Flanders day-ahead wind power forecast

Flanders is the northern region of Belgium and one of the most turbine-dense onshore areas in Europe, with 722 turbines and around 1.9 GW of installed capacity in a very small area. OpenWindCast publishes an independent day-ahead forecast of Flemish onshore wind and scores it daily against Elia's own forecast.

21.6%
lower point forecast error on average
nMAE vs each TSO's day-ahead forecast, all zones
58%
average win rate
days closer to the metered output than the TSO
20%
lower imbalance costs on average
across the 5 zones with settlement data

Averages across all zones. Every number on this page is recomputed daily from the public record below.

Coverage

Pick a bidding zone

Selected zone

Great Britain

Great Britain has one of the largest onshore wind fleets in Europe, with most capacity in Scotland. The forecast covers GB onshore wind and is scored against NESO's own forecast, published through Elexon.

Tomorrow's forecast
for —
MW
80% interval —
0 MWcapacity
Forecast

Recent forecasts vs actual output

Benchmark NESO day-ahead 11:00
Average error by window (nMAE)
Daily error
Daily detail

Every day, scored

Days marked pending are awaiting the operator's measured generation, which is published with a short delay, so the most recent day or two settle soon after.
The benchmark is captured on the day it is published and held fixed. Operators refine their forecasts nearer to delivery, so a value published later may differ.
Uncertainty

The 80% interval

Every forecast ships with an 80% interval. This shows how often the actual output landed inside it, and how wide it was.

Background

Forecasting wind in Flanders

Flanders fits a great many turbines into flat, low-lying and heavily developed ground, along the coastal polders, around the port areas and beside the motorway corridors. The terrain itself is about as simple as European wind gets, so the local wind field is comparatively straightforward. What is not straightforward is density. Turbines packed closely together sit in each other's wakes, and an intensively built landscape is aerodynamically rough, so the relationship between the free-stream wind and the power actually delivered is not the clean curve a single isolated turbine would follow.

The benchmark

The benchmark is Elia's own day-ahead forecast for Flemish onshore wind, taken from Elia Open Data. Belgium is the one place where Elia publishes early enough that the benchmark can be captured at the same time the forecast is issued rather than later in the day.

What is published for Flanders

One value per day: onshore wind output at 11:00 local time tomorrow, expressed as a load factor and published before 12:00 local, ahead of the day-ahead auction rather than after it. Once Elia's own forecast for that day is published it is captured and held fixed, and once the metered output is settled both forecasts are scored against it. Every day is added to the record below and nothing is removed or rewritten later, so the full history is there to be checked rather than summarised. How this is measured.

Why are Flanders and Wallonia separated?

Because they are different fleets in different landscapes. Flanders is dense, flat and coastal-influenced, and Wallonia is higher and more dispersed. Elia publishes them separately, so they can be scored separately.

Does this include Belgian offshore wind?

No. Belgium's offshore fleet in the North Sea is a separate system and is not part of the Flanders onshore forecast.

Compare every zone: how accurate are European TSO wind forecasts?

Method

Independent, transparent, measurable

Every morning, before the day-ahead auction closes, OpenWindCast issues tomorrow's 11:00 wind power forecast for each zone: a point estimate in megawatts plus an 80% interval, built from a multi-model weather ensemble. When the metered generation is published, the day is scored automatically and added to the public record above. The grid operator's own forecast is the benchmark and is never used as an input.

nMAE

Mean absolute error as a percentage of installed capacity. Lower is better.

Win rate

The share of days the forecast lands closer to the metered output than the operator's own forecast.

80% interval

The range the actual output should land in 80% of the time, recalibrated daily.

Benchmark

NESO's published day-ahead 11:00 forecast, a fixed yardstick that is never used as an input.

What is a day-ahead wind power forecast?

A forecast of wind generation for the next delivery day, issued before the day-ahead electricity auction closes at 12:00 CET. Market participants use it to decide how much wind energy to bid into the auction.

What exactly does OpenWindCast forecast?

One value per zone per day: the onshore wind output at 11:00 local time on the next day, expressed as a load factor and converted to megawatts with the installed capacity the grid operator publishes. Every forecast is published in the morning, before 12:00 local time.

How is forecast accuracy measured?

Each day, the absolute error of the OpenWindCast forecast and of the grid operator's own day-ahead forecast are compared on the same 11:00 target. The site reports the mean absolute error as a share of installed capacity (nMAE), the win rate, and, where official prices are available, the imbalance cost of bidding each forecast.

Where does the data come from?

Benchmark forecasts, metered generation and installed capacity come from the grid operators: Energinet, Elia, Fingrid, Red Eléctrica, the four German TSOs, PSE, REN, NESO via Elexon, and EirGrid with SONI, mostly through the ENTSO-E Transparency Platform. Weather inputs come from Open-Meteo.

Is the data free to use?

Yes. The dashboard and the daily verification record are public, and each zone's history can be downloaded as CSV from the forecast section. For anything else, email m.sahaf@live.com.