BP神经网络在台风路径预报中的应用

利用前馈型BP神经网络模型,对发生于中国沿海的热带气旋的移动路径进行了预报应用研究.根据中国<台风年鉴>发布的每个台风过程记录,对预报试验的台风个例分别选取了经度、纬度、中心气压和最大风速等81个因子,由多元回归选取了其中相关性好的因子,进行网络的学习训练,在获取前24h间隔6h的4次台风信息的基础上,用来预报了台风未来24h,48h和72h的短期路径变化.将该方法预报结果与CLIPER模式预报结果进行了比较,结果表明,BP神经网络模式的预报精度比CLIPER模式的高.

Application of BP neural network to forecasting typhoon tracks

In this paper,the forward feedback BP neural network model was used for study to forecast tropical cyclone track around the China Sea.The data of typhoon case were selected from the Tropical Cyclone Almanac pub lished by the China Meteorological Administration.For each typhoon case to be tryed for forecasting,81 factors,such as longitude,latitude,central air press,maximum wind speed,etc,were chosen.The multivariate regression arithmetic was applied to filter factors based on the correlation between the factor and typhoon track,and then the BP network was trained.After obtaining the information of typhoon every 6 hour of former 24 hours,we can predict the typhoon track for next 24,48,and,72 hours.The method can be used on ship.Comparing the way with the model CLIPER,it is shown that the accuracy of BP model is better than that of the model CLIPER.