文档介绍:第10卷 第2期 导航定位学报 ,No. 2
2022 年 4 月 Journalng and Design Group Co. Ltd., Nanchang 330029, China;
3. School of Geomatics, Liaoning Technical University, Fuxin, Liaoning 123000, China)
Abstract:In order to solve the problem that the fitting effect of Long-term and Short-term Memory Network (LSTM)
on the orbit prediction error of BeiDou navigation satellite System (BDS) predicted by dynamic model is not good, a BDS
orbit prediction error prediction algorithm based on the combination of Complementary Ensemble Empirical Mode
Decomposition and Time Convolution Network (CEEMD-TCN) is designed. Firstly, the BDS orbit prediction error sequence
is denoised by using the Complementary Set Empirical Mode Decomposition (CEEMD), then the denoised signal is
reconstructed, and finally the reconstructed orbit error sequence is used as the data training set to train the Time Convolution
Network (TCN). Taking the orbit error series of GEostationary Orbit (GEO), Inclined GeoSynchronous Orbit (IGSO) and
Medium Earth Orbit (MEO) as the research object, the short-, medium-, and long-term prediction of X, Y, and Z directions
of satellite orbit prediction error is carried out by using CEEMD-TCN, TCN, CEEMD-LSTM, and LSTM compensation
models respectively. The experimental resu