Neural network speed controller of induction motor drive based on direct MRAC method

A novel neural network speed controller (NNSC) of induction motor drive is presented, which is on-line trained by a dynamic BP algorithm based on direct model reference adaptive control (MRAC) method. In particular, a new error function is used for the dynamic BP algorithm, which can adjust the parameters of NNSC and minimize the error between motor speed and the speed provided by reference model. A 2-rank linear model is used as reference model, which can be automatically designed according to performance indices given beforehand. The induction motor drive with a NNSC based on dynamic BP algorithm and direct MRAC method is proved to be asymptotic stable. The simulation and experimental results reveal that the proposed NNSC based on direct MRAC method can ensure motor speed tracking the reference speed closely, and having robustness against load disturbances.