BDS精密钟差短期预报  

Short-term forecast of precision clock difference for BDS

作　　者：王建敏[1] 毕祥鑫 黄佳鹏[1] WANG Jianmin;BI Xiangxin;HUANG Jiapeng(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》2023年第1期30-38,共9页Journal of Navigation and Positioning

基　　金：国家自然科学基金项目(41474020)。

摘　　要：针对传统单一预报模型在钟差预报中误差积累随时间的增加而增大问题,提出一种灰度模型GM(1,1)与长短时记忆神经网络模型(LSTM)的组合模型:采用武汉大学国际全球卫星导航系统服务组织(IGS)数据中心下载的北斗卫星导航系统(BDS)3种轨道不同卫星连续2 d的精密钟差数据进行建模,首先用GM(1,1)模型进行预报,然后将GM(1,1)模型的残差利用LSTM神经网络模型进行再次预报;将2种模型的预报结果进行重构,得到最终的预报结果。实验结果表明:GM(1,1)/LSTM组合模型与单一GM(1,1)模型相比,精度提高了60%~89%;GM(1,1)/LSTM组合模型与单一LSTM神经网络相比,精度提升了30%~88%。Aiming at the problem that the accumulation of errors increases with the increase of time for the traditional single forecast model during the clock error prediction,the paper proposed a combination model of gray model GM(1,1)and long-short term mermory network(LSTM):the precision clock difference data of two consecutive days from three different satellite orbits of BeiDou navigation satellite system(BDS)downloaded by IGS(International Global Navigation Satellite Systems Service)Data Center at Wuhan University were modelled,GM(1,1)model was used to forecast firstly,and then the residual of GM(1,1)model was forecast again by LSTM neural network model;the forecast results by two models were reconstructed to obtain the final forecast results.Experimental result showed that the accuracy of GM(1,1)/LSTM combined model would be 60%~89%higher than that of single GM(1,1)model,and be 30%~88%higher than that of single LSTM neural network model.

关 键 词：钟差预报 灰度模型(GM(1 1)) 长短时记忆神经网络模型(LSTM) 组合模型 

分 类 号：P228[天文地球—大地测量学与测量工程]