Studies On Improved Particle Swarm Optimization

Particle Swarm Optimization (PSO) is a new method used in the solution of Optimization Problems. Since it was proposed in 1995, the research articles about it have increased quickly. The concept of PSO is simple and easy to implement. It randomly initializes a certain scale of particle swarm, in which each particle with certain intelligence is used to represent the candidate solution in the specific optimization problem, and then uses the information of the group and individual particles to find an optimal solution quickly through the iterative evolution. This algorithm, based on the two theories—swarm intelligence and evolutionary computation, is favored broadly because of its many advantages such as its few parameters, simple setup procedure and the fast converge.Nowadays there have been a large number of papers about the improved algorithm in order to make the PSO algorithm perform better. They have been successfully applied to the engineering optimization problems. As the research is more and more deeply, its application fields are also expanding, and its performance is also greatly improved.This paper firstly does an in-depth study of the theoretical basis, basic principle and realization process of the PSO algorithm. It analyzes the influence of related parameters on the arithmetic performance, the efficiency of the algorithm and the implementations on the basis of simulation experiments. Considering the defects of the PSO, this paper analyzes what and which aspects should be improved on the basis of the particle movement characteristics, and the paper also explains the improvement and applied scope of the algorithm. By analyzing the basic theories and the improved methods, this paper proposes a multi-agent particle swarm optimization algorithm based on the theory of multi-agents, and gives specific procedures of this algorithm. The author proves the effectiveness of its improvement through the MATLAB simulation experiment, analyzes concretely the improved results, and meanwhile points out the weak points. Finally, this paper summarizes the author’s study on the PSO algorithm, and proposes a further research plan.It is proved that the improved MAPSO is more effective than the basic PSO when solving the more complex multimodal function optimization problems. The MAPSO truly realizes the global optimization of PSO by improving the accuracy of the algorithm and solving problems such as the possibility of trapping into the local optimal, or premature convergence. However, meanwhile the improved algorithm also increases the complexity. Therefore, the theory and performance of the algorithm need to be further perfected.