Feature Selection and Explainable AI for Transparent Windmill Power Forecasting

Integrating wind power into the electrical grid is challenging because wind speeds are unpredictable. Accurate wind power forecasting is crucial for balancing electricity supply and demand. Our research aims to enhance forecasting approaches by employing a correlation-based feature selection technique to identify the most relevant features from the dataset. We evaluate several models, including Extreme Gradient Boosting (XGB), Random Sample Consensus Regressor (RANS), Partial Least Squares Regression (PLSR), Extremely Randomized Trees (ERT), Elastic Net (ENet), Least Absolute Shrinkage and Selection Operator (LASSO), Ridge Regression (RR), and Random Forest Regression (RF). Among these, the RF model exhibits outstanding performance, achieving an R2 score of 99.95%. To improve the interpretability and analysis of feature importance in these machine learning models, we utilize SHAPASH and ELI5 XAI (Explainable Artificial Intelligence) tools. This research highlights the efficacy of the RF model in forecasting wind power generation but also emphasizes the importance of model transparency and interpretability. By leveraging XAI tools, we offer valuable insights into the decision-making processes of these models, identifying the most influential features.