遗传算法在多目标优化应用中的对比研究

Genetic algorithm applied to Multi-objective optimization:A comparative study

多目标优化应用研究在过程工程领域越来越受重视。本文首先给出了多目标优化问题的一般形式,指出多目标问题求解任务:引导搜索向整个的Pareto优化范围;Pareto优化前沿上保持解集的多样性。在简要论述遗传算法求解多目标技术的基础上,对应用了遗传算法求解多目标的两种方法进行了对比研究,并给出了线性加权遗传算法和一种多目标遗传算法的计算框图。指出线性加权法求解Pareto最优解时不能不能很好地处理非凸区域、均匀分布的权重值不能生成均匀分布的Pareto 前沿等局限性,以及多目标遗传算法生成种群多样性及Pareto最优解均匀分布的优点,并用实例进行了验证说明。

The problem statement of multi-objective optimization problem （MOP） is firstly presented. The goals of problem solving are to guide the search process towards the global Pareto-optimal region and maintain population diversity in the Pareto-optimal front. After reviewing MOP techniques using genetic algorithm （GA） , two methods are presented and compared. The frameworks of these methods are introduced. And lastly, two practical examples are carried out. The resuhs show that （1） the drawbacks of weighted sum approach using GA for Pareto set generation is observed, and （2） Muhi-objective genetic algorithm have the advantages of the diversity of population and the uniform spread of Pareto-optlmal solutions.