摘  要：永磁同步电动机伺服控制中，速度环和电流环的PI参数取决于电机参数。分析电机模型后，建立电机输出量的误差函数，使之含有各种待估参数。引入单层神经网络，运用梯度方法动态更新权值，再通过权值估算电机参数。改变学习速率的大小，影响估算精度和收敛速度。实验和仿真效果均验证其有效性，PI参数自调节后，电机控制性能明显改善。
关键词：永磁同步电动机；梯度；PI自适应；神经网络

Abstract: PI parameters of velocity loop and current loop in PMSM servo control， depend on the motor’s parameters. After analyzing the motor model， the output quantities error functions which contain the parameters to be estimated， was established. The single-layer neural network using gradient method to dynamically update the weights was proposed to achieve estimation. Learning rate affects estimation accuracy and convergence rate. Both experimental and simulation results verify its effectiveness and after PI parameters self-tuning， the motor control performance is significantly improved.
Key words: PMSM; gradient; PI self-tuning; neural networks