Abstract: The uncertainty of thrust acceleration will cause enormous error to model and track the accelerated motion of maneuvered target during orbit transfer process, an on-line minimum variance tracking filter for thrust acceleration estimation is developed in this paper. The propellant mass-flow rate and exhaust velocity are selected as state variants, which are estimated by employing an integral state model and Expended Kalman Filter. The variation equations for measurement vector to state variants have been established to linearize the discrete-time measurement equations. The algorithm has applied successfully to maneuver process in commanding satellite into geo-stationary orbit. The results show the algorithm developed here can estimate thrust acceleration precisely and quickly during orbit transfer process.