Please use this identifier to cite or link to this item:
http://hdl.handle.net/11375/23732
Title: | Multiperiod Refinery Planning: Development and Applications |
Authors: | Nguyen, Alexander |
Advisor: | Swartz, Christopher L.E. |
Department: | Chemical Engineering |
Keywords: | petroleum;refinery;multiperiod;refining;lagrangean;decomposition;optimization;planning;lagrangian;nonlinear |
Publication Date: | 23-Nov-2018 |
Abstract: | The purpose of this work aims to develop and explore a nonlinear multiperiod petroleum refinery model based on a real-world model. Due to the inherent complexity and interconnected nature of petroleum refineries, various studies are implemented to describe the multiperiod model. The model is based around maximizing the profit of a petroleum refinery, starting from the crude inputs through the crude distillation unit, to the intermediate product processing through various unit operations, and finally to the blending of the final products. The model begins as a single period model, and is re-formulated as a multiperiod model by incorporating intermediate product tanks and dividing the model into partitions. In solving the multiperiod model, the termination criteria for convergence was varied in order to investigate the effect on the solution; it was found that it is acceptable to terminate at a relaxed tolerance due to minimal differences in solution. Several case studies, defined as deviations from normal operation, are implemented in order to draw comparisons between the real-world model and the model studied in this thesis. The thesis model, solved by CONOPT and IPOPT, resulted in significant gains over the real-world model. Next, a Lagrangean decomposition scheme was implemented in an attempt to decrease computation times. The decomposition was unable to find feasible solutions for the subproblems, as the nonlinear and nonconvex nature of the problem posed difficulty in finding feasibilities. However, in the case of a failed decomposition, the point where the decomposition ends may be used as an initial guess to solve the full space problem, regardless of feasibility of the subproblems. It was found that running the decomposition fewer times provided better initial guesses due to lower constraint violations from the infeasibilities, and then combined with the shorter decomposition time resulted in faster computation times. |
URI: | http://hdl.handle.net/11375/23732 |
Appears in Collections: | Open Access Dissertations and Theses |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
NGUYEN_ALEXANDER_AUGUST2018_MASC.pdf | 2.7 MB | Adobe PDF | View/Open |
Items in MacSphere are protected by copyright, with all rights reserved, unless otherwise indicated.