基于共振扩张状态观测器的内置式永磁同步电机统一全速域无位置传感器控制

Unified Full Speed Sensorless Control of Interior Permanent Magnet Synchronous Motor Based on Resonance Extended State Observer

常规的永磁同步电机全速域无位置传感器控制通过复合两种位置观测器实现全速域的转子位置估计,然而,在电机频繁宽频域调速工况下,两种位置观测器频繁地切换,易导致估计的位置和转速振荡,并且两种位置估计方法需要单独的设计和调谐,增加了系统整定难度和算法复杂度。为此,该文提出了一种基于共振扩张状态观测器的内置式永磁同步电机统一全速域无位置传感器控制方法。首先,通过共振扩张状态观测器估计基频反电动势和高频反电动势。然后,建立了统一全速域模型,通过统一全速域模型实现全速域的转子位置和转速估计。在零速和低速时,通过向d轴注入高频电压,增加统一全速域模型中转子位置信息的信噪比,从而可以准确估计零速和低速区的转子位置,消除了传统高频注入法中由滤波器和延迟引起的估计误差。当电机在中高速区运行时,统一全速域模型自动蜕变为基频模型法,不需要两种位置观测器切换控制。最后,在2.0 kW内置式永磁同步电机实验平台上验证了算法的有效性。

Conventional sensorless control of permanent magnet synchronous motors in the full speed domain achieves rotor position estimation by combining the two position estimation methods.However,in wide frequency domain speed regulation,the two position estimation methods switch frequently,which can easily cause oscillations in the estimated position and speed.Moreover,the two position estimation methods require separate design and tuning,increasing the difficulty of system tuning and algorithm complexity.For this reason,this paper proposes a unified full-speed sensorless control method of interior permanent magnet synchronous motor based on resonance extended state observer.First,the fundamental frequency back electromotive force and high-frequency back electromotive force are estimated by the resonant extended state observer.This observer adds a high-frequency disturbance estimation loop based on the traditional extended state observer.Accordingly,the high-frequency back electromotive force in the low-speed zone generated by injecting high-frequency voltage is estimated,and the fundamental frequency back electromotive force is estimated by the original disturbance estimation loop of the extended state observer.The original disturbance estimation loop of the extended state observer has the characteristics of a notch filter at the frequency of injected high-frequency voltage.In addition,it can suppress the high-frequency measurement noise,which can accurately estimate the fundamental frequency back electromotive force.The high-frequency disturbance estimation loop of the resonant extended state observer has band-pass filter characteristics with no amplitude attenuation and phase lag at the frequency of the injected high-frequency voltage,allowing it to accurately estimate the high-frequency back electromotive force.Then,a unified model in the full-speed domain is established,and the rotor position and speed estimation are realized through the unified model.At zero speed and low speed,the signal-to-noise ratio of the fundamental frequency back electromotive force is low,so it cannot accurately estimate the fundamental frequency back electromotive force.By injecting high-frequency voltage into the d-axis,the signal-to-noise ratio of the high-frequency back electromotive force is increased.However,the high-frequency back electromotive force is modulated by the injected high-frequency signal,which cannot be directly used to estimate the rotor position.A unified full-speed domain model is constructed to achieve a high signal-to-noise ratio of rotor position information in the full-speed domain.The amplitude of the unified full-speed domain model is the sum of the square of the fundamental frequency back electromotive force amplitude and the square of the high-frequency back electromotive force amplitude.The output of the unified full-speed domain model is used as the input of the phase-locked loop,and the output of the phase-locked loop is the estimated rotor position and speed.Finally,the correctness and effectiveness of the proposed method are verified on the 2.0 kW interior permanent magnet synchronous motor experimental platform.Compared with the traditional sensorless control method,the proposed method has the following advantages:(1)The full-speed domain uses the same estimator without a transition zone,simplifying the control algorithm,reducing tuning difficulty,and avoiding speed and position oscillations caused by frequent switching of the two observers.(2)Estimation errors caused by filters and delays in traditional high-frequency injection methods are eliminated.