Research On Artificial Bee Colony Algorithm For Solving Global Optimization Problems

With the rapid development of science technology and industrial technology,today's optimization problems gradually exhibit multi-modality,non-convexity,non-differentiability,and discontinuity characteristics.While bringing great challenges to traditional mathematical optimization methods,it also promotes the birth of new optimization methods.Artificial Bee Colony(called ABC)is a new swarm intelligence optimization algorithm that simulates foraging behavior of bees.Because of its simple structure,good robustness and easy implementation,this algorithm has captured much attention and research of scholars in many fields once it has been proposed,and it has been successfully applied to many practical optimization problems.In spite of this,there are still many deficiencies in ABC,such as unbalanced exploration and development capabilities,slow convergence rate,simplex search strategy and lack of information exchange between bees.To address these concerning issues and increasing the accuracy,convergence speed and robustness of ABC,in this thesis,we proposed two improved ABC variants.The main contributions are summarized as follows:(1)GABC and CABC are two outstanding ABC variants for solving global optimization problems.GABC is found to be suitable for solving unimodal and simple multimodal problems because of its exploitation search strategy.CABC works well in exploration due to the uncertain choice of individuals,and it's suitable for solving complex multimodal problems.In order to combine the advantages of both,this paper presents a hybrid artificial bee colony algorithm(called HGCABC).In the employed bee phase,CABC and Modified ABC/best/1are used to search,and the parameter p is introduced to control the frequency of use of both.In the onlooker bee phase,we give priority to seach around the good food sources and designed a new search strategy to fully exchange the location information between good food sources.The results of simulation experiments conducted on 52 benchmark functions and 7practical optimization problems proved that the accuracy,convergence speed and robustness of HGCABC are better than many other state-of-the-art ABC variants.(2)Inspired by the principle of professionalism in human society,in order to distinguish the search ability between different type of bees and different individuals in the same type,wedesigned a multi-population based artificial bee colony algorithm(called MPABC_RA).In the employed bee phase,the entire population is divided into three subgroups according to the objective function value.Then each subgroup is allocated with different search strategy to balance the exploitation ability and exploration ability.In the onlooker bee phase,the superior solutions are allocated with more resources to evolve.And onlooker bees fully exploit the area between the locations of the selected superior solutions and the current best solution by a novel search equation.Experimental results on 52 benchmark functions and 7 practical optimization problems demonstrate that MPABC_RA has better performance than HGCABC.