Study On The GPS/BDS Atomic Clock Performance And Clock Offset Prediction Model

That atomic frequency standard and time system being of high accuracy and high stability is a key to ensure high-precision navigation and positioning. Atomic clock is the kernel part of satellite navigation systems payload. Its performance directly determines the precision of navigation and time-frequency transmission. Establishing precise prediction model of satellite clock offset plays an vital role in navigation and positioning. Therefore, The time and frequency theory of atomic clock is taken as the basis. For a number of issues existing in performance of onboard atomic clocks and prediction model of satellite clock offset, the precise clock product is processed and analyzed. The researches on the spaceborne atomic clocks of developing BeiDou System(BDS) and modernizing GPS are done. Additionally, the performance level of different satellite clock and the prediction effect of models are compared. All of thses works are aimed at providing some useful reference value for the construction and application of BeiDou Navigation System. The main work of this thesis is as follows:(1)The performance of each satellite clock of GPS and BeiDou System is analyzed with precise clock product. The main performance index include frequency accuracy and day drift rate, stability and noise type. The classification analysis results suggest that the same type of satellite clock has similar performance level. BeiDou satellite clock performance analysis results show that the absolute value of the average of frequency accuracy is 3.25×10-11, the day drift rate is 1.37×10-13,the stability is 1.82×10-13 per 1000 s, 1.09×10-13 per 10000 s and 1.11×10-13 per day. All BeiDou spaceborne atomic clocks are an order of magnitude lower than GPS in frequency accuracy. The day drift of BDS MEO Rb clocks, an order of 14-10, is almost equal to the majority of GPS Rb atomic clocks, which is lower than BDS GEO and IGSO Rb clock. The day stability of BDS MEO Rb clock is an order of 14-10, which is equal to the majority of GPS atomic clocks, and of BDS GEO and IGSO Rb clocks is an order of 13-10.(2)Using a conventional prediction models and nonlinear models to do research on short-term(1d) and long-term(7d and 30d) clock prediction. The characteristics and forecast effect of each single model are analyzed and summarized from many aspects such as the type of satellite clock, the length of modeling data and of prediction. As a consequence of data changing abnormally, the prediction accuracy of GRNN(Generalized Regression Neural Network) model and RBF(Radial Basis Function) model is very unstable. Compared with these models including spectrum analysis model, gray model and time series models, polynomial model is the most appropriate model for G27 satellite. Using one day data to model polynomial model, the accuracy of 7-day forecast is 5.96 ns and of 30-day is 116.26 ns.(3)Polynomial model, gray model, time series model and GRNN are combined with classic weight and modified classic weight method to forecast. Comprehensive comparison is made between integrated forecasting model and each single model. The results show that the linear combination prediction model is easy to use, and it can improve the predicting reliability in the premise of ensuring the accuracy of the clock offset prediction.(4)Correlation analysis is done between main performance index of satellite clock and short-term forecast accuracy. The correlation level between each index with short-term(1d) prediction accuracy is as follow: the day stability is highly correlated, the frequency accuracy is low correlated and the day drift rate is micro-related to it.