Motion interval prediction of a sea satellite launch platform based on VMD-QR-GRU

Abstract

Accurately forecasting the motion of a sea satellite launch platform is crucial for decision support in selecting the optimal launch window and ensuring safety during sea launches. Hence, this paper proposes a novel hybrid prediction model that integrates gated recurrent units (GRU), variational mode decomposition (VMD), and quantile regression (QR) to predict the motion interval of the launch platform. The VMD algorithm effectively captures the multi-scale features of the data, while the QR-GRU model excels in handling nonlinear time series and providing enriched information. Combining the VMD, QR, and GRU leverages their respective strengths, resulting in enhanced prediction accuracy and stability. The experimental data were derived from field measurements of a satellite launch platform during its sea voyage and launch process. By comparing the VMD-QR-GRU model with other models, it was found that the comprehensive metrics for interval prediction of roll and pitch reduced by over 1.4% and 15.7%, respectively. Additional analysis was conducted to evaluate the robustness and transferability of the model, considering various scenarios such as different training data volumes, varying prediction durations, and different states of platform. The VMD-QR-GRU model consistently yielded satisfactory results, underscoring its effectiveness and reliability in predicting sea satellite launch platforms motion.