Improved Grey Model with Rolling Method for Wind Power Prediction

The wind power industry has achieved rapid growth in recent years. The wind power prediction is essential requirement for the safety and stability of the power system. A new method based on grey model for ultra-short term wind power prediction is presented. The grey prediction model is suitable for online operation became of few data for modeling and fast computation. The procedure of GM(1,1) and mathematical formulation are also clarified. The rolling mechanism updating data in the way of equal dimension is proposed in order to rapidly track the data change. According to the different number of raw data used in the modeling, the effect on the prediction accuracy is discussed. The error characteristic is researched by means of setting five kinds of background values. The background value is proposed to be set as the right endpoint instead of the traditional trapezoid construction value. The actual data of wind turbine's output power in some wind farm is selected for ten-minutes ahead prediction. The results demonstrate the improvement of prediction accuracy with the proposed parameters of the grey model.