摘要
【目的】对中长期径流预报模型进行比较和优选,为中长期径流预报模型的应用及提高预报精度提供参考。【方法】以黄河流域黑石关、龙门、民和水文站年径流为研究对象,利用灰色预测进行趋势分析、逐步回归进行周期分析,两者耦合建立灰色-逐步回归周期模型;利用逐步回归分析确定均生函数的周期性基函数,耦合建立均生函数-逐步回归模型;利用灰色-逐步回归中的趋势项和周期项作为预报因子,建立投影寻踪回归模型;利用前4年的实测径流数据预报当年径流,建立BP神经网络模型,并通过信息熵原理进行站点模型的综合评价和优选。【结果】建模期除黑石关水文站BP神经网络模型外,各站点拟合预报的平均相对误差均小于11.0%,合格率均大于90%。验证期除民和水文站灰色-逐步回归周期模型外,各站点平均相对误差均小于20%,合格率等于或大于80%,满足精度要求。【结论】均生函数-逐步回归径流预报模型可作为黑石关水文站的优选模型,BP神经网络径流预报模型可作为龙门和民和水文站的优选模型。
【Objective】Comparison and optimization of mid-long term runoff forecasting models were conducted to provide reference for improving model application and accuracy.【Method】Annual runoff data at Heishiguan,Longmen and Minhe hydrological stations in the Yellow River basin were selected and the grey prediction for trend analysis and stepwise regression for periodic analysis were coupled to build the grey-stepwise regression cycle model.The stepwise regression analysis was used to select the periodic function,and the mean-function-stepwise regression model was established.The trend term and periodic term in the grey-stepwise regression were used as predictors to establish the projection pursuit regression model.The BP neural network model were established using the observed annual runoff data at the first 4 years to forecast the current year runoff.The principle of information entropy was then used for comprehensive evaluation and optimization of models.【Result】During the modeling period,except the BP neural network model at the Heishiguan hydrological station,the average relative error at all sites was less than 11.0%and the qualified rate was greater than 90%.During the verification period,except the grey-stepwise regression cycle model at the Minhe hydrological station,the average relative error of all models at all sites was less than 20%and the qualified rate was greater than 80%,meeting the accuracy requirement.【Conclusion】The grey-stepwise regression cycle model was the best for runoff forecasting at the Heishiguan hydrological station,and the BP neural network runoff model was the best for runoff forecasting at the Longmen and Minhe hydrological stations.
作者
石继海
宋松柏
李航
SHI Jihai;SONG Songbai;LI Hang(College of Water Resources and Architectural Engineering,Northwest A&F University,Yangling,Shaanxi 712100,China)
出处
《西北农林科技大学学报（自然科学版）》
CSCD
北大核心
2019年第7期147-154,共8页
Journal of Northwest A&F University(Natural Science Edition)
基金
国家自然科学基金项目(51479171,51179160,50879070)
关键词
径流预报
灰色-逐步回归周期
均生函数
投影寻踪回归
BP神经网络
runoff forecasting
grey-stepwise regression cycle
mean function
projection pursuit regression
BP neural network
