Wind turbine power prediction method based on Optuna-XGBoost and SHAP interpretability analysis

Accurate forecasting of wind power is essential for enhancing the efficiency of energy scheduling strategies. Traditional physical models, characterized by high computational demands and limited predictive accuracy, are insufficient for meeting this objective. To resolve these limitations, this study introduces a novel wind power prediction approach based on an Optuna-optimized XGBoost algorithm, which achieves a remarkably high coefficient of determination R²≈ 0.999. Additionally, the SHAP method is applied to interpret the model’s outputs, quantitatively assessing the impact of each input feature and thereby reinforcing the interpretability and analytical robustness of wind turbine power forecasting.