Short-term wind power prediction with a new PCC-GWO-VMD and BiGRU hybrid model enhanced by attention mechanism.

The article focuses on the development and validation of a novel deep hybrid short-term wind power prediction model named PCC-GWO-VMD-BiGRU-Attention. This model integrates the Pearson correlation coefficient (PCC) for feature selection, gray wolf optimization (GWO) to adaptively optimize variational mode decomposition (VMD) parameters, and a bidirectional gated recurrent unit with attention mechanism (BiGRU-Attention) to enhance temporal feature extraction and prediction accuracy. Tested on a dataset from a wind farm in Turkey, the model demonstrated superior performance compared to ten other models, achieving a root mean square error (RMSE) of 0.2677 MW, mean absolute error (MAE) of 0.1509 MW, mean square error (MSE) of 0.0717 MW, and coefficient of determination (R²) of 0.9605. Additionally, the study employed deep ensemble methods to quantify prediction uncertainty, showing that the proposed model yields the most stable and reliable forecasts with the narrowest 95% confidence intervals. While the model requires longer training time due to its complexity, it offers significant improvements in prediction accuracy and robustness, making it a promising tool for wind farm operation and grid management.