Beta Free while in beta — just sign up and go. It's built and run on finite personal resources, so please use it considerately.
GDA GB GRID PUBLIC DATA

Software

The GB grid, solved.

Two solved grids from GridSim's full-history Newton–Raphson power flow, both reconstructed purely from public data (2001–2026): a reduced model solved every second, and a full 2,227-bus UK network solved across the half-hourly periods.

GridSim replays a quarter-century of published GB grid data and solves the network state behind it. It does this at two fidelities — a fast reduced model for the whole second-by-second record, and a full 2,227-bus solve for detailed network snapshots. Both are on the data lake; the per-second grid is also downloadable a day at a time, below.

Model A · Reduced
10-zone GB, every second
806,716,800converged solves
  • Cadence every second
  • Converged 100.0000%
  • Rows / day 86,400
  • Coverage 2001–2026
  • Validation pandapower ~1e-12
Download a day ↓
Model B · Full network
2,227-bus UK, half-hour sweep
6,721,086bus-level solutions
  • Buses 2,227
  • Network solves 3,018
  • Cadence ~1 / day *
  • Coverage 2001–2026
  • Per solve V, angle, line load
On the data lake →

* Model B runs one full 2,227-bus solve per day, each ~30 minutes earlier than the last, so over time it sweeps all 48 half-hourly settlement periods. 3,018 network solves × 2,227 buses = 6,721,086 bus-level results. It grows as new snapshots are added.

Download the per-second grid, a day at a time

Model A only. Pick a UTC date and get that whole day as one CSV — the smallest useful slice, not one giant file.

One day → one CSV (~86,400 rows, a few MB). Available 2001-01-01 → 2026-07 · 307 months on host.

What's in the per-second grid (Model A)

One row per second of the reduced 10-zone GB model — the converged operating point:

index, time_utc, load_factor, converged, iterations, max_mismatch, total_loss_mw, vmin_pu, vmax_pu

e.g. /download/gb/solves/day?date=2019-01-09 returns every second of 9 January 2019. Times are UTC; converged is True throughout the published record.

The full 2,227-bus UK solve (Model B)

Where Model A trades detail for a solve every second, Model B keeps the whole 2,227-bus GB transmission network and solves a full power flow for a chosen half-hour each day. Each solve stores the complete network state, split across parquet tables on the lake:

full-uk/parquet/summary → one row per solve: time_utc, converged, iterations, total_loss_mw full-uk/parquet/bus_vm → per-bus voltage magnitude (2,227 buses) full-uk/parquet/bus_va → per-bus voltage angle full-uk/parquet/line_load → per-line loading full-uk/parquet/_meta → the 2,227-bus network definition

3,018 full-network solves so far (2001–2026), each ~30 minutes offset from the last so the set walks through every half-hourly period over time — 6,721,086 bus-level solutions in total, and growing. Get it by mounting the data lake and reading GDA-SOLVES/full-uk/.

Convergence is published per solve in the summary table; unlike the per-second grid, the full-network set isn't uniformly flagged across every month, so treat its converged column as the source of truth rather than assuming 100%.