A data-driven approach for fatigue-based individual blade pitch controller selection from wind conditions

In a context of wind power production growth, it is necessary to optimize the levelized cost of energy by reducing the wind turbine operation and maintenance costs. This paper addresses these issues through an innovative data-driven approach, applied to individual pitch control and based on wind conditions clustering, from light detection and ranging (LiDAR) wind field reconstruction. A set of controllers is first designed, and a surrogate model is fitted to predict the economic fatigue cost of the wind turbine in closed-loop for each of these controllers, given a cluster of wind conditions. This allows online selection of the controller minimizing mechanical fatigue loads among the candidates for each wind condition. Preliminary tests show promising results regarding the effectiveness of this method in reducing wind turbine fatigue when compared to a single optimized individual pitch controller. The main advantages of this approach are to limit the sensitivities to controller tuning procedure and to provide an economically driven control strategy based on fatigue theory that can be effectively adapted to different wind turbine systems.