Sensorless Control of PMSM using DTC Strategy Based on Multiple ANN and Load Torque Observer

For a Permanent Magnet Synchronous Motor (PMSM), in this article, a sensorless control system is presented, with the speed controller implemented using a Multiple -Artificial Neural Network (M-ANN) and the rotor speed estimated and provided by a Model Reference Adaptive System (MRAS) type observer. The performances of the implemented and proposed sensorless control system of the PMSM are presented in comparison with the classical Direct Torque Control (DTC) system. Depending on the estimated value of the load torque, the corresponding ANN is selected, trained to obtain superior performance as compared to the utilization of the DTC strategy by adjusting the control parameters of the system for the overall range of the reference rotor speed and load torque gradient. The modeling equations of the PMSM operation, the speed, the equations of the load torque and stator resistance observer implementation, the most important blocks and control structures, their parameterizations and some results of the numerical simulations obtained using Matlab/Simulink are herein presented. By using common Simulink and Stateflow blocks, good results are obtained through the numerical simulations and this fact recommends the implementation in embedded systems of the entire proposed control structure of the PMSM.