ENTSO-E Third Probabilistic Risk Assessment Report (2025)

ENTSO-E’s third biennial progress report on Probabilistic Risk Assessment (PRA) — the EU-mandated effort to develop a probability-based complement to the deterministic N-1 criterion for transmission grid operational security. Published December 2025. Read in full (26 pp.).

• Type: Biennial progress report (third in the series; prior editions 2021, 2023)
• Full title: Biennial Progress Report on Operational Probabilistic Coordinated Security Assessment and Risk Management
• Legal mandate: CSAM (methodology for coordinating operational security analysis) Article 44(1) and 44(2)

Key findings

The legal deadline is for developing the methodology, not implementing it — this corrects a claim elsewhere in the wiki. CSAM Art. 44.1 requires: “By 31 December 2027, all TSOs shall jointly develop the methodology on common probabilistic risk assessment… and shall propose it as an amendment” to CSAM, subject to regulatory approval under SO Regulation Art. 7. The report’s own conclusion states: “It is expected that the PRA methodology will be drafted by 2027 as a collaborative effort by all TSOs, with implementation occurring post-2027.” (an expectation, not a commitment) N-1 Criterion previously cited a single SINTEF/Second Opinion interview claiming all European systems would “have implemented” probabilistic methods by end of 2027 — this report indicates that claim conflates the methodology-drafting deadline with actual implementation, which the report explicitly places after 2027 with no further date given (the report also says the timeline in its Figure 4 is subject to change).

Origin and governance: CSAM was established by ACER’s decision in June 2019. ENTSO-E’s Working Group Probabilistic Risk Assessment (WG PRA) was formed under this mandate; its terms of reference were approved April 2021 and revised November 2023. It has grown to 20 TSOs + 2 regional coordination centres (from a smaller founding group), organized into three workstreams: (1) reporting/stakeholder management, (2) infrastructure/data collection, (3) methodology/definition development (split into 3A probability computation, 3B impact assessment).

PRA is explicitly a complement to N-1, not a replacement — at least in its first implementation phase: “In its first implementation phase, PRA will be used in parallel with the ‘N-1’ criterion.” Risk is defined as probability × impact per contingency; the methodology would map all contingencies against a harmonised “acceptable risk” threshold (still undefined — see below).

Progress since 2023:

• Probability computation: WG PRA tested the VAFFEL (“varsel før feil” / warning-before-failure) concept, developed by Statnett, for wind- and lightning-driven overhead-line failure probability. MAVIR (Hungary) and TenneT NL tested it on their own data; MAVIR extended it to icing (though without enough data to validate that extension). VAFFEL uses Bayesian updating over historical failure data and medium-range weather forecasts. It computes probability only — no impact assessment.
• Impact assessment: a proof-of-concept (POC) built in Power Factory (Python API) is in testing; first results expected early 2026. The POC uses DACF common grid models from TSCNET, does not yet model SPSs, and treats a non-converging load flow as unacceptable impact; the report says the POC is needed “to conclude on the benefits” of PRA, and warns that in a highly meshed system an unsolved N-1 may nearly always cascade, in which case PRA “will not add more value than the deterministic approach.” The POC focuses first on day-ahead and intraday assessments. The POC flowchart: trigger event → N state → N-k state → load flow → special protection scheme (SPS) check → operational-limit-violation decision → impact indicators (loops back if violations found).
• Data collection: 27 of 40 TSOs provided 2023 grid disturbance data as of the 2025 survey (9 partial, 2 needing dedicated help). ~17% of reported faults still carry an “unknown” cause. Coverage is deliberately focused on 220 kV+ (100–150 kV data is sparse since many TSOs don’t own/supervise those assets, which often interlace with DSO grids).

Other TSOs’ independent PRA work (outside WG PRA, surveyed at a January 2025 workshop; WG PRA notes only VAFFEL has actually been tested, and that some of these projects lie outside its scope because they model load/generation uncertainty rather than component failure probability): Swissgrid’s ROSPRA (outage-planning risk tool, traffic-light output); German TSOs’ exceptional-contingency cascade analysis and a separate PRA project (started May 2024); RTE’s long-term outage-planning method (Antares-Simulator, 200 yearly adequacy scenarios); PSE’s long-term resilience/cascade-simulation tool; ELES’s long-term planning concept (judged too simple for WG PRA’s real-time use case).

Open questions the report itself flags: when/how PRA will actually be implemented (the report says data quality should be considered when setting an implementation deadline, and the POC comparison of PRA vs N-1 will inform the strategy); how operators will use PRA outputs in practice; how to price the cost of risk-mitigating remedial actions into the methodology. No accepted risk-acceptance threshold exists yet — most TSOs still fall back to N-1 or qualitative judgment when asked how they’d define “acceptable risk.”

Roadmap: POC results to TSOs in 2026; ACER/NRA alignment workshop planned H1 2026 (a first such workshop was already held mid-2025); next biennial report in 2027 (the Figure 4 timeline also marks a “Biennial report” at 2026), alongside “Update CSAM regarding PRA” scheduled for mid-2026 to 2027; the timeline is stated to be subject to change.

No Sweden-specific implementation status is given anywhere in the report — Svenska kraftnät is not named. The report is organized by workstream and by TSOs presenting independent projects (Statnett’s VAFFEL, ELES, RTE, Swissgrid, the German TSOs, PSE; MAVIR and TenneT NL as VAFFEL testers); Statnett is the only Nordic TSO named.

Relevance to existing wiki topics

• N-1 Criterion — corrects the page’s “Target date” claim: end-2027 is when the methodology must be drafted and proposed for regulatory approval, not when TSOs must have implemented it (implementation is explicitly “post-2027” per the report’s conclusion, with no further date stated). The page’s data gap asking for independent corroboration of the “end of 2027” figure is now answered — but the answer is that the figure needs correcting, not just corroborating.
• Grid Capacity Utilization — the same deterministic-to-probabilistic shift recurs at distribution level (hosting capacity); this report’s implementation timeline (post-2027, no fixed date) suggests the transmission-level shift is not a fast-moving near-term event.