The Speed Regulating Performance Control Of PMSM Based On Load Torque Observation And Compensation

In the AC drive servo system, the permannent magnet synchronous motor(PMSM) has an increasing application due to its high power density, excellent speedregulation performance and high reliablity. As the variables that direct measurementis difficult, the inertia and load torque have an adverse influence on the systemresponse and steady state accuracy when having a change. To improve the servosystem’s speed control performance, the inertia identification, the observation andfeedforward compensation of load torque have been researched in this paper.The mathematical model of PMSM based on the field oriented control isbuilded up firstly, for the inertia identification an offline method on the basis ofperiodic speed signal is proposed. Through the integral operation to the motormotion equation, the uncorrelated variables can be separated out. Moreover, theintroduction of the torque disturbance observer improves the stability of theidentification result. In order not to affect the normal operation of the motor, then aneffective online inertia identification is presented based on the gradient correctionmethod. It demonstrates that the online identification can achieve accurate result byselecting the adaptive gain properly.To realize the observation of the load torque, through the analysis on theLuenberger observer’s structure, the full-order and reduced-order load torqueobserver model is established respectively. The observer has good speed and torqueobservations by the reasonable poles location configuration. In order to reduce theinfluence of noise, the recursion formula of the Kalman filter is intruduced, and thenthe load torque observer based on Kalman filter is established. By means of thefeedforward compensation of observed torque to the system, the speed disturbancecan be decreased distinctly.At last, on the basis of the simulation analysis to the presented methods, thePMSM vector control platform based on the DSP processer is designed to verify thealgorithm ulteriorly. The experimental results show the effectiveness of thealgotithm very well.