对角递归神经网络永磁同步电机的无传感器控制

Sensorless PMSM control based on diagonal recurrent neural network

分析了基于对角递归神经网络观测器控制系统的动态性能和鲁棒性能.基于对角递归神经网络观测器将实际测得的电压和电流经过坐标变换后估测出电流和角速度,用估测值与实际值的差值调节神经网络观测器连接权值,直到预测误差达到设定值.该控制器具有不依赖被控对象的精确数学模型、对外界环境变化具有学习性、自适应性及鲁棒性等特点.仿真表明,该方法具有较好的转子位置和速度跟踪特性,系统具有较强的抗负载扰动性能和控制性能,能够满足精度高、反应快、鲁棒性好的要求.

The emphasis is to analyze the dynamic performance and robustness of control system based on diagonal recurrent neural network （DRNN） observer. The current and angular velocity were estimated by a neural network observer, whose inputs are voltage and current measured from motor after coordinate transformation. The errors between estimated and measured values were used to adjust the weighted values of DRNN observer until the predicted error equals the setting value. The controller has such several advantages as independent-plant mathematical model, learning ability and self-adaptive environmental impact, strong robustness. The simulation results show that the proposed method has good position and speed tracking performances. Moreover, the anti-disturbance ability is a remarkable feature, and the control performance is pretty good. It is shown that the system can satisfy the demands Of fast dynamic response, accurate speed and torque control, and strong robustness.