Using Radial Basis Neural Networks to Estimate Wind Power Production

Abstract

This paper compares two statistical methods for short-term wind power forecasting applied in a real wind farm located on complex terrain. The methods require as input past power measurements and meteorological forecasts of wind speed and direction (Numerical Weather Predictions or NWPs) interpolated at the site of the wind farm. Both methods include NWPs estimator models based on fuzzy logic and wind power forecasting models using neural networks combination.