基于自适应小波网络的永磁无刷直流电机直接电压控制  

Direct Control of Voltage Based on Adaptive Wavelet Neural Network for PM Brushless DC Motors

作　　者：史婷娜[1] 田洋[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：《电工技术学报》2007年第9期74-79,共6页Transactions of China Electrotechnical Society

基　　金：天津市科技攻关计划重大资助项目(05ZHGCGX00100);天津市应用基础研究重点资助项目(043802011)。

摘　　要：通过分析永磁无刷直流电机的转子位置与三个相电压之间的关系,提出了基于自适应小波神经网络的永磁无刷直流电机无位置传感器控制新方法。该方法根据位置检测原理和微控制器的特点构建一个以三个相电压为输入,桥路编码信号为输出的多输入单输出小波网络模型。网络隐层节点初始个数为零,在训练过程中不断地按照自适应算法添加或删除隐层节点,形成一个结构简单、紧凑的小波网络。采用梯度下降法对网络进行离线训练和在线训练,由离线训练进行参数初始化并确定网络隐层节点个数,以滤波和逻辑处理后的网络输出信号为教师对网络输出层连接权进行在线调整。从而由电机的相电压直接映射出电机的换相信号,取代传统的位置传感器,实现无位置传感器的直接电压控制。实验结果表明,该方法能得到准确的永磁无刷直流电机的换相信号。　In this paper,the relationship between rotor’s position and phase voltages of PM brushless DC motors （BLDCM） is analyzed,a novel position sensorless control method for BLDCM is proposed,which is based on adaptive wavelet neural network （WNN）. In this method a multiple inputs-single output wavelet neural network model is built according to the analysis of position sensorless and micro-controller. There is no hidden units at the beginning,and during the process of learning,they are increased or decreased according to an adaptive algorithm. The WNN is finally built with a much simpler and tighter structure. Through a gradient descent error algorithm,the WNN is trained both offline and online. In the offline training,the parameters of network are initialized and the number of hidden units is obtained. While online learning,the weights of WNN are updated based on the output signals after filter. By mapping the phase voltages to communication signals,the network can replace the traditional position sensors,and achieve direct voltage control. By the experimental results,the method is verified to obtain the exact communication signals.

关 键 词：永磁无刷直流电机 自适应小波神经网络 直接电压控制 梯度下降法 参数初始化 

分 类 号：TM383[电气工程—电机]