Leveraging Explainable Artificial Intelligence (XAI) Methods Supporting Local and Global Explainability for Smart Grids

In recent decades, Artificial Intelligence/Machine Learning (AI/ML) methods have been applied in a variety of fields, from healthcare to finance, retail, energy, and many more, with remarkable improvements. However, AI-based solutions are still questionable due to concerns regarding their trustworthiness. Explainable AI (XAI) has become an emerging research field that addresses those concerns about trustworthiness, particularly for explainability and transparency. In this study, three XAI methods supporting local and global explainability, i.e. SHAP, PFI, and LIME, are utilized to investigate the key features and their impact on the model's outputs for solar photovoltaic (PV) power generation forecasting.