Real-time neural block control for a doubly fed induction generator

This paper show the implementation of a control scheme based on the discrete-time block control technique using sliding modes, for a doubly fed induction generator (DFIG) connected to an infinity bus. In order to obtain the generator mathematical model, it is proposed to use a recurrent high order neural network (RHONN) identifier, which is trained with an extended Kalman filter (EKF) algorithm. Its performance is illustrated via a real-time implementation in a low power DFIG prototype.