Performance analysis of novel robust ANN-MRAS observer applied to induction motor drive

This paper presents a novel method for estimating the rotor speed of a sensorless indirect field-oriented control (IFOC) induction motor based on the model reference adaptive system (MRAS) scheme. As a matter of fact, this method is meant to enhance the conventional MRAS performance especially in low-speed regions, and to reduce its sensitivity to noise and system uncertainties. For this purpose, an advanced MRAS has been involved to estimate the rotor speed with artificial intelligence (AI) approach, with the aim of achieving a high-performance of vector-controlled induction machine drive. The adjustable and reference models are designed based on an artificial neural network (ANN) structure in an attempt to estimate speed and rotor flux out of the measured terminal voltages and currents. The ANN structure promised eradication of pure integration with immunity to parameter variation with extreme-precision. Accordingly, some simulation results are presented to validate the proposed method and to highlight the performance analysis of the improved Neural Network rotor flux MRAS (NN RF-MRAS) observer compared to the conventional MRAS observer. The effectiveness of the proposed observer has been carried out under different operating conditions, based on benchmark tests using MATLAB/Simulink software environment.