This paper studies the fault diagnosis problems of the satellite attitude control system. Since the satellite works in space, the attitude control system of satellite is easy to be effected by uncertainties and disturbances in the harsh working environment. Interacting Multiple Model (IMM) is one of the effective fault diagnosis methods of the satellite control system but it is subjected to strong noise or wrong data. For this reason, this paper proposes to combine the moving horizon estimation instead of the Kalman filter with the IMM algorithm. The moving horizon estimation, which adopts not the estimation error of one time but the interval estimation error of a time span, is used to estimate the system states and the transition probabilities. So it could effectively decrease the influences of the wrong data or strong noise. Finally, the mathematical simulation results for diagnosis problem of the satellite attitude control system are presented to show the effectiveness of the proposed scheme.