Eine stilisierte Importregion versorgt sich über Pipelines, LNG und Gasspeicher. Würfeln Sie viele Winter, lassen Sie Lieferanten ausfallen und drosseln Sie die Nachfrage. Der Fächer der Speicherverläufe zeigt, wie wahrscheinlich ein Engpass ist und wann er eintritt. Achten Sie auf den Unterschied zwischen Bestand und Fluss.
A stylised importing region is supplied via pipelines, LNG and gas storage. Draw many winters, knock out suppliers and curb demand. The fan of storage trajectories shows how likely a shortfall is and when it arrives. Note the difference between stock and flow.
Security-of-supply stress test for a stylised importing region · 1 Oct → 31 Mar · Monte Carlo over winters · stock vs. flow
| Lever (applied on top of the current scenario) | P(shortfall) | Δ P | Δ expected unmet |
|---|---|---|---|
| Re-runs the current scenario once per lever with that lever pushed one notch further (same seeds, common random numbers). Press the button to compute. | |||
Every winter is simulated day by day from 1 October to 31 March. Demand is a fixed non-heating block (industry, power, other) plus a heating term proportional to
heating-degree-days (max(0, 14 °C − T), saturating in deep cold). The temperature path is a seasonal mean curve shifted by the severity setting, an
AR(1) daily anomaly (φ = 0.85, σ = 2.5 °C) and Poisson-arriving cold spells of 5–14 days and 3–8 °C depth. Firm supply is the sum of pipeline A and B,
LNG (base × spot factor, capped by regasification capacity, with small arrival noise), domestic production and biomethane. Whatever demand exceeds firm supply must
come out of storage — limited by the remaining stock and by a withdrawal-rate curve that is flat at 8.5 TWh/d above 40 % fill and falls linearly below (≈ 7 TWh/d
at 30 %, ≈ 4.7 at 10 %). Any remainder is unmet demand. Surplus firm supply is injected (max 4 TWh/d). N winters are drawn with a seeded generator
(mulberry32, seed + winter index), so every reload and every lever comparison uses the same random numbers.
Stock vs. flow. Storage is a buffer, not a source: it can shift energy in time but cannot create it. A full store buys time — the days of cover tile — but if the flow deficit persists, the stock drains and, worse, the achievable withdrawal rate drops as pressure falls. That is why 30 % matters: below it the store may still hold hundreds of TWh yet be unable to deliver them fast enough on a cold day. The fan chart shows the whole distribution of trajectories; the KPIs read off its tail.
Limits of the toy model. No network topology or regional bottlenecks, no price response (demand destruction, LNG re-routing by price), no interruptible contracts or demand-side auctions, no electricity-system coupling beyond a single switching slider, deterministic capacities apart from the chosen outages, independent winters. Real security-of-supply assessments model hourly flows, infrastructure (N-1) standards and market behaviour. Treat the numbers as shapes, not forecasts.
A single scenario answers "what happens" — a Monte Carlo answers "how likely, and how bad in the tail", which is the question a board actually has to decide on. Mature analytics teams work in scenario discipline: explicit assumptions, seeded reproducibility, one-lever-at-a-time attribution, and decision thresholds agreed before the crisis (at 30 % fill, at 20 days of cover). The model may be simple; the maturity lies in treating uncertainty as the output rather than as noise around a point estimate.
Research and teaching demo — not policy or investment advice. The region is stylised ("a large importing region, roughly European scale, rounded figures"); capacities and demand are approximate, illustrative values from public sources, rounded, not authoritative statistics. All computation happens locally in your browser; nothing is stored or transmitted.