基于自适应交叉与协方差学习的改进平衡优化器算法  

Improved equilibrium optimizer based on adaptive crossover andcovariance learning

作　　者：侯新宇 鲁海燕[1,2] 卢梦蝶 胡清元 Hou Xinyu;Lu Haiyan;Lu Mengdie;Hu Qingyuan(School of Science,Jiangnan University,Wuxi Jiangsu 214122,China;Wuxi Engineering Technology Research Center for Biological Computing,Wuxi Jiangsu 214122,China)

机构地区：[1]江南大学理学院,江苏无锡214122 [2]无锡市生物计算工程技术研究中心,江苏无锡214122

出　　处：《计算机应用研究》2024年第6期1656-1662,共7页Application Research of Computers

基　　金：国家自然科学基金资助项目(12102146)。

摘　　要：针对平衡优化器算法存在的收敛精度低和易陷入局部停滞的问题,提出一种基于自适应交叉与协方差学习的改进平衡优化器算法。首先,构建外部存档来保留历史优势个体,增加种群多样性,以提高算法的全局寻优能力。其次,引入自适应交叉概率来平衡算法的全局探索能力和局部开发能力,以提高算法的寻优精度和鲁棒性。最后,采用协方差学习策略,充分利用浓度向量之间的关系来增强种群间信息交流,以避免算法陷入局部停滞。通过对CEC2019测试函数进行仿真实验,并将改进算法与反向传播(back propagation,BP)神经网络相结合用于预测新疆玛纳斯河的径流情况,实验结果表明,改进算法在收敛精度和鲁棒性方面有显著提升,且大幅提高了BP神经网络的径流预测效果。Aiming at the problems of low convergence accuracy and ease of trapping into local stagnation in the equilibrium optimizer,this paper proposed an improved equilibrium optimizer based on adaptive crossover and covariance learning.Firstly,this algorithm constructed an external archive to retain the historically dominant individuals and increase the population diversity for improving the global optimization ability.Secondly,it introduced an adaptive crossover probability to balance the global exploration ability and local exploitation ability of the algorithm,so as to improve the optimization accuracy and robustness of the algorithm.Finally,it applied a covariance learning strategy to make full use of the relationship between the concentration vectors to enhance the information exchange among the populations and thereby to avoid local stagnation.Through simulation experiments on the CEC2019 test functions and combining the improved algorithm with back propagation(BP)neural network to predict the runoff situation of the Manas River in Xinjiang,the experimental results show that the improved algorithm remarkably improves convergence accuracy and robustness,and significantly enhances the runoff prediction performance of the BP neural network.

关 键 词：平衡优化器算法 智能算法 外部存档 自适应交叉概率 协方差 径流预测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]