Modified Reference Model for Rotor Flux-Based MRAS Speed Observer Using Neural Network Controller

The poor low speed performance of rotor flux-based model reference adaptive system (MRAS) is due to the presence of integrator and parameter variation with temperature. To improve low speed performance, a novel rotor flux-based MRAS method is proposed and neural network controller (NNC) is used in place of PI controller in reference model and adaptation mechanism. In this method, a compensating voltage is added to the d-q axis rotor flux equations of induction motor (IM) by modifying voltage model to reduce DC drift and initial value problems of integrator. NNC is implemented in both modified reference model to obtain the drift voltage and in adaptation mechanism to accurately estimate the rotor speed. The proposed scheme is experimentally implemented using dSPACE ds-1104 R&D controller board with improved speed response compared to rotor flux-based MRAS.