基于RBF神经网络的永磁同步电机无位置传感器控制  

Control of PMSM based on RBF neural network without position sensor

作　　者：史婷娜[1] 王向超[1] 夏长亮[1] 

机构地区：[1]天津大学电气与自动化工程学院,天津300072

出　　处：《电工电能新技术》2007年第2期16-19,44,共5页Advanced Technology of Electrical Engineering and Energy

基　　金：天津市科技攻关计划重大项目(05ZHGCGX00100)

摘　　要：本文通过分析永磁同步电机Id=0控制策略及其间接位置检测原理,提出了基于径向基函数(RBF)神经网络的无位置传感器控制方法。该方法首先构建一个隐层节点数为零的四输入两输出RBF网络,网络的输入为电机α-β轴上电压和电流,输出为转子转角和转速,然后在离线训练过程中按照自适应算法不断添加和删除隐层节点,形成一个结构简单、紧凑的RBF网络,最后采用梯度下降纠正误差法在线训练更新网络参数。该方法通过对电机α-β轴上电压和电流的映射,得到了电机转子的转角和转速,取代了传统的位置传感器。实验结果表明了该控制方法的有效性。Firstly, the Id = 0 control strategy and the principle of control for PMSM without position sensor are analyzed Then, a new control method for PMSM is proposed, which is based on radial basis function （RBF） neural network There are no hidden units at the beginning, and the network has four inputs and two outputs. The inputs are the voltages and currents of stationary α-β reference frame, and the outputs are the position angle and speed of rotor. Then according to an adaptive algorithm, the hidden units are increased or decreased during the process of off-line learning tillthe network is built with a much simpler and tighter structure. During on-line training, the network updates the network parameters using a gradient descending error algorithm. By mapping the voltages and currents of stationary α-β reference frame to position angle and speed of rotor, the network can replace the traditional position sensors. The theory in this paper is verified by experimental results.

关 键 词：永磁同步电机 径向基函数 神经网络 无位置传感器控制 

分 类 号：TM341[电气工程—电机]