On the solution of petrochemical blending problems with classical metaheuristics

Abstract

In this paper a comparison of classical metaheuristic techniques over different sizes of petrochemical blending problems is presented. Three problems are taken from the literature and used for initial comparisons and parameter setting. A fourth instance of real world size is then introduced and the best performing algorithm of each type is then applied to it. Random search techniques, such as blind random search and local random search, deliver fair results for the smaller instances. Within the class of genetic algorithms the best results for all three problems were obtained using ranked fitness assignment with tournament selection. Good results are also obtained by means of continuous tabu search approaches. A simulated annealing approach also yielded fair results. Comparisons of the results for the different approaches shows that the tabu search technique delivers the best results with respect to solution quality and execution time for all of the three smaller problems under consideration. However, simulated annealing delivers the best result with respect to solution quality and execution time for the introduced real world size problem.