Research On Prediction Of GPS Satellite Clock Error With IGS Ultra-Rapid Products

Satellite atomic clocks is a core part of the onboard navigation system, atomic clock is the time reference of navigation’s signal forming and distance ranging, its performance directly determines the positioning accuracy and time transfer accuracy. For this reason, in order to realize the precise navigation, satellite clock model is very important content. In this thesis, we analyze the ultra-rapid satellite clock error products and then predict satellite clock error. The main contents are as follows:(1) Introduction of IGS organization, including its organizational structure and its internal department functions. Introduction IGS all kinds of products and give an explanation of the accuracy and effectiveness. Briefly introduces the satellite clock stability evaluation criteria, including frequency accuracy, frequency drift rate and frequency stability. Giving an overview of the current four major satellite navigation systems and their atomic clocks stability.(2) Compared four ephemeris accuracy and effectiveness:balancing accuracy and time delay, this thesis choose ultra-rapid ephemeris clock error data to study. Use the former 24h of the data to model, forecasting the last 24 hours clock error data, to achieve real-time satellite clock error prediction.(3) Make a detailed description of linear model, quadratic model, gray model and ARIMA model. From the results of the 24-hour forecast, new satellite clock error prediction accuracy is better than the old satellites. IGU existing prediction model of ultra-rapid ephemeris is no longer fit for the new type Block ⅡF satellite and Block ⅡR-M satellites. We need to establish a new forecasting model, and the new forecasting model should be based on satellite type and stability.(4) Introducing the basic theory of wavelet analysis and the wavelet multi-scale analysis theory, then use wavelet multi-scale to analyze satellite clock error series. We decompose satellite clock error fit residuals in four layers using DB5 wavelet. Seven satellite clock errors are selected for analysis, according to the results of the analysis:This thesis extracts significant period items of 2cycle/day and 4cycle/day. IN the A4 and D4 layer, the periodic signal of Block ⅡF satellites 25 and Block ⅡA satellites 32 are particularly significant. The harmonic cycle in D4 layer is presented in all types of satellite clock, indicating that the cycle item of 4 cycles/day existed in every satellite. The harmonic cycle in A4 layer is a major impact on the satellite clock, and then the harmonics in the D4 layer is secondary effect. Harmonic amplitude of Block IIF satellite is significantly less than the other three types of satellites’harmonic amplitude. This indicates that the new satellite is more stable and its clock error time series harmonic amplitudes are smaller, whereas the old satellites is not stable, its clock time series harmonics amplitudes is bigger than the old ones.(5) First we introduce the decomposition principles and procedures of satellite clock error modeling. Making an explanation of how to get trend item from clock error series, how to select wavelet function, how many of levels will be decomposed. Taking into account the harmonic cycle of satellite clock, each layer chose a different model fitting function for prediction. We compare the forecast results with IGU clock error, the wavelet multi-scale model prediction accuracy is better than ultra-rapid ephemeris prediction clock error, wavelet multi-scale model accuracy is improved by 18% in 6h prediction,16% in 12h prediction,31% in 24h prediction than the IGU prediction.