OL3 outage · SE1 structural shift · Nordic hydrological deficit · CHP phase-out · Transmission failure — probabilistic framework and compound scenario assessment
This note documents five concurrent structural risks in the Finnish electricity system and their compound interaction in the 2026–2030 window. Each risk is individually documented and monitored. None has been assessed in combination. The compound scenario — OL3 unplanned outage during a cold-still period, with SE1 transmission at saturation, Nordic hydrological deficit at historical low, and CHP capacity declining — produces an EPP shift from approximately 0.20 (live baseline, 4 May 2026) to 0.70–0.85 without additional load growth. With DC load growth of 1,500–3,000 MW by 2030 (Vattenfall estimate: DC share rising to 25% of total consumption; Mattila report November 2025 projects significant capacity growth through 2030), the same compound scenario approaches the BP-like threshold. This is not a forecast. It is a structural diagnosis: the system has no actor responsible for monitoring the compound state, and no instrument that aggregates these four risks into a single posture indicator. TN-009 proposes a probabilistic risk component (FRP) for integration into WEM EPP v4.
Olkiluoto 3 contributes approximately 1,600 MW of firm capacity — the largest single generation unit in the Finnish system. Its contribution to EPP FirmShare is substantial: loss of OL3 reduces FS by approximately 20 percentage points, shifting EPP by 0.15–0.25 points independently of other conditions.
Outage probability follows a Poisson process. For an EPR-class reactor in early operational years, λ ≈ 0.7 per year is a conservative estimate, giving P(outage in winter window, 120 days) ≈ 0.23. Duration distribution is heavy-tailed: 70% resolve within 24 hours, 25% within 1–5 days, 5% extend beyond 5 days. The 5% tail dominates the risk distribution. Average expected duration E[D] ≈ 43 hours.
The scheduled maintenance window for OL3 in 2026 is 10 September–30 October (50 days). OL1 is in maintenance 19 April–13 June (55 days). The risk profile for winter 2026–2027 depends on whether unplanned outages coincide with the cold period.
OL1 is currently in planned maintenance (−900 MW). OL3 is operational. WEM §03 shows nuclear W24 at 2,798 MW — consistent with OL3 full output and OL2 operational. The planned OL3 autumn maintenance window (September–October 2026) creates a structural vulnerability during the early heating season, before typical winter demand peaks.
The SE1→FI interconnection has a nominal NTC of approximately 1,500 MW. WEM §12 recorded SE1→FI flow at ~1,576 MW on 4 May 2026; corrected TRR ~68% against Aurora-updated NTC of ~2,300 MW — the connection is operating at saturation during a normal spring day. This is not a stress event; it is the new normal.
The structural cause is documented in Svenska kraftnät's Grid Development Plan 2024–2033: electricity consumption in northern Sweden (SE1 bidding zone) is growing faster than generation, driven by Stegra (700 MW electrolysis), HYBRIT, LKAB expansion. Svenska kraftnät projects that the SE1→FI flow direction will shift from export to import in 2027 — meaning SE1 will move from being a supplier to being a competitor for the same interconnection capacity Finland currently relies on.
Historical precedent: Svenska kraftnät significantly curtailed SE1→FI capacity in 2021, in violation of EU Regulation 2019/943 (minimum 70% of operationally secure capacity for cross-zonal trade). The Finnish Energy Authority documented this as an illegal restriction. The mechanism for recurrence exists — it does not require political intent, only physical congestion in northern Sweden's internal grid.
SE1→FI TRR corrected to ~68% following Aurora Line commissioning (November 2025, +800 MW NTC). High utilisation on a normal spring day — not saturation. The emergency buffer exists but is substantially committed at normal load. Under peak winter demand with compound stress (OL3 outage, cold period), the corrected NTC is reached at approximately 2,100–2,300 MW total flow — still a binding constraint. The structural trend (SE1 industrial load growth) continues to erode this buffer towards 2028.
As of March–April 2026, the Nordic hydrological balance is at its lowest level in ten years, with a deficit of approximately −27 TWh relative to the long-term average. WEM §11 records Norjan reservoaarit at 32.4% versus a median of 58.0% (NVE Week 17 2026). The change is rapid: the hydrological balance was +17 TWh (surplus) in winter 2024–2025.
Finland's domestic hydropower capacity is approximately 3,200 MW installed, producing roughly 10–14 TWh annually. Finnish reservoirs follow the Nordic pattern. WEM §03 records Finnish hydro W24 at 763 MW — consistent with spring low-flow conditions. The Finnish hydro contribution to FirmShare is real but weather-dependent: in a dry year, FS(p) — the probabilistically adjusted FirmShare accounting for hydro reservoir level — is materially lower than the nominal FS value.
The critical T–W correlation applies here: cold, calm periods (which drive highest system stress) correlate with low precipitation and low snow accumulation. A dry winter that produces the hydrological deficit also tends to produce the wind drought. These are not independent risks.
Combined heat and power (CHP) provides approximately 3,200 MW of semi-firm generation capacity with a positive temperature correlation — output is highest precisely when winter demand peaks. WEM §03 records CHP W24 at 3,200 MW and CHP/kulutus at 41.3%. WEM §10 backtest data shows CHP share declining from 36.5% (January 2022) to 25.5% (December 2024).
Helen's planned transition from CHP to electric boilers (700 MW) replaces dual-output capacity (electricity + heat simultaneously) with single-output load (heat only, consuming electricity). The net effect on the electrical system is negative: the same heat demand now requires electrical input rather than generating electrical output. The ECI semi (firm + CHP / total production) masks this transition in aggregate statistics — it appears as a CHP reduction but the system effect is a simultaneous capacity loss and load increase.
§ 01eTransmission failure risk divides into two structurally distinct categories, each with different probability distributions, durations, and system impacts. Both are relevant to the compound scenario framework.
The Baltic Sea has experienced nine submarine cable and pipeline cuts in 14 months across three separate incidents (October 2023 NewNew Polar Bear, November 2024 Yi Peng 3, December 2024 Eagle S). The Estonia–Finland Estlink 2 cable failed on 25 December 2024, reducing cross-border capacity from 1,016 MW to 358 MW for six months at a repair cost of 50–60 million euros. A separate Estlink 2 fault in January 2024 required seven months of repair. Fenno-Skan 2 (800 MW, Finland–Sweden) experienced a substation fault in March 2026 (4-hour outage) and a construction damage incident in April 2023. Estlink 1 experienced a substation fault in September 2025.
The critical structural observation: whether a transmission failure constitutes an endurance risk for Finland depends on the direction of power flow at the moment of failure. The Estlink 2 December 2024 failure occurred while Finland was exporting 650 MW to Estonia — the result was a slight price decrease in Finland and a significant increase in Estonia. The same cable failure during a Finnish import period would have had the opposite effect. With SE1→FI at ~68% TRR under normal spring conditions (corrected post-Aurora NTC ~2,300 MW), a Fenno-Skan HVDC cable or Finnish 400 kV north–south transmission failure would constitute a direct endurance event under compound winter stress — when TRR approaches 100% — but not under normal spring operation.
SE1 transmission failure is an endurance risk for Finland if and only if TRR > 0 in the import direction at time of failure. Current state: TRR ~68% (corrected for Aurora Line). The directional risk condition is active when TRR approaches 100% — which occurs under compound winter stress, not normal spring operation. The same physical failure event is benign when Finland is exporting and critical when Finland is importing. WEM §12 TRR provides real-time monitoring of this condition.
Finland's main grid comprises approximately 14,900 km of transmission lines and 138 substations. Fingrid recorded 241 disturbances in 2025, of which 151 were caused by lightning or natural phenomena. Transmission reliability in 2025 was 99.99995% — a historical record. Under normal conditions, individual line faults are self-healing or resolve within minutes.
The compound risk mechanism is different. A sustained cold period (T < −15°C, duration > 72 hours) with concurrent storm activity creates simultaneous stress across multiple components: line icing increases fault probability, repair crews face operational limitations, and demand is at maximum precisely when system margins are thinnest. The 1975 Alajärvi 400 kV fault divided Finland in two for one hour during winter conditions — Kemijoki hydro was the sole supply for northern Finland. The historical record (Tapani 2011, Eino 2013, Seija 2013) documents storms exceeding 200,000 customers interrupted. None of these coincided with a nuclear unit outage and SE1 saturation simultaneously — the compound has not been empirically tested.
The weather conditions that maximise 5b risk (cold, storm, icing) partially overlap with the conditions that maximise OL3 risk and SE1 saturation. A severe winter storm during a cold-still period reduces transmission reliability, increases demand, and occurs in the same meteorological regime as wind drought. These are not independent risks in the compound scenario.
Individual assessment of each risk understates the system vulnerability because the risks share structural correlations. The worst compound scenario is not the sum of individual worst cases — it is the intersection of conditions that are physically correlated.
The 0.5% per winter figure appears small. The correct interpretation is actuarial: over a ten-year horizon, P(at least one compound event) ≈ 1 − (0.995)^10 ≈ 5%. With DC load growth adding 1,500–3,000 MW of flat load by 2030, the same compound trigger produces a materially worse outcome — the system has less margin to absorb the shock.
| Scenario | EPP Baseline | EPP Compound | Classification | CELP¹ |
|---|---|---|---|---|
| Current state (4 May 2026, live) | 0.20 | — | Normal | ~2% |
| OL3 outage, normal conditions | 0.20 | 0.35–0.40 | Tight | ~15% |
| OL3 + cold | 0.20 | 0.50–0.65 | Elevated | ~60% |
| OL3 + cold + wind drought | 0.20 | 0.70–0.85 | BP-like | ~90% |
| Same + 1,500 MW DC (2027) | 0.30 | 0.80–0.92 | BP-like | ~95% |
| Same + 3,000 MW DC (2030) | 0.45 | 0.87–1.00 | BP-like | ~99% |
| Two-unit outage: OL1+OL3 simultaneous (planned+unplanned overlap) | 0.20 | 0.65–0.80 | BP-like (no cold needed) | ~75% |
¹ CELP (Compound Event Loss Probability): conditional probability that, given the compound trigger occurs, demand exceeds firm generation capacity for ≥1 hour within the event window. This is not an annual outage probability — it is a within-event severity indicator. Finland's annual LOLP (per official Fingrid assessment) is <0.01%. CELP 2% at baseline reflects the small fraction of normal winter hours where spot spikes require demand response activation. CELP ~90–100% under the full compound scenario means that if OL3 fails during a cold-still-low-hydro period, capacity shortfall in at least one hour is nearly certain within that event window.
Sensitivity of the compound probability estimate. The 0.5% per winter figure is sensitive to the input assumptions. OL3 failure rate λ=0.7/year is based on early EPR operating experience; TVO data from 2023–2026 would improve this estimate. P(T < −15°C) = 0.08 is a Finnish climatological approximation; the threshold definition (duration matters: a 2-hour excursion differs from a 5-day cold spell). P(wind drought | cold) = 0.25 is a rough T–W correlation estimate. Combining uncertainties, the compound probability range is approximately 0.1–2.0% per winter (factor-of-4 uncertainty). The qualitative conclusion — that the compound risk is not negligible and is structurally unmonitored — is robust across this range. The specific numbers (0.5%, EPP values) should be treated as order-of-magnitude estimates, not precise forecasts.
The fat-tail structure dominates. The P90 Expected Energy Not Served (EENS) is approximately 8 GWh; the P99 is 80–150 GWh. A standard reliability planning exercise based on expected values systematically underestimates this risk by 2–3x because it does not model the T–W–hydro correlation structure.
§ 03Each individual risk is monitored by a competent institution. Fingrid monitors OL3 outage probability and adjusts SE1 NTC accordingly. NVE publishes weekly reservoir statistics. Svenska kraftnät publishes capacity plans. TEM monitors CHP decommissioning timelines.
No institution monitors the compound state. No instrument aggregates these four risk indicators into a single posture assessment. The Mattila data centre report (November 2025) assessed the electricity system from a data centre perspective and concluded that electricity will not run out. This conclusion is defensible on its own terms — it was not the wrong answer to its question. The question did not ask what happens when OL3 is offline during a cold–still–dry period with SE1 at capacity and CHP declining. That question was not in the commission.
This is not a criticism of Mattila, Fingrid, NVE, or Svenska kraftnät. It is a structural observation: the commission architecture guarantees that the compound question is never asked, because no single actor's mandate covers the compound state. This is the measurement gap in SM-009 §09 applied to physical infrastructure risk.
The Finnish electricity system faces four concurrent structural pressures in 2026–2030, each individually monitored and each individually manageable. Their compound interaction — particularly the OL3 × cold × wind drought × SE1 saturation scenario — produces a system state that has no current monitoring instrument and no owning institution. The probability of the full compound scenario is low (~0.5% per winter) but the tail risk is severe (LOLP approaching 100%, EENS P99 at 80–150 GWh). With DC load growth of 1,500–3,000 MW by 2030, the same compound trigger operates with zero margin. The diagnostic recommendation is not a specific policy action — it is first to measure the compound state. An actor must be designated to monitor it.
The compound risk framework identifies when a system is structurally vulnerable. What is missing is an explicit decision trigger — a rule that converts the monitoring state into an action signal. Without this, TN-009 remains an analytical instrument rather than an operational one.
The Compound Alert Trigger (CAT) is a binary signal derived from the existing WEM components. It fires when the compound state crosses an operationally significant threshold:
CAT condition (A) captures the normal compound: high endurance pressure sustained over multiple hours. CAT condition (B) captures the dynamic compound: a system moving rapidly toward stress regardless of current absolute level. CAT condition (C) captures the specific compound scenario of TN-009 §02 — the three-factor intersection that produces CELP approaching 90%.
The CAT threshold values (0.65, 6h, +0.15/24h) are initial estimates requiring calibration against historical data. The structural logic is independent of specific threshold values: the trigger should fire when compound conditions are building, not after they have peaked.
§ 04The compound risk can be approximated within the WEM EPP framework by adding a Failure Risk Premium (FRP) component to EPP v4:
This extension is operationally compatible with the existing WEM §11 structure. SP_cluster and FS(p) are already computed. The NVE proxy is already integrated. FRP requires only the addition of a scalar multiplier applied to the existing EPP calculation when regime conditions are met.
§ 05OL3 λ calibration: The λ = 0.7/year estimate is based on EPR-class reactor operating experience in early commercial years. TVO's actual outage history for OL3 since April 2023 commissioning should be used to calibrate this parameter. Planned maintenance windows are known (ENTSO-E REMIT); unplanned outage rate is the key uncertainty.
SE1 NTC reduction mechanism: Svenska kraftnät's Grid Development Plan projects the flow direction change by 2027. The precise mechanism — whether through reduced NTC allocation, internal congestion management, or market design — determines how quickly the transition occurs and whether it is gradual or step-change.
Finnish domestic hydro reservoir data: WEM currently proxies Finnish hydro via Norjan NVE data. A direct Finnish reservoir level data source (Kemijoki Oy, TVO, Fortum) would improve FS(p) accuracy. The correlation between Norwegian and Finnish reservoir levels is high but not perfect.
CHP retirement timeline: Helen's electric boiler transition affects Helsinki's system. Tampere, Oulu, and other cities have their own CHP timelines. The aggregate national CHP retirement schedule is not published in a single source.
§ 06In structural testing, individual component tests — static load, dynamic load, fatigue — each produce pass results. System integration testing under combined loading reveals failure modes that no individual test predicts. The standard test protocol for a composite structure does not ask "what happens when all load cases apply simultaneously during a temperature excursion?" — because that question exceeds the scope of any single component specification.
The Finnish electricity system has passed its component tests. OL3 operates. SE1 transmits. Fingrid balances. Helen heats. Each actor within its mandate. The integration test — what happens when these systems interact under compound stress — has not been specified, and no actor has been assigned to run it.
This is not a prediction of failure. It is a diagnosis of an untested condition. In structural engineering, the appropriate response is not to assume the structure will hold — it is to run the test.
Finland is not only a recipient of transmission risk — it is also a critical infrastructure provider for the Baltic states. Following the April 2025 Baltic desynchronisation from the BRELL grid, Estlink connections to Finland are the primary security anchor for Estonian, Latvian, and Lithuanian power systems in a crisis scenario. Finnish decisions on domestic generation capacity, transmission investment, and grid resilience therefore carry systemic consequences beyond Finland's own endurance window. This interdependence runs in both directions: Baltic instability affects Fingrid's balancing operations and Estlink utilisation patterns. The compound risk framework in this note is necessarily Finland-centric; a complete regional analysis would model the mutual exposure across the Nordic-Baltic system as a whole.