计算机科学与探索 ›› 2020, Vol. 14 ›› Issue (4): 637-648.DOI: 10.3778/j.issn.1673-9418.1909020

• 人工智能 • 上一篇    下一篇

独立自适应调整参数的粒子群优化算法

张其文,尉雅晨   

  1. 兰州理工大学 计算机与通信学院,兰州 730050
  • 出版日期:2020-04-01 发布日期:2020-04-10

Particle Swarm Optimization with Independent Adaptive Parameter Adjustment

ZHANG Qiwen, WEI Yachen   

  1. School of Computer and Communication, Lanzhou University of Technology, Lanzhou 730050, China
  • Online:2020-04-01 Published:2020-04-10

摘要:

针对传统粒子群优化算法在求解复杂优化问题时易陷入局部最优和依赖参数的取值等问题,提出了一种独立自适应参数调整的粒子群优化算法。算法重新定义了粒子进化能力、种群进化能力以及进化率,在此基础上给出了粒子群惯性权重及学习因子的独立调整策略,更好地平衡了算法局部搜索与全局搜索的能力。为保持种群多样性,提高粒子向全局最优位置的收敛速度,在算法迭代过程中,采用粒子重构策略使种群中进化能力较弱的粒子向进化能力较强的粒子进行学习,重新构造生成新粒子。最后通过CEC2013中的10个基准测试函数与4种改进粒子群算法在不同维度下进行测试对比,实验结果验证了该算法在求解复杂函数时具有高效性,通过收敛性分析说明了算法的有效性。

关键词: 粒子群算法(PSO), 惯性权重, 学习因子, 粒子重构

Abstract:

Aiming at the problems that traditional particle swarm optimization (PSO) algorithm is prone to fall into local optimum and depends on the value of parameters when solving complex optimization problems, an independent adaptive parameter adjustment particle swarm optimization algorithm (IAP-PSO) is proposed. The evolutionary ability of particle, the evolutionary ability of population and evolutionary rate are redefined. On this basis, the indep-endent adjustment strategy of inertia weight and learning factor of PSO is given, which effectively balances the ability of local search and global search. In order to maintain the diversity of population and improve the speed of particle convergence to the global optimal position, a new particle reconstruction strategy is proposed in the iteration process of the algorithm. The particles with weak evolutionary ability learn from those with strong evolutionary ability and generate new particles. Finally, 10 benchmark functions in CEC 2013 are compared with four improved particle swarm optimization algorithms in different dimensions. The experimental results show that IAP-PSO algori-thm has high efficiency in solving complex functions. Convergence analysis shows the effectiveness of the algorithm.

Key words: particle swarm optimization (PSO), inertia weight, learning factor, particle reconstruction