Abstract The quadratic polynomial(QP) model in satellite clock offset prediction is susceptible to noise interference and the prediction accuracy is not high, so we construct a multi-GNSS satellite clock offset prediction model based on long short-term memory neural network. We analyze the model accuracy of different satellite systems and different clock types based on different modeling schemes. To verify the validity and feasibility of the model, we use the LSTM, QP, and QP-LSTM models to forecast the clock offset with 1 h, 3 h, 6 h, and 12 h based on 12 h and 24 h clock offset series, respectively. The results show that the LSTM model has the highest accuracy with 24 h modeling data for 1 h prediction. In the LSTM prediction model of multi-GNSS satellite clock offset, the Galileo system has the highest accuracy, followed by the BDS-2 system and GPS system, and the GLONASS system has the lowest accuracy, with STD of 0.018 ns, 0.069 ns, 0.133 ns, and 0.242 ns, respectively. The prediction accuracy of different types of atomic clocks varies. The prediction accuracy of hydrogen atomic clock is better than Rb atomic clock and Cs atomic clock. The prediction accuracy of LSTM neural network model is improved by 27% compared to QP-LSTM model and 36% compared to QP model.
JIANG Chunhua,ZHU Meizhen,XUE Huijie et al. Multi-GNSS Satellite Clock Offset Prediction Model Based on Long Short-Term Memory Neural Network[J]. jgg, 2024, 44(3): 257-262.
JIANG Chunhua,ZHU Meizhen,XUE Huijie et al. Multi-GNSS Satellite Clock Offset Prediction Model Based on Long Short-Term Memory Neural Network[J]. jgg, 2024, 44(3): 257-262.