Estimation of Rotor Flux using Neural Network Observer in Speed Sensorless Induction Motor Drive

Abstract

n speed sensorless controlled induction motor drive, the rotor flux based Model Reference Adaptive System (MRAS) for is more popular. In MRAS, the voltage model equations are used in the reference model, which leads to poor performance of the drive at low speeds. In this paper, a new stator current based MRAS is presented. In this measured stator currents of induction motor are used as a reference model to avoid the use of a pure integrator. Estimated stator currents are used as an adjustable model. The estimation of stator currents requires the information of the rotor flux which can be obtained from the voltage or current model. This encounters instability and dc drift. To overcome these problems, flux observer is proposed based on Neural Network (NN). This offline trained multilayer feed-forward NN flux observer applied to indirect vector controlled speed sensorless induction motor drive has been verified by MATLAB/SIMULINK.