基于Transformer的电离层fbEs短期预测

打开文本图片集

中图分类号：TN91-34 文献标识码：A 文章编号：1004-373X（2026）09-0107-07

Abstract：TheionosphericEslayer shielding frequency （fbEs）isakeyparameter affcting the propagationof radio waves. Theaccurate predictionoffbEs hasimportantaplicationvaluein thefieldsofcommunicationandnavigation.Inthis paper，a predictiomethodbasedonTransformerdeeplearningmodelisproposedtoimprovetheaccuracyofthetraditionalprediction modelforfbEs.Amodelwithhistorical fEsobservationdataandauxiliaryparameterssuchassolaractivityindex asinputand fbEssequenceinthenext72hoursasoutput isconstructed，andthefbEsinBeijing，HaikouandLhasaarepredictedand analyzed.Theresultsshowthatthepredictionaccuracyof themodelbasedonTransformerisbeterthanthoseofthetraditional methods suchasthe modelbasedonARIMAandthe modelbasedonLSTM，and its meanabsoluteerror（MAE）is12.5%lower onaverage thanthatofthemodelbasedonLSTM，anditsroot meansquareeror（RMSE）is13.98%loweronaveragethanthat ofthemodelbasedonLSTM.Inaddition，itspredictionerorshowsgeograpicaldiferences（Thepredictionerorsofthethree stationsinBeijing，Haikou，andLhasaincreaseinsequence.）ndseasonalchanges（highinsummerandlowinwinter）hichis highlyconsistentwiththephysical mechanismof theionosphere.Theexperimentsverifiedthefeasibilityofthemodelbasedon Transformer in the prediction of ionospheric fbEs parameters.

KeyWords：ionospheric Es layer;shielding frequency;deep learning; model basedon Transformer;encoder;decoder

0 引言

电离层作为地球高层大气的重要组成部分，其动态变化对无线电波传播、卫星通信与导航及超视距雷达等电子信息系统性能具有关键影响。（剩余8616字）