Offline Reinforcement Learning for Process Control - CDU Air Fan Cooler Control Application
Learn about offline reinforcement learning applications in industrial process control through this technical conference talk from SK TECH SUMMIT 2023. Explore how traditional online reinforcement learning faces challenges in real industrial processes due to cost and risk factors, and discover how offline reinforcement learning offers a solution by learning enhanced control policies from historical operational data. Follow along as Kim Dong-il from SK Incheon Petrochemical shares practical insights on designing state spaces, action spaces, and reward functions for model training, and implementing these models in actual crude distillation unit (CDU) Air Fan Cooler control systems. Gain valuable knowledge about bridging AI and digital technology in manufacturing processes, with specific focus on control and automation applications in petrochemical industries.