Implementation of the Adaptive Neural networks Backstepping controller for a stand-alone DFIG based WECS
Résumé
In this article, we propose an adaptive neural network Backstepping controller for a stand-alone doubly-fed induction generator (DFIG) based WECS. Our main objective is to control the stator voltage magnitude and frequency via the rotor voltages, first a new state-space model representation is developed by adding capacitive loads in parallel with the three-phase machine. The proposed technique is independent of the system model and guarantees the robustness against speed and load variations. The neural network is used to approximate online the nonlinear uncertainties of the system. The chosen architecture is a feed forward with one hidden layer, it is simple to implement and doesn't require much computational force. The Lyapunov approach is used to ensure the stability of the proposed control and neural adaptation laws. The experimental tests are conducted on a 1.5kW DIFG using a 4.5 kW wind emulator and the dSPACE 1104 card. The obtained results show the effectiveness of our strategy compared with the previous works done on the same testbed using PI and the Backstepping technique.