寻求“理想”解的改进多目标粒子群优化算法  

Improved multi-objective particle swarm optimization algorithm that can give "ideal" solution

作　　者：周黎[1] 周承恩[2] 李海滨[2] 

机构地区：[1]内蒙古工业大学管理学院,呼和浩特010051 [2]内蒙古工业大学理学院,呼和浩特010051

出　　处：《控制与决策》2015年第9期1653-1659,共7页Control and Decision

基　　金：国家自然科学基金项目(11262014);内蒙古自然科学基金项目(2014MS0709)

摘　　要：如何在众多非劣解中为决策者推荐一个合理的方案是使用多目标粒子群算法(MOPSO)所面临的问题.为此,将逼近理想解的排序方法(TOPSIS策略)引入到算法中.为了提高求解精度和均匀性,还提出了基于Pbest的变异策略和改进的k邻近距离策略.测试结论显示,仅使用TOPSIS策略确定Gbest的算法,求解精度虽好,但均匀性较差,而包含所有改进策略的算法在精度和均匀性方面都更优,并且能够按照TOPSIS方法在非劣解集中找到一个适合向决策者推荐的"理想"方案.One problem of using multi-objective particle swarm optimization algorithm（MOPSO） is how to find a recommendable solution within the non-inferior solution set for decision-makers.Therefore,the technique for order preference by similarity to ideal solution（TOPSIS） strategy is introduced into the MOPSO algorithm.In order to improve the accuracy and uniformity of solutions,two other strategies are also proposed：the mutation operator according to Pbest,and the modified method to calculate k neighbor distance.The results show that the solution uniformity of the algorithm which chooses Gbest using the TOPSIS method only is not satisfied,while the algorithm that uses all improvement strategies has better performance.An ideal solution can be found for decision-makers according to the TOPSIS strategy.

关 键 词：多目标优化 粒子群优化 TOPSIS策略 变异策略 k邻近距离 

分 类 号：TP202.7[自动化与计算机技术—检测技术与自动化装置]