ANN Based Prediction of Wind and Wind Energy

Accurate wind power energy forecasting in the electric grid supply provides a significant tool for optimizing operating costs. In this paper, wind power forecasting of one turbine at the Sidi Daoud wind farm (Tunisia) is treated by the Artificial Neural Network (ANN) technique. Measured data from this wind farm have been used to develop an ANN solution. This model shows a satisfactory estimation of wind power production under variable wind conditions. After filtering and treating measured data, the most influential are considered as ANN inputs: specifically the wind speed and direction, the temperature, the pressure and the previous power production. As weather prediction is not available at this farm, a recurrent Neural Network is proposed using a real data base to predict the wind speed and direction for two hours ahead. This prediction allows the creation of a short-term wind power energy forecast. Results thus achieved are tested with real data