Real-time sliding mode control with neural networks for a doubly fed induction generator

This paper proposes a control scheme on the basis of the block control technique using sliding modes by means of neural networks identification, for a doubly fed induction generator (DFIG) prototype connected to an infinity bus. The DFIG is widely used as a wind generator; it allows the rotor speed to vary while synchronizing the stator directly to a fixed frequency power system. This generator has one back-to-back PWM voltage-source converter between the rotor and the electrical grid. The rotor side converter (RSC) is connected via a DC-link to the grid side converter (GSC), which is in turn connected to the stator terminals directly or through a step-up transformer. A high order neural network is used in order to obtain the DFIG mathematical model; then, based on this neural model, a block control schemes using discrete-time sliding modes (NNDTSM) is proposed for the RSC and the GSC. The performance of this scheme is evaluated by implementation in real-time using a 1/4HP DFIG prototype.