Deep Reinforcement Learning of Local Trading Strategy for Grid-Scale Renewable Energy Integration
Caleb Ju, Georgia Institute of Technology
Variable renewable generation increases the challenge of balancing power supply and demand. Grid-scale batteries co-located with generation have the potential to help mitigate this misalignment. This work considers the use of deep reinforcement learning for operating grid-scale batteries co-located with solar power. We show this approach has two significant advantages compared to rules-based control methods: (1) that solar energy is more effectively shifted towards high demand periods, and (2) increased diversity of battery dispatch across different locations, reducing potential ramping issues caused by superposition of many similar actions.