一种基于模拟退火操作的混合差分进化算法  

A hybrid differential evolutionary algorithm based on the simulated annealing operation

作　　者：杨艳霞[1] 

机构地区：[1]武汉科技大学城市学院信息工程学部,湖北武汉430083

出　　处：《智能系统学报》2014年第1期109-114,共6页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金资助项目(70971020)

摘　　要：为了提高进化算法对大规模欺骗问题和等级问题这类复杂组合优化问题的求解能力,提出了一种将模拟退火操作引入到差分进化算法的改进方法。该方法对随机产生的初始个体进行模拟退火操作,对新个体进行退温操作,经过若干次迭代后,选择种群中最优解作为所求问题的解。利用模拟退火算子的突变搜索提高种群多样性,使差分进化算法能更好地利用群体差异进行全局搜索。在实验中,用各种类型的欺骗函数和具有树状结构的等级函数对算法进行仿真测试,仿真结果表明该算法在初期保持了种群多样性,在运行的后期能比较好地跳出局部最优解,收敛到全局最优解附近。In order to improve the ability of the evolutionary algorithm for solving such complicated combination and optimization problems as the massive deceptive problems and hierarchical problems , this paper proposes an im-proved algorithm , which introduces the simulated annealing operation into the differential evolutionary algorithm .U-sing this method , the simulated annealing operation is carried out for a randomly generated initial individual and the temperature-reducing operation is carried out for a new individual .After several times of iterations , the optimal so-lution in the population is taken as the solution to the question .By utilizing the mutation search of the simulated an-nealing operator to improve the diversity of the population , the differential evolutionary algorithm can better utilize colony differences for an overall search .In the experiment , various types of deceptive functions and the hierarchical functions with a tree-shape structure are applied to simulation testing of the algorithm .In the initial stage , the algo-rithm keeps diversity of the population;in the later stage , a local optimal solution may be generated , the conver-gence scope nears to the overall optimal solution .The simulation results show that the algorithm has advantage for the aspect of searching the overall optimal solution .

关 键 词：差分进化 进化算法 模拟退火 欺骗问题 等级问题 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TP18[自动化与计算机技术—计算机科学与技术]