Ultra-short-term Wind Power Combination Forecasting Model Based on MEEMD-SAE-Elman

The accurate prediction of wind power can effectively reduce the wind-free rate of wind power and improve the safety, stability and economy of the power grid. This paper proposes a modified Integrated Ensemble Empirical Mode Decomposition (MEEMD)-Sparse Auto-encoder (SAE)-Elman ultra-short-term wind power combination forecasting method, Which improves the steps of EMD in decomposition process of MEEMD algorithm. using cubic triangular Hermitian interpolation. The MEEMD is used to decompose the wind power raw data into IMF components, the sample entropy value of each IMF components are obtained and the similar of them are combined to reduce the amount of calculate in modeling; Then, a deep learning SAE prediction model is established for high frequency components, while, for low frequency components, an Elman neural network prediction model is introduced; Finally, the predicted value of wind power is obtained by overlaying the former data and analyzing error. Through the comparison of several prediction models in the example, the prediction accuracy of the combined prediction model is improved, which can be applied to the ultra-short-term prediction of wind power.