一种极坐标下的改进多目标粒子群优化.pdfVIP

第 24 卷第 8 期 系统仿真学报© Vol. 24 No. 8 2012 年 8 月 Journal of System Simulation Aug., 2012 一种极坐标下的改进多目标粒子群优化 逄 珊，邹海林，苏庆堂 (鲁东大学信息与电气工程学院，烟台 264025) 摘 要：对现有多目标粒子群优化算法的全局最优解选择机制进行分析，指出其不足。在此基础 上设计一种全新的极坐标下的选择机制：利用极坐标下解和粒子的角度信息计算适应度角度，选 择适应度角度最大的解作为粒子的全局最优解。并针对多目标粒子群优化算法在迭代后期收敛变 慢的问题改进位置更新公式：将位置更新过程产生的中间点也作为粒子新位置的候选解，有效提 高算法收敛速度。对测试函数的仿真试验表明，所提出的改进算法在解集的分布性和收敛性上较 其它典型算法有明显提高。 关键词：多目标优化；粒子群优化；极坐标；分布性 中图分类号：TP301 文献标识码：A 文章编号：1004-731X (2012) 08-1576-06 Improved Multi-objective Particle Swarm Optimization in Polar Coordinates PANG Shan, ZOU Hai-lin, SU Qing-tang (College of Information Science and Engineering, Ludong University, Yantai 264025, China) Abstract: Representative global best solution selecting strategies for Multi-objective Particle Swarm Optimization was analyzed first. Defects of these strategies were pointed out. A new selecting strategy in polar coordinates was proposed. The new strategy used inclination angles of particles and solutions to calculate “fitness angle”. Solution with maximum fitness angle was selected as particle’s global best. To overcome slowdown of convergence during MOPSO’s later iterations, a new position update equation was also designed: the relay points produced during position updating were selected as the candidates for new position. The new position update equation helps to improve convergence speed . Results on test functions show that the proposed algorithm can improve the distribution and convergence performance effectively for MOPSO algorithms. Key words: multi