自回归神经网络的电离层总电子含量预报  

TEC prediction of ionosphere based on nonlinear auto regressive model

作　　者：吉长东[1] 王强 沈祎凡 潘飞 JI Changdong;WANG Qiang;SHEN Yifan;PAN Fei(School of Geomatics,Liaoning Technical University,Fuxin,Liaoning 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：《导航定位学报》2018年第4期96-101,共6页Journal of Navigation and Positioning

摘　　要：为了进一步提高TEC的预报精度,针对TEC时间序列高噪声、非平稳、包含线性和非线性动态序列的特性,运用经验模态分解和非线性自回归动态神经网络,基于分解-预测-重构的思想构建EMD-NAR预测模型;并对比分析EMD-NAR组合模型和单一模型的预报精度,同时运用EMD-NAR预测模型分析不同环境下的电离层TEC时间序列。实验结果表明EMD-NAR动态神经网络模型能很好地反映电离层TEC的变化特性,平静期和活跃期的预测平均相对精度分别为94%和88.3%,预报残差小于1个TECu的分别占71%和68.5%,小于3个TECu的分别占90.3%和87.5%。In order to further improve the predition accuracy of TEC, in view of the characteristics of highly noisy, non stationary, and containing linear and nonlinear dynamic sequences for TEC time series, the paper used the method of empirical mode decomposition（EMD）and the model of nonlinear auto regressive（NAR）dynamic neural network to establish the EMD-NAR model based on the thought of decomposing predicting-rebuilding, and contrasted the prediction accuracy between EMD-NAR model with the single model. Finally the predition model was used to analyze the ionospheric TEC time series under different enviornments. Experimental result showed that the proposed model could reflect the change of TEC well, with the average relative prediction accuracy 94% in quiet period and 88.3% in active period respectively, and 71% and 68.5% of the prediction residuals in the two periods respectively could be less than 1 TECu, while 90.3% and 87.5% respectively less than 3 TECu.

关 键 词：非线性自回归神经网络 电离层预报 时间序列 经验模态分解 总电子含量 

分 类 号：P228[天文地球—大地测量学与测量工程]