摘要
含路径和终端约束的动态多目标优化是过程系统工程的一个重要研究方向,难度较高。传统蚁群算法仅适于离散问题,今采用混合正态分布描述信息素分布,并设计相应的解构造操作,使之拓宽至连续优化问题。通过对目标函数和约束矩阵的非劣排序,确定解的等级,用以克服传统约束处理方法的局限性。借鉴了免疫系统的浓度概念,将其与解的等级结合,共同确定解的适应度,有助于保持种群的多样性。在更新信息素时将利用外部优解库和种群信息,可加快收敛速度。基于拥挤度距离更新外部优解库可更均匀地逼近Pareto最优解集。由此构建了一种基于免疫机制的多目标蚁群算法(Immune Mechanism based Multi-Objective Ant Colony Algorithm,IM-MOACA),并用于间歇反应器的动态多目标优化问题,效果良好,显示出较强的全局优化性能,能以较快的速度逼近真实的Pareto最优前沿,可为用户进行合理的决策分析提供有效的支持。
Dynamic multi-objective optimization with both path and terminal constraints is an important research direction of process systems engineering. Traditional ant colony algorithm is just suitable to discrete problems, in this paper, a mixture of normal distribution was adopted to describe the pheromone distribution, and the corresponding solution construction operation was designed to make it suitable for continuous optimization problems. The non-dominated sorting was used to both objective function and constraints matrix, and the rank of solutions was determined, which can overcome the limitation of traditional constraint handling methods. The concentration concept of immune system was adopted, and then it was combined with the rank to determine solution fitness, which is propitious to maintain the population diversity. Pheromone update is based on both the external archive and the current population, which helps to improve the convergence speed. A crowding distance based external archive update strategy was adopts to uniformly approximate the Pareto optimal solution set step by step. Finally, immune mechanism based multi-objective ant colony algorithm (IM-MOACA) was proposed, and it was applied to dynamic multi-objective optimization of batch reactor. The satisfactory result illustrates that IM-MOACA has well global optimization performance, and it can approximate the true Pareto optimal front in a reasonable time. The obtained Pareto optimal solutions can contribute to rational and effective decision-making for the users.
出处
《高校化学工程学报》
EI
CAS
CSCD
北大核心
2009年第2期326-332,共7页
Journal of Chemical Engineering of Chinese Universities
基金
国家自然科学基金(20276063)
关键词
多目标蚁群算法
免疫机制
PARETO最优集
间歇反应器:动态优化
multi-objective ant colony algorithm
immune mechanism
Pareto optimal solution set
batch reactor
dynamic optimization