摘要
以广西西南部前汛期5、6月25个气象站平均逐日降水量作为预报对象,采用自然正交分解方法和模糊化方法对输入因子预处理后,结合Modular模糊神经网络建立了一种新的降水预报模型,并进行了逐日业务预报应用试验。结果表明,该降水预报模型比常规Modular模糊神经网络方法及逐步回归方法有更高的预报精度,具有较好的业务应用前景。
A new forecast model of precipitation is exemplified by use of modular fuzzy neural network to daily forecasts of 25 weather stations＇average precipitation of southwest Guangxi in May and June. The 25 weather stations＇ rainfull forecast model is established based on the pretreatment of input factors by empirical orthogonal function and fuzzification approach. Daily operational forecast trial of the new model is made. The results show that the new forecast model is superior in prediction accuracy to the general modular fuzzy neural network model or the stepwise regression procedure in daily precipitation forecast in the period. The prospects for the application of the new model are proved.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第18期4797-4800,共4页
Computer Engineering and Design
基金
国家科技部社会公益性研究专项基金项目(2004DIB3J122)
关键词
模糊化
模块化模糊神经网络
自然正交展开
逐日降水量
预报建模
fuzzification
modular fuzzy neural network
empirical orthogonal function
daily precipitation
forecast model
