A Global Optimization Scheme for Mixed-Integer Nonlinear Programming Problems
| Autor: | Cheng-Tien Hsu, 許政天 |
|---|---|
| Rok vydání: | 2005 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 93 This thesis aims at the development of a global optimization algorithm for the solution of mixed-integer nonlinear programming (MINLP) problems. A novel two-stage global optimization scheme is proposed. In the first stage, the modified Simulated Annealing (SA) algorithm is used to allocate the exact values of the discrete variables, while in the second stage the global optimal solution is obtained through making use of the information theory, a chaotic algorithm and a feasible point strategy. The applicability and effectiveness of the proposed global optimization scheme have been tested with some typical MINLP problems and extensive comparisons with existing SA and/or information theory-based algorithms have also performed in this work. Simulation results reveal that, due to the advantages of the chaotic algorithm and the two-stage solution approach, the proposed global optimization scheme is more efficient and outperforms the conventional SA and/or information theory-based algorithms. To extend the proposed global optimization scheme to the solution of the dynamic MINLP problems, we introduce the orthogonal collocation strategy for converting the original dynamic problem into a conventional MINLP problem. With this conversion, each of the dynamical constraint is reformulated into a set of equivalent discrete constraint forms with decision variables being at the pre-specified collocations points. This effort leads the proposed two-stage global optimization scheme directly applicable to the solution of dynamic MINLP problem and makes the solution procedure quite easy. For demonstration, we applied the solution scheme to solve several optimal control problems of dynamic chemical processes having simultaneously the continuous and discrete variables. Extensive simulation results corroborate again the effectiveness and advantage of the proposed global optimization schemes for the solution of MINLP problems. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|