Temperature Prediction of Satellite Flywheel Based on LightGBM

Abstract

In order to ensure the stable operation of satellites, it is important for the ground system to monitor and predict the satellite state, especially the monitoring of flywheel temperature. As an important component of attitude control system of a satellite, the temperature of flywheel is important to identify the state of the system. The prediction of flywheel temperature is of great significance to the stable operation of satellites in orbit. In this paper, based on the LightGBM machine learning framework, a gradient boosting decision tree model is established by using spatial environmental data and in-orbit telemetry data of a satellite, to predict the temperature change of satellite flywheel. By comparing with the actual flywheel temperature, the prediction accuracy can meet the monitoring requirement of satellite flywheel temperature. This model can be applied to warn the ground system the possible temperature anomalies of attitude control system, so that controllers can avoid risks ahead of time and ensure the safe operation of satellites. The research results have certain universality for other satellite flywheel systems.