基于ESMD-NNBR耦合的月径流预测模型  

Monthly Runoff Forecasting Based on ESMD-NNBR Hybrid Model

作　　者：刘尚东 孙东永[1] 许晶晶 袁业鹏 赵洋 LIU Shangdong;SUN Dongyong;XU Jingjing;YUAN Yepeng;ZHAO Yang(School of Water and Environment,Chang'an University,Xi'anan 710064,China)

机构地区：[1]长安大学水利与环境学院,陕西西安710064

出　　处：《水文》2023年第1期52-56,共5页Journal of China Hydrology

基　　金：国家自然科学基金资助项目(51979005);陕西省自然科学基础研究计划项目(2020JM-250)。

摘　　要：为提高对非线性非平稳径流序列的预测精度,以“分解-预测-重构”为思路,将极点模态分解法(ESMD)与最近邻抽样回归模型(NNBR)结合构建耦合预测模型,并将其应用于渭河流域华县站、张家山站的月径流预测。对比发现,ESMD-NNBR模型的性能最好,预测精度最高。具有自适应平稳化序列能力的ESMD法分解径流序列,体现了不同时间频率下的径流变化规律,将其与NNBR模型结合可以显著提高模型的预测性能。In order to improve the prediction accuracy of nonlinear non-stationary runoff series,based on the idea of decomposition,prediction and reconstruction,this paper constructed the hybrid forecasting model by combining ESMD and NNBR,and predicted the monthly runoff of Huaxian station and Zhangjiashan station.Compared with single NNBR、EMD-NNBR and EEMD-NNBR models,ESMD-NNBR model has the best performance and the highest forecasting accuracy.The decomposition of runoff series by ESMD method with self-adaptive stationary sequence can not only reflect the variation of runoff at different frequencies,but also improve the performance of forecasting model by combining it with NNBR model.

关 键 词：ESMD NNBR 耦合模型 径流预测 渭河流域 

分 类 号：P338.2[天文地球—水文科学]