An Automated Data Validation Approach for Power Distribution Networks using Grid Partitioning and Multi-faceted Quality Scoring

Abstract. Accurate Geographic Information System (GIS) data is fundamental to the reliable management, planning, and operation of modern power distribution networks. Conventional validation methods, however, often rely on network-wide rule-based checks or manual inspections, which are inefficient at identifying and localizing errors within vast, heterogeneous infrastructures. These approaches frequently fail to detect complex spatial or topological inconsistencies, leading to significant operational challenges and costly data remediation efforts. To address these limitations, a novel, automated validation pipeline has been developed with a modular, two-stage approach. The first stage, Smart Grid Partitioning, spatially divides the network into manageable cells using either fixed-size grids or a density-aware dynamic partitioning. This dynamic mode employs a bottom-up, clustering-inspired algorithm that adapts grid sizes to the local intensity of network equipment, effectively resolving issues of data sparsity and overload. The second stage, AI-Assisted Grid Validation, calculates a comprehensive Correctness Score for each resulting grid. This score provides a quantitative measure of data quality by synthesizing four weighted factors: (1) configurable rule-based attribute checks, (2) connectivity file conformance, (3) topological integrity assessed via advanced network trace functions, and (4) a series of representative graph-theoretic metrics. By generating an intuitive, color-coded map of data health, our framework allows utility providers to precisely localize data quality issues and prioritize remediation efforts. This targeted approach significantly enhances the efficiency of data maintenance, improves the integrity of foundational GIS data for critical power infrastructure, and streamlines integration with essential platforms like SCADA, OMS, and DMS.