气象干旱组合预测模型研究及其应用  

Study on the Combined Forecast Model and Its Application of Meteorological Drought

作　　者：马建琴[1] 毕静静[1] 赵晓慎[1] 

机构地区：[1]华北水利水电大学水利学院,河南郑州450011

出　　处：《人民黄河》2015年第4期10-13,共4页Yellow River

基　　金：国家自然科学基金资助项目(41071025);河南省教育厅自然科学研究基金资助项目(2009A170004);河南省科技攻关基金资助项目(092102310197)

摘　　要：为提高干旱预测精度,克服单一预测模型的不足,在分析灰色理论和遗传神经网络模型特点的基础上,构建了气象干旱的多尺度组合预测模型。该模型首先提取灾变序列,利用GM(1,1)模型进行拟合和预测,然后采用遗传神经网络对拟合值进行修正,得到训练好的网络结构,最后修正GM(1,1)模型的预测值。利用郑州市1951—2012年月降水数据进行的干旱预测结果表明:针对不同尺度的灾变序列,组合预测模型的预测效果优于GM(1,1)模型和遗传神经网络模型,且模型的平稳性较好。In order to enhance the prediction accuracy of the drought and overcome the shortcomings of a single forecasting model,this paper put forward the combined forecast model of meteorological drought as the object,based on the analysis of grey theory and genetic neural net-work. The combined forecast model was used to analyze the monthly rainfall data of Zhengzhou City from 1951 to 2012. Firstly using GM (1, 1)model to fit and forecast the extracted sequences. Secondly using genetic neural network to correct the fitted values of GM (1,1)model and get the trained network structure. Finally using the trained network structure to modify the forecasting value of GM (1,1)model. The re-sults show that for the different scales of catastrophic sequence,the accuracy of the combined forecast model is much higher than that of the GM (1,1)model and genetic neural network and the stability of the model has been enhanced.

关 键 词：组合预测模型 GM(1 1)模型 遗传神经网络 干旱预测 

分 类 号：P457[天文地球—大气科学及气象学]