Roadmap

Roadmap

The strategy is unchanged since day one: a small, correct core, grown deliberately into a complete grid model, with every step validated before the next.

Done

1. Power flow — steady state, with generator Q-limits (PV→PQ switching) and OLTC voltage regulation
2. Live simulation — the UK model runs continuously, demand following a day/night curve, re-dispatching and re-solving each tick
3. Frequency data — streaming NESO CSV reader + RoCoF upsampler (1 s → 20 ms)
4. Frequency dynamics — centre-of-inertia swing equation with governor droop and load damping; trip a unit and watch RoCoF, nadir and recovery
5. Boundaries — ETYS-style transmission boundaries with MW transfer limits (B6 Scotland–England, etc.) and branch loading
6. Sparse solver — direct sparse LU (RCM + partial pivoting), scales to the ~6,000-bus GDA replica
7. The GDA replica — GridSim.Gda builds the real NGED network from the LTDS archive; all 52 GSP supply areas solve to steady state (16.56 GW, 8.4 s)
8. The national fleet — TEC/interconnector/embedded registers → 943 units, 134.8 GW, real system inertia driving the dynamics model

In progress

• Whole-licence-area meshed solve. Solving an entire DNO licence area as one meshed block is bounded but not yet fully converged — a raw multi-GSP distribution model needs OLTC voltage support and careful infeed conditioning to settle as one system.

Next

• Network detail — line ratings in the solve, richer N-1 contingency analysis, shunts.
• Scale — the ILGPU/CUDA backend behind ILinearSolver (see Architecture), then many-scenario runs (N-1 across the full replica, year-round half-hourly studies).
• Real data, deeper — GDA adapters for UK demand profiles and generation-mix time series, so the live simulation runs on real shapes rather than the synthetic daily curve.
• Other DNOs — UKPN/SPEN/SSEN/NPg/ENWL sources are wired (ImpedanceConvention handles the per-DNO units); they activate as their LTDS data is populated on the archive.

The long game

A national model — transmission backbone + every DNO's distribution network — solvable fast enough to ask "what if" questions in bulk: inertia scenarios, boundary stress, contingency sweeps, storage placement. The physics core is deliberately small and validated so that scaling it is an engineering problem, not a correctness one.