Numerical design of a wind observer and feedforward control of wind turbines

Abstract

Wind speed is an unknown variable that is very difficult to measure relying on instruments that are commonly present on-board the wind turbine nacelle. The alternative to this approach is to extract a wind speed estimate from other measured quantities. The so-obtained wind speed could then be integrated in common power control algorithms as an additional scheduling variable or it can be exploited to realize feedforward actions to enhance the machine power capture and fatigue life. This work deals with the design of an observer to estimate the spatially-averaged wind speed on the rotor swept area. A linear model of the wind turbine drivetrain based on steady aerodynamics is introduced and used as the base for the wind speed observer. The observer is implemented on the DTU 10MW RWT and a feedforward action based on the estimated effective wind speed is added to the standard feedback controller. FAST co-simulations are performed to evaluate the performances of the implemented system in several wind turbine operating conditions. Its responsiveness is shown to be greater when operating in presence of above-rated winds due to the higher rotor aerodynamic sensitivity. The observer response is instead slower in below-rated winds, where its capability of predicting high-frequency wind gusts is limited by the combination of rotor inertia and generator torque control.