Implementation of an enhanced M5p tree controller based on ANN data applied to a doubly-fed induction generator

The present paper concerns the indirect control of the stator powers of a wind system based on a doubly-fed induction generator (DFIG). The DFIG is controlled by its rotor through an association of a grid side converter with a rotor side converter, and the stator makes it possible to supply a resistive load. The current 1.5 kw generator is driven by a wind emulator based on a DC motor. Simulation and experimental studies are carried out using, first, a conventional proportional integral, a neural network controller (NNC), and then a M5P decision tree algorithm (M5P-DTA) is proposed to bring improvements to the control. The M5P-DTA is obtained from a learning process via the dataset provided by NNCs. The proposed algorithm allows a fast and less complex control scheme for the DFIG. The simulation study and its results are obtained through the MATLAB/SIMULINK software, while the experimental test is carried out via the dSPACE DS1104 interface card ordered by MATLAB and the graphical interface of the Control-Desk software.