计算机应用 ›› 2010, Vol. 30 ›› Issue (2): 472-475.

• 人工智能 • 上一篇    下一篇

利用前两代信息的改进粒子群优化算法

欧旭1,梁京章2,罗德相3,张新华4   

  1. 1. 广西医科大学
    2.
    3. 广西民族大学
    4. 山东省莱阳卫生学校
  • 收稿日期:2009-08-05 修回日期:2009-09-25 发布日期:2010-02-10 出版日期:2010-02-01
  • 通讯作者: 欧旭

New modified particle swarm optimization on basis of two latest generations

  • Received:2009-08-05 Revised:2009-09-25 Online:2010-02-10 Published:2010-02-01

摘要: 针对粒子群算法(PSO)在寻优后期尤其在高维搜索空间中无法得到满意结果的问题,提出了一种利用前两代信息的改进粒子群优化算法。在速度更换公式新加了一部分,该部分表示了粒子前两代的信息对自己下一步行为的影响。该部分主要根据当前粒子前两代位置求解出其前两代的中心位置,其作用类似于当前全局最优位置。同时深入探讨新加部分的学习因子范围及其对新改进算法的影响。仿真实验结果表明,新算法在全局搜索能力、收敛速度、精度和稳定性方面均有了显著提高。

关键词: 粒子群算法, 中心位置, 学习因子, 收敛速度, 稳定性

Abstract: A modified Particle Swarm Optimization (PSO) on the basis of the two latest generations was proposed to solve the problem that no satisfactory results can be reached during later period of PSO, especially in high-dimensional search space. A new part was added to the velocity of replacement formula,suggesting that the particle comprehensively utilized the information from the previous two acts to instruct its next step. Primarily based on the record of recent changes of the current particle in the two latest generations, the central location of the previous two generations of the particle was calculated,the role of which was to point out the current global optimal position. The paper,at the same time, discussed deeply a new learning factor and their impact on the new modified algorithm. The experimental simulation results show that global searching ability, convergence rate, accuracy and stability of the new algorithm have been improved significantly.

Key words: Particle Swarm Optimization (PSO), central location, learning factor, convergence rate, stability