物理-经验混合知识模型驱动的感应电机工况辨识方法  

Condition Identification Method for Induction Motors Driven by a Hybrid Physical-empirical Knowledge Model

作　　者：凌云飞 刘志亮 谢川[3] 左明健 LING Yunfei;LIU Zhiliang;XIE Chuan;ZUO Mingjian(School of Mechanical and Electrical Engineering,University of Electronic Science and Technology of China,Chengdu 611731;Glasgow College,University of Electronic Science and Technology of China,Chengdu 611731;School of Automation Engineering,University of Electronic Science and Technology of China,Chengdu 611731)

机构地区：[1]电子科技大学机械与电气工程学院,成都611731 [2]电子科技大学格拉斯哥学院,成都611731 [3]电子科技大学自动化工程学院,成都611731

出　　处：《机械工程学报》2024年第23期320-328,共9页Journal of Mechanical Engineering

基　　金：国家自然科学基金(52475091);四川省科技计划(2024NSFSC0185、2024JDHJ0057)资助项目。

摘　　要：工况辨识是感应电机状态监测的基础,然而现有的非侵入式方法在精确度、鲁棒性和泛化性方面存在不同程度的问题。针对这些挑战,提出了一种对感应电机转速、负载转矩两种工况参数进行辨识的新方法,该方法基于物理-经验混合知识模型,融合了物理模型法的明确机理和经验模型法的易于实施两方面的优势。本文方法引入定子电流的频率与有效值两个维度的经验信息,从感应电机物理模型出发,分别推导了定子电流频率与有效值这两个因变量和感应电机转频、转矩负载之间的物理关系式,并以此作为后续分析的拟合函数形式,结合先验工况辨识经验数据集进行最小二乘拟合确定了该拟合函数中的待定系数,从而实现了仅使用定子电流信号对感应电机工况参数进行虚拟测量。本文方法引入了定子电流中包含的多维经验信息,并且采用了具有物理机制的拟合函数,因此具有辨识精确度高、鲁棒性强的天然优势。实验结果表明,采用本文方法辨识的工况参数与真实值吻合较好,同时与其他已有工况辨识方法进行了对比,验证了本文方法在工况辨识精确度上的先进性。Condition identification is the cornerstone of induction motor state monitoring.However,existing non-intrusive methods suffer from varying degrees of challenges concerning precision,robustness,and generalizability.In response to these challenges,this paper proposes a new approach for identifying two operating parameters of induction motors:speed and load torque.This method is based on a hybrid physical-empirical knowledge model,amalgamating the advantages of clear mechanistic understanding from physical modeling and the practicality of empirical modeling.Our method introduces empirical information from two dimensions of stator current:frequency and RMS value.Commencing with the induction motor’s physical model,we derive mathematical relationships between the stator current's frequency and RMS value—these are the dependent variables—and the physical relationships between the induction motor's speed and torque load.These relationships are employed as a fitting function in subsequent analyses,and they are determined through least-squares fitting using prior empirical data on condition identification.Consequently,our method enables the virtual measurement of induction motor operating parameters solely using stator current signals.By incorporating multi-dimensional empirical information from stator current and utilizing a fitting function grounded in physical principles,our approach inherently possesses a high level of accuracy and robustness.Experimental results demonstrate that the operating parameters identified using our method closely match real values.Furthermore,comparative analyses with other existing condition identification methods validate the superior accuracy of our approach in condition identification.

关 键 词：感应电机 转矩测量 转速测量 状态监测 物理-经验混合知识模型 

分 类 号：TG156[金属学及工艺—热处理]