NORDIC GRID ROBUSTNESS

Methodology & validation

Public model definition, data provenance, validation status and reproducibility information for the indicator shown on elnatet.nu.

Back to dashboard

What the headline number means

The current 0-100 initial-disturbance margin is a model-derived indicator. Higher is better. The headline value is the weakest score among five fixed event-response cases when the required catalogue is complete.

It is not a probability or an official security classification. A value such as 80/100 does not mean that the grid is "80% safe". It is not an official Svenska kraftnät, Fingrid, Statnett, JAO or SMHI assessment, a blackout probability, an N-1 compliance result or a control-room decision variable.

Current model identity

Model
2.1.0
Profile
fixed-five-event-response-v2.1
Scoring construct
fixed-five-event-response-v2.1
Score-profile SHA-256
fba67e0308b51c8c28526456a327e6e31638852f7a73b492e856dda3a7546b11

The machine-readable stable parameterization is published as model-v2.1-config.json .

The complete implementation source code is not published on this site. The active scoring configuration, version identity, runtime artifacts, provenance rules, method analysis, decision record and validation limits are public.

Five score-driving cases

Case Magnitude Interpretation
Nordic reference incident up to 1450 MW Generic reference-incident proxy, capped by Swedish nuclear availability, including a synchronous lost-inertia proxy.
Hydro unit loss up to 500 MW Generic hydro-unit loss proxy with synchronous lost-inertia proxy.
Finland import dependency up to 1450 MW Measured aggregate Finnish net-import dependency. Zero import is a valid inactive state.
Wind common-mode loss 35% of current Swedish wind view Generic common-mode wind-generation loss stress.
Large-load disconnection 800 MW Generic large-load disconnection stress.

Fourteen checks are visible, but only these five drive the headline. Mapped HVDC cases, direction summaries, the extreme stress case and JAO network constraints remain context/stress checks.

Scoring method

For a simple non-synchronous active-power disturbance the modeled initial rate of change of frequency follows:

|RoCoF| = 50 x |P| / (2 x E x 1000)

Synchronous generator-loss proxies reduce post-event kinetic energy by the configured lost-inertia estimate before RoCoF is evaluated.

The continuous response transform is anchored at:

The headline is the minimum complete score among the five fixed cases. Nordic kinetic energy is also displayed as physical context, but is not applied a second time as a separate score limiter.

Data provenance

Input Provider Role
Nordic kinetic energy Fingrid Open Data, dataset 260 Score input. Current/post-event kinetic-energy basis.
Finland net exchange Fingrid Open Data, dataset 194 Score input. Measured Finnish import dependency.
Swedish production/consumption view Statnett operational data Score input. Nuclear, hydro and wind event sizing, with published source timestamp/freshness retained.
Nordic frequency Fingrid Open Data, dataset 177 Displayed/validation context, not a direct v2.1 headline input.
Physical area-to-area flows Statnett PhysicalFlowMap Mapped HVDC-link and directional context only.
Nordic Flow-Based CNE/PTC JAO Publication Tool Nordic API Day-ahead grid-model context, not treated as a real-time emergency margin.
Reserve procurement Svenska kraftnät Mimer FCR-D/aFRR/mFRR diagnostics only, not scored.
Weather SMHI Open Data, SNOW1gv1 Storm/wind-exposure diagnostic only, not scored.
Historical production/consumption eSett Open Data Retrospective pre-live reconstruction. Early 2017 remains partial where source data are unavailable.

Current source freshness and degraded-state information is available in v2-source-health.json . The live snapshot also retains source timestamps, quality and provenance fields.

Historical and live evidence policy

Before 2026-08-28T08:18:48.589150Z , public long-term evidence uses frozen retrospective provider-record reconstruction.

From that cutover onward, canonical collected v2.1 live evidence is authoritative and retrospective reconstruction is not allowed to replace post-cutover observations in public statistics.

Post-cutover statistical intervals count only when both bounding observations are current, complete catalogue scores and collector spacing is at most 7.5 minutes. Degraded, stale, carried-forward, cold-start and collector-gap periods remain explicit missing statistical coverage rather than silently receiving statistical weight.

Validation status

Implementation verification: the stable v2.1 release passed the automated regression suite, fixed-catalogue invariants, frozen-source checks, provenance gates and exact rc2-to-stable numerical parity. All 325,035 score-bearing predecessor state vectors reproduced the exact same floating-point score, with maximum difference 0.0 .

Release verification also checks that canonical collected-live evidence remains append-only, earlier version/profile provenance is preserved, the frozen pre-live baseline contains no post-cutover retrospective statistical evidence, and prior annual summaries remain unchanged when score physics has not changed.

Scientific validation limit: the composite 0-100 robustness indicator has not been independently prospectively validated as a predictor of Nordic power-system security outcomes. Stable denotes implementation/release maturity, not official or empirical certification of the 0-100 scale.

Reproducibility and audit trail

The visual dashboard is not the only public evidence. Versioned machine-readable artifacts expose the active assumptions, current runtime state, historical evidence policy and release decision.

Stable model configuration Current live snapshot 72-hour evidence Long-term evidence Source health Method analysis JSON Method analysis Markdown Decision record JSON Decision record Markdown

Scope exclusions

Return to Nordic Grid Robustness