摘要
为了提高紫外线预报准确率,应用BP(Back Propagation Learning Algorithm)神经网络模型和支持向量机(Support Vector Machines,简称SVM)回归方法建立重庆市主城区紫外线辐射强度客观预报模型。统计相关分析结果显示,不同季节影响紫外线辐射强度的主要因素并不相同。对所有相关分析因子用逐步回归方法,按方差贡献大小筛选出预报因子,以每日紫外线平均辐射量为预报对象,分季节建立预报模型。比较用不同方法建立的预报模型发现,两种非线性模型(BP模型和SVM模型)的拟合能力优于线性逐步回归模型,但独立样本检验结果表明,3种模型的预报准确率基本相当。将3种方法所建预报模型应用T213数值预报资料进行业务试报,得到较好预报效果。
The ultraviolet radiation forecast models were built using two artificial intelligence techniques： Artificial neural network about Back Propagation Learning Algorithm and Support Vector Machines regression model in order to improve on the forecast accuracy in Chongqing city.At first,the statistics correlation analysis results exhibited that UV radiation had shown a regular change with season and the main factors to affect UV radiation had varied with season.Then,it considered the daily mean UV radiation as the dependent variable,the predictors were selected by stepwise regression analysis for all the factors of the correlation analysis according to their variance contribution,and the forecast models were established to every season.The comparison results from three different forecasting techniques manifested that the fitness results of two nonlinear models（BP and SVM） were better than stepwise regression equations,but there was no significant difference in the forecast accuracy from three techniques.At last,BP models,SVM models and stepwise regression forecasting equations were applied in operational prediction every day by use of the interpretation of numerical forecasting products from T213,and the forecasting results showed a good reference value.
出处
《高原气象》
CSCD
北大核心
2010年第2期539-544,共6页
Plateau Meteorology
基金
重庆市气象局"重庆主城区紫外线客观预报方法研究"项目资助
关键词
BP神经网络
支持向量机
紫外线
预报准确率
BP Neural Network
Support vector machines
Ultraviolet radiation
Forecast accuracy
