Hybrid nature-inspired computation methods for optimization

OF DOCTORAL DISSERTATION HELSINKI UNIVERSITY OF TECHNLOGY P.O. BOX 1000, FI-02015 TKK http://www.tkk.fi Author Xiaolei Wang Name of the dissertation Hybrid Nature-Inspired Computation Methods for Optimization Manuscript submitted 23/2/2009 Manuscript revised 28/4/2009 Date of the defence 29/5/2009 Monograph Article dissertation (summary + original articles) Faculty Faculty of Electronics, Communications and Automation Department Department of Electrical Engineering Field of research Industrial Electronics Opponent(s) Prof. Mark J. Embrechts Supervisor Prof. Seppo J. Ovaska Instructor Docent Xiao-Zhi Gao Abstract The focus of this work is on the exploration of the hybrid Nature-Inspired Computation (NIC) methods with application in optimization. In the dissertation, we first study various types of the NIC algorithms including the Clonal Selection Algorithm (CSA), Particle Swarm Optimization (PSO), Ant Colony Optimization (ACO), Simulated Annealing (SA), Harmony Search (HS), Differential Evolution (DE), and Mind Evolution Computing (MEC), and propose several new fusions of the NIC techniques, such as CSA-DE, HS-DE, and CSA-SA. Their working principles, structures, and algorithms are analyzed and discussed in details. We next investigate the performances of our hybrid NIC methods in handling nonlinear, multi-modal, and dynamical optimization problems, e.g., nonlinear function optimization, optimal LC passive power filter design, and optimization of neural networks and fuzzy classification systems. The hybridization of these NIC methods can overcome the shortcomings of standalone algorithms while still retaining all the advantages. It has been demonstrated using computer simulations that the proposed hybrid NIC approaches are capable of yielding superior optimization performances over the individual NIC methods as well as conventional methodologies with regard to the search efficiency, convergence speed, and quantity and quality of the optimal solutions achieved.