利用人工神经网络预测电离层f_0F_2月中值

提出了利用人工神经网络技术预测电离层临界频率月中值的方案．在利用人工神经网络技术研究电离层月中值隔月变化规律的基础上，考虑足够的周年和黑子周期变化的数据训练网络，使 ｆ０Ｆ２月中值预测值与实测数据比较平均误差为 ０．３４ＭＨｚ，预测精度有了较大改进．最后采用分形学的基本理论得到电离层 ｆ０Ｆ２月中值的分数维为 ３，选用 ３个输入量，分别预测高、低年 ｆ０Ｆ２月中值，与实测数据比较平均误差为 ０．３ＭＨｚ．

PREDICTING THE MONTHLY MEAN VALUES OF f_0F_2 BY USING THE ARTIFICIAL NEURAL NETWORK

A method to predict the monthly mean Values of f0F2 is presented in this paper. The analysis results of monthly values of f0F2 indicate that the characteristic of this ionospheric parameter change with different month and different year. Based on it, the f0F2 monthly mean value is predicted by taking sufficient data of many yeas into account and improved the predicting method. Compared with the conserved data, the average error is less than 0.34MHz. Then, the fraction characteristic of ionosphere has been set forth by using the theory of fraction and the fraction dimension of the f0F2 monthly mean value has been acquired. Based on it, 3 parameters are selected to predict the f0F2 monthly mean value for different year, thus, the predicted technology is improved further. Compared with the conserved data, the average error is less than 0.3 MHz.