改进的粒子群优化算法及应用.pdf

摘 要
粒子群优化算法(PSO)是近年来被广为关注和研究的一种智能优化算法，作为智能
优化算法的重要分支，粒子群优化算法具有许多不可比拟的优越性如：实现简单、收敛
速度快、具有较强的全局搜索能力、较少的参数设置。目前，该算法已成功应用于函数
优化、神经网络训练、模式识别、模糊系统控制等诸多领域。
本文分析了粒子群优化算法的原理、流程、改进和应用，总结目前 PSO 算法研究的
成果，对比分析了目前对粒子群优化算法的多种改进。针对现有 PSO 算法容易陷于局部
极值、收敛速度慢和精度差，而其他的一些改进方法，往往是在改动中使得算法变得更
加的复杂，提出了一种简化的算法 MSPSO，主要是针对粒子群算法的特点及其公式本身
的特点，改进速度更新公式，使粒子更快的获取与当前全局最好位置的差异，增强粒子
的学习能力，并将新算法与标准粒子群算法和线性递减权重粒子群算法进行比较，实验
结果表明改进的粒子群优化算法保留了基本粒子群优化算法的优点，并具有更好的优化
效果。
最后，用改进的粒子群优化算法优化神经网络的权值，并将其应用到信用评估中。


关键字：粒子群；优化算法；进化计算













I
Abstract
Particle swarm algorithm(PSO),has been paid attention and researched an
important branch of intelligent optimization algorithms,it has many advantages,such as:easy
to be realized,converging quickly,strong ability to global search,less ,PSO
have been successfully applied in many areas such as function optimization,neutral network
training,model identification,fuzzy system control,etc.
In this paper,we talking about the basic principle of PSO,rocesses of PSO algorithm,we
also describes its’ various improvements and paper analyses the elements of
PSO,for the existing local extremum vulnerable,slow convergence and poor accuracy
deficiencies and some other method,often making changes in more complex
algorithms,proposes a simplified SPSO,mainly for the characteristics of particle swarm
optimization and the characteristies of the formula velocity updating formula is
changed,so one particle can acquire more information of the global best position,the
experimental results show the novel PSO reserve many advantages of the basic PSO,and it
has better performance when compared with the standard PSO(SPSO),and the linearly
decreasing weight(LDW) of the particle swarm.
Finally we use the MSPSO algor