Data sources
EPF is built entirely on public data. No proprietary feed is needed to produce a forecast.

Public prices for 45 European bidding zones; 4 are forecast.
| Category | Providers | What it gives |
|---|---|---|
| Day-ahead and intraday prices | OMIE, ENTSO-E Transparency Platform, REE/ESIOS | Cleared prices for 45 European bidding zones (4 forecast); Spain’s intraday auction and imbalance prices |
| Grid demand and generation | REE/ESIOS (Spain), ENTSO-E (all markets) | Actual and forecast demand, generation by technology, cross-border flows |
| Official renewable forecasts | REE/ESIOS, ENTSO-E | Day-ahead wind and solar forecasts published by system operators |
| Weather | Open-Meteo | Temperature, wind, solar irradiance and cloud cover, observed and forecast, for weighted locations in each country |
| Fuel and carbon | Public market quotes | Natural gas (TTF), crude oil (Brent) and EU carbon allowance prices |
| News | Public news feeds | Energy headlines with a sentiment score, shown in the app; not a model input |

Spain's day-ahead and intraday auction prices, as collected. Forward-market prices are shown on the Markets page for context.
Coverage differs by market
Section titled “Coverage differs by market”
Not every input exists for every country, and Spain has the richest public data.
Germany is the clearest gap: no public day-ahead wind and solar forecast is published for it on the European transparency platform. That is one reason its forecast is the hardest of the four.
Data quality checks
Section titled “Data quality checks”| Check | What it does |
|---|---|
| Run tracking | Every collection run records whether it succeeded and how many rows it wrote. A run that writes nothing is a failure, so a provider outage cannot hide behind a green status; the System page shows it. |
| Publication timing | Providers publish on their own schedules, sometimes late. A value is used only if it would really have been available at forecast time; publication times are measured, not assumed. |
| Gap handling | Short holes are bridged within strict limits. Longer gaps stay missing, and the tree models handle missing values natively. |
| Sanity bounds | A forecast run stops instead of publishing if its output falls outside plausible price bounds. |
| Scope checks | Every price series is tied to its own market, and automated tests guard against one market’s data being used for another. |