CEEMD与GRNN神经网络电离层TEC预报模型  

Ionospheric TEC forecast model of based on CEEMD and GRNN

作　　者：高清文 赵国忱[1] GAO Qingwen;ZHAO Guochen(School of Geomatics,Liaoning Technical University,Fuxin 123000,China)

机构地区：[1]辽宁工程技术大学测绘与地理科学学院,辽宁阜新123000

出　　处：《全球定位系统》2021年第4期76-84,共9页Gnss World of China

摘　　要：针对电离层电子总含量(TEC)存在非线性、非平稳,由多因素影响导致高噪声的问题,建立了补充集合经验模态分解(CEEMD)和广义回归神经网络(GRNN)模型相结合的CEEMDGRNN电离层TEC预报模型.以解决直接使用原始数据进行预测会导致拟合效果差、预测精度低的问题.采用国际GNSS服务(IGS)中心提供的2019年电离层数据对高、中、低纬度磁暴和非磁暴的不同年积日数据进行实验,低纬处均方根误差(RMSE)最优可达到0.97 TECU,相对精度为91.28,验证了CEEMD-GRNN预报模型精度高于EMD-GRNN以及单一的GRNN模型.Aiming at the problem of non-linear and non-stationary electrons in the ionospheric total electron content(TEC),and high noise caused by multiple factors,a CEEMD-GRNN ionospheric TEC prediction model combining the complementing ensemble empirical mode decomposition(CEEMD)and generalized regression neural network(GRNN)to solve the problem of poor fitting and low prediction accuracy caused by direct use of raw data for prediction.The ionospheric data of 2019 from IGS center for high,middle and low latitude are used.Different day of year data with magnetic storms and without magnetic storms are tested.Results show that the RMSE of low latitude is 0.97 and the relative accuracy is 91.28,which verifies that the accuracy of the CEEMD-GRNN forecast model is higher than that of the EMD-GRNN and the single GRNN model.

关 键 词：电离层 电子总含量 补充集合经验模态分解(CEEMD) 广义回归神经网络 预报精度 

分 类 号：P228.4[天文地球—大地测量学与测量工程]