Building Power Grid Models from Open Data: A Complete Pipeline from OpenStreetMap to Optimal Power Flow

A team of researchers from academia has released a complete pipeline that builds realistic, OPF-solvable transmission grid models entirely from open data. The five-stage process extracts power infrastructure from OpenStreetMap via a local Overpass API instance, reconstructs bus-branch topology using voltage inference and transformer detection, estimates electrical parameters with EIA plant-level data, allocates hourly demand from EIA-930 using US Census population as a spatial proxy, and solves both DC and AC optimal power flow using a progressive relaxation strategy that automatically loosens constraints on imprecise models. The pipeline was validated on all 48 contiguous US states and six multi-state regions, including the full Western Interconnection (5,076 buses) and Eastern Interconnection (21,697 buses).

Results show 88% of single-state models converge at the strictest relaxation level for AC-OPF at peak hour, rising to 92% off-peak. Median dispatch costs of $22/MWh and system losses of 1.0% align closely with real wholesale-market outcomes. The pipeline relies exclusively on public data sources, making it fully reproducible. All 54 models (48 single-state and 6 multi-state) are publicly released at the provided GitHub link, enabling researchers, utilities, and policymakers to conduct grid analysis without needing proprietary, restricted datasets. This work addresses a long-standing barrier in power systems research — access to realistic transmission network models under critical-infrastructure regulations.