基于滤波器的迭代学习最小二乘辨识方法

Filter based iterative learning least square identification method

针对永磁同步直线电机作为控制对象的参数辨识问题,在对直线电机的数学模型模型以及最小二乘辨识算法等内容进行了研究,发现了批处理最小二乘辨识算法虽然能够成功辨识伺服系统模型参数但是模型不够精确,尤其是针对伺服系统中存在的对振荡环节的辨识。根据传统辨识模型的不足,提出了一种基于滤波器的迭代学习最小二乘辨识方法,该算法通过滤波器来减小辨识过程中的噪声干扰,用迭代的方式来求解辨识模型带来的非线性问题。仿真和实验结果表明:与批处理最小二乘算法相比,基于滤波器的迭代学习最小二乘辨识方法能够有效提高辨识精度,振荡环节的辨识结果与功率谱分析得出的图线更加吻合。

Aiming at the parameter identification of the permanent magnet linear synchronous motor（ PMSLM）,the model parameters of servo system can be successfully identified by the block least squares algorithmon the basis of the study of the linear motor＇smathematical model and the least square identification algorithm. However,its model is not accurate enough,especially for the oscillating element identification in the servo system. According to the shortcomings of the traditional identification model,a filter-based iterative learning least square identification method was proposed. The noise interference was reduced by the filter in the identification process,andnonlinear problem of the identification model was solved by iterating in this algorithm. The simulation and experimentresults indicate that the filter-based iterative learning least square identification method can effectively improve the accuracy of identificationcompared with the block least squares algorithm,and the identification results of the oscillating element are more consistent with those obtained by power spectrum analysis.