WIND TURBINE PITCH ANGLE CONTROL WITH ARTIFICIAL NEURAL NETWORKS

Abstract

Pitch angle control in wind turbines is required to obtain maximum efficiency from wind turbines at variable wind speeds. Since the wind turbine pitch control structure is not linear, the control cannot be fully achieved, and oscillations occur at the power output. This oscillation can increase because the pitch angle cannot be adjusted stably. This study employs pitch angle control using artificial neural networks, a proportional-integral-derivative (PID) controller, and adaptive neuro-fuzzy inference systems (ANFIS) methods. When the artificial neural network, PID, and ANFIS outputs are compared, it is evident that the system created using the artificial neural network yields better results than the PID. However, the best output is obtained with ANFIS pitch angle control. Two types of performance indices are used in the performance comparison: the error performance indices and the time response performance indices. Considering the control performance parameters, the maximum overshoot of the PID-controlled system is 0.68 %, while the maximum overshoot of the artificial neural network-controlled system is 0.48 %. The maximum overshoot of the ANFIS-controlled system is 0.46%. As a result, better system performance and a more stable power output are obtained compared to the studies in the literature.