利用神经网络预报中国地区电离层f0F2

摘要

By using artificial Neural Network (NN) and considering the effects of the solar and geomagnetic activities on the ionosphere in this paper we brought out a method to forecast the ionospheric critical frequency, f0F2, in China up to 5 hours ahead.The inputs of the NN are time, day of the year, geographical latitude, solar zenith angle, the twelve recent past observations of f0F2 (F-23, F-22, F-21, F-20, F-19, F-18, F-5, F-4, F-3, F-2, F-1, F0) and the 30-day mean moving values of f0F2 (A-24, A-23, A-22, A-4, A-3, A-2, A-1, A0) from the target location.The outputs of the NN are F+1, F+2, F+3, F+4, F+5, representing the values of f0F2 up to 5 h ahead.Data from Wulumqi, Changchun, Chongqing and Guangzhou stations spanning the period 1958-1968 are used for training the NN.Historical data at nine different stations in China are used to checkout the network respectively (Not including the training set).The performance of the NN is measured by calculating the Root-Mean-Square error (RMS) difference between the NN outputs and measured station data.The results indicate that the prediction of NN has good agreement with measured data.Taken into account those temporal and spatial inputs mentioned above, the NN can be applied successfully in the short-term forecasting of the ionospheric f0F2 in China.%利用神经网络技术并考虑太阳和地磁活动对电离层的影响,提出了一种提前5 h预报中国地区电离层临界频率f0F2的方法.网络输入包括时间、季节、地理纬度、太阳天顶角、最近一天的12个观测值(F-23,F-22,F-21,F-20,F-19,F-18,F-5,F-4,F-3,F-2,F-1,F0)和前30天滑动平均值(A-24,A-23,A-22,A-4,A-3A-2,A-1,A0),网络输出分别为未来5 h的f0F2值F+1,F+2,F+3,F+4,F+5.选取乌鲁木齐、长春、重庆和广州站1958-1968年间的数据训练网络,利用中国9个电离层站的历史数据检验网络,根据均方根误差衡量网络性能的好坏.结果表明,神经网络的预报结果能较好地符合实测数据.这说明利用神经网络实现中国地区电离层f0F2的时空预报是可行的.