永磁同步电机神经网络谐波注入转矩脉动抑制方法

Torque Ripple Suppression of PMSM with Harmonic Injection and Linear Neuron

摘要:

为了抑制永磁同步电机转矩脉动,提出一种基于偏差电流输入的自适应线性神经网络谐波注入方法。该法以交轴和直轴电流偏差作为神经网络环节输入,应用两个自适应线性神经网络环节拟合电压谐波补偿分量,并将其注入到电流环控制器输出,用以抵消交轴电压和直轴电压中的谐波分量,进而抑制转矩脉动。为验证本方法有效性,对比进行了三种控制方法的仿真和实验。结果表明,与传统双闭环控制相比,本方法可将转矩脉动降低约97%。相比于传统神经网络谐波注入法,本方法可获得相近的转矩脉动抑制效果,并且控制结构更为简单,参数整定方便,易于工程实现,算法运行时间缩短约60%。

Abstract:

In order to suppress the torque ripple of permanent magnet synchronous motor (PMSM), based on the current deviation input and adaptive linear neuron (CDI-Adaline), a new harmonic injection method is proposed. The q-axis and d-axis current deviations are input into the neural network. Moreover, the voltage harmonic components from the two Adaline parts are injected into the controller output of the current loop, which can reduce the voltage harmonics and suppress the torque ripple. The simulation and experimental results show that, compared with the double closed loop control, the torque ripple can be reduced by about 97% when the CDI-Adaline method is used. Furthermore, the CDI-Adaline method has the advantages of simple structure, convenient parameter setting and easy engineering implement. In addition, compared with the traditional Adaline method, the similar effect about torque ripple suppression can be obtained by using the CDI-Adaline control, and the algorithm running time is shortened by about 60%.