作者:Huijie Xue¹,Tianhe Xu¹,Wenfeng Nie¹,Yuguo Yang¹ and Qingsong Ai²

Abstract：The prediction accuracy of clock offset is very important for real-time precise point positioning (PPP). By considering the influence of the gross error, periodic error and uncertainty error on model fitting, an enhanced prediction model for BeiDou Navigation Satellite System (BDS) ultra-rapid clock offset by combining robust estimation, singular spectrum analysis and gray model is proposed. Firstly, the singular spectrum analysis is used to decompose the clock offset sequence into two parts: a certain part and an uncertain part. Secondly, the robust quadratic polynomial model with additional period terms is adopted to model the certain part. Thirdly, the fitted residual from the second step and the uncertain part from first step are combined and modeled by the robust gray model. Finally, the prediction parts through the robust quadratic polynomial model with additional period terms and the robust gray model are added together with the initial deviation corrected to get the final prediction value. The proposed enhanced model is verified by the ultra-rapid clock bias products from international Global Navigation Satellite System Monitoring and Assessment Service (iGMAS) by taking the final clock bias products from iGMAS as reference. The results show that the proposed model is able to improve the prediction accuracy by 6.7%, 19.5%, 31.7%, and 42.2% over the iGMAS ultra-rapid prediction (ISU-P) products at 3h, 6h, 12h, and 24h prediction span respectively, when using clock bias products of two-day arc for modeling.