利用BP神经网络改进电离层短期预报模型

Improved Prediction Model of Ionospheric TEC by BP Neural Network

电离层总电子含量TEC(Total Electron Content)是电离层的一个重要特征参数。对TEC的预报也已经成为电离层研究的一个热点。根据JS CORS中心提供的GPS观测数据,建立了区域实时多站多项式模型;并分别以模型计算得到的南京地区的电离层电子含量数据和苏州地区的电离层电子含量数据为样本,采用时间序列和BP神经网络融合模型进行了预报。结果表明,采用融合模型在短期预报中能够取得较好的效果,精度比时间序列模型提高20%左右。

Total Electron Content（ TEC） is one of the most important characteristic parameters of ionosphere, and TEC prediction also is a hot spot in the research for ionosphere. A local real-time polynomial model is established based on the data provided by JS CORS. And the TECs of Nanjing and Suzhou are calculated by using this model,to compare with the prediction effect of time series model and BP neural network fusion model. The result indicates that the fusion model achieves a higher precision in short-term forecast and the accuracy of fusion model is about20 percent higher than that of the time series model.