基于ELM实现的IPMSM转矩观测器  

IPMSM Torque Observer Based on ELM Neural Network

作　　者：厉亚强 张文涛[1] 李涉川 LI Ya-qiang;ZHANG Wen-tao;LI She-chuan(Guilin University of Electronic Technology,Guilin 541004,China)

机构地区：[1]桂林电子科技大学

出　　处：《微特电机》2019年第5期50-54,共5页Small & Special Electrical Machines

基　　金：国家自然科学基金项目(61565004);国家科技重大专项(02专项)子课题(2017ZX02101007-003)

摘　　要：针对内置式永磁同步电机(IPMSM)参数非线性及不确定性导致的电机转矩难以准确估测和控制等问题,研究了基于ELM(极限学习机)神经网络的IPMSM的矢量控制转矩观测器。利用ELM神经网络较强的泛化能力和逼近能力,同时根据IPMSM的矢量控制要求,设计出电流到转矩的非线性映射;将神经网络输出的估测的转矩作为反馈转矩输入PI转矩控制器,并通过PI电流控制器调节q轴电流;使实际电机转矩等于转矩命令,从而完成对电机的转矩控制。实验结果表明,该观测器有效地减少了90%的转矩脉动,具有良好的动态和静态性能,同时对系统参数不确定性和非线性具有较好的适应性。Because of the nonlinearity and uncertainty of the parameters of the interior permanent magnet synchronous motor,it is difficult to accurately estimate and control the motor torque. A vector control torque observer of the interior per-manent magnet synchronous motor based on the ELM (extreme learning machine) neural network was studied. Due to the strong generalization ability and approaching ability of extreme learning machine neural network,the current-to-torque non-linear relationship was mapped according to the vector control requirements of the interior permanent magnet synchronous motor. And the estimated torque from the neural network was fed into the PI torque controller to adjust q-axis current. The experimental results show that the observer effectively reduces the torque ripple by 90%,has good dynamic and static per-formance,and has good adaptability to the uncertainty and nonlinearity of system parameter.

关 键 词：内置式永磁同步电机 转矩观测器 极限学习机 矢量控制 闭环控制 

分 类 号：TM464[电气工程—电器] TM351