Soft Starting of Induction Motors

Starting inrush current and pulsations in the induced torque affect the performance of an induction motor. Artificial neural networks (ANNs) and adaptive neuro fuzzy inference system (ANFIS) can enhance the performance of the motor by making a control system which would provide smooth starting to induction motor. Dynamic model of induction machine in different frames of reference was implemented using Matlab Simulink. Feed forward back propagation based and radial basis neural networks were trained, with data obtained using simulations, to estimate different parameters required by ANFIS to adjust firing angle of back-to-back connected pairs of thyristors in AC voltage controller. Inrush current and pulsations in torque were reduced significantly. Radial basis and feed forward neural networks were compared for off-line and on-line training, training time, memory required for implementations, number of neurons, computational procedures and algorithms, reliability of the system and most important cost of implementation. Artificial neural networks and Adaptive neuro fuzzy inference system were developed using tool boxes in Matlab Simulink.