Forecasting of total electron content (TEC)/global positioning system (GPS) signal delays in storm conditions is considered the most challenging task for accurate position estimation, especially in critical applications. Therefore, a storm-time ionospheric model is proposed to forecast TEC based on the artificial neural network (ANN) using singular spectrum analysis (SSA). The study area covers four Global Navigation Satellite System (GNSS) stations located in the low-latitude and two GNSS stations located in the mid-latitude ionosphere. The geographical area extends between 11–43°N latitude and 77–93°E longitude. The selection of GPS–TEC data is based on the storm criterion of Dst ≤ −50 nT, and the storm day data sets from 2009 to 2017 period are used for model development. The proposed algorithm is tested with the GPS–TEC data sets of three geomagnetic storm days (i) severe, (ii) moderate storm and (iii) strong at low and mid-latitudes. The average precision and mean absolute error of the proposed SSA-ANN model is 1.41 and 1.06 TECU (strong storm), respectively. The prediction performance of the proposed SSA-ANN model is compared with the standard principal component analysis-ANN model. The improvement factor of the SSA-ANN is improved by 43.82%.

中文翻译：

---

利用SSA–ANN模型实现风暴时间电离层预报算法

预测风暴条件下的总电子含量（TEC）/全球定位系统（GPS）信号延迟被认为是精确位置估计的最具挑战性的任务，尤其是在关键应用中。因此，提出了一种基于奇异谱分析（SSA）的基于人工神经网络（ANN）的风暴时间电离层模型来预测TEC。研究区域涵盖了位于低纬度的四个全球导航卫星系统（GNSS）站和位于中纬度电离层的两个GNSS站。地理区域在北纬11–43°和东经77–93°之间。GPS-TEC数据的选择基于Dst≤−50 nT的风暴标准，并将2009年至2017年期间的风暴日数据集用于模型开发。在三个地磁风暴日（i）严重，（ii）中度暴风雨和（iii）在低纬度和中纬度的强地震天的GPS-TEC数据集上测试了提出的算法。所提出的SSA-ANN模型的平均精度和平均绝对误差分别为1.41和1.06 TECU（强风暴）。将提出的SSA-ANN模型的预测性能与标准主成分分析-ANN模型进行了比较。SSA-ANN的改进因子提高了43.82％。