A pipeline for building a large-scale realistic transmission grid dataset from public data

Microsoft Research has released a pipeline and dataset that builds an approximate model of the US transmission grid using only public data. It uses OpenStreetMap's transmission corridors, substations, and power plants as the skeleton, and combines them with EIA energy statistics and Census data to create a model spanning 48 states and multi-state power systems.

The scale ranges from small systems with 11 buses to the Eastern Interconnection with 21,697 buses. The core validation criterion is AC-OPF convergence. Unlike toy networks that only look plausible in topology, this model must simultaneously satisfy power supply-demand balance and line, voltage, and generator constraints, demonstrating that computationally feasible solutions can be found at continental scale using only public data. It is not a replica of the actual operating grid, nor is it intended for market forecasting or real-time operations.

This model is used for analyses such as:

• Where new transmission lines could physically be installed
• Which lines should be prioritized for expansion
• Where large-scale datacenter loads could be sited

In the contiguous US, of 31,488 transmission corridors, 27,506 were single-circuit, and the densest corridor had 10 circuits (30 conductors). This shows that some segments in urban areas already have limited room for physical expansion.