Implementation And Application Of A Modified Multi Objective Particle Swarm Optimization Algorithm

The multi-objective optimization problem has always been a popular topic, it often need to balance the interests of many aspects when solving a problem in real life. Particle swarm optimization algorithm has received comprehensive attention because of its brief and valid algorithm model, less regulation parameters, fast convergence competence and good global optimization competence. However, with the improvement of the complexity of the optimization objective, the traditional particle swarm optimization algorithm has some limitations. For example, it is difficult to ensure the variety of particles when the convergence rate of the algorithm performance is good. The limitations of the algorithm restricts the competence of the algorithm to look for the optimal solution in the case of solving the problem which has multi peak objective functions and many local optimum points.Through the analysis of each detail, such as the influence of the initial distribution of particles, the renewal mechanism for speed and location, the file maintenance mechanism, the characteristics of the decision vector's changes, and the influence of the non-inferiority relationship on the preservation of particles and so on. This article seeks factors which hinder the convergence speed and optimizing ability of the algorithm and study factors which destroy the diversity of the algorithm. And then puts forward the improvement strategy according to the limitation of different mechanism.In view of the diversity of particle swarm optimization, Some of the changes are proposed. First, the grid mechanism and crowding distance are respectively used to maintain the external file, in this paper, the complexity of the algorithm is simplified, particle diversity are maintained meanwhile the iteration speed of the algorithm is not affected. Second, the particle velocity update model is modified and a small amount of disturbance is added in order to avoid together phenomenon because swarm may get excessive dependence on the global optimal solution. Third, the non-inferior particle selection mechanism, which is to real time monitor of particles in an external file and mix new particles into the population to produce more possibility of solutions in the decision space when local extreme points may appear. A modified IMOPSO algorithm is proposed.In solving the complex multi peak optimization objectives, the search competence of the particle swarm optimization algorithm is weak. Based on the idea of direct searching method and Gauss mutation operator, a local search algorithm is proposed. This method combines the partly search method with the universal search method organically. The search competence of the particle is enhanced, simultaneously, the particle maintains a high convergence rate. A modified DSMOPSO algorithm is proposed.In this paper, the validity of the improved algorithm is proved by comparing with the other algorithms on the test function. The algorithm is applied to solve the practical problem, and proved the practicability of the algorithm.