改进萤火虫算法及其在全局优化问题中的应用  

An improved firefly algorithm and its application in global optimization

作　　者：刘畅[1,2] 刘利强[3] 张丽娜[3] YANG Xinshe 

机构地区：[1]哈尔滨工程大学动力与能源工程学院,黑龙江哈尔滨150001 [2]海军驻哈尔滨汽轮机厂有限责任公司军事代表室,黑龙江哈尔滨150046 [3]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001 [4]Design Engineering and Mathematics,Middlesex University

出　　处：《哈尔滨工程大学学报》2017年第4期569-577,共9页Journal of Harbin Engineering University

基　　金：国家自然科学基金项目(51109041;51009036)

摘　　要：针对标准萤火虫算法容易陷入局部最优的问题,本文提出一种改进的萤火虫算法。在标准萤火虫算法的位置移动公式中,利用指数分布和韦伯分布对吸引力项进行改进,以增强算法的全局探测能力;同时利用步长单调递减模式对随机项进行改进,以增强算法后期的局部挖掘能力。通过13个测试函数对本文提出的改进算法、模拟退火算法、粒子群算法和差分进化算法进行算法性能的比较。实验结果表明,本文提出的改进算法能较好地平衡算法的全局探测能力和局部挖掘能力,使算法跳出局部最优,从而提高算法的收敛速度和精度。Escape from the local minimum of the standard firefly algorithm exhibits low probability,and hence an improved firefly algorithm w as proposed in this article. In this paper, exponential distribution and weibull distribution were used for randomizing the firefly algorithm＇s attractive mechanism to enhance the exploration a-bility of the algorithm. At the same time, randomized m o v e terms can be changed by monotonous decreasing stratagem to improve the exploitation ability of the later iteration. In our experiments, extensive experiments were conducted between the modified firefly algorithm and the simulated annealing algorithm, particle swarm optimization, differential evolution algorithm on 13 benchmark functions. The results of these experiments indi-cate that the modified firefly algorithm can fairly balance the exploration and exploitation ability in the sense of avoiding the local optimum. More over, the convergence rate and the precision of the firefly algorithm can be improved significantly.

关 键 词：萤火虫算法 随机分布 元启发式算法 随机性算法 全局优化 模拟退火算法 粒子群算法 差分进化算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]