基于STL-LSTM的唐山市地表径流预测研究  

Research on Surface Runoff Prediction in Tangshan City Based on STL-LSTM

作　　者：时玉涛 Shi Yutao(Tangshan Hydrological Survey and Research Center of Hebei Province,Tangshan 063000,Hebei)

机构地区：[1]河北省唐山水文勘测研究中心,河北唐山063000

出　　处：《陕西水利》2025年第4期26-28,31,共4页Shaanxi Water Resources

摘　　要：采用STL方法,将1958年~2023年的月度地表径流量数据分解为趋势项、季节项和残差项,并利用LSTM模型进行预测,得到最终的径流量预测值。结果显示,LSTM模型在训练集和验证集上的拟合优度分别为0.95和0.94,均方根误差分别为4.9 mm和1.4 mm,预测准确性较高。此外,对2024年地表径流量的预测表明,夏季和秋季径流量显著增加。该方法可有效提升地表径流量预测的准确性,也可为唐山市的水资源管理和防洪减灾提供参考依据。The STL method is used to decompose the monthly surface runoff data from 1958 to 2023 into trend term,seasonal term and residual term,and the LSTM model is used to predict the final runoff prediction value.The results show that the goodness of fit of the LSTM model on the training set and the verification set is 0.95 and 0.94,respectively,and the root mean square error is 4.9 mm and 1.4 mm,respectively,and the prediction accuracy is high.In addition,the prediction of surface runoff in 2024 shows a significant increase in summer and autumn runoff.This method can effectively improve the accuracy of surface runoff prediction,and can also provide reference for water resources management and flood control and disaster reduction in Tangshan City.

关 键 词：唐山市 STL LSTM 地表径流量 

分 类 号：TV121.2[水利工程—水文学及水资源]