基于投影面的多目标优化问题进化算法MOEA/P  

An Evolutionary Algorithm for Multi-Objective Optimization Problem Based on Projection Plane MOEA/P

作　　者：杨爽 陈未如 YANG Shuang;CHEN Wei-ru(Shenyang University of Chemical Technology,Shenyang 110142,China)

机构地区：[1]沈阳化工大学计算机科学与技术学院,辽宁沈阳110142

出　　处：《沈阳化工大学学报》2022年第5期468-475,共8页Journal of Shenyang University of Chemical Technology

摘　　要：大多数多目标优化问题研究更关注面向整个目标解集的求解收敛性、多样性、鲁棒性等方面的性能.大量、分散的解影响算法的求解效率和求解质量,过多的解使得用户无从抉择,这在超多目标优化问题中显得尤为突出.本文提出一种基于投影面的多目标优化问题的求解算法,根据决策需求将目标空间分成投影面和自由维,再把投影面分割成多个投影格,由各个投影格决定求解方向,在各个投影格上求解自由维的最优值,从而得到多目标优化问题的最优解.投影面的划分将高维多目标优化问题简化成求解低维多目标优化问题,即仅在投影面上求解自由维目标的优化.投影格的分割将求解确定在由决策者指定的目标值具体范围内,提高了求解精度和效率.通过分别对多目标和超多目标问题进行求解实验,结果表明本算法能够有效求解多目标优化问题,并在超多目标优化问题的求解上能为特定方向的目标决策提供有效支持.Most researches on multi-objective optimization problems pay more attention to the performance of convergence,diversity,robustness and so on.A large number of scattered solutions affect the efficiency and quality of the algorithm.Too many solutions make users have no choice,which is particularly prominent in the super multi-objective optimization problem.In this paper,a multi-objective optimization algorithm based on projection surface is proposed.According to the decision requirements,the target space is divided into projection surface and free dimension,and then the projection surface is divided into multiple projection lattices.The solution direction is determined by each projection lattice,and the optimal value of free dimension is solved on each projection lattice,so as to obtain the optimal solution of the multi-objective optimization problem.The division of projection plane simplifies the high-dimensional multi-objective optimization problem to solve the low-dimensional multi-objective optimization problem,that is,the optimization of free-dimensional objects is solved only on the projection plane.The division of projection lattice determines the solution within the specific range of the target value specified by the decision-maker,which improves the accuracy and efficiency of the solution.Through the experiments of solving multi-objective and super multi-objective problems,the results show that the algorithm can effectively solve multi-objective optimization problems,and can provide effective support for the goal decision-making in a specific direction.

关 键 词：多目标优化 投影面 进化算法 目标决策支持 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]