An adaptive robust strategy based on hypothesis testing for satellite clock bias short-term forecasting

Abstract

The study of short-term forecasting of satellite clock bias is important to promote the development of real-time precise point positioning, and the Kalman filter model has certain advantages over other models in the single forecast model. However, the filtering performance will be degraded by the model bias, and a fading factor is constructed to control the effect brought by the model bias. Considering that the fading factor may cause the covariance matrix to be abnormally inflated or even invalid when the observed information is unreliable, the chi-square test is introduced as a test indicator. At the current epoch moment, the fading factor or robust estimation is conditionally applied based on the result of the hypothesis test. An improved IGGIII weight function is used by the robust estimation, which simplifies the solution of the parametric robust solution. BDS-3 satellite clock bias data with different clock types were selected for the analysis of fitting accuracy and forecasting accuracy, and the experimental results showed that the proposed algorithm can well eliminate the anomalous perturbation phenomenon caused by model deviation and improve the fitting accuracy. The comparison of forecast accuracy between the proposed strategy and GM (1,1), QP, and ARIMA models at 3, 6, and 12-hour intervals reveals that the proposed strategy consistently demonstrates optimal forecast accuracy.