Research on the Influence of Prediction Model and Modeling Sequence Length on Short-term Prediction Accuracy of Clock Bias

Aiming at the problem that different types of GPS space-borne atomic clock shave different adaptability to different prediction models in clock difference prediction, QP model, GM(1,1) model and GM(1,1)+AR model are used to predict the clock bias of different types of atomic clocks. The analysis focuses on the prediction accuracy of different types of atomic clocks, the effect of clock bias series with different lengths and the effect of clock sequence fluctuations on model establishing and forecasting. The results show that the accuracy of the clock bias prediction has a certain relationship with the length of the modeling series. The QP model is most affected, and the GM(1,1)+AR model is least affected. The optimal modeling sequence length for different types of atomic clocks is different under different prediction models. The rubidium clock is less affected by the length of the modeling series than the cesium clock. QP model has poor prediction effect on cesium clock, and the prediction effect on rubidium clock is comparable to GM (1,1) model and GM (1,1)+AR model. When the clock bias series fluctuates, the accuracy of modeling prediction decreases, and the prediction results of different models are affected differently by the amplitude of the clock difference fluctuation.