Abstract: Aiming at the problem that the accuracy of the speed estimation of the traditional model reference

adaptive system（MRAS）method without speed sensor decreases after the speed load mutation in the low speed

range，a speed identification method based on the combination of back propagation（BP）double-layer artificial

neural network（ANN）and MRAS was proposed using the super online estimation and adaptive ability of the

double-layer neural network，which improved the dynamic performance of the speed response in the low speed

range. Through Matlab simulation and PMSM drive control physical platform，the ANN-MRAS observer and the

traditional MRAS observer were compared and analyzed. The results show that the proposed method can still

maintain good dynamic performance and has strong robustness after sudden changes in speed and load torque.

      Key words: speed sensorless；model reference adaptive system（MRAS）；speed load mutation；double-layer

artificial neural network（ANN）；speed identification