Effective angular momentum combined with EWT-LSTM to improve ERP medium- and long-term prediction accuracy

High-precision prediction of Earth Rotation Parameters (ERP) is the basis for spacecraft orbit determination and deep-space exploration. The accuracy of conventional forecasting methods decreases rapidly over time. Research shows that considering the influence of Effective Angular Momentum (EAM) in ERP forecasting can significantly improve forecasting accuracy. The Empirical Wavelet Transform (EWT) can split original data into subsequences with different frequencies before feeding them into the Long Short-Term Memory (LSTM), thereby enhancing LSTM performance on time series. Based on EOP C04 (Combined Series of Earth Orientation Parameters provided by IERS) data and EAM, ERP were predicted using EWT-LSTM. ERP forecasting rounds were conducted using historical data to verify the algorithm's performance. The proposed method has a notable effect on medium- and long-term forecasts. Compared with Bulletin A of IERS (International Earth Rotation and Reference Systems Service), the PMX (Polar Motion in X Direction), PMY (Polar Motion in Y Direction), and UT1-UTC (Universal Time1 minus Coordinated Universal Time) forecasted on the 100th day were 6.51 mas, 3.88 mas, and 4.87 ms, representing improvements by 20.51%, 3.96%, and 10.64%, respectively. On day 365, the forecasted PMX, PMY, and UT1-UTC were 12.17 mas, 9.65 mas, and 21.18 ms, respectively, which are improved by 25.32%, 34.88%, and 44.02% respectively. LSTM has advantages in medium- and long-term forecasting of ERP. Incorporating EAM information with the EWT classified sequence can improve the forecasting effect.