Research On Improvement Of Multi-objective Particle Swarm Optimization Algorithm

At present,there are several problems that are the key to the research of multi-objective particle swarm optimization.The first problem is how to choose an excellent global optimal solution to guide the flight of other particles in the population,so that the particles can obtain effective flight experience.The second problem is how to balance the ability of global exploration and local development of the algorithm,so that the algorithm can get rid of the interference of local extremum.The third problem is how to control the size of the external archive,improve the calculation efficiency of the algorithm,and save the calculation cost.Based on the above key issues,the main research work of this article is as follows:(1)In order to improve the convergence accuracy of algorithm,a multi-objective particle swarm optimization algorithm based on crowded distance(CDMOPSO)is proposed in this paper.This algorithm uses adaptive grid technology and crowded distance to propose a strategy for selecting global learning samples through proportional allocation.At the same time,considering the size of the external archive and the distribution of the population,the algorithm proposes a dynamic external archive maintenance strategy.This strategy also maintains the good distribution of non-inferior solutions while maintaining the size of the external archive.(2)To flexibly adjust particle's flying velocity,jump out of the local optimum.This paper proposes a fuzzy multi-objective particle swarm optimization algorithm based on hybrid learning(HFMOPSO).This algorithm adjusts the flying speed of particles by calculating the crowding degree of particles.In HFMOPSO,to improve the quality of the candidate solutions in the external archive,a hybrid learning strategy is proposed to update the non-dominated solutions in the external archive to provide higher quality global learning samples for the particles in the population.(3)To balance the ability of global exploration and local development of the algorithm.This paper proposes a multi-objective particle swarm optimization algorithm with Cauchy-Gaussian dynamic mutation(MOPSO-CGDM).In MOPSO-CGDM,the particle position update formula is adjusted by a linear decreasing strategy,in the early stage of the search,a stronger global search ability is needed to find a better solution,and in the late stage of the search,a stronger local search ability is needed to explore a better local solution region.At the same time,the algorithm uses Cauchy mutation and Gaussian mutation to alter the position of particles alternately to adjust the particle's ability to explore in the objective space and improve the convergence accuracy of the algorithm.This paper compares the proposed algorithms with other multi-objective optimization algorithms in the same experimental environment.The proposed algorithms all show good performance.