A New Data Preprocessing Method for Satellite Clock Bias and Its Application in WNN to Predict Medium-term and Long-term Clock Bias

In order to improve the prediction precision of navigation satellite clock bias in the medium and long term, we design a new prediction model using a wavelet neural network based on a new data preprocessing method, aimed at processing the single difference sequence of satellite clock bias data. Specifically, this model firstly makes difference between two values of adjacent epoch for the given clock bias data, thus obtaining the corresponding single difference sequence, and then uses the proposed preprocessing method to process the sequence, and adopts the preprocessed sequence when modeling a wavelet neural network to predict the following medium- and long-term sequences. Finally,the proposed model restores the predicted sequences to the corresponding prediction clock bias. Using clock bias data from satellite-bone rubidium clocks in GPS, we conducted medium- and long-term prediction tests for the new method, simultaneously comparing it with three common prediction methods ;the quadratic polynomial model, grey model, and Kalman filter model. The results show that the new method can effectively reduce the prediction error in the medium- and long-term satellite clock bias prediction.