神经网络汛期降水短期气候预测模型  

Short-Term Climate Prediction Model of Neural Network on Flood Season Rainfall Amount

作　　者：吴有训[1] 程雪生[1] 胡安霞[1] 

机构地区：[1]安徽省宣城市气象局,安徽宣城242000

出　　处：《数学的实践与认识》2010年第3期103-106,共4页Mathematics in Practice and Theory

摘　　要：用1960-2005年每年5-9月安徽宣城7县(市)平均降水量作为预报对象,在对应降水前期逐月74项大气环流特征量资料、500hPa月平均高度场和月平均海温场资料中选取因子.采用主分量分析方法构造网络学习矩阵,降低矩阵维数,提高网络预测模型泛化性能.建立的神经网络汛期降水短期气候预测模型对历年样本拟合精度高,试报效果较好,可在气候预测业务中使用.Based on the seven station of the average precipitation in Xuancheng With May to September each year from 1960 to 2005 as predictand,the selection factor from the monthly atmospheric circulation characteristic value data,the average monthly 500hPa height field data and the average monthly sea surface temperature field data.Establishment by means of making a low-dimension ANN leafing matrix through principal component analysis, Improve the network prediction model of generalization performance.The results show that the prediction model,has fitting accuracy on historial samples and testing effect as well,and can be applied in operational climate prediction.

关 键 词：神经网络 降水预测 主分量分析 

分 类 号：P457.6[天文地球—大气科学及气象学] O242.1[理学—计算数学]