With the continuous progress of science and technology, the problems encountered in scientific research and engineering practice are becoming more and more complicated, and the traditional calculation methods are facing the facts such as high complex calculation, long computing time and etc. Recently, the extensive research on swarm intelligent algorithms has provided a new idea to solve such problems, because they do not need any specific mathematical models and any special assumptions to the problems. Artificial Bee Colony(ABC) algorithm is put forward by Turkish scholar Karaboga in 2007 based on mimicking the intelligent behavior of bee populations. Among intelligent algorithms on bee swarm proposed in the same period, it is the most widely applied and researched algorithm, and now has become an emerging research branch within the field of swarm intelligent research. But up to now, the architecture of ABC algorithm is not mature enough, and some problems still exist in it, such as simple structure,single operator form and so on. In addition, its existing research mostly focuses on the single-objective optimization problem in some related fields, while the research on multi-objective problems is just starting. The research method is simple; the quality to solve such problems is not high; and the speed of convergence is slow. Therefore, to study how to improve ABC algorithm, especially according to different optimization problems how to adopt some mature solution strategies and operators in the-state-of-the-art algorithms to overcome the above shortcomings in ABC algorithm, has important theoretical significance and potential application value.After sufficient studying and deep exploring the current improved ABC algorithm, this thesis proposes several hybrid strategies against the shortcomings of the algorithm according to the specific optimization problems, designs several improved artificial bee colony algorithms,and performs a large number of numerical experiments in typical test functions. The main results of this thesis include the following:1、In view of the deficiencies of the original ABC algorithm in exploiting capacity, especially in the position near the optimal solutions the search ability of ABC becoming weak, and the convergence rate becoming slow, the chaotic map in this thesis is used to improve the local search performance of ABC algorithm. Wherein, the Logistic chaotic map is a very simple classical model adopted in many improved algorithms, but it has many shortcomings including the map results influenced greatly by the initial conditions, the values near “0” and “1”distributed more than other regions. Chaotic sequences generated by the Tent map have the periodicity and its distribution is more uniform than Logistic map, but due to the existence of the unstable periodic points and the fixed points, the distribution of some values is worse. To solve this problem, in this thesis, the improved Tent map is designed. Finally, Hennnon map is a two-dimensional chaotic mapping model, which has the characteristics of larger mapping space,complex dynamic characteristics and easy to implement, and it can be used to improve the search scope of the ABC algorithm.2、This thesis proposes two kinds of ABC algorithm based on chaotic search to solve the single-objective unconstrained optimization problem, of which one is GTENTABC based on the improved Tent chaotic mapping, and the other is HENABC based on Hennon chaotic mapping.In the experiment of seven benchmark functions, the above two algorithms, the commonly used Logistic chaotic search ABC algorithm, the Tent chaotic search ABC algorithm, and the basic ABC algorithm are compared with each other. The results show that the GTENTABC algorithm has better convergence speed and higher solution precision than the basic ABC algorithm and other chaotic search algorithms, whether in the unimodal or multimodal problems. And with the increase of the solution dimension, it also can keep better effectiveness and robustness. So the GTENTABC algorithm not only has the ability of global optimization, but also has strong ability of local search. Besides, the HENABC algorithm based on the two-dimensional Hennon chaotic map model wins the better results in high-dimensional and multi-modal functions, which shows that the HENABC algorithm can expand the search space of the algorithm, and it is more suitable to solve such complicated problems with higher dimension.3、Because of ABC algorithm having poor performance in the constrained optimization problems,inspired by the Memetic algorithm, this thesis proposes the MGT_ABC algorithm which is an ABC algorithm with the framework of Memetic algorithm under feasible rules to solve theconstrained optimization problems. Based on the front GTENTABC algorithm, the MGT_ABC algorithm uses Differential Evolution(DE) algorithm to perform the global search and the feasible rules to deal with the constraints conditions. In the beginning of MGT_ABC,the bee swarm individuals are proportionally selected to search the location of the nectar source according to the DE algorithm in order to enhance the diversity of the population. Then the following bees mine the neighborhood of the current nectar position based on a certain probability model to dynamically allocate the computing resource according to the current performance of the two update strategies(DE and GTENTABC), in order to fit the characteristics of constrained optimization problems. Experiments are carried out on 9constrained optimization problems, including the difficult Bump problem, and the results are compared with other algorithms in the literatures, which verify the effectiveness of the proposed MGT_ABC algorithm.4、A new multi-objective artificial bee colony(MOABC) algorithm is proposed by decomposing a multi-objective optimization problem(MOP) to many aggregate optimization problems. In the meantime, we adopt penalty-based boundary intersection(PBI) method to generate the aggregate functions through the minimization of the distance and the direction between the current solution and the ideal point, and therefore the algorithm could obtain the good distribution and convergence of the optimal solutions. Besides, the fitness is the improvement degree of every aggregate optimization problems, which could overcome the two shortcomings of the weighted sum of multi-objective function value usually in the conventional multi-objective artificial bee algorithm. In order to dynamically adjust the selection pressure of unemployed bees in the optimizing process of unemployed bees and avoid getting trapped in a local optimum, the Boltzmann mechanism is used to obtain the probability of unemployed bees following the employed bees. The algorithm is validated on CEC2009 problems and the problems with complicated Pareto set shapes in terms of four indicator: IGD, HV, SPR, and EPS. Experimental results show that our proposed algorithm can performs better than other state-of-the-art algorithms with competitive convergence and diversity, and can be considered as a promising alternative to solve MOPs. |