GNN deep ensembles for N-1 contingency decisions

Graph Neural Networks (GNNs) have shown promise in modelling complex graph-structured data within power distribution systems. However, their application remains limited by the absence of uncertainty quantification (UQ) techniques, which are crucial for decision-making. This paper presents a deep ensemble framework for GNNs that generates uncertainty estimates in n-1 contingency criterion classification predictions, an important task for distribution system operators (DSOs) to evaluate network reliability. This paper demonstrates that our approach enhances prediction reliability by providing uncertainty measures, showcasing the effectiveness of deep ensembles for uncertainty quantification (UQ) in GNNs. This capability makes it a valuable tool for supporting decision-making in power system operations.