Computer Engineering and Applications ›› 2014, Vol. 50 ›› Issue (6): 255-260.

Previous Articles     Next Articles

Multi-innovation stochastic gradient identification for permanent magnet synchronous motor

XU Peng1,2, XIAO Jian1, ZHOU Peng2, LI Shan2   

  1. 1.School of Electrical Engineering, Southwest Jiaotong University, Chengdu 610031, China
    2.School of Electronic and Automation, Chongqing University of Technology, Chongqing 400054, China
  • Online:2014-03-15 Published:2015-05-12

基于多新息随机梯度永磁同步电机参数辨识

徐  鹏1,2,肖  建1,周  鹏2,李  山2   

  1. 1.西南交通大学 电气工程学院,成都 610031
    2.重庆理工大学 电子信息与自动化学院,重庆 400054

Abstract: Permanent Magnet Synchronous Motor(PMSM) has some excellent features, such as fast response, better accuracy, high torque to current ratio. Based on analysis of PMSM mathematical model, the system regression model is proposed, and multi-innovation stochastic gradient algorithm for PMSM parameters identification is derived. Simulation and real-time experiments results show that MISG algorithm has more outstanding performance on parameter estimate convergence than SG algorithm because of reusing measurable output and input information. Meanwhile, with multi-innovation length increased and forgetting factor affected, convergence performance of MISG algorithm is close to RLS.

Key words: Permanent Magnet Synchronous Motor(PMSM), multi-innovation, stochastic gradient, convergence performance

摘要: 永磁同步电机(Permanent Magnet Synchronous Motor,PMSM)具有响应快、高精度、高转矩比等诸多优点。在永磁同步电机系统数学模型基础上,构建系统回归模型,推导得永磁同步电机多新息随机梯度参数辨识算法(MISG),仿真和实时实验结果表明由于MISG算法重复利用可测输入输出信息,较单新息随机梯度算法(SG)有着更好的参数估计收敛性,并且随着新息长度[p]的增加及遗忘因子引入,MISG算法辨识效果与最小二乘(RLS)算法接近。

关键词: 永磁同步电机, 多新息, 随机梯度, 收敛性