GIS-Driven Graphs and KD-Trees for Synthetic Distribution Network Construction

Distribution networks were not designed to accommodate high penetration of distributed energy resources (DERs), microgrids, and electric vehicles (EVs), resulting in operational challenges. Studying new topologies is essential, and synthetic distribution networks (SDNs) are key tools for modeling and validating present or future grid architectures. In this context, this paper presents a four-stage GIS-graph methodology that integrates graph topology with geospatial layers. A GIS-based graph is constructed from a street map and load points corresponding to customer coordinates. In the first step, the process validates street geometries and repairs disconnections. Then, it segments edges to respect realistic span limits, creating candidate support nodes aligned with typical pole placements. In the third stage, a two-dimensional KD-tree links each load point to its nearest support node. Finally, loads are grouped at support nodes and unnecessary elements are removed. The resulting graph-based network is topologically, spatially, and statistically congruent with the existing network; moreover, it is approximately 2.6 times smaller than networks produced with geometric approximations such as Delaunay triangulation and K-nearest neighbors (KNN).