Adaptive Speed Observer using Artificial Neural Network for Sensorless Vector Control of Induction Motor Drive

This paper presents an adaptive speed observer for an induction motor using an artificial neural network with a direct field-oriented control drive. The speed and rotor flux are estimated with the only assumption that from stator voltages and currents are measurable. The estimation algorithm uses a state observer combined with an intelligent adaptive mechanism based on a recurrent neural network (RNN) to estimate rotor speed. The stator and rotor resistances are estimated by a simple Proportional-Integrator (PI) controller, which reduces sensitivity to variations, due essentially to the influence of temperature. The proposed sensorless control scheme is tested for various operating conditions of the induction motor drive. Experimental results demonstrate a good robustness against load torque disturbances, the estimated fluxes and rotor speed converge to their true values, which guarantees that a precise trajectory tracking with the prescribed dynamics.