A Three-Stage Ensemble Short-Term Wind Power Prediction Method Based on VMD-WT Transform and SDAE Deep Learning

Accurate wind power prediction (WPP) will contribute not only to the economic dispatching of power system but also to the safe and stable operation of the power grid. A novel three-stage ensemble short-term WPP method is proposed in this paper to effectively improve the WPP accuracy. First, in stage one, variational mode decomposition (VMD) and wavelet transform (WT) are applied to decompose the original wind power sequence into multiple subsequences. Second, in stage two, multiple stacked denoising auto-encoders (SDAE) are constructed based on the subsequences to perform WPPs separately. Third, in stage three, and the support vector machine (SVM) is applied to assign weights to each sub-prediction value to obtain the final ensemble prediction result. The case study shows that, compared with back propagation neural network (BPNN) and SVM, the average normalized root mean square error (NRMSE) and normalized mean absolute error (NMAE) of proposed ensemble WPP method, in the step range of 12 hours, are reduced by 19.66% and 19.91% compared to BPNN, and 14.43% and 14.65% compared to SVM, respectively, which illustrates the effectiveness and advancement of the proposed method.