The Improvement Of Differential Evolution Algorithm And Gravitational Search Algorithm

Global optimization problems often appear in science and engineering field-s and so on.With the development of science and technology,the scale of the optimization problem becomes more bigger,the structure is more complicated,and the requirements for optimization technology are also more higher.Then,it has very important theoretical and practical value to design efficient algorithms for solving complex optimization problems.In recent years,the intelligence opti-mization algorithms are designed by simulating the behavior of groups in nature.Because of simple principle and operation,good optimization performance,no requirement of the continuity or differentiability of objective function,the intel-ligence optimization algorithm have attracted the attention of many researchers and achieves in many fields.As two typical swarm intelligence optimization algorithms,differential evolution algorithms and gravitational search algorithm-s have good performance on dealing with complex optimization problems and nonlinear problems,but they exist slow convergence and falling easily to local optimum and so on.To effectively balance the exploration and exploitation and overcome premature convergence and other shortcomings,thesis propose a d-ifferential evolution algorithm with group strategy and dynamic parameter set-ting,and a gravitational search algorithm based on dynamic gravitational con-stant and population size reduction.The main content of this thesis describes as follows:1.In order to prevent from falling into local optimum,a differential evolu-tion algorithm with group strategy and dynamic parameter setting is designed.A strategy is first designed to dynamically adjust the range of elite solutions a-mong the whole population during the evolutionary process to balance the glob-al exploration and local exploitation effectively.The population is then grouped according to individual fitness values,and different adaptive scaling factors are adopted,respectively.Finally,an adaptive cross rate is dynamically set to help the search jump out of the local optimal.2.A gravitational search algorithm based on dynamic gravitational constant and population size reduction is proposed to further enhance the global explo-ration of standard gravitational search algorithm and overcome its premature convergence and other shortcomings.First,the proposed algorithm effectively balance the global search and local development capabilities of the algorithm by adding sine terms and random terms to dynamically adjust the gravitational con-stants.Second,the population size and individual information are dynamically adjusted and screened by using population size decline to improve the perfor-mance and convergence speed of the algorithm.The proposed method is extensively evaluated on 30 benchmark problems from CEC2014.In contrast to other evolution algorithms,experimental results show that the effectiveness of the proposed algorithms.