In this paper, an estimation method of vehicle yaw rate based on the fusion of kinematic models is proposed which outperforms the existing method in accuracy and adaptability to large lateral acceleration conditions and is practical in engineering. First, a two-track kinematic model of the vehicle is established according to steering geometry for the condition of no wheel slips under small lateral accelerations. Four yaw rate estimations are calculated and averaged with four wheel speeds and the front wheel angle as inputs. Second, a circular kinematic model is established considering the vehicle as a particle, and the yaw rate is estimated with the inputs of lateral acceleration and longitudinal speed under large lateral accelerations. Third, a piecewise weighted algorithm is designed to fuse the above two yaw rate estimations to adapt to all driving conditions. Finally, constant steering wheel angle tests and slalom tests are carried out based on a distributed drive electric SUV to verify the proposed yaw rate estimation method. The results show that the estimated yaw rate of the existing method contains obvious noise and deviates from the real value when the lateral acceleration exceeds 5 m/s2, while the proposed fusion-based method can accurately estimate the yaw rate under static and dynamic conditions within lateral acceleration of −8 m/s2–8 m/s2, with a root mean square error of less than 0.0678 rad/s.