The Short-Term Forecast of BeiDou Satellite Clock Bias Based on Wavelet Neural Network

According to nonlinear and nonstationary characteristics of BeiDou satellite clock bias time series, this paper proposed a method using the wavelet neural network (WNN) based on the first-order difference of adjacent epoch to predict the satellite clock bias. Experimental data with sampling interval of 15 min rapid and ultra-rapid satellite clock bias provided by Wuhan University is used to test the validation of the method. The results show that the forecast precision of 6 h for BeiDou satellite can reach 1-2 ns, and the 24 h can reach 2-4.6 ns using the proposed method. The test results also show that the accuracy and stability of the model prediction can be improved greatly using the proposed method compared to the traditional gray model and quadratic polynomial model.