Combined Probabilistic Prediction of Distributed Wind Power Based on WRF

At present, with the increase in the proportion of distributed wind power, the difficulty of its acceptance by the power grid has gradually increased, and the problem of wind abandonment has become prominent. Therefore, accurate prediction of distributed wind power is necessary. However, most of the current wind power forecasting technologies are aimed at centralized wind farms. Considering the characteristics of distributed wind power, this paper proposes a distributed wind power forecasting method based on WRF. This method obtains historical numerical weather prediction (NWP) data of distributed wind farms through WRF model back calculation, then analyzes the cross-correlation of NWP data and historical wind power and the autocorrelation of historical wind power to select the appropriate explanatory variables, and then trains the combined model of sparse Bayesian learning (SBL), kernel density estimation (KDE), and beta distribution estimation (BDE). Case study confirms the effectiveness of this method.