. The reason of this positive phenomenon
is that the proposed secondary decomposing strategy decreases
the instability of the original wind speed data considerably. Since
the raw wind speed data have been decomposed, the built Elman
neural network can realize the final high-precision wind speed
predictions; (b) the hybrid WPD-Elman model outperforms the
single Elman neural network. This indicates that the combination
of the WPD algorithm and the Elman neural network is a right
way to promote the nonlinear capacity of the single Elman neural
network in the wind speed multi-step forecasting; (c) the hybrid
WPD-FEEMD-Elman model precedes the hybrid WPD-Elman
model. The obtained results justify the proposed strategy consisting
of decomposing the wind speed time series using two successive
signal processing stages is effective to improve the
forecasting performance of the Elman model; and (d) the proposed
new WPD-FEEMD -Elman model can be programmed and realized
conveniently by adopting the Matlab platform