基于iInformer的电离层TEC短期预测  

Short-Term Prediction of Ionospheric TEC Based on iInformer

作　　者：田晓鹏 罗亦泳[1,2] 张紫怡 TIAN Xiaopeng;LUO Yiyong;ZHANG Ziyi(School of Surveying and Geoinformation Engineering,East China University of Technology,330013,Nanchang,PRC;Key Laboratory of Mining Environment Monitoring and Treatment around Poyang Lake,Ministry of Natural Resources,330013,Nanchang,PRC)

机构地区：[1]东华理工大学测绘与空间信息工程学院,南昌330013 [2]自然资源部环鄱阳湖区域矿山环境监测与治理重点实验室,南昌330013

出　　处：《江西科学》2025年第1期52-58,210,共8页Jiangxi Science

基　　金：江西省自然科学基金项目(20202BABL204070)。

摘　　要：电离层总电子含量TEC的监测与预报是近地空间环境研究的重要内容,对卫星通讯和导航定位等有重要意义。使用基于Transformer(变形金刚)的iInformer(告密者)模型,提出中国区域电离层TEC短期预报新方法,且分别对磁静期与磁暴期电离层进行预测。为了分析短期电离层新模型预测效果,选取神经网络模型、线性模型、长短时记忆模型进行对比。结果表明,磁静期选定区域内iInformer模型有效适用于短期预测任务且预测精度明显优于其他对比模型,均方根误差在3个区域均低于1.45 TECU(total electron content units,总电子含量单位)。iInformer模型在应对不同数据量时,均能保持稳定的预测性能。特别是在数据集数量相对有限(少于2个月)的情况下,iInformer模型的预报精度显著优于其他模型。相较于单一数据源,多数据源下的iInformer模型预测精度有显著提升,提升幅度在2%~7.4%。The monitoring and prediction of total ionospheric electron content(TEC)is an important aspect of near-Earth space environment research,which is of great significance for satellite communication and navigation positioning.This article proposes a new method for short-term prediction of TEC in the Chinese regional ionosphere using the Transformer-based iInformer model,and predicts the ionosphere during geomagnetic period and storm periods respectively.In order to analyze the predictive performance of the new short-term ionospheric model,comparisons were made with the PatchTST model,Dlinear models,and long short-term memory(LSTM)models.The results show that the iInformer model within the selected area of magnetostatic period is effectively applicable to short-term prediction tasks and the accuracy of predicting TEC is significantly better than other comparative models,with RMSE(Root mean square error)below 1.4 TECU(total electron content units)in all three regions.The iInformer model can maintain stable predictive performance when dealing with different data volumes.Notably,when the number of datasets is relatively limited(less than two months),the forecasting accuracy of the iInformer model is significantly better than other models.Compared to a single data source,the iInformer model under multiple data sources has significantly improved prediction accuracy,with an improvement range from 2%to 7.4%.

关 键 词：电离层总电子数(TEC) TRANSFORMER iInformer 线性模型 磁静期 磁暴期 

分 类 号：P352[天文地球—空间物理学]