Neural Network Observer for Sensorless Direct Torque Controlled Induction Motor Drive

Direct torque control (DTC) allows control of torque and flux which involves estimation of torque and flux from measured values of current and voltage. Speed measurements are necessary for generation of reference torque and flux for control action. Sensors are disadvantageous as they increase cost and decrease reliability. In this paper henceforth, a neural network observer is presented for torque/speed and flux estimation in a direct torque-controlled induction motor drive. Neural network on one hand improves cost by eliminating speed sensor and on the other hand improves reliability by making speed/torque and flux estimation independent of motor parameter variations. MATLAB simulation is carried out on a 15Hp, 400V, 50Hz, three phase induction motor (IM) drive to verify the efficiency of proposed control strategy.