一种基于模拟退火的参数自适应差分演化算法及其应用  

A Differential Evolution Algorithm with Self-Adaptive Parameters based on Simulated Annealing and Its Applications

作　　者：李亚楠[1,2,3] 郭海湘[1,2,3] 黎金玲[1,4] 刘晓[1,2,3] 

机构地区：[1]中国地质大学经济管理学院 [2]中国地质大学数字化商务管理研究中心 [3]中国地质大学中国矿产资源战略与政策研究中心,武汉430074 [4]武汉工程科技学院,武汉430200

出　　处：《系统管理学报》2016年第4期652-662,共11页Journal of Systems & Management

基　　金：国家自然科学基金资助项目(71103163,71103164,71573237);教育部新世纪优秀人才支持计划(NCET-13-1012);教育部人文社会科学研究规划基金资助项目(15YJA630019);中央高校基本科研业务费专项资金资助项目(CUG120111,CUG110411,G2012002A,CUG140604,GUG160605)

摘　　要：提出了一种改进的自适应差分演化算法,该算法对差分演化算法中控制参数的生成方式进行了改进,在算法的运行过程中保留表现较好的控制参数,通过反馈机制调整下一代控制参数的生成,使算法能够在运行的过程中自适应的调节控制参数,并且算法在选择操作中融入了模拟退火的思想,以提高算法的全局搜索能力。在17个典型的Benchmarks函数上对算法进行了测试,并将该算法与jDE、JADE、SaDE、EPSDE和CoDE等5种改进的差分演化算法进行了比较,实验结果表明,该算法的性能整体上优于其他5种算法。另外,测试了模拟退火中初始温度参数对算法性能的影响,结果表明,当初始温度参数为1 000时,算法的性能较优。最后,将该算法应用于TSP这一组合优化问题,结果表明,该算法能够有效解决TSP问题。An enhanced self-adaptive differential evolution（ESADE）algorithm is proposed,aiming at improving the way that control parameters generate.The control parameters with good performance in the algorithm are retained to adjust the generation of the next generation control parameters through feedback mechanism,making it possible for the algorithm to self-adaptively adjust control parameters in the running process.In order to enhance its global search capability,simulated annealing（SA）is incorporated into the selecting operation.By using 17 typical benchmark functions,extensive computational simulations and comparisons between this algorithm and other algorithms such as jDE,JADE,SaDE,EPSDE and CoDE were conducted.The computational results show that the ESADE algorithm generally outperforms the other five algorithms. Furthermore,how initialized temperature parameter in SA influences the performance of ESADE is tested and the results show that the optimal performance of the algorithm appears when the temperature is set as 1000.How to apply the algorithm to solve TSP problem is demonstrated and the results show that TSP problem can be solved effectively by this algorithm.

关 键 词：差分演化 自适应 模拟退火 控制参数 

分 类 号：N945.15[自然科学总论—系统科学]