Improved Artificial Bee Colony Algorithm And Its Applications

Inspired by the foraging behavior of bees, Karaboga proposed a novel swarmintelligence optimization algorithm, i.e., artificial bee colony (ABC) algorithm, in2005.ABC algorithm possesses the characteristics of simplicity, less control parameters androbustness. ABC algorithm is widely applied in various fields such as functionoptimization, path planning, job shop scheduling, artificial neural network training andsystem identification. However, ABC has some drawbacks including inefficiency inexploiting, slow convergence speed and easily trapped into local optima. In this thesis,some works have been done to improve the performance of ABC algorithm, and itsapplication in chaotic system parameter identification, fractional controller design.To enlarge the searching scope of ABC algorithm and speed up its convergence, themulti-exchange neighborhood structure is incorporated into ABC algorithm, and thus twoimproved ABC algorithms with different neighborhood are proposed. First, path-exchangeneiborhood ABC is presented by introducing the path exchange neighborhood structureinto employed bees. Second, CNC-ABC algorithm is presented utilizing the cyclicexchange neighborhood structure and chaos. The proposed algorithms are tested usingbenchmark functions.Chaos is a class of complex dynamic system and it is an important problem to obtainthe parameter value of chaoc system. In this thesis, the parameter identification of chaoticsystem is sovled from the viewpoint of optimizatioin, where the parameter identificationproblem is converted to a parameter optimization problem. The proposed PN-ABCalgorithm is used to solve this optimization problem. Experiments on Lorenz chaoticsystem and Logistic system verify the effectiveness of the proposed method.A parameter optimization method for fractioal order PID (FOPID) controller isproposed based on CNC-ABC. In the proposed method, Bode model is used as refenrencemodel, the optimal parameter of FOPID controller is obtained by minimizing the outputerror between the actual sytem and that of reference model. The CNC-ABC algorithm isused to solve this optimization problem. The proposed method is applied to design FOPID controller for AVR system. Experimental results show that the designed FOPID controllerhas good performance.A nonlinear system identification method is propsed based on quantum-inspiredartificial bee colony (QABC) optimized multi-kernel least square support vector machine(MKLSSVM). First, QABC is presented utilizing the concept and principle of quantumcomputation. Then, the proposed QABC algorithm is used to optimize the parameters ofMKLSSVM. Experimental results show that the proposed method can obtain goodidentification results.A hybrid ABC algorithm is proposed by introducing the difference evolutionmutation strategy into ABC algorithm, i.e., DE-ABC algorithm. In DE-ABC algorithm,the DE/best/1mutation strategy is utilized to employed bees. The proposed DE-ABC isused to identify the parameters of Wiener model. Experimental results show thatDE-ABC can accurately identify the parameters of Winer model.