基于递归神经网络的永磁同步电机参数辨识研究

电机参数的变化会加大永磁同步电动机(PMSM)的控制难度,所以研究参数辨识对于闭环控制系统的稳定运行有着重大的意义。在采用变分理论实现最小绝对值偏差法(LAD)的基础上,研究了一种基于递归神经网络(RNN)的辨识方法。仿真结果表明,该方法具有很快的收敛速度,能准确地辨识PMSM的定子电阻、d,q轴电枢电感及转子磁链等参数,并且具有良好鲁棒性,在出现参数变化或异常值情况下仍能辨识到正确结果。

Research on Identification of PMSM Based on Recurrent Neural Network

The change of motor parameters will increase the control difficulty of the permanent magnet synchronous motor. So the research of parameter identification is great significant for the stable operation of the closed loop control system. Using the variational theory to realize least absolute deviation(LAD),a new identification method based on recurrent neural network was researched. It could accurately identify PMSM’s parameters,such as the stator resistance,the d/q-axis armature inductance and the rotor flux. Simulation results show that the method has lots of advantages such as fast convergence and good robustness and it can still get right results in the presence of outliers.