The Application Of Hybrid Method Based On Functional Network For Navigation Satellite Clock Error Prediction

As we all know, time is very important in our life, and the atomic clock is a main part of thenavigation satellite. Establishing accurate clock error prediction model has considerably practicalsignificance for the navigation constellation. This paper emphasizes on the method which can improvethe accuracy of navigation satellite clock error prediction. The main contents and innovations aresummarized as follows:1. In this paper, the main technical standards of the atomic clock are summarized, especially thefrequency stability of a clock. It can be represented both in time-domain and frequency-domain. Inthis paper, the frequency stability of Rb clocks, Cs clocks and H clocks are all calculated by themodified Allan variance, which is a time-domain method. And the characteristics of the output data ofthe clock is also analysed.2. Four kinds of traditional methods to predict the clock error are presented in this paper. Theyare polynomial model, gray model, Kalman filtering model and ARIMA model. Both the GPS satelliteclock error released by IGS and the laboratory clock error measured are used for examples to showhow to be predicted by the four methods. Also, in connection with the prediction results, a briefanalysis is given out.3.The basic theories of the functional network are introduced in detail, which includes the basicelements, the functional network topology, and the learning process. A hybrid model to predict theclock error based on functional networks is proposed in this paper. The method is firstly fitting thedeterministic part of the clock error by polynomial model. Then the periodic term of the clock error ismodeled by functional network joined with spectrum analysis, and finally the stochastic term is alsomodeled by functional network combined with the chaos theory phase reconstruction. Some GPS andlaboratory clock error data are used to test and verify that the method is very useful.