An Improved Particle Swarm Optimizer For Solving Numerical Optimization Problems

• Main Authors: Chen, Po-Han, 陳泊翰
• Other Authors: Hsieh, Sheng-Ta
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/33293790769072323018

碩士 === 亞東技術學院 === 資訊與通訊工程研究所 === 101 === In this paper, an improved particle swarm optimizer algorithm is proposed for solving single-objective numerical optimization problems. The proposed method can add breadth and depth of search capability. In general, particle swarm optimization algorithm utilizes entire swarm to search optimal solution in solution space. Once a particle fall into local optimal, the others will be clustered to search around this particle. It will reduce particles’ exploring capability. In order to improve traditional PSO, four mechanisms are proposed. First, grouping mechanism is joined to randomly divide the swarm into two sub-swarms. It can avoid all particles fall into a local optimum. Second, in order to enhance particles’ exploring capability a new particle movement mechanism is involved. Third, in order to speed up solution searching procedure, an information sharing mechanism for two sub-swarms is also proposed. Finally, the global best particle will be mutated for increasing the chance to escape from local optimum. In the experiments, 25 test functions of CEC 2005 are used to compare the proposed method with the recent PSO works. From the results, it can be observed that the proposed method can find better solutions than other PSO variants for solving most test function. The proposed method can quickly exploration to a better solution than related works.