Wind power output exhibits strong randomness, volatility, and intermittency, and high-precision short-term wind power forecasting is urgently needed to ensure the safe and stable operation of the large power grid under the new power system. The time-series characteristics of wind power data were fully utilized, and a short-term wind power stacking learning ensemble forecasting method was proposed, based on pearson correlation coefficient, variational mode decomposition, and fuzzy entropy. First, the pearson correlation coefficient between wind power and meteorological variables was used to identify the main meteorological variables. Then, the wind power sequence was decomposed into subsequences using the variational mode decomposition algorithm. Fuzzy entropy values were used to classify subsequences into high-frequency or low-frequency subsequences, and stacking ensemble learning forecasting models suitable for high-frequency or low-frequency subsequences were established separately. Finally, the predictions were obtained through aggregation and reconstruction. Practical examples demonstrate that the proposed method reduces the mean absolute error by 6.0% to 27.5%, significantly improving the accuracy of short-term wind power forecasting.