J4 ›› 2011, Vol. 29 ›› Issue (2): 231-235.

• 业务技术应用 • 上一篇    下一篇

基于BP 人工神经网络的区域温度多模式集成预报试验

李倩1,2 ,胡邦辉2 ,王学忠2 ,顾锦荣2   

  1. 1. 解放军93550 部队,河北涿州072757; 2. 解放军理工大学气象学院,江苏南京211101
  • 收稿日期:2010-11-30 修回日期:2011-03-07 出版日期:2011-06-30 发布日期:2011-06-30
  • 作者简介:李倩( 1977 - ) ,女,汉族,河北保定人,硕士研究生,主要从事天气分析与预报. E - mail: wxiaozhuzhu@ yahoo. com. cn

Multi - model Integration Forecast Experimentation for Regional Temperature Based on BP Artificial Neural Network

LI Qian1,2,HU Banghui2,WANG Xuezhong2,GU Jinrong2   

  1. 1. 93550 Army,Zhuozhou 072757,China; 2. College of Meteorology,PLA University of Science and Technology,Nanjing 211101,China.
  • Received:2010-11-30 Revised:2011-03-07 Online:2011-06-30 Published:2011-06-30

摘要:

基于T213L31、T106L19 和欧洲中期预报中心数值预报产品,应用BP 人工神经网络技术,建立了850 hPa 高度区域温度集成预报模型,并进行了检验。结果表明: 该模型能比较准确地预报强冷空气活动过程中冷中心及温度槽脊的位置和强度,预报结果的平均绝对误差明显小于3 个子模式,预报场与实况场的相关程度明显高于3 个子模式,预报误差在我国华北北部、东北地区较小,在蒙新高地和帕米尔高原地区误差较大。模型实现了多模式产品的最优综合

关键词: 人工神经网络, 温度场预报, 多模式集成

Abstract:

Based on the production of T213L31,T106L19 and ECMWF,the regional multi - model integration temperature forecast model at the level of 850 hPa was developed using artificial neural network method. Forecast testing result indicated that the model could forecast the center of cold air developing,the position and intensity of temperature trough and ridge,and the mean error of forecast result was obviously smaller than that of three single model,the correlation between the forecasted and observed field was better than that of three single model. Forecast error was small over north and eastern China,but big over Mengxin highland and Pamirs. The
multi - model integration based on artificial neural network was a kind of effective method to temperature forecast.

Key words: artificial neural network, temperature field forecast, multi - model integration

中图分类号: