Artificial Neural Network Based Speed Estimator for Sensorless Field Oriented Control of Three Phase Induction Motor

Abstract

The presented work proposes an effective speed estimation scheme based on time series neural network for field oriented controlled three phase induction motor drives. This method utilizes the d and q axis voltage supplied by the SVPWM inverter, electromagnetic torque and rotor flux as inputs to the neural network. The proposed scheme reduces the amount of computational complexity than the Model Reference Adaptive System (MRAS) based speed estimation and provides a better result due to universal function approximation capability of neural networks. Accurate speed estimation at zero and 10 rad/sec proved the effectiveness of the proposed scheme at low speed region. The proposed estimator performs satisfactorily at all the speed levels within the specified motor speed range. At all the four quadrants, the proposed estimator showed stable performance. The performance and stability evaluation of the proposed neural network-based speed estimator was performed in MATLAB/Simulink.