小样本情况下需水量预测模型研究  

Research on Water Demand Forecasting Model under Small Sample

作　　者：曾蒙秀[1,2] 宋友桂[1] 

机构地区：[1]中国科学院地球环境研究所,黄土与第四纪地质国家重点实验室,陕西西安710075 [2]中国科学院研究生院,北京100049

出　　处：《节水灌溉》2011年第11期13-18,共6页Water Saving Irrigation

基　　金：国家科技支撑计划项目(2007BAC30B06);国家重点基础研究发展规划项目(2010CB833406)

摘　　要：针对传统的GM(1,1)模型,分析了其预测结果与实际过程存在的偏差,通过GM(1,1)模型与自回归滑动平均模型相结合的方法以弥补偏差。以钦州市1999-2009年及2005-2009年城市供水总量这两组基础数据对本文所建模型进行验证,并利用此模型预测了钦州市2010-2018年需水量。通过与其他预测模型的对比,进一步证明了本文所提新方法具有更高的预测精度、更强的数据拟合能力及更强的适应性,在小样本的预测上显示出新模型特有的优越性。在时间序列短、原始数据序列波动大的情况下,对于需水量预测这类复杂的问题,本文所提模型具有较高的可靠性。There is some deviation between the predicted value and the actual value in the traditional GM（l,l） model,which is based on grey theory.The Auto Regressive Moving Average（ARMA） model is introduced to the new method to reduce some error in the forecasting processes.The new model is validated based on the forecasting results obtained by GM（l,l） model and the new model,which utilize the water consumption in Qinzhou city from 1999 to 2009 and from 2005 to 2009 as original data series.The water consumption from 2010 to 2018 in Qinzhou city predicted by this new model shows good agreement with the development plan of Qinzhou city.The comparison between the new model and the other models indicates that the new model has higher precision,better applicability and stronger capability of fitting,especially for the small sample,the new model displays its stronger superiority.This method has higher reliability for water demand forecasting in the circumstance of shortage of relevant data or the fluctuation data.

关 键 词：小样本 需水量预测 灰色GM(l l)模型 自回归滑动平均模型 模型对比 

分 类 号：TV211[水利工程—水文学及水资源] TV213