On November 21, we hosted a scientific seminar at Ruhr University Bochum titled “Multi-Model Predictive Control of a Chemical Reactor”. The seminar was led by Miroslav Fikar, a professor at the Slovak University of Technology in Bratislava. It was organized in the “Research Seminar on Smart Cybernetics” series with a live online stream.
Abstract:
We study optimal control of a nonlinear continuous stirred tank reactor. The process is modeled using multiple models. The first approach uses a combination of a linear state-space model and da ata-based nonlinear feedforward neural network. The second approach uses a combination of multiple piece-wise affine models obtained from multiple operating points. A multimodel-based offset-free model predictive control (MM-MPC) strategy is proposed to deal with input and output constraints. The method predicts future states and considers the constraints of all models. Several cost function formulations are proposed. The MM-MPC approach effectively outperforms single-model MPC strategies. We show that the proposed MM-MPC exhibits enhanced robustness and operational efficiency, evidenced by its reduced performance indices and low average CPU time. This innovative controller proves highly effective for nonlinear systems with variable dynamics and stringent constraints.
This project has received funding from the European Union’s Horizon under grant no. 101079342 (Fostering Opportunities Towards Slovak Excellence in Advanced Control for Smart Industries).
0 Comments