摘要
基于极值中心三次样条插值法对集合经验模态分解(EEMD)技术进行改进,在此基础上将改进的EEMD与最近邻抽样回归模型(NNBR)结合,提出了改进的EEMD-NNBR耦合模型。改进的EEMD能对上、下极值点进行较好的拟合,且将序列均值延拓到序列两端以减缓端点效应。首先由改进的EEMD分解年径流序列,得到各本征模函数IMF和趋势项,然后分别对各本征模函数和趋势项建立最近邻抽样回归模型得到各分解序列的预测值,最后将预测值累加即为年径流预测值。将改进的EEMD-NNBR耦合模型用于屏山站年径流预测,并与EEMD-NNBR耦合模型对比,其预测值的平均相对误差由10.08%提高到8.59%,表明建议模型能提高径流预测精度。
The ensemble empirical mode decomposition(EEMD)technique is improved by the extremum center cubic spline interpolation.On this basis,the improved EEMD is combined with the nearest neighbor bootstrapping regressive model(NNBR)to obtain an improved EEMD-NNBR coupling model.The improved EEMD can better fit the upper and lower extreme points,and extend the sequence mean to the two ends of the sequence to reduce the end effect.This paper firstly decomposes the annual runoff series by the improved EEMD to obtain the intrinsic mode function(IMF)and trend term,and then establishes the NNBR for each IMF and trend term to obtain the predicted values of decomposed series,and accumulates the predicted values to get the annual runoff,finally applies the improved EEMD-NNBR coupling model for the annual runoff prediction of Pingshan Station.The results show that compared with the original coupling model,the mean relative error of the improved EEMD-NNBR coupling model have reduced from 10.08%to 8.59%,so the proposed model can improve the accuracy of runoff prediction.
作者
郑芳芳
王文圣
张岚婷
ZHENG Fangfang;WANG Wensheng;ZHANG Lanting(College of Water Resources&Hydropower,Sichuan University,Chengdu 610065,China)
出处
《人民珠江》
2021年第2期1-6,共6页
Pearl River
基金
国家自然科学基金(51679155)。