【摘要】 利用人工神经网络研究低纬电离层参量的预测，首先我们研究从某一个月的电离层月中值预测下一个月的月中值。由于低纬电离层昼夜遵从不同的变化规律，我们将一天２４小时分成两到三个时间段进行分别预测，达到降低预测误差的目的。平均预测误差一般可以小于５－８％。其次我们将电离层看成一个系统，太阳辐射通量作为这个系统的输入，利用人工神经网络寻求太阳辐射通量与电离层Ｆ层参量之间的非线性关系，实现直接从太阳辐射通量预测电离层的月中值的目的。我们利用海南和广州两个台站１１年资料训练网络，采用训练后的网络预测电离层Ｆ层的临界频率的月中值，预测结果优于ＩＲＩ－９０的和更接近观测值。初步研究结果表明，人工神经网络能够充分利用大量的观测资料训练网络，训练后的网络不仅学习一些具体的例子，而且学会了从这些例子中所概括出的一般变化规律，寻求电离层复杂的非线性行为。更多还原

【Abstract】 In the present paper, variations of ionorpheric parameters at lower latitudes are studied by using theArtificial Neutal Net (ANN). First, predicting the monthly median values of f0F2 from the monthly median of thelast month mean values is conducted. Analysis of calculated errors indicates that tile variation characteristic of ionospheric parameter changes with time during different parts of day and night. They may be controlled by different factors and obey different variation rules. In order to get a high prediction precision I we have predicted the monthly median values of f0F2 in two or three seperated time ranges. If we use a single neural net to predict ionospheric parameters for a day, mean errors are less than 10% I but the errors may decrease to 5-8% if the time ranges are subdivided into two or three. Next, we use data of the solar radiation flux as the input of the ANN, and seek the nonlinearrelation between ionospheric parameters and the solar radiation flux. The data of both stations, Hainan andGuangzhou, for 11 years are used to train the ANd. The predicted results are better than one of IRI-90, and moreclose to the observed ones. The first results show that the ADN can use a large amount of ionospheric data, and obtain complicated nonlinear relation in the ionosphere. At the same time, it is also a very useful, and an artificial interligence technology for predicting ionocpheric parameters.更多还原