Abstract: At present, the widely used statistical prediction model for wind power needs a lot of the historical data for training to achieve acceptable accuracy, so it is not suitable for the new wind farms which are lack of historical data. This paper presented a wind power prediction method for power grid based on small sample set. Firstly, based on few historical data of wind farm, the general wind power prediction model was built with using support vector machine （SVM） method, and was used for the preliminary forecasting of wind power. Then, on the basis of the general model prediction, the characteristic parameters of regional wind farm were used to adjust and modify the general model for power grid, so as to obtain the prediction results of wind power for regional power grid. Finally, the practical examples verified the feasibility of prediction method based on small sample set. The results show that the prediction method based on small sample set with good accuracy is suitable for the power prediction of wind farm with few historical data, and can reduce the dependence of statistical prediction method on data during power forecasting.

Key words: wind power forecasting, power grid, small sample set, support vector machine, general model