BP人工神经网络法在大同市日极大风速预报中的应用

The Application of BP Neural Network to Daily Maximum Wind Velocity Forecast in Datong

文章利用2010年1月1日—2013年12月31日逐日NCEP再分析资料(1°×1°)和大同地区地面常规观测资料,采用BP人工神经网络法建立大同市分站点、分季节日极大风速人工神经网络预报模型并且在对T639数值预报产品和EC细网格数值预报产品释用基础上建立了台站日极大风速的客观预报系统,对2015年9月1日—2016年7月31日进行了24h预报,试用结果显示,各季模式平均绝对误差在3.2~5.7m·s^-1之间,因此,该系统可以为预报员快速做出日极大风速的预报提供客观参考依据。

Based on the NCEP re-analysis data(1^0×1^0)and the observed data in Datong from January 2010 to October 2013,an BP neural network model with three layers is constructed by different stations and seasons to forecast daily maximum wind velocity and the independent samples can be tested in it.Based on the interpretation of numerical forecasting products of the T639 and ECTHIN,a simulation test for predicting maximum wind velocity for 24 hours at Datong station in 2015 is carried out.The results show out that the mean absolute errors of seasonal models are between 3.2 and 5.7,so the system can provide objective and reference evidence for forecasters.