Dynamic optimisation by a modified bees algorithm

Abstract

A modified bees algorithm was applied to dynamic optimisation problems in chemical engineering. A two-level factorial experiment was used to tune the settings of the population parameters, based on the premise that it is most important to avoid those configurations that cause the worst performances than to look for those that reach the best performance. Tested on eight well known benchmark problems, the tuned algorithm outperformed the standard bees algorithm and other two well known optimisation methods. The performance of the proposed algorithm was also competitive with that of the state-of-the-art in the literature, and the solutions produced were very close to the known optima of the benchmarks. The results demonstrate the efficacy of the modified bees algorithm as a tool for the solution of dynamic optimisation problems. The results also proved the effectiveness of the proposed statistical parameter tuning algorithm, and indicated its competitiveness as an alternative to the standard complex and subjective trial-and-error methods.