Reference Models

A 90-bus, 254-line synthetic network for an invented service territory. Corresponds to no real system.

What this network is rated for

• RobustDirection of change in new build by technologySame agreement study: direction agreed in 44 of 44 paired cases. Magnitude did not, and is not claimed.
• RobustRelative system cost between two capacity portfoliosAgreement study, OpenGrid 2026: the sign of the cost difference between portfolio pairs survived transfer to the synthetic network in 41 of 44 paired cases. The three that did not were all within 0.4% on the real network.
• FragileLocal congestion and nodal price outcomesThe topology is plausible at system scale and arbitrary at any particular node. A congestion result here is about this network and transfers to nothing.
• UnknownResource adequacy and loss-of-load expectation

The GB transmission network as mapped in OpenStreetMap — real substations, real corridors — with every electrical parameter estimated, because the source has none.

Share-alike · ODbL-1.0 This model's licence obliges derived works to carry the same terms. ODbL share-alike operates at the database level, so joining it to your own data may bring that data into scope — an open question for adopters, not a settled one. Read the licence before you build on it.

What this network is rated for

• RobustWhere transmission corridors run, and which substations they connectThis is the part that is measured rather than derived. Routes, terminals, nominal voltage and circuit counts are OSM survey data carried through unchanged; the coordinates are the mapped substation locations, not synthesised positions.
• RobustWhich locations are topologically peripheral or singly fedA property of the graph, and the graph is the measured part. Reading it does require the transformers: over lines alone the extract falls into 48 pieces, because OSM models a substation as one bus per voltage level, and a reader who cut the network on lines alone would find isolation everywhere and believe it.
• FragilePower flow magnitudes and line loadingsFlow divides in inverse proportion to impedance, and every impedance here is a class average from a per-kilometre table. Two parallel corridors of the same voltage and length get identical values whatever their real conductors are, so the split between them is an artefact of the estimator. The 45 underground circuits are worse: they carry an overhead type, whose capacitance is an order of magnitude too low and whose reactance is too high.
• UnknownContingency severity and reserve requirements