Deterministic global flowsheet optimization for the design of energy conversion processes
Bongartz, Dominik; Mitsos, Alexander (Thesis advisor); Chachuat, Benoit (Thesis advisor)
Aachen (2020) [Book, Dissertation / PhD Thesis]
Page(s): 1 Online-Ressource (XIV, 196 Seiten) : Illustrationen, Diagramme
Reducing anthropogenic greenhouse gas emissions requires a new generation of energy conversion processes that make efficient use of renewable resources. Flowsheet optimization is a tool that can aid the design of such processes. Since the resulting optimization problems are nonconvex, deterministic global optimization is desirable. However, many problems remain computationally intractable for global optimization. In this thesis, two approaches for expediting the global solution of flowsheet optimization problems are investigated and applied to power-to-fuel processes as an example for a new type of energy conversion processes. First, factorable reduced-space formulations are considered as a way of reducing the size of the optimization problems arising from flowsheet optimization. In these reduced-space formulations, optimization variables are eliminated from the problem using equality constraints that can be rearranged to compute the variables as a factorable function of other variables. In flowsheet optimization, such formulations can be achieved at the modeling stage in analogy to established methods from flowsheet simulation and local optimization, where they can be interpreted as hybrids between equation-oriented and sequential-modular methods. The reduced-space formulations enable significant savings in computational time compared to fully equation-oriented approaches, both in the state-of-the-art global solver BARON and the newly developed open-source solver MAiNGO. The conducted analyses suggest that the savings are due to effects resembling selective branching and constraint propagation, as well as the reduced size of the subproblems for lower and upper bounding and bound tightening. Second, tight relaxations are developed for two classes of thermodynamic property models that occur in flowsheet optimization problems: general pure component models that are used when modeling multicomponent systems, and the IAPWS-IF97 model for water. The developed relaxations are significantly tighter than the relaxations obtained with general purpose methods and result in significant reductions in computational time for several case studies. Finally, methylal (also known as OME1) is considered as an example product of power-to-fuel processes because of its attractive combustion properties. In detailed process simulations, OME1 production in a power-to-fuel process chain based on a combination of existing processes is found to be less efficient than the production of other fuels. Therefore, an alternative process based on direct oxidation of methanol to OME1 is considered and globally optimized with the above methods. The results demonstrate that the developed methods make optimization of relatively complex flowsheets with few degrees of freedom tractable. However, they also highlight remaining limitations regarding the complexity of certain unit operation models as well as the importance of using realistic boundary conditions, in particular regarding heat integration and available utilities.