人工神经网络在感热通量计算中的应用

摘要

The evaluation of sensible heat flux is crucial in the boundary layer study. Based on the boundary layer observation data at Wenjiang Station in the eastern edge of the Sichuan Basin, the method for calculating sensible heat flux with ANN (artificial neural network) is preliminarily studied and compared with the empirical method. The corresponding correlation analysis and error analysis are conducted with the true values. Two caseson April and May in 2009 are studied, and the results show that the correlation coefficient and variation trend of the ANN method are better than those of the empirical method; the root mean square error (RMSE) of ANN is slightly larger than that of the empirical method in April 2009 but obviously smaller in May 2009.%感热通量计算方法的研究是边界层研究中最重要的内容之一.基于中日JICA计划项目中青藏高原东缘四川盆地温江站的边界层铁塔观测资料初步研究了使用人工神经网络(ANN)计算边界层感热通量的方法,并将ANN和经验公式法计算得到的感热通量分别和真值作相关和误差分析.对2009年4月和5月的两个个例研究的结果表明:ANN计算结果和真值的相关性都高于经验公式法且趋势变化和真值更加吻合,ANN计算的2009年4月的感热通量与真值的均方根误差(RMSE)稍大于经验公式法,但2009年5月的RMSE明显小于经验公式法.