Research on Very Short-Term Nonparametric Probabilistic Prediction of Wind Power Based on Enhanced LSTM and Improved Bootstrap

It is reliable for high-precision wind power generation prediction method to provide reference for the scheduling and control of power systems. The unsteadiness and uncertainty of wind energy have adverse effects on the stable operation of power system. To quantify the uncertainty of wind power, this paper proposes a new probabilistic prediction method, based on enhanced long short-term memory (E-LSTM) and the improved Bootstrap. Firstly, a new wavelet threshold function is proposed to denoise wind power data. Secondly, a deterministic power prediction method is proposed, based on the self-adaptive inertia weights, learning factor and LSTM. Then, an improved Bootstrap method is proposed to expand the resampling range of the observation points, for the improved accuracy of the sample distribution. After that, a nonparametric probabilistic prediction model based on the proposed deterministic prediction is established. Finally, the proposed method is validated through multistep prediction intervals.