摘要
为了提高基于Internet的网络控制系统中随机时延的预测精度,提出了基于经验模式分解(empirical mode decomposition,EMD)与最小二乘支持向量机(Least Squared Support Vector Machines,LS-SVM)的一步时延预测方法.首先利用EMD将时延序列分解成若干个本征模式函数分量,分解后的分量去除了原始时延序列的长相关性,同时突出时延序列不同的局部特征.然后根据各个分量的变化规律,选择不同的LS-SVM模型分别进行预测.最后将各分量的预测值叠加得到最终的预测值.仿真结果表明本文方法具有较高的预测精度.
In order to predict the random time-delay of Intemet-based networked control system effectively, a hybrid one step time-delay forecasting meth~ based on Empirical Mode Decomposition（EMD） and Least Squared Support Vector Machines（ LS- SVM）is presented in the paper. Firstly, EMD algorithm can decompose time-delay sequences into some intrinsic mode functions （IMF） ,the IMF after decomposed remove the long-range dependence of original time-delay sequences and prominent the different local feature of the time-delay sequences. Secondly, according to the change law of each IMF, chose different LS-SVM model to predict time-delay.At last, all the forecasted values corresponded to these partitions are superposed to get the forecasted time-delay. Simulation results show that the proposed method has higher prediction accuracy.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2014年第5期868-874,共7页
Acta Electronica Sinica
基金
国家自然科学基金重点项目(No.61034005)
关键词
网络控制系统
经验模式分解
最小二乘支持向量机
时延
预测
networked control system
empirical mode decomposition
least squared support vector machines
time-delay
prediction
