摘要
为提升多目标粒子群算法粒子初始化解的分布性与全局最优解的精度,提出一种新的基于网格密度的混合多目标粒子群算法(GDHMOPSO)。该算法通过利用Henon映射来初始化粒子在决策空间中的位置来增加粒子的初始搜索范围,提高粒子的分布多样性。针对选取全局最优解,首先利用求和策略,统计出外部存档中各个粒子所有函数值的和值,选择每个网格中和值最小的那个粒子作为该网格中的最优解,并引入反向选择策略选取一个全局最优解,降低粒子群陷入局部最优解的概率。通过实验证明,GDHMOPSO算法在测试函数上有较好的收敛性与分布性。
For the purpose of improving the distribution of particles and the precision of the global optimal solution, a newhybrid multiobjectiveparticle swarmoptimization based on grid density is proposed (GDHMOPSO). The algorithmmakes use of the Henon map toinitialize the particle in the decision space in order to increase the initial search area of the particle and improve the distribution of particles.Next, in order to select the global optimal solution, its first step is to use the summation strategy to calculate the value of all thefunction values of the various particles in the the external archive, and then select the smallest particle of one grid as the optimal solutionin the grid. Additionally, the inverse selection strategy is used to select a global optimal solution respectively to decrease the probabilitythat the algorithmtraps into a locally optimal solution. When the network density is smaller, the probability that the optimal solution inthe grid is chosen as the global optimal solution of the particle is greater. Besides, each particle uses this strategy to select a global optimalsolution respectively to decrease the probability that the algorithmtraps into a locally optimal solution. The simulation results showthatthe GDHMOPSO algorithmhas good effect in convergence and distribution.
作者
吴耀威
刘衍民
WU Yao-wei;LIU Yan-min(College of Mathematics and Computational Science,Wuyi University,Jiangmen 529000,China;College of Mathematics,Zunyi Normal University,Zunyi 563006,China)
出处
《遵义师范学院学报》
2021年第5期63-67,共5页
Journal of Zunyi Normal University
基金
国家自然科学基金项目(71461027)
贵州省科技创新人才团队项目(黔科合平台人才[2016]5619)。
