Particle Swarm Optimization Algorithm And Performance Analysis Of Multi-objective Optimization Problems

Multi-Objective Optimization problems widely exist in the practical work andscientific research in many fields, so for the Multi-Objective Optimization Algorithmresearch has always been highly attention. With the continuous development ofscience and technology era, especially the rapid development of computer science andtechnology, and promote the rapid development of solving optimization method. Andone of a mature kind of algorithm is Evolutionary Algorithm.According to the laws of the nature of biological evolution, people designs outthe Evolutionary Algorithm, which as Genetic Algorithm, Simulated AnnealingAlgorithm, Neural Network and Ant Colony Algorithm and Particle SwarmOptimization (PSO) Algorithm, is through constant research by scholars, experiencingthe improvement, generation after generation to solve more areas, more complexoptimization problems.In the Genetic Algorithm the most classic of the algorithm is the NSGA-II. thealgorithm reduces the computational complexity than before, at the same time alsocan improve the non dominated solution of uniform distribution of the real Paretofront. However, the algorithm still has some deficiencies and defects, for example: itis not high accurate in the high-dimensional data. And PSO algorithm has someadvantages such as simple process, fast convergence speed and high searchingefficiency etc, which can not only solve the single objective optimization problem, atthe same time on solving the multi-objective optimization problem it also works well.Multi-objective Particle Swarm Optimization Algorithm based on the designconcept of the basic PSO algorithm, combining with the Pareto dominance criterion tocompare between different solutions. Through the external file set to storenon-dominated solutions, and select global wizard from it. And by limiting the size ofthe file, according to the crowding distance descending order, beyond the size of thenon dominated solution will be deleted, that will reduce the operation cost. Keep thepopulation’s versatility by adding disturbance factor. And, in order to further improve the running speed of Particle Swarm Optimization algorithm, the use of competitionmechanism to quickly get the non-dominated solutions and external file sets.According to the analysis of experimental results before and after the improvedmulti-objective particle swarm algorithm, and at the same time the two algorithms arecompared with classical genetic algorithm the NSGA-II, analysis of advantages anddisadvantages of the algorithm.This article mainly under the environment of Matlab numerical experimentanalysis, by comparing the algorithm solving a multi-objective optimization problemof the average elapsed time, and the contrast of each algorithm to calculate theaverage distance of the Pareto front and the real Pareto front and adjacent targetvector distance variance measure to reflect the stand or fall of each algorithm.Through the experimental results concluded that the running speed of multi-objectiveParticle Swarm Optimization algorithm and the convergence and distribution ofsolutions are better than the NSGA II algorithm. And the improved multi-objectiveParticle Swarm algorithm greatly improves the search efficiency, reduces the runningtime. The chosen standard test functions are respectively dual objective function,ZDT-1, ZDT-2, ZDT-3and three objective functions DTLZ-2, DTLZ-7.