Wind Turbine Inflow Estimation Using Blade Loads and Scada: Steps Towards Real-World Viability

Abstract Accurate wind field estimation is a valuable input to wind farm control. This paper advances a previously developed blade-load-based wind sensing methodology, bringing it closer to real-world application. First, an additional feature is introduced to the localized linear regression model, and its improved performance is verified through comparison of the estimation error with the original method. Second, the method’s practical usefulness is assessed by testing it with higher-fidelity inputs than those used for training. The training data uses medium-fidelity Mann turbulence and the Dynamic Wake Meandering model for wind fields, combined with an aero-servo-elastic solver based on a Blade Element Momentum method. The high-fidelity testing data uses Large Eddy Simulations for wind and a coupling of the Actuator Line method with the servo-elastic part of the previous solver for turbine response. Wind field estimates from three implementations of varying fidelity are analyzed qualitatively and quantitatively. Mean error values show that the change in wake representation between training and implementation is the primary source of inaccuracy, whereas differences in ambient turbulence structure have a less significant impact. Based on these findings, a potential training setup being a trade-off between computational cost and accuracy, thereby supporting the method’s viability for field deployment, is discussed.