计算机应用 ›› 2010, Vol. 30 ›› Issue (05): 1290-1292.

• 数据挖掘与人工智能 • 上一篇    下一篇

新的混合智能优化算法及其多目标优化应用

张汉强1,卢建刚2,陈金水3   

  1. 1. 浙江大学工业控制研究所
    2. 浙江大学 信息科学与工程学院
    3. 浙江大学
  • 收稿日期:2009-11-17 修回日期:2010-01-13 发布日期:2010-05-04 出版日期:2010-05-01
  • 通讯作者: 张汉强
  • 基金资助:
    国家自然科学基金资助项目;国家863计划项目;国家863计划项目;浙江省科技计划资助项目;浙江省科技计划资助项目;浙江省自然科学基金资助项目

New hybrid intelligent optimization algorithm and its application in multi-objective optimization

  • Received:2009-11-17 Revised:2010-01-13 Online:2010-05-04 Published:2010-05-01
  • Supported by:
    the National Natural Science Foundation of China under Grant

摘要: 针对人工鱼群算法后期收敛速度较慢、解精度不高的不足,按照分阶段寻优和变参数寻优的改进策略,并结合禁忌搜索算法中的相关规则,提出一种新的混合智能优化算法。该算法将寻优过程分为锁定最优解或者局部解邻域和求得高精度最优解两个阶段,每个阶段设置不同的参数并结合禁忌搜索算法以提高收敛速度和最优解精度。典型函数验证表明,该算法收敛速度快、精度高;同时,对于多目标优化问题,该算法可以提高Pareto最优解集质量,扩大决策分布范围,维持决策多样性,有利于决策者作出决策。

关键词: 人工鱼群算法, 分阶段寻优和变参数寻优, 禁忌搜索, 多目标优化

Abstract: After analyzing the disadvantages of the slower convergence property and lower accuracy in Artificial Fish-Swarm Algorithm (AFSA), a new hybrid intelligent optimization algorithm was proposed based on phased optimization and variable parameter optimization as well as some relevant rules in tabu search algorithm. In this algorithm, the optimization process was divided into two phases, one was to lock the neighborhood of the optimal solution or partial solution, the other was to obtain the optimal solution of high-precision. Each phase set different parameters and combined the tabu search algorithm to improve convergence speed and accuracy of optimal solutions. The simulation results show that the proposed algorithm can greatly improve the ability of seeking the global excellent result, convergence speed and accuracy. As for the multi-objective optimization problem, the proposed algorithm can also improve the quality of Pareto optimal solutions, enlarge the distribution area of decisions and maintain diversity in decision-making.

Key words: Artificial Fish-Swarm Algorithm (AFSA), phased optimization and variable parameter optimization, Tabu Search (TS), multi-objective optimization