Scalable Architectures for Real-Time Outage Inference in Power Distribution Networks Using Dynamic Outage Graph Models

This paper introduces a comprehensive and scalable approach to outage inference in modern electric power distribution systems. As utilities increasingly face the challenge of managing complex network topologies – characterized by dynamic configurations, multiple power sources, and evolving operational states - traditional outage management systems often fall short in delivering timely and accurate diagnostics. The proposed method addresses these limitations by leveraging a network topology aware framework that enables real-time, parallelized processing of outage data, while also supporting on-the-fly updates to the network model. The solution centers on an advanced outage inference engine operating on a dynamically updated outage graph that captures the essential structure of the distribution network, enabling effective performance across complex, multi-source and non-radial topologies. Its innovation lies in combining four capabilities: a preprocessing stage that classifies graph edges to precisely trace outage origins, an incremental real-time graph-update mechanism that avoids costly full rebuilds, integration of real-time engineering data streams from AMI, SCADA, and protection devices, which improve the inference process by aligning customer trouble calls with actual grid conditions and reducing false positives while improving crew dispatch prioritization.