| dc.description.abstract | The increasing threat of climate change due to human emissions of global greenhouse gases has led to investigation of several technologies to mitigate such emissions from large point sources. One of these technologies is post-combustion carbon dioxide capture by the use of membranes. In this work, the goal has been to develop a model which finds the membrane system, and its operating conditions, that will minimize the cost of carbon dioxide capture for a given gas feed and separation requirements.
The design of a gas separation process with membranes involves many process parameters that interact with each other in complex ways, while having varying impacts on the cost of capture. MINLP models that can handle multicomponent gas flow was developed to optimize important variables simultaneously and achieve a desired gas separation to the lowest cost. However, due to lack of convergence to a finite solution, the model was simplified to multiple NLP models.
For solving this problem, the models was implemented in GAMS and the global optimization solver BARON was utilized for solving the models. Baron was able to find the optimal solution early in the solution process, and depending on model and parameter inputs, used from a few minutes to several hours to prove optimality.
The models were tested with a binary flue gas, consisting of carbon dioxide and nitrogen, from three different sources, and resulted in three different membrane networks and configurations, which all performed the desired gas separation of 90\% carbon dioxide capture and purity. The results shows the complexity of the process, and that different conditions and membrane types will yield different optimal networks. Further research should be done to develop a model that is able to handle different network conditions and possible membrane types simultaneously. Incorporating temperature in the models, such that heating and cooling is also evaluated, is another direction that could show benefits. | en |
