支持向量机在太阳辐射预报中的应用  

Applications of Support Vector Machines in the Solar Radiation Forecasting

作　　者：张礼平[1] 陈正洪[2] 成驰[2] 王晓莉 

机构地区：[1]武汉区域气候中心,武汉430074 [2]湖北省气象科技服务中心,武汉430074 [3]湖北省公众气象服务中心,武汉430074

出　　处：《暴雨灾害》2010年第4期334-336,355,共4页Torrential Rain and Disasters

基　　金：科技部公益性行业(气象)科研专项(GYHY201006036);华中区域气象中心推广项目(QY-T-200902);华中区域气象中心重点项目(QY-Z-201010);中国气象局工作任务(气预函[2010]76号)

摘　　要：利用EOF能分解数据场和SVM回归分析可建立因子与预报量非线性关系的优势,设计预报方案:(1)将多因子和多预报量分别方差标准化,EOF场展开,提取主分量;(2)用SVM回归分析实现多因子主分量对多预报量主分量非线性预测;(3)由预报的多预报量主分量与对应空间函数反演原预报量。选用武汉预报日同一天气类型的上一日逐时(05—18时)总辐射、日最高温度、温度日较差、日天气类型观测值以及预报日的日最高温度、温度日较差、日天气类型预报值为因子,对预报日逐时辐射量进行预报。独立预报试验表明,预报与实况接近。With the superiority of both of EOF（Empirical orthogonal functions） separating fields and nonlinear SVM regression forecasting,a program is projected considering the following：（1） Both of factors and predictors are variance standardized,then EOF,and principal components of both are extracted respectively;（2） With SVM regression,principal components of predictors are estimated by factors;（3） The original predictors are recovered by linear combination of principal components and the eigenvectors. Solar radiation over Wuhan is forecasted hour by hour with Solar radiation of previous day,the maximum temperature,temperature range and weather type of both previous day and forecasting day.It is indicated that Solar radiation forecasted hour by hour were approximate to the observations.

关 键 词：支持向量机 回归分析 太阳辐射 

分 类 号：P461.1[天文地球—大气科学及气象学]