An EMD-RF Based Short-term Wind Power Forecasting Method

Wind power forecasting of wind field is a common problem recently. Forecasting wind power accurately is a challenge for dispatchers who need to establish dispatching strategies due to the randomness and volatility of wind power. This paper presents a kind of wind power forecasting method based on empirical mode decomposition (EMD) and random forest (RF). EMD is applied to this method to decompose wind power sequence into several intrinsic mode functions (IMF) and a residual component, and then RF is used to train each component. Finally, the predicting results of each component are summed together to obtain the wind power forecasting values. The proposed method is tested on actual data from a wind farm in America. The results show that the EMD-RF method reduces the forecasting error and tracks the change of wind power more accurately compared with the traditional forecasting model.