Public defence in Automation, Systems and Control Engineering, M.Sc. Jifei Deng

The title of the thesis: Reinforcement Learning Methods for Setpoint Optimization and Control Method Design in Process Industry with Case Studies in Steel Strip Rolling and District Heating 

Thesis defender: Jifei Deng
Opponent: Prof. Thilo Sauter, Danube-University Krems, Austria
Custos: Prof. Valeriy Vyatkin, Aalto University School of Electrical Engineering, Department of Electrical Engineering and Automation

This study explores the applications of reinforcement learning (RL) in the process industry, aiming to enhance product quality and energy efficiency through the development of intelligent RL-based methodologies for precise setpoint optimization and adaptive control solutions. 

In setpoint optimization, accurately computing equipment parameters to meet quality specifications is crucial. For process control, implementing real-time control methods directly impacts product quality. Traditional approaches, such as first-principles models, empirical models, and trial-and-error methods, often involve simplification and linearization to handle the complex and dynamic nature of industrial processes. To improve product quality and energy efficiency, there is a growing need for intelligent and adaptive methodologies that compute optimal solutions for industrial processes. 

Recognizing RL's potential to learn from interactions, RL techniques have been adopted to develop policies for both setpoint optimization and process control. In the context of setpoint optimization in strip rolling and fuel cost reduction in district heating, RL methodologies have been investigated to calculate and optimize system setpoints. By leveraging environment models of the processes, RL agents generate optimal solutions based on machine capacity to meet customer demands. Additionally, RL-based adaptive control methodologies have been developed for the steel strip rolling process, allowing dynamic responses to evolving conditions. 

The results of case studies indicate that the proposed RL-based methods outperform traditional approaches. This research demonstrates RL's effectiveness in optimization and control, suggesting significant improvements in automation, efficiency, and product quality within the process industry, with potential for broader industrial application.

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact:

Doctoral theses in the School of Electrical Engineering: https://aaltodoc.aalto.fi/handle/123456789/53