Economic Optimizing Control for a U-loop Reactor Modelling, Estimation, and Economic Nonlinear Model Predictive Control

In this thesis, we present models and algorithms for monitoring and economic model predictive control applied to a fermentation process for single-cell protein (SCP) production in a U-loop bioreactor. We describe a modelling framework for compactly describing stoichiometry and kinetics in chemical and biochemical systems. In a technical report, we outline and apply the modelling framework for batch (BR), fedbatch (FBR), continuous stirred tank (CSTR), plug flow reactors (PFR), as well as their combinations, and present numerical examples of chemical and biochemical systems. We apply the modelling framework to existing and novel models describing growth of the bacteria Methylococcus capsulatus (Bath) for single-cell protein (SCP) production. Based on the modelling framework, we present a continuous U-loop fermentor model combining CSTR and PFR compartments. We describe model-based monitoring and control techniques for continuous-discrete nonlinear systems involving stochastic differential equations (SDEs). In a FOCAPO/CPC 2023 paper, we present the extended Kalman filter (CD-EKF), unscented Kalman filter (CD-UKF), ensemble Kalman filter (CD-EnKF), and a particle filter (CD-PF), test the estimations algorithms on a nonlinear test system, and discuss their performance. In papers presented as CCTA 2019 and CDC 2020, we apply the CD-EKF, present economic optimal control for optimal production economy, and perform a numerical experiment of economic nonlinear model predictive control (ENMPC) for SCP production in a U-loop bioreactor based on an SDE model. To include measurement information from commonly available online sensors, e.g. dissolved oxygen, temperature, and pH-value, we present monitoring and control techniques for continuous-discrete systems involving stochastic differential algebraic equations (SDAEs). In a ECC 2023 paper, we present an SDAE model describing cell growth and chemical equilibria, i.e. pH dynamics, and economic optimal control for SCP production in a laboratory-scale CSTR. The laboratory-scale CSTR model is based on a physical laboratory-scale bioreactor and is intended for systems identification experiments. In chapter 8, we present the CD-EKF for SDAE models and apply it to the laboratory-scale CSTR in chapter 11. As such, we outline the main components necessary for the implementation of ENMPC for SCP production in a U-loop bioreactor based on available online sensors. Finally, we present a high-performance Monte Carlo simulation toolbox for uncertainty and performance quantification in closed-loop systems. In a paper presented at CDC 2021, we describe and apply the Monte Carlo toolbox in a numerical experiment of SCP production in a fedbatch reactor with open-loop, PID, and model predictive control strategies.