Research On An Improved Filled Function Method And Hybrid Optimization Algorithm

The global optimization problem permeates in all aspects of life,and the powerful tools for solving the problem are emerging.The existence of multiple local minima is a challenge in solving global optimization problems,but the conventional optimization methods are difficult to achieve good results,the proposed intelligent algorithm solves this difficulty well,the Particle Swarm Optimization(briefly,PSO)algorithm has been widely used to solve such problems because of its advantages,such as,easy implementation and rapid calculation.However,in the later evolution,PSO fall into a local optimum easily for resolving the problems with multi-extreme value and high-dimensional.Therefore,it is of both theoretical and practical significance to study more efficient method that can refrain the algorithm sinking into a local optimal solution.In this paper,it focus on defect that PSO is easy to jump into local optimal solutionswhen handling the problems with a lot of local optimal solutions,and then,the mechanism of jumping out of the local solution is introduced.On the basis of framework of PSO,it fuses the filled function method(briefly,FFM)with the advantages of deviating from local solution,and a hybrid algorithm based on both is proposed.Firstly,for the FFM,a new type of simple form parameter filled function without exponent term is constructed,the function refrains the cumbersome adjustment of multiple parameters and prevents the optimal point from losing due to the index term,at the same time,the function is continuously differentiable and has good analytical properties in theory.In view of the idea of better initial points could enhance local search,a local search with uniform strategy is designed.A new filled function algorithm(briefly,NFFA)based on the above two points is proposed,and the numerical experiments show that the NFFA is effective and efficient.Secondly,for the hybrid optimization algorithm,the adaptive PSO is the framework of the algorithm and the NFFA is embedded into the PSO,based on the mixture of algorithms and the improved PSO(briefly,IPSO)is constructed to intensify their search capabilities.The IPSO uses the filled function method at a local solution to help the algorithm jump out this point,and it can avoid the phenomenon of falling into local optimum in the iterative process to find a better one,and enhance the efficiency of solving the multi-peak global optimization problem.Finally,experiments and algorithm comparisons are performed on the 30 and 50 dimensions of six benchmark functions in CEC'2013 test set,and can verify the performance of the IPSO.The results indicate that the IPSO is feasible and more efficient,moreover,it solves multimodal function optimization more efficiently.