基于ARIMA模型的补远江含沙量预测

ARIMA-based Sediment Concentration Forecasting for Buyuanjiang River ZHONG Ronghua,FU Kaidao,HE Daming,XING Yumin,SU Bin

流域输沙的估算是水资源管理中广泛面临的问题。基于时间序列自回归滑动平均(ARIMA)预测模型,分别对补远江曼安水文站1993～2008年雨季、旱季月平均含沙量资料进行建模拟合。综合AIC值、相对误差,确定模型的阶数,运用Marquardt非线性最小二乘法估计模型参数,建立ARIMA预测模型。经检验,雨季AIC=-61.046,旱季AIC=-131.785,相对误差低于20%的合格率分别为92.1%、76.9%,残差序列均为白噪声序列,表明旱季ARIMA(1,1,1)、雨季ARIMA(1,1,2)模型较为合理。应用模型对2009～2011年曼安水文站的雨季、旱季平均含沙量进行了预测,实现了河流输沙状况的短期预报。

Estimate of sediment load is widely required in water resources management. Based on auto-regressive integrated moving average （ARIMA） forecasting models, we simulated the trends of monthly average sediment concentration at the Manan Station on the Buyuanjiang River in dry and wet seasons from 1993 to 2008. The order of the model was tested with AIC value, while relative error and the parameter of the model were estimated using Marquardt least square method for results assessment. The calculation shows that the AIC are 61.046 and -131.785 in rainy and dry season respectively, with errors less than 20 % and residual errors character by white noise series. So it is argued that both ARIMA （1, 1, and 1） for rainy season and ARIMA （1, 1, and 2） for dry season are feasible in this basin. Then the model was employed to predict the annual sediment concentration at the Manan Station from 2009 to 2011, so that short-term forecast for river sediment transportation was successfully realized in the studied reach.