Wind power estimation using recurrent neural network technique

The estimation of wind power generation at 1-hour horizon based on the time series wind data obtained on a 50m hub height at Slangkop site have been presented. It is believed that an hourly wind power forecast at Slangkop site will help for developing well functioning day-ahead markets, unit commitment decisions, maintenance and transmission scheduling etc. The Layer Recurrent Network (LRN) is used to predict the long term wind speed and power output of a 1.3MW wind turbine in real time over 1-hour horizon of up to 288 hours ahead (12days). The LRN provides an accurate prediction of the wind speed and power outputs, as the predictions are compared to the expected wind speed and power. The forecast model returns a mean square error value of 0.156% and a mean square error value of 0.009%.