摘要
为了提高基于ARIMA模型的地球自转参数预报精度,着重分析了ERP数据序列预处理中周期项成分的精确获取技术、ARMA模型阶次的优化选择问题,以及ERP数据序列不同差分次数对模型精度的影响。利用国际地球自转和参考系统服务组织公布的ERP最终产品,365组预报结果表明,UT1-UTC预报30d精度可达3.398 6ms,10d以内预报应选取2次差分模型,10d以上选取1次差分模型;日长预报30d精度可达0.337 5ms,应选取零次差分预报模型;Xp预报30d精度为11.189 1mas,Yp预报30d精度为7.932 1mas,两者均选择1次差分模型。
Some critical problems are successfully resolved to improve the Earth rotation parameters(ERP)prediction accuracy based on the auto-regressive integrated moving average(ARIMA)model,which are the precise extraction of periodic components in the ERP,the order optimization of ARIMA model and the effects of model order on the ERP prediction accuracy.The prediction experiments are implemented on the final ERP products released by the International Earth Rotation and Reference Systems Service(IERRSS).Some key conclusions can be drawn from the total 365 groups of prediction results.Firstly,the 30-day prediction error of UT1-UTC is 3.398 6ms.The two-order difference model can be selected for the prediction in less than 10 d,but the one-order difference model is selected for the prediction in more than 10 d.Secondly,the 30-day prediction error of length of day(LOD)is 0.337 5 ms and the difference model order is zero.Finally,the 30-day prediction error is11.189 1mas and 7.932 1mas for the x-axis and y-axis component of polar motion respectively.Additionally,the difference model order is one for the above two cases.
出处
《导航定位学报》
2016年第1期68-74,87,共8页
Journal of Navigation and Positioning
关键词
差分自回归滑动平均模型
地球自转参数预报
差分
模型阶次
优化
auto-regressive integrated moving average model
Earth rotation parameters prediction
difference
model order
optimization