使用指数平滑预测的动态多目标优化算法  

Dynamic Multi-objective Optimization Algorithm Based on Exponential Smooth Prediction

作　　者：陈满丽 马永杰[2] CHEN Manli;MA Yongjie(School of Media Engineering,Lanzhou University of Arts and Sciences,Lanzhou 730070,China;College of Physics and Electronic Engineering,Northwest Normal University,Lanzhou 730070,China)

机构地区：[1]兰州文理学院电子与通信工程学院,兰州730070 [2]西北师范大学物理与电子工程学院,兰州730070

出　　处：《自动化与仪器仪表》2024年第12期20-27,32,共9页Automation & Instrumentation

基　　金：兰州文理学院校级科研项目(2020QNRC10);国家自然科学基金项目(62066041)。

摘　　要：为进一步加快种群的收敛速度,获得分布均匀的pareto解集,提出了一种基于指数平滑预测的动态多目标优化算法。首先,对新环境下的种群中心点进行指数平滑预测,计算种群进化方向。其次,结合种群进化方向预测种群,并对预测后的种群采用切比雪夫聚合方法筛选出80%的个体。其余20%的个体通过对上一时刻最优解集多项式变异产生。最后,对新种群进行边界检测。另外,在种群进化操作中,提出了一种新的交叉算子。通过与现有动态多目标优化算法在多个测试函数上比对分析,所提算法在大多数测试函数上的逆世代距离指标和收敛指标优于对比算法,展现了所提算法处理动态多目标优化问题的优越性。A dynamic multi-objective optimization algorithm based on exponential smoothing prediction is proposed to further accelerate the convergence rate of population and obtain the uniform pareto solution set.Firstly,the exponential smoothing prediction of the population center point under the new environment is carried out to calculate the evolutionary direction of the population.Secondly,the population was predicted based on the evolutionary direction of the population,and 80%of the individuals were screened out by Chebyshev polymerization method.The remaining 20%of individuals are generated by variation of the polynomial of the optimal solution set at the last moment,so as to avoid large influence of prediction error on the algorithm convergence speed due to irregular changes.Finally,the boundary detection of the new population is carried out to avoid the new individuals consuming the iteration time of the algorithm for the infeasible solution.In addition,a new crossover operator is proposed for population evolution operation.By comparing with existing dynamic multi-objective optimization algorithms on multiple test functions,the inverse generation distance index and convergence index of the proposed algorithm are superior to the comparison algorithm on most test functions,which shows the superiority of the proposed algorithm in dealing with dynamic multi-objective optimization problems.

关 键 词：动态多目标优化 指数平滑预测 切比雪夫聚合方法 多项式变异 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP29[自动化与计算机技术—控制科学与工程]