基于热网络法的永磁电机温度在线估计  

On-Line Temperature Estimation of Permanent Magnet Motor Based on Lumped Parameter Thermal Network Method

作　　者：师蔚[1] 骆凯传 张舟云 Shi Wei;Luo Kaichuan;Zhang Zhouyun(College of Urban Rail Transit Shanghai University of Engineering Science,Shanghai 201620 China;Shanghai Electric Drive Co.Ltd,Shanghai 201806 China)

机构地区：[1]上海工程技术大学城市轨道交通学院,上海201620 [2]上海电驱动股份有限公司,上海201806

出　　处：《电工技术学报》2023年第10期2686-2697,共12页Transactions of China Electrotechnical Society

基　　金：国家“十三五”科技部重研发计划资助项目(2016YFB0100700)。

摘　　要：将集总参数热网络(LPTN)模型应用于永磁电机温度在线估计能够预防电机因过温而造成的电机损坏,特别是降低永磁体的不可逆退磁风险,提高永磁电机使用的安全性。该文首先针对LPTN模型应用于温度估计过程中存在灰度模型的不确定性,对永磁电机LPTN模型进行比较,建立不同节点的低阶灰箱永磁电机LPTN模型。搭建永磁电机变工况温度实验平台,基于实验数据进行模型对比分析,得到用于永磁电机在线温度估计的最优节点数LPTN模型。同时为克服扩展卡尔曼滤波算法出现的滤波发散及热阻参数在变工况下非线性导致温度估计累积误差问题,利用加权多新息强跟踪扩展卡尔曼滤波(WMI-STEKF)算法对永磁电机LPTN模型进行在线参数辨识及温度估计,将温度估计误差降低至3℃,提高非线性系统在变工况条件下参数辨识过程的鲁棒性。最后,通过仿真和温度实验结果对比分析,证明该方法的正确性和准确性。The online temperature estimation of permanent magnet motors(PMM)can prevent motor damage caused by over temperature,especially reduce the risk of irreversible demagnetization of permanent magnet and improve the safety of PMM.For traditional online temperature estimation of lumped parameter thermal network(LPTN),the influence of LPTN topology with different scales on the speed and accuracy of on-line estimation has yet to be deeply studied.At the same time,the thermal resistance unit nonlinearity and the error accumulation from loss calculation generate the deviation between the estimated temperature and the actual value.Therefore,after comparing the low-order gray box LPTN models of PMM with different nodes,this paper proposes the weighted multi-innovation strength tracking extended Kalman filter(WMI-STEKF)algorithm to carry out on-line parameter identification and temperature estimation with high accuracy under variable working conditions.Firstly,the LPTN gray box models of PMM with different nodes are established.Next,the experimental temperature platform of variable condition temperature of PMM is built.Based on the experimental temperature data,the models are compared and analyzed.The optimal LPTN model for online temperature estimation of PMM is obtained.Third,the time-varying fading factor is introduced into the extended Kalman filtering algorithm to adjust the real-time gain matrix to improve the robustness of the model and the accuracy of the extended Kalman filtering algorithm identification.At the same time,the residual scalar in the extended Kalman filtering algorithm is extended to the innovation matrix to improve the adaptability of the system to nonlinear systems.WMI-STEKF algorithm is used in on-line parameter identification and temperature estimation.The grey box thermal network models with four different node models are evaluated according to the experimental temperature data.The five-node model has the best accuracy,convergence speed,and stability.The average error and maximum error of the temp

关 键 词：集总参数热网络 永磁电机 温度估计 加权多新息强跟踪扩展卡尔曼滤波(WMI-STEKF) 

分 类 号：TM351[电气工程—电机]