Differential Evolution Algorithm With Control Parameter Co-Evolution and Its Application

doi:10.3696/j.issn.1006-396X.2010.01.022

Journal of Petrochemical Universities ›› 2010, Vol. 23 ›› Issue (1): 93-98.DOI: 10.3696/j.issn.1006-396X.2010.01.022

Key Laboratory of Advanced Control and Optimization for Chemical Processes of Ministry of Education, East China University of Science and Technology, Shanghai 200237,P. R. China

Received:2009-02-14 Revised:2009-10-10 Published:2010-03-25 Online:2010-03-25

控制参数协进化的差分进化算法及其应用

范勤勤,颜学峰

化工过程先进控制和优化技术教育部重点实验室, 华东理工大学, 上海200237

作者简介:范勤勤(1986 -), 男, 浙江宁波市, 在读硕士
基金资助:
国家自然科学基金(20776042);国家863 项目(2007AA04Z164);教育部博士点基金(20090074110005);上海市曙光
计划(09SG29);上海市重点学科建设项目资助(B504)

Abstract

Abstract:

In order to implement dynamic and self-adaptive adjustment of control parameter with population evolution, a novel differential evolution algorithm with control parameter co-evolution (DE-CPCE) was proposed. In DE-CPCE, control parameter were designed as the symbiotic individual of original individual, and each original individual had its own symbiotic individual. Differential evolution operator was applied to search the global optimization solution of problem; meanwhile, it was also employed to co-evolve the population consisting of symbiotic individuals according to the evolution efficiencies of original individuals. Thus, with the evolution of the population consisting of original individuals in DE-CPCE, control parameter were dynamically and self-adaptively adjusted and the real-time optimum control parameter were obtained. The results of the experiments show that control parameter of DE-CPCE have dynamic and self-adaptive property; the probability and precision of obtaining the global optimization solution, and the convergence are all better than those of DE/rand/1, DE/best/1, DE/rand-to-best/1, DE/rand/2, DE/best/2, self-adaptive Pareto DE and self-adaptive DE. Further, DE-CPCE is applied to estimate the kinetic model parameter of oxidation, and the result is better than that of the reference.

Key words: Differential evolution algorithm , Co － evolution , Self － adaptive , Kinetics parameters

摘要： 提出一种控制参数协进化的差分进化算法(DE-CPCE),实现算法控制参数随种群搜优进展,自适应动态调整。D E-CPCE算法将控制参数作为原始个体的共生个体,且每一个原始个体都有各自的共生个体;算法在对原优化问题进行差分进化搜优的同时,以原始个体进化效率作为共生个体(即控制参数)的评价,并通过共生个体的差分进化操作实现其协进化。D E-CPCE算法能随优化问题搜优进展,自适应动态调整算法控制参数,实时为算法搜优提供最优的控制参数。仿真研究表明,DE-CPCE算法的控制参数具有动态自适应性;并且在与文中所提及的算法(DE/rand/1,DE/best/1,DE/rand-to-best/1,DE/rand/2,DE/best/2,self-adaptive Pareto DE and self-adaptive DE)比较中,该算法能以较高概率求得全局最优值,且收敛速率快,求得最优解的精度高。同时,应用 DE-CPCE算法估计 SO2催化氧化反应动力学模型参数,结果优于文献报道。

关键词: 差分进化算法 , 协进化 , 自适应 , 动力学参数

FAN Qin-qin, YAN Xue-feng. Differential Evolution Algorithm With Control Parameter Co-Evolution and Its Application[J]. Journal of Petrochemical Universities, 2010, 23(1): 93-98.

范勤勤,颜学峰. 控制参数协进化的差分进化算法及其应用[J]. 石油化工高等学校学报, 2010, 23(1): 93-98.