渭河干流月径流的时历相依性研究

Time Dependency of Monthly Hydrologic Data Series of the Main Channel of Weihe River

目前的径流预测方法大都要求数据序列具有时历相依性,因此在应用相应的方法进行时间序列的预报时,应先判别时间序列的时历相依性。以渭河干流林家村水文站和咸阳水文站为研究对象,采用线性自回归模型对相邻两个月的径流量进行了相关性分析,研究了其月径流序列的时历相依性,并采用BP神经网络模型进行了月径流预报。结果表明:当相邻两个月径流量的相关度≥0.7时,数据序列时历相依性较好,可以进行月径流量预报建模;当BP神经网络模型的预报合格率≥80%时,可用于月径流量预报。

Presently,most methods of runoff forecast require data sequence to have nature of time dependency,therefore,before carrying on the forecasting of time series,the first step should distinguish the time dependency. Taking Linjiacun Hydrologic Station and Xianyang Station of Weihe River as the object,using linear auto-regression model,it analyzed the degree of correlation of neighboring two monthly runoff series,studied their time dependency and using BP neural network model to carry on the monthly runoff forecast. The result shows that when degree of correlation of two neighboring monthly runoff ≥0. 7,dependency of data sequence is good,can be used for forecasting;when the pass rate of forecasting≥80%, monthly runoff series can be used for forecasting by using neural network model.