F. Haghgoeian, M.A. Ouhrouche, and J.S. Thongam (Canada)
Induction Motor, Neural Networks, Vector Control, Sensorless Speed Estimation.
This paper presents a sensorless speed estimation method of an induction motor using recurrent neural networks in an indirect field -oriented control system. The measured voltages and currents of the motor are used as the inputs to the neural network, while the feedback of the estimated speed realizes the structure of a Jordan network. The proposed network is trained online using backpropagation algorithm, and the error between the back electromotive force (emf) obtained from adaptive model and the one obtained directly from reference model is used for training. This method achieves a wide bandwidth of speed control and is verified using real-time simulation using `Matlab Simulink' on PC platform.
Important Links:
Go Back
