Guest Seminar - Machine Learning Koopman Modeling for Predictive Control and State Estimation of Nonlinear Processes

Industries are placing greater emphasis on enhancing process safety, ensuring production consistency, improving efficiency, minimizing waste and emissions, and optimizing profitability. This evolving landscape demands smarter, more efficient, and highly adaptable integrated automation solutions that offer seamless monitoring, advanced control, and beyond. Modern industrial processes, however, exhibit highly nonlinear and dynamic behaviors, adding to the complexity of achieving these objectives.
Nonlinear models built based on physical knowledge of highly nonlinear industrial processes have been extensively used as the basis of state estimation and advanced control systems development. However, this type of solution has the following limitations: First, the design and analysis of nonlinear control and estimation algorithms are much more challenging as compared to those based on linear models. Second, in the presence of constraints, the use of a nonlinear model for optimization-based control and state estimation can incur a much heavier computational load as compared to the linear counterpart. More importantly, it can be expensive or hard to build a high-fidelity physical nonlinear model for complex industrial processes in practice.
To address these challenges, we have made attempts to address optimization-based control and state estimation for nonlinear processes within an alternative framework of Koopman modeling, which can be leveraged to construct data-driven linear dynamic models to predict the dynamical behavior of nonlinear processes. This presentation will discuss the following points:
1)    A concise overview of the Koopman modeling concept, and its connection with optimal control and estimation of nonlinear systems.
2)    Koopman modeling for model predictive control.
3)    Machine learning-based Koopman modeling for efficient estimation, model predictive control (MPC), and economic MPC of general nonlinear systems.
4)    Applications of the proposed methods to various industrial processes, including water treatment, post-combustion carbon capture, and crystallization.
Additionally, we plan to present our recent findings on integrating machine learning with the data-enabled predictive control (DeePC) framework. Our goals are twofold: a) to facilitate economic operation within the DeePC framework, and b) to reduce or eliminate the need for optimization during online control implementation.