In order to solve the speed estimation problems of speed sensorless vector-controlled induction motor drives, the paper presents two speed estimation schemes based on neural network mode identification theory. The advantages of each scheme are discussed and the simulation results show that the estimated speed can trace the actual speed better (even under the circumstances of load variation or speed step variation). Also, these schemes are not sensitive to the variations of motor parameters and the effect of iron loss. Therefore, the proposed neural network based on speed sensorless vector-controlled induction motor drives have good performance in stady-state and transient-state operation.