Explainable Artificial Intelligence for Real-Time Stability Assessment in Smart Power Systems

Modern electrical grids face big operational challenges because many renewable sources and distributed generators are highly variable. To keep the grid reliable, real‑time stability assessment is essential. This project investigates a hybrid deep‑learning framework that combines a Graph Transformer (GT) and a Temporal Fusion Transformer (TFT) to evaluate smart‑grid stability. The approach builds a feature‑correlation graph that captures complex spatial relationships among grid nodes. The Graph Transformer processes this graph to learn spatial dependencies. At the same time, the Temporal Fusion Transformer models the time‑based dynamics from sequential smart‑grid measurements. A feature‑fusion layer merges the spatial and temporal representations, then passes the combined data to a neural‑network classifier that predicts

stability. Because deep‑learning models can act like black boxes, the framework adds SHapley Additive exPlanations (SHAP) to provide both local and global interpretability of the model’s decisions. This research‑focused implementation shows how to structure spatiotemporal models and offers a systematic way to evaluate model transparency, helping control‑center operators make informed, trustworthy decisions during transient events.

Message: Hi FE, I need help with “Explainable Artificial Intelligence for Real-Time Stability Assessment in Smart Power Systems” in “Electrical Engineering”