利用CEEMDAN-ARIMA-BiLSTM模型预报电离层总电子含量  

Ionospheric Total Electron Content Prediction Based on CEEMDAN-ARIMA-BiLSTM Model

作　　者：沈豫 管辉 王杰 孙晨辉 SHEN Yu;GUAN Hui;WANG Jie;SUN Chenhui(Zhoushan Natural Resources Mapping and Design Center,Zhoushan 316021,China;Hangzhou Tiantu Geographic Information Technology Co.,Ltd.,Hangzhou 311100,China;Zhejiang Institute of Surveying and Mapping Science and Technology,Hangzhou 310030,China)

机构地区：[1]舟山市自然资源测绘设计中心,浙江舟山316021 [2]杭州天图地理信息技术有限公司,浙江杭州311100 [3]浙江省测绘科学技术研究院,浙江杭州310030

出　　处：《地理空间信息》2025年第3期92-95,105,共5页Geospatial Information

摘　　要：为了提升电离层总电子含量(TEC)的预报精度,在现有预报模型的基础上提出了一种新的组合预报模型。首先采用自适应噪声完备集合经验模态分解(CEEMDAN)方法分解TEC序列,并进行排列与重组;再分别利用差分自回归移动平均模型(ARIMA)和双向长短期记忆网络(BiLSTM)对高、低频分量进行建模与预报;最后重构不同分量预报结果,得到最终预报值。根据地磁活动情况,分别选取磁平静期和磁暴期的低、高纬度地区电离层TEC序列进行实验,结果表明该模型在磁平静期预报结果的均方根误差为0.61 TECu,比单一BiLSTM、ARIMA模型分别减少了0.11 TECu、0.05 TECu;磁暴期的均方根误差为0.87 TECu,比单一BiLSTM、ARIMA模型分别减少了0.32 TECu、0.18 TECu,验证了该模型的稳定性与优越性。In order to improve the prediction accuracy of ionospheric total electron content(TEC),we proposed a new combined prediction model based on existing prediction models.Firstly,we used the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)method to decompose the TEC sequence,and arranged and reassembled the decomposed components based on the calculated permutation entropy algorithm.Then,we used the differential autoregressive integrated moving average(ARIMA)model to model and predict high-frequency unstable components,and used the bidirectional long short-term memory(BiLSTM)network to model and predict low-frequency components.Finally,we reconstructed the predicted results of different components to obtain the final predicted values.Based on the geomagnetic activity,we selected the low and high latitude ionospheric TEC sequences during the magnetic quiet period and the magnetic storm period for experiments.The results show that the root mean square error(RMSE)of proposed model during the magnetic quiet period is 0.61 TECu,which is 0.11 TECu and 0.05 TECu less than the single BiLSTM model and ARIMA model,respectively.RMSE of the predicted results during the magnetic storm period is 0.87 TECu,which is 0.32 TECu and 0.18 TECu less than the single BiLSTM model and ARIMA model,respectively,verifying the stability and superiority of this model.

关 键 词：电离层 TEC预报 CEEMDAN方法 排列熵 ARIMA模型 BiLSTM模型 

分 类 号：P228[天文地球—大地测量学与测量工程]