摘要
反电动势(BEMF)的准确估计是实现永磁同步电机(PMSM)无位置传感器控制的关键。由于BEMF的动态包含未知转速信息,现有估计方案通常设计自适应项在线估计转速以确保观测器稳定性。传统自适应观测器方法通常需构造严格正实(SPR)条件,显著增加了系统设计复杂度。为简化观测器设计并确保其稳定性,本文提出基于参数估计的BEMF观测器。首先,将BEMF估计问题转化为开环观测器设计和未知常数辨识问题,克服对SPR条件的依赖,简化整体设计流程。其次,采用动态回归扩展与混合(DREM)技术在线估计未知常数,相比传统梯度下降法,有效消除估计参数间的耦合,提升了观测器的动态性能。最后,通过严格理论分析验证所提观测器稳定性,并通过实验结果验证该方案的有效性。
The accurate estimation of back electromotive force(BEMF)is crucial for achieving sensorless control of permanent magnet synchronous motor(PMSM).Since the dynamics of BEMF contain unknown speed information,existing estimation schemes usually design adaptive terms to estimate the speed online to ensure the stability of the observer.Conventional adaptive observer methods typically require the construction of strictly positive real(SPR)conditions,which significantly increases the complexity of system design.To simplify the design while ensuring the stability of the observer,this paper proposes a BEMF observer based on parameter estimation.Firstly,the BEMF estimation problem is transformed into the design of an open-loop observer and the identification of unknown constants,overcoming the dependence on SPR conditions and simplifying the overall design process.Secondly,the dynamic regression extension and mixture(DREM)technique is used to estimate the unknown constants online.Compared with the traditional gradient descent method,it effectively eliminates the coupling between the estimated parameters and improves the dynamic performance of the observer.Finally,the stability of the proposed observer is verified through strict theoretical analysis,and the effectiveness of the scheme is verified through experimental results.
作者
白振毅
陈永进
邓美玲
江伟程
许景尚
Bai Zhenyi;Chen Yongjin;Deng Meiling;Jiang Weicheng;Xu Jingshang(Guangdong Power Grid Co.,Ltd.Shaoguan Renhua Power Supply Bureau,Shaoguan 512300,China)
出处
《电力电子技术》
2025年第11期136-143,共8页
Power Electronics
基金
中国南方电网有限责任公司科技项目(GDKJXM20230364)。
关键词
永磁同步电机
无位置传感器控制
参数估计观测器
动态回归扩展与混合
permanent magnet synchronous motor
sensorless control
parameter estimation-based observer
dynamic regression extension and mixture
