摘要
针对原始差分进化算法在求解约束全局优化问题时存在陷入局部最优的缺陷,提出一种改进的差分进化算法.该算法在保留原始差分进化算法全局搜索能力的基础上,采用基于规则的方法进行约束处理和种群个体的比较及选择,并利用种群相似度和最优变异操作改善种群进行全局范围搜索的多样性,提高算法跳出局部最优的能力.数值实验表明,该算法稳定性较好,目标函数评价次数较少,收敛速度较快,全局寻优能力较强,不仅能有效求解连续变量约束优化问题,也适用于离散变量或混合变量优化问题.
A modified differential evolution algorithm is provided for constrained global optimization problems instead of the original one that is probably trapped in local optima. Keeping up the global searching ability of the original algorithm, the modified one introduces a rule-based way to handle constraints and select comparatively the individuals from population. The diversity of population in global search is improved via population similarity and best mutation operation, thus enabling the algorithm to jump over any local minimum trap. Numerical experiments reveal that the modified algorithm is reliable, efficient, fast and robust in global optimization. It is able to solve not only the constrained optimization problems with continuous variables but also the optimization problems with discrete or mixed continuous-discrete variables effectively.
出处
《东北大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2009年第7期936-939,共4页
Journal of Northeastern University(Natural Science)
基金
解放军总装备部武器装备预研基金资助项目(9140A18010106LN0101)
辽宁省博士启动基金资助项目(20071022)
关键词
全局优化
差分进化
约束处理
最优变异
数值模拟
global optimization
differential evolution
constraint-handling
best mutation
numerical simulation
