Numerical Methods for Optimal Boundary Control of Advection-Diffusion-Reaction Systems

Marcus Johan Schytt,John Bagterp Jørgensen
2024-04-14
Abstract:This paper considers the optimal boundary control of chemical systems described by advection-diffusion-reaction (ADR) equations. We use a discontinuous Galerkin finite element method (DG-FEM) for the spatial discretization of the governing partial differential equations, and the optimal control problem is directly discretized using multiple shooting. The temporal discretization and the corresponding sensitivity calculations are achieved by an explicit singly diagonally-implicit Runge Kutta (ESDIRK) method. ADR systems arise in process systems engineering and their operation can potentially be improved by nonlinear model predictive control (NMPC). We demonstrate a numerical approach for the solution to their optimal control problems (OCPs) in a chromatography case study. Preparative liquid chromatography is an important downstream process in biopharmaceutical manufacturing. We show that multi-step elution trajectories for batch processes can be optimized for economic objectives, providing superior performance compared to classical gradient elution trajectories.
Numerical Analysis,Optimization and Control
What problem does this paper attempt to address?
This paper mainly discusses how to use numerical methods to solve the optimal boundary control problem of chemical systems described by the Advection-Diffusion-Reaction (ADR) equations. These equations are widely used in process systems engineering, such as chromatographic separation, which is an important downstream process in biopharmaceutical manufacturing. The paper proposes the use of the Discontinuous Galerkin Finite Element Method (DG-FEM) for spatial discretization and the multiple shooting method for direct discretization of the optimization control problem. Time discretization and sensitivity computation are implemented using the explicit singly diagonally implicit Runge-Kutta (ESDIRK) method. In the study, the author demonstrates how multi-step elution trajectories can be optimized for economic objectives using the example of pre-processing liquid chromatography, providing better performance than traditional gradient elution trajectories. The paper also discusses how to estimate the upstream component composition online using a nonlinear model predictive control (NMPC) strategy for optimal predictive control of downstream processes. Overall, this paper aims to improve the efficiency of model predictive control for chemical processes through mathematical optimization methods. It provides a numerical solution method for solving the optimal control problem of ADR systems and verifies and applies it through a specific chromatographic case.