An Ensemble Learning Model for Ultra‐Short‐Term Wind Power Prediction Based on Variational Mode Decomposition and Adaptive Weighting

ABSTRACT The volatility and randomness of wind power generation pose significant challenges to grid integration and dispatching. Accurate wind power forecasting is crucial for ensuring grid security and optimizing dispatching strategies. This study proposes an ultra‐short‐term wind power forecasting model based on Variational Mode Decomposition (VMD) and an adaptive weighting integration mechanism. The model first employs VMD to decompose wind power time series into multiple Intrinsic Mode Functions (IMFs) to extract latent multiscale features. Subsequently, Convolutional Neural Networks (CNN) and Long Short‐Term Memory networks (LSTM) are utilized to independently model the IMFs, capturing spatial features and temporal dependencies. The outputs of CNN and LSTM are then dynamically weighted using an attention mechanism to adaptively enhance the representation of critical features. Finally, the predictions of CNN and LSTM are integrated using a Gradient Boosting Decision Tree (GBDT) meta‐learner to produce the final output. Experimental results on wind farm data from Xinjiang, China, demonstrate that the proposed model significantly outperforms traditional and state‐of‐the‐art forecasting methods in terms of accuracy and robustness, especially in handling the nonlinear and nonstationary characteristics of wind power.