Optimal design of power-to-x processes

  • Optimales Design von Power-to-X-Prozessen

Burre, Jannik; Mitsos, Alexander (Thesis advisor); Martin, Mariano Martin (Thesis advisor)

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

In: Aachener Verfahrenstechnik series AVT.SVT - Process systems engineering 25
Page(s)/Article-Nr.: 1 Online-Ressource : Illustrationen

Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2022


The increasing share of renewable energy sources in the electricity grid causes curtailments, which prevent exploiting the full environmental and economic potential of renewable electricity. Power-to-X processes can utilize this electricity to produce certain products that would have been otherwise produced from fossil-based sources. To benefit the most, these Power-to-X processes need to be optimized for a maximum resource-efficiency. We demonstrate that the sole replacement of raw materials for industrial process concepts is not expedient. We therefore develop optimization-based methods to identify sustainable process concepts and support their optimal design. These methods are applied to the production of dimethoxymethane (referred to as DMM or OME1)—a promising synthetic fuel candidate and intermediate for the production of longer-chain oxymethylene ethers (OME3-5). To analyze DMM and OME3-5 production using established process concepts, we implement process models with detailed thermodynamic models from the open literature. Even by considering their maximum potential for heat integration, these process concepts have been found to be much less efficient than those for the production of other synthetic fuel candidates. Therefore, fundamentally new processes need to be designed. Emerging Power-to-X processes are usually on a very different stage of development. To enable a fair comparison and support process design, we develop a methodology that incorporates optimization-based methods on different hierarchy levels. The methodology allows a systematic way to design and evaluate each candidate regarding three key performance indicators: production costs, exergy efficiency, and global warming impact. Applied to five reaction pathways for DMM production, we identified the direct reduction of CO2 to be the most suitable one for sustainable DMM production at its current state. For a successful implementation, detailed process models are necessary. As the complicated form of such models often cause difficulties for deterministic optimization, we develop a hybrid process model for reductive DMM production incorporating Gaussian processes and artificial neural networks. For solving the resulting nonconvex program, we use a reduced-space formulation and a hybrid between the McCormick and the auxiliaryvariable method implemented in our deterministic global solver MAiNGO. Only with these measures on both the modeling and algorithm level, convergence was possible. As Power-to-X design problems often contain discrete decisions, we analyze different problem formulations regarding their suitability for global superstructure optimization and applied the most suitable one to the design problem for reductive DMM production. For mixed-integer nonlinear programming problems containing nonconvex functions, we identified such formulations as particularly promising that reduce the number of optimizationvariables. Although they introduce nonconvex terms, corresponding relaxations remain comparably tight for our example problems. However, a large library with benchmark problems of different complexity would be necessary to derive generally valid statements. The application of optimization-based methods to DMM production has demonstrated great potential. However, also limitations and further improvement potential was identified—for both the methods and DMM production as a Power-to-X process.


  • Chair of Process Systems Engineering [416710]