时间序列与神经网络组合方法在电离层TEC预报中的应用

An application of time series and neural network method in ionospheric TEC prediction

时间序列和神经网络是广泛应用于各个领域的两种经典预测方法,将这两种方法结合起来可充分发挥其各自的优势。利用组合方法和IGS发布的2006年长沙地区某点处前11个月的电离层TEC(Total Electron Content,电子总含量)数据,对12月份的电离层TEC值进行了分析预报。结果表明,这种组合预测方法进行30天内电离层TEC预报相对精度在90%左右,比使用单个方法进行预报精度提高了2%～3%。

Time series and neural networks are two classic forecasting methods,which are widely used in a variety of fields.Combination of the two methods can fully adopt their individual advantages.Ionospheric TEC（Total Electron Content） value of December 2006 is predicted with the combined method using the TEC data provided by the IGS from January to November 2006 in Changsha.The results indicate that the relative TEC prediction accuracy within 30 days is about 90% and an accuracy improvement of 2% to 3% can be obtained using the combined method,comparing with using individual method.