第7卷第6期 智能系统学报 Vol.7 No.6 2012年12月 CAAI Transactions on Intelligent Systems Dec.2012 D0I:10.3969/i.issn.16734785.201205011 网络出版t地址：htp://www.cnki.net/kcma/detail/23.1538.TP.20121116.1700.002.html 交互学习的粒子群优化算法 秦全德，李丽，程适23，李荣钩 (1.深圳大学管理学院，广东深圳518060；2.英利利物浦大学电气电子工程系，英国利物浦L693G;3.西交利 物浦大学电气电子工程系，江苏苏州215123：4.华南理工大学工商管理学院，广东广州510640) 摘要：分析基本的粒子群优化学习机制的缺陷，启发于人类社会不同群体之间可以交互学习的特点，提出了一种 改进粒子群优化算法一LPS0.在LPS0算法中，粒子由2个种群构成.当2个种群中最佳的全局最优位置在连续 一定的迭代次数内没有改善时，执行交互学习策略.依据每个种群的全局最优位置的适应值，运用模拟退火的机制 和轮盘赌的方法确定学习种群和被学习种群.提出了一个基于适应度排序的经验公式，计算学习种群中的每个粒子 向被学习种群学习的概率.为了摆脱选择压力，采用了一种速度变异的方法.多个测试函数的数值实验结果表明，L PS0具有较好的全局搜索能力，是一种求解复杂问题的有效方法. 关键词：粒子群优化算法；交互学习；学习策略；学习行为：群体多样性 中图分类号：TP18文献标志码：A文章编号：16734785(2012)06054707 Interactive learning particle swarm optimization algorithm QIN Quande',LI Li',CHENG Shi23,LI Rongjun (1.College of Management,Shenzhen University,Shenzhen 518060,China;2.Department of Electrical Engineering and Electron- ics,Liverpool University,Liverpool L69 3GJ,UK;3.Department of Electrical and Electronics Engineering,Xian Jiaotong-Liverpool University,Suzhou 215123,China;4.School of Business Administration,South China University of Technology,Guangzhou 510640, China) Abstract:Analyzing the drawbacks of learning mechanism in the basic particle swarm optimization(PSO),an in- teractive learning particle swarm optimization (ILPSO)is presented,which is inspired by the phenomenon in hu- man society that individuals in different groups can learn each other.Particles are composed of two populations in ILPSO.When the best particles fitness value of two populations does not improve within a certain number of suc- cessive iterations,interactive learning strategies are implemented.According to the best particle's fitness value of each population,a simulated annealing mechanism and roulette method are used to identify the learning population and the learned population.This paper proposes an empirical formula of sorting fitness value to calculate the proba- bility of each particle in the learning population learning from the learned population.In order to escape selection pressure,a speed mutation method is used.The numerical experimental results of some benchmark functions show that ILPSO has good global search capability and is an effective method for solving complicated problems. Keywords:particle swarm optimization algorithm;interactive learning;learning strategy;learning behavior;popu- lation diversity 粒子群优化(particle swarm optimization,PSO) 了鸟群觅食过程中的迁徙和群集行为.PS0算法 算法是一种基于种群搜索的随机优化技术，其模拟 具有概念简单、控制参数少、收敛速度快和易于编程 收稿日期：20120507.网络出版日期：2012-11-16 实现的优点21，自提出以来受到广大学者的关注 基金项目：国家自然科学基金资助项目(71071057,71001072)；广东 但PS0算法同其他的随机搜索方法类似，在求解复 省自然科学基金资助项目(S2011010001337). 通信作者：秦全德.E-mail:qinquande@mail.com. 杂多峰函数时，容易陷入局部最优3].为了提高算