Winter Endurance Monitor · Compound Risk Analysis (TN-009) · Layered Economic Loss Function (TN-010)
Live data as of 5 May 2026. EPP values reflect a 24h rolling average and update on page refresh. Seasonal variation is normal — summer EPP 0.15–0.22, winter EPP 0.25–0.45 without stress events.
The system is in balance — not because the structure is strong, but because the load profile is easy. No cold demand, no wind drought, no nuclear outage. Under winter conditions (−15°C, OL3 unavailable, wind <500 MW), the same instrument moves to EPP 0.70–0.85 without additional structural change.
| Risk factor | Current state (2026) | Trend 2027–2030 |
|---|---|---|
| OL3 (1,600 MW) — largest single firm unit | Autumn maintenance planned | Single point of failure structure unchanged |
| SE1→FI transmission | TRR ~68% (spring), ~100% (winter peak) | SE1 industrial load growth — flow direction shifts 2027 |
| Nordic hydrology | NVE 32% vs median 58%; deficit −27 TWh | Lowest in 10 years — spring 2026 |
| CHP phase-out | 25.5% of consumption (down from 36.5%) | Electric boilers add load, remove firm capacity simultaneously |
| Data centre load | +500 MW flat (2026) | 1,500–3,000 MW by 2030; competes for diesel in crisis |
| Transmission faults (5a/5b) | Estlink 2 (Dec 2024), Fenno-Skan 2 (Mar 2026) | Baltic desynchronisation (Apr 2025) makes Estlink existential |
| T–W–hydro correlation | Cold + calm + dry = same weather regime | Three risks move in the same direction simultaneously |
Compound probability (OL3 + cold + wind drought + SE1 at capacity): ≈0.5%/winter (range 0.1–2.0%)
| Sector | Loss range | Mechanism |
|---|---|---|
| Continuous industry | €400–550M | Process damage + extended restart |
| Water and wastewater | €250–550M | Contamination risk; biological reset (β=1.5); recovery 3–5× outage |
| Households | €200–600M | Heating loss; freeze damage; insurance claims |
| Cold chain | €150–300M | Near-total warehouse spoilage after 24h |
| ICT, retail, healthcare, logistics | €600–1,200M | Cascade effects via fuel and telecom failure |
| Total 48h CEL | €1.6–3.2B | Central estimate ~€2.4B |
P90 (€17.5M/year) ≈ a typical electricity price VAR budget (€20M). P99 (€178M/year) = 9× VAR budget. Standard deviation (€118M) is 10× the mean — the distribution is fat-tailed and dominated by rare extreme events. Capital provisions based on expected value are structurally inadequate; the correct stress capital reference is P99.
Adding 2,000 MW of DC load doubles the 72h CEL from €2.5B to €5B. The mechanism is not load volume — it is cascade acceleration. DC→generator switching is always worse than keeping DCs on grid under compound stress: 2,000 MW of DC consumes approximately 450 m³/hour of diesel, directly competing with hospitals, water systems, and telecommunications for the same limited fuel supply.
Optimal strategy: 50–80% load reduction ("sleep mode") at 10–14 hours into the event. Not switching to generators. Not immediate shutdown (DC direct losses exceed cascade savings when action is taken before threshold crossings begin).
Missing instrument: Finnish data centre grid connection agreements do not include a CAT-1 sleep protocol. This is a coordination instrument requiring contract text revision, not a technical investment.
| State | CEL₉₀ | EPP* | Action | Time | Actor |
|---|---|---|---|---|---|
| CAT 0 | <€50M | <0.50 | Standard monitoring | — | WEM |
| CAT 1 slope | €50–200M | 0.50–0.65 | Readiness: fuel pre-position, DC sleep alert | 2h | Fingrid, HVK |
| CAT 1 level | €200M–1B | 0.65–0.75 | Active: DC →50%, demand response | 1h | Fingrid, TEM |
| CAT 1 compound | >€1B | >0.75 | Emergency: DC →20%, load control, civil authority | 30min | Fingrid, VNK |