Abstract：In the sensorless control of the traditional permanent magnet synchronous motor, the motor parameters determine the accuracy of the rotor position estimation, and the motor parameters are easily changed by the external environment, resulting in the decrease of the control performance. In this paper, an on-line identification method of motor parameters based on Interior Permanent Magnet Synchronous Motor (IPMSM) is presented, and the parameters are estimated by affine projection algorithm, which can accurately estimate the stator resistance, rotor flux, straight shaft inductance and cross-axis inductance. As the variation rate of stator resistance and rotor flux and stator inductance varies with the external environment, two different interrupt cycles are used to identify stator resistance and rotor flux and stator inductance,which can greatly reduce unnecessary calculation and improve the efficiency of the algorithm in practical application. Aiming at the disadvantage of the slow convergence speed estimation of the traditional model ref-erence adaptive algorithm, an improved rotor position and velocity estimation method based on the current adaptive state observer is proposed,which improves the speed identification convergence speed and reduces the steady state error. The experimental results show that the on-line identification method of motor parameters can accurately identify the motor parameters, and has good stability and dynamic performance.