AN MRAS BASED SPEED ESTIMATION METHOD WITH A LINEAR NEURON FOR HIGH PERFORMANCE INDUCTION MOTOR DRIVES AND ITS EXPERIMENTATION

This paper presents a new speed observer for high application of an MRAS speed observer based on Artificial ??performance FOC (Field Oriented Control) and DTC (Direct Torque Control) induction motor drives. It is an MRAS (Model Reference Adaptive Systems) observer which employs a linear ANN (Artificial Neural Network) for the estimation of both the rotor speed and the flux-linkage. The training of the ANN based adaptive model of the MRAS observer is performed on-line by means of an OLS (Ordinary Least-Squares) algorithm. The sensorless algorithm has been verified experimentally both in a FOC and in a DTC control system. The experimental results neural network model, for stability reason, is employed not as show that the dynamical performances of the sensorless drive a simulator as in [10][11][12] but as a predictor. Finally this are definitely comparable to those obtainable with the corresponding FOC and DTC drives with encoders in medium to high speed ranges. With regard to low speed ranges, this sensorless algorithm works properly down to a speed as much as 10rad/s.