基于混合预测策略的动态多目标优化算法  

Dynamic Multi-objective Optimization Algorithm Based on Hybrid Prediction Strategy

作　　者：冯劲宇 陈得宝[2] FENG Jinyu;CHEN Debao(School of Computer Science and Technology,Huaibei Normal University,235000,Huaibei,Anhui,China;School of Physics and Electronic Information,Huaibei Normal University,235000,Huaibei,Anhui,China)

机构地区：[1]淮北师范大学计算机科学与技术学院,安徽淮北235000 [2]淮北师范大学物理与电子信息学院,安徽淮北235000

出　　处：《淮北师范大学学报（自然科学版）》2024年第4期53-61,共9页Journal of Huaibei Normal University：Natural Sciences

基　　金：国家自然科学基金项目(61976101);高校优秀拔尖人才培育项目(gxbj ZD2022021);安徽省学术与技术带头人及后备人选科研活动经费资助项目(2021H264);智能计算理论及应用优秀科研创新团队(2023AH010044)。

摘　　要：在处理具有可预测性规律的动态多目标问题时,预测策略发挥着重要作用。但对于一些复杂的变化环境,仅使用单一的预测策略来响应环境变化,算法的性能往往不高。因此,提出一种基于混合预测策略的动态多目标优化算法。采用四分位点方法对目标空间进行划分,从而避免空域的形成。根据不同时刻子区域中位点的信息,分别采用线性预测策略和振荡序列灰色预测策略生成新个体,同时设计一种基于变量相关系数的选择策略,确定新环境下初始种群的部分个体。设计一种自适应群体多样性维持策略,生成部分新个体,确保良好的种群多样性。为证明所提出算法的有效性,使用3种经典比较算法在9个不同动态特征的测试函数上进行仿真实验。结果表明,该算法在大多数动态优化问题上具有更好的性能。Prediction strategies play a crucial role in dealing with dynamic multi-objective problems with pre-dictability laws.However,for some complex changing environments,the performance of the algorithms is of-ten not high when only a single prediction strategy is used to respond to the environmental changes.There-fore,a dynamic multi-objective optimization algorithm based on a hybrid prediction strategy is proposed.The objective space is segmented using the quantile method to avoid the generation of null regions.Then,accord-ing to the information of the median point in the sub-region at different moments,a linear prediction strategy and an oscillating sequence grey prediction strategy are used to generate new individuals,respectively,and a selection strategy based on variable correlation coefficient is designed to determine some individuals of the initial population in the new environment.A proportional strategy based on non-dominant solution is de-signed to generate some of the new individuals to ensure good population diversity.To demonstrate the effec-tiveness of the proposed algorithm,three classical comparison algorithms are used to perform simulation ex-periments on 9 test functions with different dynamic characteristics.The results show that the algorithm has better performance on most dynamic optimization problems.

关 键 词：动态多目标优化 预测 进化算法 灰色模型 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]