Autonomous orbital maneuvering of an electric propulsion satellite in geosynchronous transfer orbit (GTO) requires more accurate autonomous orbit determination (OD). Implementing the traditional autonomous OD method for the electric propulsion satellites may result in the accumulation of measurement errors and dynamic errors from orbit extrapolation, which significantly restrict the OD accuracy. Fortunately, an electric propulsion satellite in GTO can sense the transit time in or out of the Earth shadow according to the output power change of the solar array. Therefore, an Earth shadow time/accelerometer/GNSS-based geosynchronous transfer OD method for electric propulsion satellites is proposed in this paper. First, a system state model with dynamics is established for the entire orbit transfer process. Second, measurement models that use the measurement information from the accelerometer, GNSS, and the transit time into or out of the Earth shadow are established. Third, a sequential filtering estimation approach is designed to solve the problem of asynchronous data fusion for autonomous OD. Comparison simulations demonstrate the effectiveness and superiority of the proposed method.