ML Driven Thermal Analysis for a Real-Time Control System in Satellites

This paper presents a new method of a faster and precise thermal control in space with an emphasis on the need of real-time prediction in dynamic environments. The total irradiation on sun sensors is computed by considering different environmental factors as well as satellite dynamics within the proposed approach. A neural network-based model is trained on this data, which is then coupled with the control system of the satellite for quick thermal adjustment. Successful validation lays the foundation for utilizing simulation-based physics-informed thermal analysis in conjunction with machine learning to influence the control system in near real-time for satellite subsystems.