Abstract: Aiming at the practical engineering problems of complex spacecraft operation environment and unknown errors in ground modeling, a method of modeling spacecraft control system by fusing mechanism and data is proposed based on the idea of digital twin. The deviation between mechanistic models and physical entities is modeled by mining information from ground test data and in orbit telemetry data through LSTM networks. Then the mechanistic models and data driven models are combined to obtain highly accurate systemlevel and component –level models. Simulation results of typical components of spacecraft control systems show that the model constructed by this modeling method can provide more accurate portrayal of multi dimensional characteristics, effectively improve the accuracy of control system modeling， which lays an important foundation for accurate performance evaluation and intelligent health management of spacecraft control systems in orbit.

Key words: aerospace control system, digital twin, mechanistic model, data driven, LSTM network