Study On Performance Analysis And Modeling And Forecasting Method Of BD-2 Terminal Satellite Clock

The Beidou Navigation Satellite System(BD2)started the construction of the Beidou Navigation Satellite System in 2004.After eight years of research and development,it was successfully put into operation on December 27,2012,providing users in the Asia-Pacific region with regional navigation,positioning and timing services.As the key equipment of satellite navigation system,spaceborne atomic clock provides high precision time reference for satellites.With BD2 operating smoothly and accurately for nearly a decade,some satellites in orbit have reached or are approaching their design life.It is a research subject of theoretical significance and practical value to carry out the corresponding clock performance analysis and clock error prediction for the satellite entering the end of life,and it is of great significance to optimize the service performance of the Beidou system.Based on this,this paper focuses on studying the performance of the spaceborne atomic clock at the end of the life of BD2 satellite and establishing a clock difference prediction model that can adapt to the operating state of the atomic clock at the end of the life of BD2 satellite.The main achievements of this paper are as follows:1.A clock data preprocessing method based on wavelet analysis is proposed.In this method,the original clock data is decomposed by wavelet,the decomposed low frequency and high frequency wavelet coefficients are processed by certain criteria,and then the processed clock data is obtained by reconstruction.In this paper,the effects of different wavelet functions and different decomposition levels on the preprocessing effect are analyzed emphatically,and the most suitable wavelet combination for Beidou clock data processing is selected.2.A more comprehensive clock data quality assessment method is designed.The quality of Beidou satellite clock data is evaluated from four aspects of data continuity,internal coincidence accuracy,external coincidence accuracy and data cycle characteristics.From the results,the Beidou satellite clock data has a high accuracy,the accuracy of one hour span is better than 0.1ns.At the same time,there are significant period characteristics in the clock data,the main period of GEO and IGSO satellite is 12 hours and 24 hours,the main period of MEO satellite is 6 hours and 12 hours.The period term of C11 satellite is obviously different from that of other satellites.3.Using three years of satellite clock difference data,the long-term performance of the airborne atomic clock of the BD2 satellite in orbit is analyzed.From the results,the overall frequency accuracy of the BDS satellite atomic clock is at the magnitude of 10^-11,the frequency drift rate is basically at the magnitude of10^-13/d,and the frequency stability is basically at the magnitude of 10^-14.However,the performance of some satellites fluctuates.Combined with the on-orbit operation time,it is preliminarily judged that the fluctuation of clock performance is related to the proximity of the design life of the satellite.4.Introduced the principle of using primary satellite clock bias data for clock error prediction,and compared the prediction accuracy of several commonly used clock error prediction models using primary satellite clock bias data and phase data.The results show that the accuracy of clock error prediction can be improved obviously by using the single difference data.5.The BP neural network model based on primary satellite clock bias data was studied,and the accuracy of satellite clock error prediction was improved by constructing the BP neural network suitable for satellite clock error data.The experiment shows that the BP neural network can overcome the influence of frequency change effectively when the satellite performance fluctuates.However,the model itself also has some limitations.6.Study the LSTM neural network model based on the primary satellite clock bias data,and build the LSTM neural network suitable for the satellite clock bias data.Through comparative experiments,the results show that the LSTM neural network can overcome the influence of frequency variation and has a good accuracy of clock error prediction when the satellite performance fluctuates.