Transparency
Inside the model
Our European power-price outlook is built by a real market model, anchored to observed market data. This page shows what goes into it — the architecture, the named data sources, the scenario dials — and the limitations we publish alongside every figure.
Last reviewed July 2026
Why we publish this
Most commercial power-price outlooks are a black box: a number arrives, and you are asked to trust it. We think a forecast you cannot interrogate is a forecast you cannot use in diligence. So we publish what goes in, how it is validated, and where it is weak — and we attach a source and a quality grade to every stored figure, so any number can be traced to its origin: measured (an official TSO or regulator figure), derived (our percentage from a published volume) or estimate (a flagged screening basis where no official figure exists).
How well it actually performs is a separate page: model accuracy & backtests.
Two models, one outlook
A fast zonal market model
Reproduces each country's hourly price shape and how prices respond to fuel, carbon and renewable build-out. It sets the price levels and the scenario deltas, solved as a real cost-minimising dispatch optimisation — not a spreadsheet extrapolation.
A continental nodal grid model
A full weather-driven optimisation of the European transmission system at high time resolution. It supplies the locational structure — how prices and congestion differ from node to node within a country.
The two are stitched: headline numbers stay the calibrated market values, while the map inherits the grid model's spatial detail. Hydro-dominated markets that price off their neighbours — Switzerland, Austria, Norway, Sweden — are modelled at import parity with their actual coupling partners and at hydro's water value (the opportunity cost of stored water), so the physics carries the price level and calibration only trims it.
28 markets, named
Every market below carries three scenario price paths across six milestone years to 2060, with merchant capture and curtailment per technology. Markets marked nodal additionally carry locational price structure from the continental grid model.
Luxembourg is part of the DE-LU bidding zone and is modelled as its share of that zone. Nodal locational detail covers 20 of 28 markets today; the remainder carry zonal structure until their nodal solve is added.
What goes in
Named sources with vintages. No anonymous “proprietary datasets” — if a number feeds the model, we say where it came from.
ENTSO-E Transparency Platform & Elexon
Historical hourly wholesale prices for every covered market — the calibration anchor. Every headline figure starts from the most recent full year of observed outturn.
National TSOs & regulators
Curtailment outturn per country and technology — EirGrid/SONI, NESO, Bundesnetzagentur, RTE, Red Eléctrica, Terna, IPTO/ADMIE, URE/PSE, Fingrid and others. Each figure stored with its source and a quality grade.
netztransparenz.de, Bruegel, Ricardo/WSP
Independent public capture-rate records used to cross-check our per-technology merchant capture against observed outturn.
Live TTF gas, coal and EU-ETS carbon prices
Spot commodity levels feeding the near-term marginal-cost stack.
Open European transmission dataset
Grid topology for the continental nodal model, run against a recent full weather year of generation profiles.
EMHIRES / JRC capacity-factor record
30 weather years (1986–2015) of per-country, per-technology capacity factors — the basis of the weather-risk ensemble.
ENTSO-E hydrology feeds
Weekly reservoir-filling and hydro-generation series conditioning the coupled hydro markets.
Three scenario paths — and what actually drives them
Each market carries a conservative, baseline and aggressive price path. They are not competing stories about the energy transition — they are a deliberate price band around the single variable that moves European thermal costs most: the EU-ETS carbon price. The baseline path rises from 80 €/t today to 200 €/t by 2050; the conservative and aggressive paths bracket it from 75→130 to 90→300 €/t. Every path is re-solved by the model — never scaled from another path — and fleet build-out effects are being added through the nodal scenario solves, a limitation we state rather than blur.
Weather risk from 30 real years, not one
Much of the industry runs on a single weather year. We measure inter-annual weather risk across the full 30-year EMHIRES capacity-factor record (1986–2015, per country and technology), giving each market a weather-P90 downside yield and a variability figure. One robust finding: wind carries roughly two to three times the weather-year risk of solar — which is why the same P50 revenue is not the same risk.
What we publish
- Scenarios are price paths that bracket uncertainty — not predictions, and not a full enumeration of technology futures.
- The locational map is indicative — best read as a ranking of connection locations, not as precise euro-per-MWh differentials.
- Curtailment is a prudent downside that grows with build-out to a capped ceiling; late-horizon figures in the most-constrained grids read as a stress level, not a central expectation.
- Locational (nodal) detail currently covers 20 of the 28 markets; the rest carry zonal structure until their nodal solve lands.
- Some markets rely on neighbour proxies or clearly-flagged screening estimates where no official figure is published — the flag ships with the number.
- The outlook is screening-grade, for information: it exists to inform your own revenue diligence, not to replace a bankable price study.
What we don't
The recipe. Calibration constants, algorithm internals, exact tolerances and per-country fleet assumptions stay in the engine — they are the result of continuous measurement against live markets, and they are what you subscribe for.
Subscribers get the full methodology: the complete method notes per product domain, the validation record, and the honest limitations — updated on every calibration cycle, inside the platform.
See how it performs
We back-test the model against observed market prices on a held-out year and publish the error — including the markets where it is weakest.