Study on the multi-step forecasting for wind speed based on EMD

The multi-step forecasting for wind speed is more significant than the single-step one in practice. Aiming to improve the forecasting precision, a novel forecasting approach for wind speed based on empirical mode decomposition(EMD) was presented and its validity was investigated in this paper. The original wind speed sequences were decomposed into some intrinsic mode functions(IMF) in advance employing EMD. As a result, the embedded random, periodic or trend components can be extruded from the original wind speed sequences. Further, these components were restructured as three partitions with different frequency to raise the efficiency of building models. According to the exhibited characteristics of the three partitions, the corresponding forecasting models were built. The forecasted results for each partition were superposed as the future wind speed. The proposed approach was applied to the wind speed from a given wind farm and its advantage in multi-step forecasting has been obviously displayed.