In variable-pitch wind turbines, pitch angle control is implemented to regulate the rotor speed and power production. However, mechanical loads of the wind turbines are affected by the pitch angle adjustment. To improve the performance and at the same time alleviate the mechanical loads, a gain-scheduling fractional-order PID (FOPID), where a trained RBF neural network chooses its parameters is proposed. The database, which the RBF neural network is trained based on, is created via optimization of a FOPID in several wind speeds with chaotic differential evolution (CDE) algorithm. The simulation results are compared to an RBF based PID controller that is designed via the same method, a conventional gain-scheduling baseline PI controller developed by NREL, an optimal RBF based PI controller, and a FOPI controller. The simulations indicate that the RBF based FOPID improves the control performance of the benchmark wind turbine in comparison to the other controllers, while the applied loads to the structure are mitigated. To validate the performance and robustness, all controllers are implemented on FAST wind turbine simulator. The superiority of the proposed FOPID controller is depicted in comparison to the other controllers.
- Gain-scheduling fractional-order PID
- Wind turbine pitch control
- Chaotic differential evolution
- RBF neural network
- BLADE PITCH CONTROL
- ENERGY-CONVERSION SYSTEM