PowerBench Dataset – Part 1: Line Outage and State Estimation

PowerBench: Line Outage and State Estimation Datasets This dataset is part of the PowerBench benchmark suite for evaluating machine learning methods in power distribution networks. It include two of the three core learning tasks: Line Outage Detection Simulates realistic single and multi phase line outages across the IEEE 34-bus, 123-bus, and 8500-node distribution feeders. Includes outage event labels, network topology and sensor metadata. Supports graph classification, branch/node level localization. State Estimation Focuses on voltage magnitude estimation using sparse bus and branch sensor data under varying operating conditions. Suitable for node-level regression and graph-based imputation. All scenarios are generated using OpenDSS with Python automation via OpenDSSDirect.py. Networks are exported as .gml graphs compatible with NetworkX. Metadata on sensor placements is included. See the included README.md for individual directory structure and data loading instructions.