A Study of Particle Swarm Optimization with Random Particles and Fine-Tuning Mechanism for Multimodal Functions Problems

• Main Authors: Yu-hao Su, 蘇昱豪
• Other Authors: Sen-Lin Lu
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/9r4kux

碩士 === 國立臺灣科技大學 === 機械工程系 === 94 === This thesis aims at developing a modified particle swarm optimization (PSO) algorithm. The proposed method called PSO-RPFT will introduce two operators, “Random Particles” and “Fine-Tuning”, into the PSO algorithm. The former can prevent the population from trapping into the local optimum and the latter can promote the ability of local search to modify the defects of high similarity of individual particles on the late period of search following PSO algorithm. The complete architecture of random particles and fine-tuning operators is established in this thesis. At last the performance of PSO-RPFT and PSO-CF which was used widely in this filed will be compared by optimizing seven massively multimodal functions with varying complexities. The results show that the performance of PSO-RPFT is better than PSO-CF on both the search success rate and the average convergence generations.