Chaotic learning based ANN for improved rotor flux estimation of induction motor drive

Local minimum is an integrated problem in training of artificial neural networks (ANNs) and the speed of convergence is very slow due to this effect. To avoid this problem, the chaotic variations of learning rate (LR) are included with the conventional learning rate. In this paper, chaotic variations of LR have been included with the learning rate of three algorithms such as backpropagation (BP), Real Time Recurrent Learning (RTRL) and Correlated Real Time Recurrent Learning (CRTRL) algorithms to estimate the rotor flux components of induction motor drive accurately. All the algorithms mentioned above generate a chaotic time series with logistic map. This paper is an initiative application of chaotic learning based ANNs for rotor flux estimation of induction motor drive. It is found that chaotic learning based ANNs is very much effective to estimate the rotor flux components of induction motor drive at both of transient and steady state conditions.