Abstract:Based on the 15-day precise satellite clock bias (SCB) data from different aspects, we comprehensively analyze the prediction effect before and after using the SCB’s single difference prediction principle for six common prediction models, including the linear polynomial (LP), the quadratic polynomial (QP), the grey (GM), the spectrum analysis (SA), the time series ARIMA (ARIMA), and the Kalman filtering (KF) models. Accordingly, the following conclusions are drawn: 1) Using the principle can improve prediction accuracy. Specifically, they are the LP model, SA model, GM model and KF model used in the 3 h prediction for GPS SCB, the QP model and ARIMA model used in 3 h prediction for ⅡF Rb clock SCB, the LP model and GM model used in 6 h and 12 h prediction, the ARIMA model used in 6 h, 12 h and 24 h prediction. 2) The prediction results of the principle are related to the types of satellites and their onboard clocks, and for GPS BLOCK ⅡF Rb clocks, based on the prediction principle, the short-term prediction accuracy of the six models can be improved, especially for GM, LP and ARIMA models. 3) For the prediction of 3 h and 6 h, the RMS value corresponding to the DLP model (the LP model after using the principle) is the smallest, that is, the prediction accuracy of the DLP model is the highest, indicating that the single difference data of SCB is more in line with the short-term prediction of the LP model.
WANG Yupu,XUE Shenhui,WANG Wei et al. Effect Analysis of Common Prediction Models Specific to Satellite Clock Bias Based on the Principle of Single Difference Prediction[J]. jgg, 2021, 41(4): 336-341.