动态多目标优化的预测遗传算法  

Predictive multiobjective genetic algorithm for dynamic multiobjective optimization problems

作　　者：武燕[1] 刘小雄[2] 池程芝[2] 

机构地区：[1]西安电子科技大学理学院,西安710071 [2]西北工业大学自动化学院,西安710072

出　　处：《控制与决策》2013年第5期677-682,共6页Control and Decision

基　　金：国家自然科学基金项目(61105065);航空科学基金项目(20100753009);西安电子科技大学基本科研业务费项目

摘　　要：为了在动态环境中很好地跟踪最优解,考虑动态优化问题的特点,提出一种新的多目标预测遗传算法.首先对Pareto前沿面进行聚类以求得解集的质心;其次应用该质心与参考点描述Pareto前沿面;再次通过预测方法给出预测点集,使得算法在环境变化后能够有指导地增加种群多样性,以便快速跟踪最优解;最后应用标准动态测试问题进行算法测试,仿真分析结果表明所提出算法能适应动态环境,快速跟踪Pareto前沿面.Dynamic multiobjective optimization problems require an algorithm to continuously track a changing Pareto optimal solutions over time. Therefore, a new predictive multiobjective genetic algorithm（PMGA） is proposed, in which the centroid of Pareto optimal is soluted by clustering. And Pareto optimal solutions are described by applying the centroid points and reference solutions. Then the prediction set is generated by using the inertia predict and Gauss mutation. After an environment changed, the prediction set is incorporated in the current population to increase the population diversity by guided fashion. Finally, experimental studies on dynamic multiobjective optimization problems are carded out. The simulation results show that PMGA can quickly adapt the dynamic environments and track Pareto optimal solutions.

关 键 词：动态多目标优化 遗传算法 预测 种群多样性 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]