Modern wind turbine controllers use wind speed information to improve power production and reduce loads on the turbine components. The turbine top wind speed measurement is unfortunately imprecise and not a good representative of the rotor effective wind speed. Consequently, many different model-based algorithms have been proposed that are able to estimate the wind speed using common turbine measurements. In this paper, we present a concise yet comprehensive analysis and comparison of these techniques, reviewing their advantages and drawbacks. We implement these techniques and compare the results on both aero-servo-elastic turbine simulations and real turbine field experiments in different wind scenarios.

The time and measurement update for the discrete time Kalman filter can be formulated in terms of conditional means and covariances. The unscented Kalman filter can be interpreted as calculating these conditional means and covariances by using the unscented transform. This approach can also be directly applied to nonlinear models as an alternative to the discrete time extended Kalman filter. In this paper, a novel method for computing the unscented Kalman filter for a nonlinear model with continuous time dynamics and discrete time measurements is presented. Compared to the existing approaches, this method is far simpler and less computationally demanding, and it performs at least as well.