基于多策略麻雀搜索算法的径向基神经网络时延预测方法  

Delay Prediction Method of RBF Based on Improved Sparrow Search Algorithm

作　　者：游达章[1,2] 杨润 张业鹏 李存靖[3] YOU Dazhang;YANG Run;ZHANG Yepeng;LI Cunjing(School of Mechanical Engineering,Hubei University of Technology;Hubei Key Lab of Manufacture Quality Engineering;Hubei 3611 Special Equipment Co.,Ltd)

机构地区：[1]湖北工业大学机械工程学院 [2]湖北省现代制造质量工程重点实验室 [3]湖北三六一一应急装备有限公司

出　　处：《仪表技术与传感器》2024年第6期86-92,120,共8页Instrument Technique and Sensor

基　　金：国家自然科学基金项目(51875180)。

摘　　要：网络诱导时延序列存在随机性和不稳定性,单一的预测算法难以准确预测,针对该问题,提出一种基于多策略麻雀搜索算法(ISSA)的径向基神经网络(RBF)时延预测方法来准确预测网络时延。首先,为了应对麻雀种群初始分布不均匀的问题,增加改进Tent混沌映射来提高种群早期的分布质量,并引入新的惯性权重因子改进发现者位置更新策略,扩大麻雀早期寻优范围;随后引入正余弦优化算法(SCA)更新跟随者位置,避免种群后期陷入局部最优。其次,考虑到RBF隐含层节点中心,幅值以及输出层权值不确定的问题,提出利用ISSA算法寻优求取。最后,搭建改进的时延预测模型并输入实测时延数据得到预测时延值。实验研究结果表明:相较于传统SSA-RBF模型,文中提出的ISSA-RBF时延预测模型的MSE、MAE和MAPE分别提高了69.46%、32.83%、34.43%,可有效预测网络时延,为之后的时延补偿提供基础。The randomness and instability of the network induced delay sequence make it difficult for a single prediction al-gorithm to predict the network delay accurately.To solve this problem,a radial basis neural network(RBF)delay prediction method based on the improved sparrow search algorithm(ISSA)was proposed to predict the network delay accurately.Firstly,to deal with the problem of uneven initial distribution of sparrow population,an improved Tent chaotic map was added to improve the distribution quality of the early population,and a new inertial weight factor was introduced to improve the discoverer location up-date strategy and expand the early optimization range of sparrow.Then the sine-cosine optimization algorithm(SCA)was intro-duced to update the follower position to avoid the population falling into local optimal.Secondly,considering the uncertainty of RBF hidden layer node center,amplitude and output layer weight,ISSA algorithm was proposed to optimize and obtain.Finally,an improved delay prediction model was built,and the predicted delay value was obtained by inputting the measured delay data.The experimental results show that compared with the traditional SSA-RBF model,the proposed ISA-RBF delay prediction model im-proves the MSE,RMSE and MAE by 69.46%,32.83%and 34.43%respectively,which can effectively predict the network delay and provide a basis for the later delay compensation.

关 键 词：网络化控制系统 时延预测 改进Tent混沌映射 正余弦优化算法 惯性权重因子 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]