaluating carbon emissions. Future work will likely focus on improving training stability and adapting the system to different energy profiles.