摘要
利用分布估计算法(EDA)的全局搜索性能及差分进化(DE)算法的局部优化能力,提出了一种多目标优化问题的混合智能求解方法DE-EDA。DE-EDA的子代个体由两部分构成,一部分按差分进化算法生成,另一部分则是通过对分布估计算法的概率模型进行随机采样生成。利用模拟退火技术在线调整尺度因子Pr,即在进化的初期选择较大的Pr,以保证EDA起主导作用,由EDA引导DE搜索向Pareto前端,增加全局搜索能力,然后在进化的过程中逐渐降低Pr,使得DE逐渐占据主导作用,确保解精确收敛到Pareto前端。通过4组基准函数来测试算法性能,并与NSGA-II和DE算法进行实验比较,结果表明该方法不仅解的多样性和分布性好,而且能够有效提高种群进化的收敛速度,是一种求解多目标优化问题的有效方法。
A hybrid intelligent algorithm called DE-EDA for solving multi-objective optimization problems was proposed by taking advantage of the global searching capability of estimation of distribution algorithm (EDA) and the local optimizing capability of differential evolution (DE). The offspring population of DE-EDA is composed of two parts, one part of a trial solution generated comes from the DE, and the other part is sampled in the search space from the constructed probability distribution model ofEDA. A scaling factor Pr used to balance contributions of DE and EDA can be adjusted in on-line manner using a simulated annealing method. At the initial evolutionary phase, a larger Pr should be adopted to ensure the dominant function of EDA and to enhance the global searching capability. EDA directs DE to search along the Pareto front. The scaling factor should be reduced during the evolutionary process to make DE take up the dominant function gradually and to ensure solutions converge to true Pareto front. The hybrid algorithm was validated using four benchmark cases. The experimental results show that DE-EDA, compared with NSGA-II and DE algorithms, can find many Pareto optimal solutions distributed onto the Pareto front and can improve convergence speed effectively.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2009年第16期4980-4985,共6页
Journal of System Simulation
基金
国家自然科学基金(60804022)
教育部新世纪优秀人才支持计划(NCET-08-0836)
江苏省自然科学基金(BK2008126)
高等学校博士学科点专项科研基金(20070290537
200802901506)
关键词
多目标优化
差分进化
分布估计算法
PARETO最优解
multi-objective optimization
differential evolution
estimation of distribution algorithm
Pareto optimal solution