摘要
复杂过程全局进化算法是一种具有类似分散搜索的通用框架结构,能够高效完成全局搜索的新型进化算法。在该算法的基础上,提出了差分型复杂过程全局进化算法。差分型算法采用拉丁超立方体抽样生成多样性种群,并应用"最小欧几里德距离的最大值法"产生参考集Refset2,以保证参考集的多样性。采用差分变异和交叉策略替代原算法的线性合并,兼顾算法的收敛速度和种群的多样性。应用Nelder-Mead直接搜索法进行局部搜索,防止搜索过程在局部最优点附近反复。仿真结果表明差分型复杂过程全局进化算法,具有较高的搜索效率。
Evolutionary algorithm for complex-process optimization is a new global search evolutionary algorithm which has a similar flexible framework structure of scatter search. On this basis, evolutionary algorithm for complex-process optimization based on differen- tial evolutionary strategy is proposed. The set RefSet2 is built by selecting those individuals from diverse vectors which is generated by Latin hypercube uniform sampling with minimum Euclidean distance to set ReJSetl is the highest. To take account of convergence speed and population diversity, differential mutation and crossover strategy is used to replace linear combination method of the original algorithm. Nelder-Mead simplex algorithm is adopted to improve the trial solution generated at ＂go-beyond strategy＂ stages. The simu- lation results show that evolutionary algorithm for complex-process er search efficiency. optimization based on differential evolutionary strategy has high-
出处
《计算机工程与应用》
CSCD
2012年第8期24-27,共4页
Computer Engineering and Applications
基金
国家自然科学基金(No.60374032)
北京市教委重点学科控制理论与控制工程(No.XK100080537)
关键词
差分
复杂过程
优化:进化方法
differential
complex-process
optimization
evolutionary algorithm
