(429f) Model Predictive Control of an Experimental Electrochemical Reactor

The electrochemical reduction of carbon dioxide (CO₂) has gained significant attention recently as an alternative source of manufacturing carbon-based fuels attributed to the rise in clean energy supply [1, 2]. With CO₂ reduction becoming an increasingly viable option for producing chemicals in an environmentally friendly manner, it has become necessary to model, optimize, and control the electrochemical CO₂ reduction process. In order to better comprehend the impact of mass transfer and reaction kinetics on productivity, a rotating cylinder electrode (RCE) reactor was constructed at UCLA [3], which allows for the analysis of these factors independently. Specifically, the RCE reactor employs two manipulated control variables to separate the kinetic and mass transfer effects on the reaction, based on the physical-chemical phenomena where the applied potential strongly influences reaction energetics, while electrode rotation controls hydrodynamics and convective mass transfer. In our recent work, a multi-input multi-output (MIMO) control scheme is developed on the RCE reactor to control the production rate of ethylene and carbon monoxide, utilizing techniques of neural network modeling, nonlinear optimization, and proportional integral (PI) control [4]. However, the RCE system generates sixteen liquid/gas products, which outnumber the available control inputs. This implies that the PI control method may have limitations in achieving comprehensive control over the production rate of all product species.

This paper aims to develop a model predictive control (MPC) to implement the MIMO control scheme that is capable of handling the situation with an unbalanced number of control states. To achieve this target, the production rate of hydrogen is included as an additional output state of the control system, which means there are three outputs (i.e., production rate of CO, C₂H₄, and hydrogen) and two inputs (i.e., surface potential and rotation speed) in the MIMO system. Considering the computational efficiency of solving the optimization problem in the MPC, a quadratic programming based MPC is chosen to carry out the real-time control process. Lastly, the performance of the developed MPC will be supported by experimental evidence and evaluated in comparison to the previously developed PI control system.

References:

[1] Morales-Guio, C.G.; Cave, E.R.; Nitopi, S.A.; et al. Improved CO₂ reduction activity towards C2+ alcohols on a tandem gold on copper electrocatalyst. Nat Catal 1, 764–771 (2018).

[2] Jang, J.; Shen, K.; Morales-Guio, C.G.; Electrochemical direct partial oxidation of methane to methanol. Joule. 2019, 3(11), 2589-93.

[3] Morales-Guio, C.G.; Jang J; Rüscher M; Winzely M.; Gastight Rotating Cylinder Electrode: Towards Decoupling Mass Transport and Intrinsic Kinetics in Electrocatalysis. Authorea Preprints. 2021 Oct 18.

[4] Çıtmacı B; Luo J; Jang J; Morales-Guio, C.G.; Christofides PD; Machine learning-based ethylene and carbon monoxide estimation, real-time optimization, and multivariable feedback control of an experimental electrochemical reactor. Chemical Engineering Research and Design. 2023 Mar 1;191:658-81.