Assessment of power curves in models of wind power forecasting

The modeling of wind power curve is important in turbine performance monitoring and in wind power forecasting. There are several techniques to fit the power curve of a wind turbine, which can be classified into parametric and nonparametric methods. This paper explores the advantages offered by the nonparametric methods for modeling of wind turbine power curve, because the application of nontraditional techniques enhances the accuracy and is easy to implement. We study two nonparametric methods to estimate the wind turbine power curve: Artificial Neural Networks (ANN) and Fuzzy Inference System (SIFs). Actual case studies were carried out from two wind farms in northeastern Brazil. The models were adjusted to forecast the wind power with steps from 1 to 24 hours ahead. When compared to the reference models, the developed models' gains lay in the range of 29 to 60%, for forecasts in a period of twenty-four hours ahead. The results were compared and have shown that the models created have a very promising performance.