UA-Grid-Graph: an open graph dataset of the high-voltage transmission grid of Ukraine (OSM-derived) with topology metrics and a node-classification study

A graph dataset of the high-voltage transmission grid of Ukraine (the IPS of Ukraine backbone: 220 / 330 / 400 / 500 / 750 kV), extracted from the OpenStreetMap-based PyPSA-Eur prebuilt electricity network (Zenodo 10.5281/zenodo.14144752, ODbL) and restricted to Ukraine. 294 substations (nodes), 387 transmission lines forming 341 unique graph edges (parallel circuits between the same substations are merged in the simple graph; the raw 387 line records are shipped in ua_grid_edges.csv), 20,558 km, 15 connected components, 75 articulation points. The deposit ships the node table (with computed degree, betweenness, closeness, clustering, component id, articulation flag), the edge table, a graph-metrics summary, and a ranking of the most critical substations by betweenness centrality. A node-classification mini-study (is a substation part of the 330 kV+ extra-high-voltage network?) compares base_v1 (geography only) vs improved_v2 (+ graph-topology features): topology features lift every model (GBM F1 0.76 to 0.95, ROC-AUC 0.81 to 0.98), i.e. the grid topology encodes its voltage hierarchy. This instantiates the "full graph model of the energy system for anomaly detection in specific segments" called for in the author's digital-twin thesis (ITMM-2025). Real OSM-derived topology; not a simulation. Small graph (294 nodes) — ML numbers are indicative, the consistent cross-model gain is the result.