Modification And Application Of Particle Swarm Optimization Algorithm

Particle Swarm Optimization Algorithm (PSO) is a famous intelligence algorithm, whichalgorithm that finds the most advantages by exchanging information between groups ofparticles. Compared with many intelligence algorithm, it has many advantages, such as：simple principle, less parameters, easy to be realized, so it has attracted increasing attentionand has been a new hotspot in the filed of artificial intelligence in the past decade. Now, PSOalgorithm has become a widely used optimization tools. It has been successfully applied inmany areas such as function optimization, neutral network training,intelligent control, etc.In this paper, we talking about the basic principle and processes of PSO algorithm, wealso describe its’ various improvements and application. The paper analyses the elements ofPSO, since it get into local extremum easily, get convergence slowly and get poor accuracydeficiencies, we proposes two improved particle swarm optimization algorithm. The firstmodified algorithm is chaos particle swarm optimization, for the features of optimizationquestions, we use chaos theory to initialize the position and speed of particles, not to changethe random nature of initialized particle, the use of chaos enhances the diversity of speciesand ergodicity of particle search, at a large amount of the initial group on the basis of meritfrom the initial groups. In the process of iteration, when the optimal particle of populationappears stagnation, we use chaos to mutate the particle, it can help escape from localminimum points and improve the result of optimization. The second modified algorithm ismulti-population particle swarm algorithm, by using different developed and detectiveabilities of different populations to search the optimal solution paralleled, getting the higheraccuracy by global search and local search the space. We propose an adaptive strategy updatespeed on the basis of multi-population particle to make the convergence faster. By doingexperiments through five standard test functions verify the effectiveness and superiority of themodified algorithm, we found that chaotic particle swarm algorithm has some performanceimprovements for low-dimensional function optimization, but performance is poor forhigh-dimensional function, multi-swarm particle swarm optimization algorithm onlow-dimensional and high-dimensional function optimization algorithm can get better result,the algorithm can improve the accuracy, stability and convergence rate. Finally, we use the multi-population particle swarm algorithm to solve softwarereliability allocation questions based on solving the constraint problem with PSO. Usingmulti-population particle swarm algorithm to solve practical problem we can prove that themodified algorithm has better performance.