A PMSM Speed Control Strategy of LADRC based on Neural Network Load Torque Estimation

This paper introduces a novel speed control method for Permanent Magnet Synchronous Motors (PMSM) utilizing Linear Active Disturbance Rejection Control (LADRC). The initial step involves collecting data on the motor’s d-axis and q-axis currents, speed, and load torque during stable operation. A small-scale neural network, characterized by low computational demands, is then trained to provide a non-exact load torque estimation. Subsequently, this estimated value of the load torque is incorporated into the LADRC, resulting in a reduction of the pressure calculated by the Extended State Observer (ESO). The discrepancy between the actual and estimated load torque can be viewed as the total disturbance experienced by the system. Finally, a LADRC speed controller that utilizes an imprecise load torque estimation is constructed. To enhance the flexibility and adaptability of the method, a weight coefficient is added to the estimated load torque. The efficacy of this speed control strategy is confirmed through Matlab/Simulink simulation, demonstrating superior control performance when faced with sudden changes in load torque.