永磁同步电机在多参数辨识会出现参数耦合导致辨识模型欠秩、数据的利用率较低以及辨识参数不精准问题。本文提出一种基于改进递推最小二乘算法对永磁同步电机进行参数辨识。首先在两相同步旋转正交坐标系上建立永磁同步电机数学模型，其次，通过结合递推最小二乘算法和扩展卡尔曼滤波算法实现对电机参数分步辨识，避免了参数耦合导致辨识模型欠秩，同时，两种算法分布辨识时引入了多新息理论，从而提高数据利用率。最后将本文算法与原算法以及模型参考自适应算法进行参数辨识对比。在恒定负载，转速时辨识电阻,电感,磁链相比原算法精度提高了4.89%，1.86%，3.80%,比模型参考自适应算法提高了11.32%，1.21%，2.03%。负载，转速发生突变时本文算法相比于模型参考自适应算法产生的波动更小，表明本文算法参数辨识时精度更高，稳定性更好。

Permanent magnet synchronous motor in the multi-parameter identification will appear parameter coupling leads to identification model under-rank, the utilization rate of data is low, and identification parameter imprecision problem. In this paper, we propose a parameter identification of permanent magnet synchronous motor based on improved recursive least squares algorithm. Firstly, a mathematical model of permanent magnet synchronous motor is established on two synchronous rotating orthogonal coordinate systems, and secondly, the stepwise parameter identification is realized by combining the recursive least squares algorithm and the extended Kalman filtering algorithm, which avoids the parameter coupling resulting in the under-ranking of the identification model, and at the same time, the two algorithms introduce the multi innovation theory in the distribution of the identification, so as to increase the utilization of the data. Finally, the algorithm of this paper is compared with the original algorithm and the model reference adaptive dystem algorithm for parameter identification. The accuracy of identifying resistance, inductance and magnetic chain at constant load and speed is improved by 4.89%, 1.86% and 3.80% compared to the original algorithm, and 11.32%, 1.21% and 2.03% compared to the model reference adaptive algorithm. The fluctuation of this algorithm is smaller than that of the model reference adaptive algorithm when the load and rotational speed change suddenly, which indicates that this algorithm has higher accuracy and better stability in parameter identification.