The Experimental Analysis Of Multi-Objective Particle Swarm Optimization

The optimization problem is a branch of Computational Science. It has a wide rangeof applications in engineering, manufacturing, finance and many other fields.Multi-Objective Optimization is one of the major research fields in optimizationproblems, many problems in our life need to optimize more than one object that areusually contradictory.Traditional methods usually transform Multi-Objective Problems into singleobjective problems by weighted summation, but they need information about the problem,it is difficult to deal with the problems in real life. Some of them can only get onesolution once. With the development of evolutionary algorithms, they make a good effectin Multi-Mbjective Problems and some of them are excellent, such as NSGA-Ⅱ andSPEA-Ⅱ.Particle Swarm Optimization Algorithm (PSO) is proposed by Kennedy andEberhart in1995, which based on social influence of social psychological model andsocial learning. Basic Particle Swarm Optimization Algorithm which has beensuccessfully applied to many problems is simple, and it is easy to come true, but it isinadequate in the convergence and the diversity of the population. It cannot be used tosolve the multi-objective problem directly. It needs to be improved.In this paper, a Multi-Objective Particle Swarm Optimization Algorithm is described.It uses the dominance relationship to construct the exterior set which saves thenon-dominated solutions from beginning to now and guides the algorithm close to Paretofront and maintains solution distribution. It makes effects by using rapid sequencingmethods. New methods have been used to select the local optimal position and globaloptimal position. It gets new mutation operation to make disturbance, and prevents thealgorithm being trapped in local optimal position.NSGA-Ⅱ is a multi-objective genetic algorithm, and it is a classic way for solvingmulti-objective optimization problem. In order to verify the effectiveness of the algorithm,this paper compares the improved Particle Swarm Optimization Algorithm with NSGA-Ⅱin the same experimental environment. The experiment chooses standard test functions ZDT and DTLZ, and use specific evaluation methods to compare two algorithms. Theresults show that, the Particle Swarm Optimization Algorithm is efficient, which runs lesstime than NSGA-Ⅱ. Also it is better than NSGA-Ⅱ in convergence. And it makes goodeffect in keeping the solution set of distribution. The algorithm is simple and the result isstable, and is more efficient to solve the multi-objective optimization problems.