An implementation of neural network and multi-fuzzy controller for permanent magnet synchronous motor direct torque controlled drive

Abstract

To compensate voltage difference between the reference and the actual output voltages caused by dead-time effects, a novel compensation method for permanent magnet synchronous motor (PMSM) direct torque controlled (DTC) drive based on neuro-fuzzy observer is proposed. This method presents the implementation of a voltage distortion observer based on the artificial neural network (ANN). Using the output of the fuzzy controller (FC1), online training is carried out to update the weights and biases of the ANN. To reduce torque and flux linkage ripples, another fuzzy controller (FC2) is adopted to replace the conventional hystersis controller. The proposed control scheme combines the capability of fuzzy reasoning in handling uncertain information and the capability of neural network in learning from processes. Results of simulations and experiments are provided to demonstrate the effectiveness of the proposed method even under the occurrence of different reference speed and load torque. Index Terms -PMSM, DTC, ANN, FC.

The Research of Adaptive Robust Predictive Current Control Algorithm of Permanent Magnet Synchronous Motors

“…In order to obtain high control effect, on-line access to accurate motor parameters such as fault diagnosis, condition monitoring is important in many areas. At present domestic and foreign scholars have made a lot of online parameter estimation methods include: extended Kalman filter [1][2][3][4], neural networks [5][6], Adaptive techniques [7,8], a method based on models [9][10][11], or based on least mean squares method [12][13][14]. These methods are the same as the EKF, due to the large matrix inverse and matrix calculation , which require higher processing power.…”