Gearbox is one of most critical components of speed-increasing wind turbines. During long-term operation, the dynamic parameters of the gear transmission system may deviate from the design specifications due to changes in service conditions, resulting in significant differences in the system's dynamic response. These differences may lead to significant errors in the identification of dynamic parameters based on initial design parameters. In this study, a torsional dynamic model of a three-stage wind turbine gearbox transmission system was developed using the lumped parameter approach, accounting for internal and external excitations such as working conditions, time-varying meshing stiffness and meshing damping. A dynamic identification model for the transmission system parameters was proposed based on the least squares method, incorporating dynamic feedback signals of rotor angular acceleration, angular displacement, torque, and generator load. The results indicate that the proposed method can effectively predict the dynamic response of the transmission system by estimating the gear mesh stiffness and rotational inertia.