Techniques for improving differential evolution and artificial bee colony algorithms

Abstract

The differential evolution (DE) algorithm was widely studied in the past decade. DE is simple to implement, easy to use and computationally fast. The paradigm of DE is shown to be very powerful. Its variants (improved DE algorithms) secure competitive rankings in all IEEE Congress on Evolutionary Computation (CEC) competitions. The artificial bee colony (ABC) algorithm is a typical swarm intelligence (SI) method. It is one of the most prominent approaches in the field of bee-inspired methods. Applications of ABC can be found in classification, clustering, image processing and engineering design, etc. This thesis concentrates on the development of new algorithms based on the paradigm of DE and ABC respectively. The developed method for DE is a directional mutation operator. The purpose of this method is to speed up the convergence of DE. The developed methods for ABC are one-position inheritance mechanism, opposite directional search and hybridization with migration operator of biogeography-based optimization algorithm. Three improvements of ABC are obtained using these methods. The principle of information exchange amongst solutions is the main core behind these improvements. The pros and cons of these improvements are analyzed and compared through testing a large number of benchmark functions. All the results are statistically examined and compared with those of other SI and evolutionary algorithms. It is found that the proposed methods can improve the performance and speed up the convergence of the DE and ABC algorithms. We then apply our improved ABC method to two heating, ventilating and air conditioning (HVAC) applications: (1) design of duct system, and (2) energy management of HVAC heat rejection system. The experimental results indicate that our methods can improve the efficiency of the two systems. The two applications confirm the significance of our research. Finally we apply our improved ABC method to two electromagnetic design applications: (1) testing electromagnetic analysis method (TEAM) workshop problem 22, and (2) design of loudspeaker. Again, the experimental results show the effectiveness and efficiency of our method over other algorithms in terms of both solution quality and robustness.

Date of Award	2 Oct 2013
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Shiu Yin Kelvin YUEN (Supervisor)