The terms used across this site, defined once, in plain language.
A forecast of how much a wind fleet will generate on the following delivery day, produced before that day's electricity is traded. In most of Europe the day-ahead auction closes at 12:00 CET, so a forecast only counts as day-ahead if it exists before that gate. Anything published later may well be more accurate, but it is no longer a forecast anyone could have traded on.
Generation divided by installed capacity, expressed as a share between 0 and 1. A 10 GW fleet producing 3 GW is running at a load factor of 0.30. Load factor is the natural unit for a forecast record that has to stay comparable over years, because fleets grow. A record kept in megawatts would appear to improve every time a new wind farm is connected, while the load factor keeps the scale fixed.
The average size of the forecast miss, expressed as a percentage of installed capacity. If a 10 GW zone has an nMAE of 4%, the typical day-ahead miss is around 400 MW. Normalising by capacity rather than by output is deliberate: a percentage of actual output becomes unstable on calm days, when a small absolute miss turns into a very large relative one, even though the cost to a market participant is much the same either way.
The geographic area within which electricity trades at a single price. Zones are the unit that matters for a wind forecast, because that is the level at which power is bought, sold and settled. Some zones are whole countries, and some countries are split: Denmark into DK1 and DK2, Belgium reported by region. A forecast that does not match the zone boundary cannot be compared against anything the market uses.
The organisation that runs a country's high-voltage grid and keeps supply and demand in balance. Elia in Belgium, Energinet in Denmark, Fingrid in Finland, Terna in Italy, Red Eléctrica in Spain, PSE in Poland, NESO in Great Britain, and their counterparts elsewhere. Transmission system operators publish their own day-ahead wind forecasts as a public good, and those published forecasts are the benchmark OpenWindCast is scored against.
What it costs when the volume sold day-ahead is not the volume actually delivered. The difference is settled at the imbalance price, which is usually worse than the day-ahead price and can be considerably worse. It is the mechanism that turns a forecast error into money, and it is why forecast accuracy carries a price rather than only a percentage.
Output deliberately held below what the wind would allow. It happens when the grid cannot absorb the power, or when prices fall so low that generating is not worth it. Curtailment matters for any forecast record because metered generation on a curtailed day falls short of what the weather alone would have produced, so the shortfall appears in the measured error of every forecast of that day.
The combined rated output of every turbine in a fleet, the figure a fleet would produce if every machine ran at its maximum at once. It is published by the grid operator and it is the denominator behind both the load factor and nMAE, which is why a forecast record has to track it as the fleet grows.
These terms are all used in the methodology and on every zone page.