GPS Satellite Clock Bias Prediction Based on Metabolic Grey Model

To address the issue of accuracy degradation caused by error accumulation over time in satellite clock bias forecasting using the grey model, a metabolic grey model is proposed. This model continuously updates satellite clock bias data sequences by removing outdated data, maintaining the system in a state of constant renewal to enhance forecasting accuracy. Forecasting trials were carried out utilizing satellite clock bias data with a 30-second sampling frequency, provided by the GNSS Analysis Center at Wuhan University. Forecasting approaches such as the linear polynomial model, quadratic polynomial model, grey model, and metabolic grey model were employed to perform 6-hour-ahead predictions, with actual clock bias data used as the reference benchmark for validation. Experimental results demonstrate that the metabolic grey model achieves significantly improved forecasting accuracy and stability. Achieving an average 6-hour prediction accuracy and stability of 0.17 ns and 0.32 ns, respectively, the proposed model demonstrates significant improvements over the linear polynomial, quadratic polynomial, and conventional grey models. Specifically, the average prediction accuracy is enhanced by 50.00%, 83.67%, and 39.29%, while prediction stability is improved by 41.82%, 83.51%, and 28.89% compared to these models.