Wind Speed Prediction in Bingol Region: A Deep Learning and Stacking Ensemble Approach

This study aims to develop a novel model for wind speed prediction by integrating advanced deep learning techniques with ensemble methods using wind speed data collected from various districts of the Bingol region. The methodology includes rigorous data preprocessing, time‐based feature engineering, STL decomposition, and standardization – all mathematically modeled. A hybrid deep learning model comprising Conv1D, LSTM, and attention mechanisms is implemented alongside a stacking ensemble approach that integrates predictions from Ridge, Random Forest, XGBoost, LightGBM, CatBoost, SVR, and MLP regressors. Model performance is evaluated using RMSE, MAE, R², and EVS, with each district’s data supported by specific mathematical analyses.