Study On Multi-populations For Multi-objective Optimization PSO Algorithm And It’s Application

The study on multi-objective optimization problems has a wide range of particle value, particle swarm optimization algorithm is an effective method to solve multi-objective optimization problem. The current researches of multi-objective optimization are aimed to solve the convergence and distribution of the algorithm, some existing algorithms in solving multi-objective optimization problem has difficulty in comparing the two dominant solutions, especially for those high dimensional complex problems. To solve these problems, we proposed the multi-populations for multi-objective optimization problems, each population collaborated with each others.The algorithm includes following works:(1) The algorithm uses multi-populations to solve multi-objective optimization problem, a population optimize only one objective, we use an external archive to store pareto solution, meanwhile we change the velocity updating formula, so that each target’s not only influenced by the particle in their own population but also by the particles in the external archive, with these operations we can not only make every populations find their optimal solution, but also avoid the conflicts of problem solutions between multi-objective optimization problem solutions.(2) When selecting the external file values to guide all populations flight, we used the mean optimal particle for each target to boot, so that all particles are able to be flying towards the Pareto foreword.(3) In order to avoid premature of the algorithm, we introduce the elite variation strategies in external files to improve the population diversity and avoid to be trapped in local minima due to the convergence of the algorithm.Finally, we adopt the standard test functions to prove the algorithm, and we compared it with classical multi-objective optimization algorithm NSGA-Ⅱ. By comparing the distribution and convergence of the algorithm, we find the algorithm in this paper has a better performance. At the same time, we use the algorithm to solve multi-objective knapsack problem, and compared the results with the NSGA-Ⅱ, by contrast, we can find the algorithm in the paper can find solutions that get closer to the solution we need.