CAJ | 학술논문

为解决多目标粒子群优化算法存在解的多样性差、分布不均等问题，提出一种混合择优机制：在迭代过程中每个粒子依概率，根据解集信息熵或Sigma值确定其全局极值；并直接对解集进行基于信息熵的克隆选择，根据支配关系更新解集，充分发掘分布性更好的解。测试函数的仿真实验结果表明，该算法在保持较好的收敛性能的同时，其求解的分布性指标要明显优于其他算法，这说明混合择优机制能够有效地提升多目标粒子群优化算法求解的多样性和分布性。
위해결다목표입자군우화산법존재해적다양성차、분포불균등문제，제출일충혼합택우궤제：재질대과정중매개입자의개솔，근거해집신식적혹Sigma치학정기전국겁치；병직접대해집진행기우신식적적극륭선택，근거지배관계경신해집，충분발굴분포성경호적해。측시함수적방진실험결과표명，해산법재보지교호적수렴성능적동시，기구해적분포성지표요명현우우기타산법，저설명혼합택우궤제능구유효지제승다목표입자군우화산법구해적다양성화분포성。
In order to solve the problems of loss in diversity and poor distribution of Pareto solutions in Multi-Objective Particle Swarm Optimization(MOPSO), a hybrid global best selecting strategy is proposed. Each particle’s global best is selected according to information entropy or Sigma value of solutions with a varying selecting probability. And clone selection strategy is used to update Pareto solution set according to dominance relationships. As a result, the better distributed solutions are exploited. Results on several benchmark functions show that the proposed algorithm has better distribution performance while maintains a good convergence. This indicates that the proposed hybrid strategy is effective in improving the diversity and distribution of MOPSO.