基于BP神经网络的汛期降水预测模型研究  

STUDY ON RAINFALL PREDICTION MODEL INRAIN SEASON BASED ONBP NEURAL NETWORK

作　　者：李永华[1] 刘德 金龙[3] 

机构地区：[1]南京气象学院,南京210044 [2]重庆市气象台,重庆400039 [3]江苏省气象科学研究所,南京210008

出　　处：《气象科学》2002年第4期461-467,共7页Journal of the Meteorological Sciences

基　　金：国家自然科学基金项目(40075021)资助

摘　　要：采用误差反传前向网络（简称BP神经网络）建立汛期（6—8月）降水预测模型,并与采用逐步回归方法建立的模型进行对比试验。结果表明,基于BP神经网络的汛期降水预测模型不仅对历史样本的拟合精度更高,而且其独立样本的预测检验效果也比逐步回归方法更好,是一种具有一定应用前景的汛期降水预测方法。A prediction model for June-August rainfall is established in terms of a neural networkand and a stepwise regression expression is formulated with the. aid of the same factors for comparision. Results show that the model based on BP neural network is superior in fittings and predictions compared to the model stepwise regression, and it is a important and useful model for the actual operational forecasting of rainfall in the rainy season.

关 键 词：BP神经网络 逐步回归 汛期 降水预报 对比试验 

分 类 号：P457.6[天文地球—大气科学及气象学]