Predictive model of yaw error in a wind turbine

The yaw position of a wind turbine is adjusted in response to the changing wind direction for maximum energy extraction. A data-mining approach is proposed to predict wind direction. To accommodate the full range of yaw motion, the wind direction data is transformed into two time series (sine value and cosine values). Parameters of the time series are selected for predictive modeling. Four data-mining algorithms are applied to build prediction models. Industrial data is used to develop, validate, and test the proposed models. Computational experience with data representing four seasons and four sampling frequencies is reported in this paper.