Novel Adaptive Learning Rate Back Propagation Neural Network-Based Online Rotor and Stator Resistance Estimator for Sensorless Induction Motor Drives

With consideration of low cost due to the absence of permanent magnets, the field-oriented control (FOC)-based induction motor drives are the most suitable choice for industrial applications. The knowledge of machine parameters is essential for excellent dynamic response of the sensorless vector-controlled induction motor drives. Machine parameters do not last constantly during the operation. Both stator and rotor resistances vary with temperature due to the temperature coefficient of resistance of the material. The accurate online estimation of stator and rotor resistances is needed for the precise control of sensorless vector-controlled induction motor drives. An artificial neural network-based rotor and stator resistance estimator suitable for vector-controlled induction motor drives is proposed in this paper. For the online estimation of the rotor resistance, the rotor flux linkages obtained from the voltage model are compared with that estimated from the neural network model. The weights of the neural network are adjusted by back propagating the obtained error till the error is minimized. For the online stator resistance estimation, the d–q axes stator currents are compared with that obtained from the neural network model, and error is back propagated through adjusting the weights of the neural network till the error is minimized. The obtained weights after learning were used to estimate stator and rotor resistance values. The simulation results justify an adaptive learning rate with added momentum instead of a constant learning rate throughout the learning giving faster convergence of error and hence suitable for online estimation of stator and rotor resistances.