基于主成分分析与BP神经网络的雾天能见度等级预报.PDF

第9 卷第12期 Vol.9 No.12 2016 年 6 月 June 2016 基于主成分分析与 BP 神经网络的雾天 能见度等级预报 黄 政，包云轩* （南京信息工程大学应用气象学院，南京 210044 ） 摘要：利用江苏省昆山市2012 年至2014 年逐时常规气象观测数据、空气质量监测数据和能见度数据，分析雾 天能见度与各要素的相关性，并通过主成分分析提取影响主成分，建立基于主成分的三层BP 神经网络模型。 结果表明，雾天能见度不仅与气象要素呈现较好的相关性，空气污染物（如NO 、O 、PM10 ）对雾天能见度 2 3 也有较大影响；主成分神经网络模型能够较准确地预测雾天能见度等级[大雾、浓雾、（特）强浓雾] ，对提高 雾天能见度精细化预报效果具有良好的参考价值。 关键词：应用气象学；雾；能见度等级；主成分分析；神经网络 中图分类号：P468.0+28 文献标识码：B 文章编号：1674-2850(2016)12-1262-06 Forecasting model for visibility levels in foggy weathers based on principal components analysis and BP neural network HUANG Zheng, BAO Yunxuan (Applied Meteorology School , Nanjing University of Information Science and Technology, Nanjing 210044, China) Abstract: Based on the hourly weather data, the environmental atmosphere quality monitoring data and visibility data from 2012 to 2014 in the Kunshan city of Jiangsu province, the correlation coefficient between visibility and different variables was analyzed and a three-layer BP neural network model was constructed based on the principal components. The results show that visibility in the foggy weathers not only has a good correlation to meteorological factors but also is influenced by air pollutants such as NO2, O3 and PM10; the model can accurately predict the visibility levels (heavy fog, dense fog, extremely dense fog), and has a good reference value to improve the ability of fine fog forecast. Key words: applied meteorology; fog; v