Research On Modeling And Forecast Method Of The LSTM Satellite Clock Bias Forecast Model

The time system is the basis of the Global Navigation Satellite System(GNSS).The accuracy and performance of the time system will directly determine the accuracy of navigation and positioning.The satellite clock bias(SCB)is the difference between the system time of the satellite navigation system and the measured face value of the satellite-borne atomic clock.The related research on the satellite clock bias plays an important role in the fields of navigation message parameters optimization,real-time precision single-point positioning and satellite autonomous navigation.The establishment of modelling and forecast method of SCB are an important part of satellite clock bias forecasting.It is of great research significance to establish a stable and reliable SCB forecast model and compare the forecast results and characteristics of different models in SCB forecasting.Based on this,this paper has researched on the model establishment and forecast results of SCB forecasting.The main achievements and innovations of the paper arc summarized as follows:(1)Based on the Long Short-Term Memory network(LSTM),a new SCB forecast model is proposed:LSTM satellite clock bias forecast model.Through theoretical analysis and experimental comparison,the structure and parameters of the LSTM satellite clock bias forecast model are determined,and a reasonable and effective LSTM satellite clock bias forecast model is constructed.(2)Based on the original SCB sequence and the one-time difference SCB sequence,the LSTM satellite clock bias forecast model was used to make short-term and mid-long-term forecast of SCB,and the forecasting effect of the model when using the original SCB sequence and the one-time difference SCB sequence is compared.It is believed that the LSTM satellite clock bias forecast model can have better forecasting effects when using one-time difference data.Based on this conclusion and the established LSTM satellite clock forecast model,the method and process of using the LSTM satellite clock bias forecast model for SCB forecasting research are determined.(3)Based on different forecasting strategies,SCB forecasting experiments were conducted using QP model,GM model,BP model,WNN model and the established LSTM model,and the specific forecasting effects of each model were compared to verify the LSTM satellite clock bias forecast model is effective and reliable.In short-term forecasting,the LSTM satellite clock bias forecast model has achieved ideal forecast results.Its forecast accuracy and forecast stability are similar to other commonly used models and can be applied to satellite clock forecast.In mid-long-term forecasting,regardless of forecast accuracy or forecast stability,the overall effect of the LSTM model in the four forecasting schemes is superior to the other four models.In the 7-day forecast,the average forecast accuracy of the LSTM model is improved by 65.9%than the accuracy of the optimal model,and the forecast stability is improved by 35.6%than the stability of the optimal model.In the 30-day forecast,the average forecast accuracy of the LSTM model is improved by 24.8%compared with the optimal model,and the forecast stability is improved by 19.6%compared to the optimal model.