Study On Particle Swarm Optimization Algorithm Based On The Combinatorial Strategy Of Phased Improvement

Optimization problems exist in many different fields such as production planning,vehicle path planning,and controller parameter setting.With the development of science and the advancement of information technology,more and more swarm intelligence algorithms and evolutionary algorithms are used to help solve practical optimization problems.Among them,Particle Swarm Optimization(PSO)has become the research focus of academic circles in recent years because of its simple principles,few parameters,and fast convergence speed.However,PSO cannot overcome the shortcomings of local optimality and poor algorithm diversity during application.It is of great significance to research on particle swarm optimization with better optimization effect.Based on the practical requirements of resource scheduling optimization problems,the research progress and trends of particle swarm optimization at home and abroad,the structural characteristics of particle swarm optimization is analyzed in this thesis and a phased improved combination strategy is proposed to improve the overall performance of particle swarm optimization algorithm.Aiming at the single-objective constrained optimization problem and the multi-objective optimization problem,the corresponding particle swarm improvement algorithm is proposed by this thesis.In order to test the effectiveness of the improved method,this thesis conducts simulation experiments through related problem models and classical functions.In order to verify the practicability of the improved method,the improved algorithm is applied to industrial practical cases.The main research contents are as follows:For the particle swarm optimization algorithm to overcome the problems of local optimization,improve the accuracy of the solution and shorten the overall performance of the solution time,on the basis of the particle swarm algorithm structure is divided into two stages of population "exploration" and "exploitation",the research idea of combining improvement from three directions of algorithm parameters is adopted,algorithm structure and multi-algorithm fusion in stages,and a framework for improving combination in stages is proposed.In the "exploration" stage of the population,through particle evacuation operations and learning from elite particles,the search range of the population is expanded,and the algorithm's ability to jump out of the local optimum is improved;In the "exploitation" stage of the population,the directional migration of particles can improve the quality of the population,accelerate the convergence of the population,and shorten the time for the algorithm to output.For the elite particles of the population,a competitive local search is used to improve the final solution quality.By balancing the improvement strategies at various stages,the overall performance of the algorithm is improved.Aiming at the single-objective constrained optimization problem with constrained characteristics,a single-objective constrained particle swarm optimization algorithm is proposed to improve the combination strategy in stages under the framework of improved combination in stages.Based on a feasible initialization strategy,in the "exploration" stage,the population particle entropy is calculated,and elite particles are selected to guide the evacuation of particles in dense areas.This strategy fixes particles that do not meet constraints.In the "exploitation" stage,the poor particles are migrated into the optimal area,and improve the accuracy of the output solution by globally optimal competitive search.Through the actual cloud computing task optimization scheduling model,the results are compared with the classical single-objective evolutionary algorithm to verify the improvement effect of the proposed phased improved combination strategy on the performance and solution time of the particle swarm optimization algorithm.Aiming at the multi-objective optimization problem with the characteristics of output non-single optimal solution,a multi-objective particle swarm algorithm with improved combination in stages is proposed by this thesis.Based on combination method,this article focuses on expanding the search range of populations by evacuating dense particles during the "exploration" stage.In the "exploitation" stage,through the improvement of the quality of non-dominated particles and the competitive search of the archives,in order to enrich the distribution of the frontier of Pareto.Through the combination of improved strategies,the Pareto generated by the algorithm is close to the real front.The classic multi-objective test function simulation experiment is used to prove that the optimization strategy significantly improves the performance of the multi-objective particle swarm algorithm on the frontier distribution,result diversity and other indicators.Appling the proposed combined improvement method in stages to a deep learning parameter optimization project of a locomotive bolt assembly management platform,an improved ISPSO-CNN algorithm is proposed.This algorithm help improve the accuracy of image recognition and solve practical problems in industrial assembly applications.Through practical application,the algorithm proves the practicability of improving the combination strategy in stages.