Prediction of power generation from a wind farm

Abstract

Wind farms produce a variable power output depending on the wind speed. For management of power networks and for bidding for the supply of power, the future power available needs to be predicted for time intervals ahead of a few minutes to about 24 hours.

This project used data from a wind farm and three meteorological stations to determine methods and ability to predict wind speed. Analyses using regression, neural networks, and a Kalman filter were examined. Prediction using a combination of local wind measure-ments and meteorological data appears to give the best results.