A Hybrid Method Based on Empirical Mode Decomposition and Random Forest Regression for Wind Power Forecasting
Gamze Ogcu Kaya
As a renewable energy source, wind energy is abundant, environmentally and friendly. But due to intermittent nature of wind, accurate prediction of wind power is quite difficult. On the other hand, accurate prediction of wind energy production is very significant for integration of wind power to the power grid. This study addresses hourly wind energy production forecasting by proposing a hybrid method based on two powerful techniques: empirical mode decomposition (EMD) and random forest regression (RFR). This is the first study which uses EMDRFR hybrid method for wind power forecasting. The performance of the proposed method is tested on real wind power data of one of the pioneering energy companies in Turkey. The proposed model was verified by comparing it with other intelligent models. The experimental results demonstrate that the EMD-RFR model outperforms the single RFR model, the single SVMR model and the EMD- SVMR hybrid model based on different performance measures.
Keywords: Wind power, forecasting, empirical mode decomposition (EMD), random forest regression (RFR)