21-09-2017, 12:58 PM
Indirect vector control of the induction motor is simulated and velocity is estimated using the conventional Adaptive Model Reference System (MRAS). It is modified by the PI controller of the neural network. A MRAS speed estimator based on a conventional mathematical model can give a relatively accurate speed estimation result, but an error will occur during the low frequency operation. It is also very sensitive to variations of machine parameters. Therefore, a two-layer neural network PI controller (NNPIC) is used instead of PI controller. With the help of the projection algorithm, the NNPIC parameters are automatically adjusted and the difference between the two MRAS models is minimized for speed estimation. MRAS estimator based on neural networks gave a robust performance during low frequency and parameter variation. In addition, this scheme reduced the work of the tuning mechanism of the PI controller. The estimated velocity was taken as feedback and velocity was controlled by indirect vector control using spatial vector pulse width modulation (SVPWM). The results of the simulation showed an improvement in the performance of an induction motor.