Low-Earth Orbital Lifetime Prediction Based on Parameter Sensitivity Analysis and Deep Learning

Abstract

This article proposes an intelligent prediction method for the orbital lifetime of resident space objects (RSOs) in low-Earth orbit. This method is intended to satisfy the computational efficiency and accuracy requirements when predicting the large-scale RSOs. It helps to analyze the distribution and uncertainty of RSOs’ lifetime. First, based on the semianalytical orbit prediction model, a parameter sensitivity analysis method for the orbital lifetime is proposed, and the parameters that affect the orbital lifetime are investigated. Second, using a fully connected neural network, an orbital lifetime prediction model is established based on deep learning. The hyperparameters of the deep-learning model are optimized using the black-winged kite algorithm. Finally, by combining the sensitivity analysis with the deep-learning model, an intelligent prediction method is proposed for the rapid prediction of orbital lifetime. Simulation results show that the proposed intelligent prediction method accurately determines the orbital lifetime of RSOs and has higher prediction efficiency than the traditional prediction method. The research results provide an effective calculation and analysis tool for predicting the orbital lifetime of large-scale RSOs in the future.