自回归移动平均模型的电离层总电子含量短期预报  

Short-term TEC Prediction of Ionosphere Based on ARIMA Model

作　　者：张小红[1] 任晓东[1] 吴风波[1] 芦琪[1] 

机构地区：[1]武汉大学测绘学院,湖北武汉430079

出　　处：《测绘学报》2014年第2期118-124,共7页Acta Geodaetica et Cartographica Sinica

摘　　要：在充分考虑乘积性季节模型的情况下,采用时间序列分析中的求和自回归移动平均模型(autoregressive integrated moving average,ARIMA)对TEC值序列进行预报分析。以欧洲定轨道中心(CODE)提供的2008—2012年电离层TEC值为样本数据,重点分析该方法在不同电离层环境(电离层平静期和活跃期)和不同纬度下的预报性能以及影响该方法预报精度的因素分析等。试验结果表明:在预报精度方面,电离层平静期和活跃期预报6d的平均相对精度可达83.3%和86.6%;而平均预报残差分别为0.18±1.9TECU和0.69±2.6TECU,其中预报残差小于3TECU分别达到90%和81%以上;而且两个时期都具有纬度越高相对精度越低而绝对精度越高的规律。在影响因素方面,预报精度会随TEC样本序列长度增加而提高,但随着样本序列增加到一定值(约30d左右)后,其相对精度提高不大;而相同样本数据的预报精度则会随预报长度的增加而减小,初期并不明显,但超过30d其相对精度将随时间明显降低。With a fully consideration of the multiplicative seasonal model,seasonal time series for ionspheric total electron content （TEC）was transformed into a stationary time series by seasonal differences and regular differences firstly, then use autoregressive integrated moving average （ARIMA） model from the time series analysis theory to model the stationary TEC values so as to predict the TI：C series. Using the TEC data from 2008 to 2012 provided by the Center for Orbit Determination in Europe （CODE） as sample data, the precision of this method in the prediction of ionosphere TEC value was analysed which varies from high latitude to low latitude in both quiet and active period. The effect of TEC sample＇s length on the predicted accuracy is analyzed, too. Results from numerical experiments show that in ionospheric quiet period the average relative prediction accuracy of 6 days are up to 83.3% with an average prediction residual errors of about 0.18_：I.gTECu, while it changes to 86.6% with an average prediction residual errors of about 0.69±2.6TECu in ionospheric active period. For the former, above 90% of predicted residual is less than ±3TECu while the latter is only about 81%. Two periods show the same law that the higher the latitude, the higher the absolute precision, and the lower the predicted relative accuracy. In addition, the prediction accuracy will improve with the increase of TEC sample sequences length, but it will gradually reduce if the length exceed the optimal length about 40 days. On the other hand, with the same TEC sample, the predicted days increase, the predicted accuracy decreases. Though it＇s not obvious in the beginning, it will be significantly reduced over 30 days.

关 键 词：自回归移动平均模型 电离层预报 时间序列 预报精度 总电子含量 

分 类 号：P228[天文地球—大地测量学与测量工程]