永磁同步电机因体积小、功率密度高、起动和控制性能优越等特点，应用越来越广泛，尤其适用于机器人、新能源电动车等对电机运行性能要求高的机电系统中。近年来，随着控制技术的不断发展，高端机电系统对电机运行性能的要求在不断提高。利用精确的电磁参数（定子电阻Rs，直轴电感Ld，交轴电感Lq，永磁体磁链φf），借助先进的控制算法可以实现永磁同步电机的高性能控制。但电机电磁参数在实际运行过程中会发生变化，进而影响控制性能。因此，研究电机电磁参数的在线辨识技术对实现永磁同步电机高性能控制具有重要意义。 论文以永磁同步电机在线参数辨识理论及技术为主要研究内容。首先，建立两相旋转坐标系下的永磁同步电机数学模型，以其为参数辨识模型，构建基于矢量控制技术的永磁同步电机参数辨识系统，并借助Matlab软件对永磁同步电机控制系统的动态和稳态性能进行仿真先验。其次，为解决最小二乘法中“数据饱和”和抗干扰能力弱的问题，将多新息方法与限定记忆最小二乘法结合，对基于帕德逼近法离散化永磁同步电机数学模型的参数辨识系统辨识性能进行了研究。针对电压偏差对参数辨识系统精度的影响，深入分析了电压偏差产生的机理并研究了抑制电压偏差的死区补偿方法；对于死区补偿过程中电流极性难以准确判断、特性参数难以获取的问题，将交、直轴给定电流变换至三相静止坐标系下，明确电流极性的判定依据，并简化特性参数的处理。仿真验证表明：死区补偿后电流畸变率由11.42%降至3.76%，电压偏差被控制在1.5%以内，辨识精度提高至0.6%，说明死区补偿方法对提高参数辨识精度的有效性。最后，设计了永磁同步电机参数辨识系统并搭建了实验平台。基于实验平台，对死区补偿方法和多新息限定记忆最小二乘参数辨识算法进行了实验验证，结果表明：死区补偿后电流畸变率降至3.658%，电磁参数的辨识误差在5%以内，进一步验证了死区补偿方法和参数辨识算法的合理性和可行性。

Permanent magnet synchronous motor is more and more widely used because of its small size, high power density, excellent starting and control performance. It is especially suitable for mechanical and electrical systems with high-performance requirements for motors such as robots and new energy electric vehicles. In recent years, with the continuous development of control technology, the requirements of upmarket electrical and mechanical system systems for motor performance are also increasing. The precise electromagnetic parameters (stator resistance Rs, linear inductance Ld, quadrature inductance Lq and permanent magnet flux φf) can be used to realize the high performance control of permanent magnet synchronous motor, but the parameters of the motor are changing in the actual operation process. It is of great significance to study the on-line identification technology of motor electromagnetic parameters in order to achieve high-performance control of permanent magnet synchronous motor. The main content of this paper is the theory and technology research of on-line parameter identification of permanent magnet synchronous motor. Firstly, the mathematical model of permanent magnet synchronous motor in two-phase rotating coordinate system is established and used as the parameter identification model. The parameter identification system of permanent magnet synchronous motor based on vector control technology is constructed. The dynamic-state and steady-state performance of permanent magnet synchronous motor control system is simulated with Matlab software. Secondly, in order to solve the problem of "data saturation" and weak anti-interference ability in least squares method, the performance of parameter identification system of permanent magnet synchronous motor discretized by Pade approximation method is studied by combining multi-innovation method with limited memory least squares method. Aiming at the influence of voltage deviation on the accuracy of parameter identification system, the mechanism of voltage deviation is deeply analyzed and the dead-time compensation method for suppressing voltage deviation is studied. For the problems that current polarity is difficult to accurately judge in dead-time compensation process and characteristic parameters are difficult to obtain, the currents of i* d and i* q are transformed into abc coordinates to determine the basis of current polarity. The processing of characteristic parameters is simplified. The simulation results show that the current distortion rate is reduced from 11.42% to 3.76%, the voltage deviation is controlled within 1.5%, and the identification accuracy is improved to 0.6%. This shows that the dead-time compensation method is effective in improving the identification accuracy of parameters. Finally, a parameter identification system is designed and an experimental platform is built. Based on the experimental platform, the dead-time compensation method and multi-innovation limited memory least squares parameter identification algorithm are experimentally validated. The results show that the current distortion rate is reduced to 3.658% after dead-time compensation, and the identification error of electromagnetic parameters is less than 5%, which is in good agreement with the simulation analysis. The rationality and feasibility of the dead-time compensation method and parameter identification algorithm are further verified.