摘要
系统在经大量数据训练、运算基础上，选用了ＢＰ网络模型。为适应台风路径预报，对ＢＰ网络进行了改进，应用改进后的ＢＰ网络，自动训练到２１５９次后，由神经网络输出的判定台风移向趋势──西进，北上，西北移与实际历史移行路径概括率达９７％。在业务实用时，只需在指定硬盘路径中给出系统运行的３个文件，包括天气系统主要特征参数，系统即可以人机对话形式给出当前台风移动路径概率值。
hrough lots of data training and calculations,the selected BP network is improved to beadapted to the typhoon track prediction. After 2159 times of training,the neural network output of the typhoon track-westward,northward and northwestward,is compared with thehistorical data. Its fitted ratio is up to 97%,The probability of the typhoon moving trend isgiven by means of human-computer dialogus.
出处
《气象》
CSCD
北大核心
1996年第1期18-21,共4页
Meteorological Monthly
