Satellite orbit prediction model based on VMD-RC-LSTM

Orbital prediction is a critical task for ensuring space situational awareness. Satellites, during their orbital operations, are affected by a multitude of complex perturbation forces, leading to positional deviations. Accurately modeling these forces in real-time remains a significant challenge. To enhance the precision of orbital position predictions, a spacecraft orbit time series prediction method integrating Whale Optimization Algorithm (WOA), Variational Mode Decomposition (VMD), Residual Compensation(RC), Long Short-Term (LSTM) Memory (denoted as VMD-RC-LSTM) is proposed. Variational mode decomposition (VMD) algorithm is applied to decompose forecast residuals in the frequency domain, while LSTM compensation model leverages historical satellite orbital parameters to predict satellite position coordinates. Additionally, whale optimization algorithm (WOA) is employed to optimize LSTM model parameters. Experimental results demonstrate that VMD-RC-LSTM model achieves a reduce the root mean square error (RMSE) approximately 98.89%, 86.74%, and 81.25% in the X, Y, and Z axes in the J2000 coordinate system, respectively, compared to WOALSTM model. Furthermore, VMD-RC-LSTM model reduces RMSE in X, Y, and Z axes in the J2000 coordinate system by 44.4%,19.11%, and 61.17%, respectively, when compared to RC-LSTM model. Overall, the prediction and compensation capabilities of VMD-RC-LSTM model significantly outperform both WOA-LSTM and RC-LSTM models.