摘要
准确的径流预测在水资源规划和管理中发挥着重要作用.然而,受气候变化和人类活动等因素的影响,径流形成过程十分复杂,具有高度的非线性和非平稳性,更增加了径流预报的难度.为提高月径流预测精度,提出了基于时变滤波器的经验模态分解(TVF-EMD)和结合粒子群优化算法(PSO)的门限循环单元(GRU)的混合模型(TEPG).首先利用TVF-EMD将原始月径流序列分解为若干个固有模态函数(IMF),然后再利用PSO-GRU模型分别对每一个IMF进行预测,最后将每个IMF的预测结果相加得到原始月径流序列最终的预测结果.以黄河干流4个代表性水文站(包括唐乃亥站、头道拐站、花园口站、利津站)为研究对象,应用该模型对这4个测站的月径流进行单步预测研究,并与PSO-GRU(PG)模型、基于互补经验模态分解(CEEMD)的PSO-GRU(CPG)模型和基于经验模态分解(EMD)的PSO-GRU模型(EPG)进行对比分析.选用纳什效率系数NSE、相关系数R、均方根误差RMSE、预报合格率QR及预报精度等级等评价指标对模型预测精度进行评价.结果表明,与PG模型、CPG模型、EPG模型相比,TEPG模型具有更高的预测精度和更好的泛化能力,4个水文站的NSE均达到0.981及以上,R均达到0.992及以上,RMSE最大仅为64.031 m^(3)/s,QR均达到84.7%及以上,预报精度等级均为乙等及以上.因此,TEPG模型在预测非平稳和非线性月径流序列中具有较好的应用前景.
Accurate runoff forecasting is important in water resources planning and management.However,because of the influence of climate change and human activities,the runoff formation process is very complex,with a high degree of nonlinearity and nonstationarity,making runoff forecasting challenging.A hybrid model TEPG based on time-varying-filter-based empirical mode decomposition(TVF-EMD)and gated recurrent unit(GRU)combined with particle swarm optimization(PSO)is proposed to improve the accuracy of monthly runoff forecasting.First,the original monthly runoff series is decomposed into several intrinsic mode functions(IMFs)by the TVF-EMD,and each IMF is subsequently predicted by the PSO-GRU(PG)model.Finally,the final prediction result of the original monthly runoff series is obtained by adding the prediction results of each IMF.The model was used to predict the monthly runoff of four representative hydrological stations in the mainstream of the Yellow River(including the Tang-naihai station,Toudaoguai station,Huayuankou station,and Lijin station)as the research object,and it was compared to the PG model,the complementary ensemble EMD(CEEMD)-based PG(CPG)model,and the EMD-based PG(EPG)model.To evaluate the prediction accuracy of the model,the evaluation indicators,including the Nash-Sutcliffe efficiency coefficient NSE,the correlation coefficient R,the root mean square error RMSE,the qualified rate of prediction QR,and the prediction accuracy grade,were selected.The results show that the TEPG model had higher prediction accuracy and better generalization ability than the PG,CPG,and EPG models.The four hydrological stations had an NSE of 0.981 and above,an R of 0.992 and above,a maximum RMSE of only 64.031 m^(3)/s,a QR of 84.7%and above,and prediction accuracy of grade B and above.Therefore,the TEPG model has a good application prospect in predicting nonstationary and nonlinear monthly runoff series.
作者
王秀杰
张帅
田福昌
苑希民
曹鲁赣
Wang Xiujie;Zhang Shuai;Tian Fuchang;Yuan Ximin;Cao Lugan(State Key Laboratory of Hydraulic Engineering Simulation and Safety,Tianjin University,Tianjin 300072,China;School of Civil Engineering,Tianjin University,Tianjin 300072,China)
出处
《天津大学学报（自然科学与工程技术版）》
EI
CAS
CSCD
北大核心
2022年第8期802-810,共9页
Journal of Tianjin University：Science and Technology
基金
国家重点研发计划资助项目(2018YFC1508403)
科技部重点领域创新团队项目(2014RA031).
