基于子群混合与变异的微粒群算法及其应用  

Particle Swarm Optimizer with Shuffled Subpopulations Mutation and Its Application

作　　者：王辉 钱锋[1,2] 

机构地区：[1]华东理工大学化工过程先进控制和优化技术教育部重点实验室,上海200237 [2]华东理工大学信息科学与工程学院,上海200237

出　　处：《系统仿真学报》2009年第12期3603-3608,共6页Journal of System Simulation

基　　金：国家杰出青年科学基金(60625302);国家973项目(2009CB320603);国家863计划课题(2008AA042902;2007AA041402);国家科技支撑计划(2007BAF22B05);国家自然科学基金面上项目(60804029;20876044);上海市国际科技合作基金项目(08160710500);高等学校博士学科点专项科研基金新教师基金项目(200802511011);长江学者和创新团队发展计划(IRT0721);高等学校学科创新引智计划(B08021);上海市重点学科建设项目(B504)

摘　　要：提出一种子群混合与变异的微粒群算法(SSMPSO),按适应值大小将微粒均匀分为两个子群分别进行目标优化,当子群在一定进化代数内未满足收敛条件时重新混合为一个种群。对混合种群中部分适应值差的微粒实施变异,用适应值好的微粒替代。微粒群反复进行分群进化、子群混合和种群变异操作,直至算法满足终止条件。SSMPSO算法通过几种典型高维非线性函数进行了测试,结果表明其性能明显优于基本微粒群算法。将SSMPSO算法用于常压塔汽油干点软测量,实验表明SSMPSO算法建立的汽油干点软测量模型比用基本微粒群算法所建软测量模型有了很大提高。Particle Swarm Optimizer （PSO） with the idea of SSMPSO （shuffled subpopulations and mutation PSO） was improved. In this algorithm, the swarm is divided into two subpopulations by the fitness of particles and all particles will be shuffled together to be a new swarm again if the terminal conditions don＇t satisfied after certain iterations. Some of the particles with poor position will be mutated to other particles with better position of the swarm. The process will be repeated until the terminal conditions to be satisfied SSMPSO was tested with some benchmark problems and compared with traditional PSO. The results indicate that SSMPSO performs better. Furthermore, SSMPSO is applied to train artificial neural network to construct a soft-sensor of gasoline endpoint of crude distillation unit. The results show that the model constructed by SSMPSO is better than that of PSO.

关 键 词：微粒群算法 子群 混合 变异 软测量 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]