Optimal biorefinery processes using differential-algebraic equation systems with optimization criteria embedded

  • Optimale Bioraffinerieprozesse unter Verwendung differential-algebraischer Gleichungssysteme mit eingebetten Optimierungskriterien

Ploch, Tobias Michael; Mitsos, Alexander (Thesis advisor); Wiechert, Wolfgang (Thesis advisor)

Aachen : RWTH Aachen University (2021, 2022)
Book, Dissertation / PhD Thesis

In: Aachener Verfahrenstechnik series - AVT.SVT - Process Systems Engineering 19 (2022)
Page(s)/Article-Nr.: 1 Online-Ressource : Illustrationen, Diagramme

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2021


Biorefineries comprise a variety of process steps to convert renewable feedstocks into desired products. The dynamic simulation and optimization of detailed process models support the analysis and control of biorefinery processes shortly before or during implementation in pilot plants. The dynamic flux balance analysis (DFBA) enables detailed modeling of biotechnological processes. The algebraic variables are calculated as the solution of an optimization problem leading to a differential-algebraic equation system with embedded optimization criteria (DAEO). The dynamic modeling of separation processes based on thermodynamic equilibrium is another important example for DAEOs. In this thesis, a heuristic solution method for general DAEOs is presented that is based on replacement of the embedded optimization problem with its first-order optimality conditions (KKT conditions). The method is implemented in a DAEO toolbox which can also be used for other DAEO applications. In addition, a Modelica library of dynamic models describing typical parts of biorefinery processes is presented. Together with the DAEO toolbox, the dynamic simulation of plant-wide biorefinery models is carried out. Furthermore, the optimization of DAEOs is considered. Mathematically, this leads to a dynamic bilevel optimization problem. In the first step, the DAEO is replaced with its KKT conditions yielding a nonsmooth differential-algebraic equation system (DAE). The extension of a solver for nonsmooth DAEs enable the gradient-based optimization in a single-shooting approach. Finally, limitation are discussed that arise when a relaxed KKT approach from literature is applied to solve dynamic phase equilibrium problems with multiple liquid phases. An alternative modeling approach is proposed that enables the dynamic simulation of such processes.