Blend Optimization through Recommendation using Value-Based Reinforcement Learning Agent

In his session, “Blend Optimization through Recommendation using Value-Based Reinforcement Learning Agent,” Suman Senapati explores how AI-driven reinforcement learning can transform refinery blending operations through data-driven optimization and real-time recommendations. The session highlights opportunities to reduce product quality giveaway (QGA) while maintaining on-spec product quality, improve coordination across refinery units, and use dynamic forecasting to support operational decisions. It also examines how reinforcement learning can be scaled across multiple product streams to enhance refinery profitability and operational efficiency.