组合模型在水库泥沙淤积量预测中的应用

基于指数平滑、模糊移动、线性回归、灰色预测四种单一模型, 建立了组合预测模型, 并根据某水库1997 年- 2012 年泥沙淤积量实测数据, 构建了该水库泥沙淤积量组合预测模型。实证研究结果表明, 组合预测模型在有效 综合各单一模型的有用信息方面具有较强的优势, 模拟结果平均绝对误差百分比为2. 539 3%, 均高于四种单一模 型的预测精度, 因此组合模型具有较好的实用价值。 A combining for ecast mo del w as developed based on four individual models, including the ex ponential smoothing , fuzzy moving , linear r egr essio n, and g rey fo recast models. Mor eover , a combining f orecast mo del was developed fo r a r eser voir based on the measured sedimentatio n amo unts from 1997 to 2012. T he r esult s show ed that the combining for ecast model has advanta2 g es in combining the useful informat ion fr om individua l mo dels, and the mean absolute err or percent age is 2. 539 3% fo r the combining for ecast mo del, w hich has hig her pr edict ion accuracy than the indiv idual mo dels. T her efor e, the combining fo recast model has pr act ical values.