Prediction models for wind speed at turbine locations in a wind farm

In wind farms, individual turbines disturb the wind field by generating wakes that influence other turbines in the farm. From a control point of view, there is an interest in dynamic optimization of the balance between fatigue and production, and an understanding of the relationship between turbines manifested through the wind field is hence required. This paper develops models for this relationship. The result is based on two new contributions: the first is related to the estimation of effective wind speeds, which serves as a basis for the second contribution to wind speed prediction models. Based on standard turbine measurements such as rotor speed and power produced, an effective wind speed, which represents the wind field averaged over the rotor disc, is derived. The effective wind speed estimator is based on a continuous–discrete extended Kalman filter that takes advantage of nonlinear time varying turbulence models. The estimator includes a nonlinear time varying wind speed model, which compared with literature results in an adaptive filter. Given the estimated effective wind speed, it is possible to establish wind speed prediction models by system identification. As the prediction models are based on the result related to effective wind speed, it is possible to predict wind speeds at neighboring turbines, with a separation of over 700 m, up to 1 min ahead reducing the error by 30% compared with a persistence method. The methodological results are demonstrated on data from an off-shore wind farm. Copyright © 2011 John Wiley & Sons, Ltd.