Deep RL Optimizes Solar-Battery Hybrid Design for Multi-Market Trading Under Uncertainty

The rapid growth of variable renewable energy has increased the need for flexible, coordinated energy resources. Hybrid resources combining solar generation with battery storage can participate in multiple electricity markets—energy and ancillary services—but limited power/capacity makes optimal sizing and bidding a complex design problem. Uncertainty in weather and prices further complicates profitability assessment, as conventional optimization methods struggle to remain effective under stochastic conditions.

To address this, authors Hoshino, Mantani, and Furutani propose a deep reinforcement learning-based co-optimization framework that jointly optimizes hybrid system sizing and multi-market bidding strategies within a unified stochastic formulation. By embedding design variables directly into the policy learning process, the framework handles uncertainty using historical renewable generation and market data. Case studies demonstrate economically rational system designs, offering a path to maximize revenue from multiple market streams while reducing operational complexity. This AI-driven approach could accelerate deployment of cost-effective solar-battery hybrids.