Energiforsk 2026-1165 Stokastisk Driftsäkerhet Transmissionsnät (2026)

Energiforsk research report presenting the mathematical methodology for calculating operational risk (driftrisk) in transmission networks — the technical foundation for moving from the deterministic N-1 criterion toward stochastic/probabilistic operational security assessment.

• Title: Stokastisk driftsäkerhet för transmissionsnät
• Authors: Math Bollen, Zunaira Nazir, Sarah Rönnberg (Luleå tekniska universitet)
• Report number: 2026:1165
• ISBN: 978-91-89918-48-1
• Type: PhD-project research report (part of Energiforsk’s Risk- och tillförlitlighetsanalys programme)

Core thesis

N-1 has been very successful in transmission operation but never quantifies risk; the risk can be too high but also too low, because a state that fails N-1 triggers changes in operation (a producer paid to start, a producer disconnected, effects on the electricity price) that, since the risk is unquantified, may turn out not to have been needed. Stochastic methods can give additional information alongside N-1 and might in the longer term support operating decisions, but the authors say the methods are far from ready to take over from the deterministic methods (§1.3).

Key methodology

Operational risk equation: R = Σ P(c)·F(c) over all considered contingencies c — probability of the contingency times its severity factor, summed.

• Contingency (oförutsedd händelse): an unpredictable component failure. First-order = one component; higher orders = combinations, which need not occur simultaneously — what matters is the joint probability of all occurring within the lead time.
• Unavailability (Q): probability a given component is unavailable; used to build contingency probabilities via simplified independent-component formulas (which are shown to double-count overlapping states and can be off by roughly a factor of three at scale — see worked examples below).
• Observation point vs. period of interest: the point at which the grid’s current state is known, versus the future window being assessed — the gap between them is where forecast/outage uncertainty accumulates. Central to short-lead-time operational risk assessment; largely irrelevant to classic multi-year reliability analysis.
• Severity factor (allvarlighetsfaktor): quantifies consequences along three axes — technical (voltage/current/frequency vs. thresholds), customer-related (customers/kW/kWh disconnected, outage duration), economic (cost of disruption and of mitigation).
• Stochastic dependence: common-mode/common-cause outages, cascading failures, and protection failures (skyddsfel, including “hidden failures” only revealed by a second, unrelated fault) — the independent-component assumption used in the basic formulas can significantly underestimate risk where dependence exists (§7.1.1: outages in the transmission grid are often caused by common-mode, cascading or protection failures; under a storm, independent outages are less dominant).

Key worked numbers

• A 250-component network has ~1.81×10⁷⁵ possible availability combinations (2²⁵⁰) — described as more than all the world’s computers running in parallel could evaluate within the age of the universe. Table 1 in the report gives combination counts by contingency order (2nd/3rd/4th) for networks of 25/50/100/250/1000 components.
• Independent three-component example (Q₁=0.1, Q₂=0.2, Q₃=0.3): summing all simplified contingency probabilities plus the “no outage” probability gives 1.22 — mathematically impossible for genuine probabilities, illustrating the double-counting in the simplified formulas.
• 100-component network at 1% unavailability each: the simplified second-order probability formula overestimates true probability by roughly a factor of three.
• Worked calculation examples use the IEEE 39-bus New England test network, varying load 100%–200%, showing different risk components rise and fall independently with loading — the reason multiple severity factors are needed for a complete picture.

Practical barriers to adoption

Four barriers identified: (1) computation time, binding hardest for short-lead-time (real-time/hour-ahead) applications and worsening with network size; (2) presentation of results — needs research and cross-disciplinary collaboration (data visualisation, grid and risk expertise) so it is immediately clear which events and components contribute most; (3) interpretation of results — not enough experience yet to develop automatic methods such as machine learning, and no experience of what counts as high or low risk (§7.3.3); (4) realistic data, especially failure and repair rates, with statistics lacking for most component types in high-failure-rate periods such as storms (§7.3.4). Separately, §7.2.3 names severity-factor standardisation — defining suitable severity factors, needing collaboration between researchers and method users — as the authors’ biggest challenge, and §7.4 calls the absence of methods or criteria for judging whether a computed risk is acceptable a serious barrier.

Explicit flexibility connections

• Foreword: stochastic operational-risk methods are named as important infrastructure for implementing “smart-grid approaches like flexibility markets, large- and small-scale battery storage, demand curtailment, and end-customer participation in electricity markets.”
• §7.1.2: forecast errors also exist on the consumption side; “especially for new types of consumption such as vehicle charging, demand flexibility and customer battery storage” they can be large (the report does not say they are growing) — noted alongside wind/solar forecast error, which matters most for day-ahead planning and maintenance planning rather than the operating hour.
• §2.3: one of the listed possible applications of day-ahead operational risk assessment is to support operational planning, to assess the need to intervene in the electricity market, or to decide whether activation of flexibility markets is needed — a listed use case, not a described implementation.

EU/ACER context (independently confirmed)

§1.2 traces the same European thread as Source - Säkerhetskrav Hinder För Elektrifieringen (2024): ACER raised the need for probabilistic transmission risk methods in 2019, leading to an ENTSO-E working group that has published two “state of play” reports (2021, 2023), with a third expected late 2025 — work so far has focused mostly on data-collection methodology, with risk-calculation methods still at an early stage. The report’s authors note their project developed independently of the ACER/ENTSO-E process and reached “quite similar conclusions,” treating the convergence as reassuring. This report does not itself state a 2027 implementation deadline — that figure comes only from the SINTEF/Second Opinion interview.

Scope note

The report’s own definition of “transmission network” (§1.3) covers not just Svenska kraftnät’s own grid but more generally networks at 130 kV and above in a Swedish context — though the methods are noted as applicable at lower voltage levels and in industrial networks too (per Article [F], a CIRED 2023 conference paper on industrial networks).

Relevance to wiki

• N-1 Criterion — primary technical source; the operational-risk equation, contingency/severity-factor definitions, combinatorics example, and the practical barriers are all drawn from this report
• Congestion Management — day-ahead risk assessment as an input to flexibility-market activation decisions
• Flexibility Need Assessment — demand-side forecast error (EVs, flexibility, batteries) as a factor operational risk assessment must account for

Data gaps

• The third ENTSO-E “state of play” report (expected late 2025) — check for updated implementation timeline once published
• Whether any Swedish TSO/DSO control room has since piloted these methods operationally