基于VMD-TCN模型的渭河流域月径流量预测研究  

Monthly Runoff Prediction in Weihe River Basin Based on VMD⁃TCN Model

作　　者：张上要 罗军刚[1] 石国栋 景鑫 连亚妮 左岗岗 ZHANG Shangyao;LUO Jungang;SHI Guodong;JING Xin;LIAN Yani;ZUO Ganggang(State Key Laboratory of Eco-Hydraulics in Northwest Arid Region of China,Xi’an University of Technology,Xi’an 710048,China;Shaanxi Weihe Ecological Zone Protection Center,Xi’an 710004,China)

机构地区：[1]西安理工大学西北旱区生态水利国家重点实验室,陕西西安710048 [2]陕西省渭河生态区保护中心,陕西西安710004

出　　处：《人民黄河》2023年第10期25-29,共5页Yellow River

基　　金：国家自然科学基金资助项目(51679186);陕西省自然科学基础研究计划项目(2019JLZ-15)。

摘　　要：为了有效减小径流序列非线性、非平稳性对径流预测精度的影响,提出一种将变分模态分解(VMD)与时间卷积网络(TCN)相耦合的VMD-TCN模型,并用于渭河流域咸阳和华县水文站的月径流量预测。采用多种模型性能评价指标对VMD-TCN模型与其他模型(TCN、EEMD-TCN、ARIMA)的预测性能进行比较,并分析了VMD-TCN模型在不同预见期下的预测表现。结果显示VMD-TCN模型较其他模型具有预测误差更小、预测精度更高以及峰谷值拟合更优的特点,且随着预见期的增大,模型的预测性能会逐渐降低但预测效果保持良好。In order to reduce the influence of nonlinearity and nonstationarity of runoff series to the runoff prediction accuracy,a VMD-TCN model coupled with variational mode decomposition(VMD)and time convolution network(TCN)was proposed and applied to the monthly runoff prediction of Xianyang and Huaxian hydrological stations in the Weihe River Basin.The prediction results of VMD-TCN were compared with those of other models(TCN,EEMD-TCN,ARIMA)by using a variety of model performance evaluation indexes,and the prediction per-formance of VMD-TCN under different leading times was analyzed.The results show that VMD-TCN has smaller prediction error,higher pre-diction accuracy and better peak-valley fitting than other models.With the increase of leading time,the prediction performance of the model will gradually decrease but the prediction effect remains good.

关 键 词：径流预测 变分模态分解 时间卷积网络 渭河流域 

分 类 号：P338[天文地球—水文科学]