Scalable Topology Inference for Power Distribution Grids

Maintaining accurate topology records for power distribution grids is crucial for efficient operations like outage localization and voltage analytics, but it's often challenging due to diverse and imperfect utility data. Existing methods for topology identification frequently rely on electrical similarity or spatial records alone, which can become unreliable in dense urban feeders or with inconsistent metadata. To address this, a new framework formulates distribution topology identification as a constrained inference problem. Instead of rebuilding connectivity from scratch, it refines an existing base topology using various data sources while enforcing spatial feasibility and physical operational constraints. This approach detects inconsistencies, performs localized reconnections for scalability, and iteratively ensures operational consistency. Validated with operational data from a large U.S. utility, the framework achieved over 95% accuracy, significantly reducing computational effort compared to global inference methods, and proved more robust than correlation-based methods in complex environments.