Wind P50 Modeling Approach
Overview
This page explains how Re-Twin estimates a P50 wind generation profile for your site when you do not upload one. It is the wind counterpart to the PV P50 approach.
A P50 profile is the median expected yield under typical weather conditions — there is a 50% probability that actual generation exceeds it, and a 50% probability it falls short. It is the standard reference for energy yield assessments and financial models.
If you already have a measured or consultant-provided profile, upload it instead — see P50 Generation Profile.
What the Profile Contains
The result is one value for every hour of a full year: the capacity factor, meaning the turbine's output as a share of its maximum rated output. A value of 1.0 is full nameplate output, 0.5 is half, and 0 is standstill.
Because these are shares rather than absolute megawatts, the same profile scales to a farm of any size — multiply by your installed capacity to get expected generation.
How the Profile Is Built
The estimate combines two things: historical weather for your site's coordinates, and the chosen turbine's own performance characteristics.
1. Historical Wind Data
Ten years of hourly wind speed are collected for the site's location, drawn from the European reanalysis datasets published by the Copernicus Climate Change Service. A reanalysis is a gap-free reconstruction of past weather, produced by combining historical measurements with a weather model, giving consistent coverage everywhere — including locations with no weather station nearby.
2. Adjusting to Turbine Hub Height
Wind blows faster the higher you go, and the historical data is reported at two fixed heights above ground: 10 metres and 100 metres. Comparing the speeds at those two heights reveals how quickly wind accelerates with height at your site — and it does so hour by hour, since the effect is far stronger on a calm night than on a windy afternoon.
That site-specific relationship is then used to carry the wind speed up to the hub height of your actual turbine. Safeguards keep the calculation stable in the rare hours when the near-ground reading is close to zero.
3. Selecting a Typical Year
Rather than blending a decade of weather into a single average, the model assembles a typical year month by month. For each calendar month it looks across all ten years and picks the one real historical January, the one real February, and so on, whose weather was statistically most representative of the long-term norm for that month.
The comparison uses the Finkelstein–Schafer statistic, the standard measure in solar and wind resource assessment for how closely one month's spread of wind speeds resembles the long-term spread.
The benefit is that the profile is built from genuinely observed weather rather than an average that flattens out real conditions.
4. Building the Month-by-Hour Pattern
The selected year's wind speeds are organised into a table covering every combination of month and hour of the day — 12 months across 24 hours, so 288 entries. Each entry averages that month's readings for that hour: every 2 pm in January, every 3 am in June, and so on.
This captures both the seasonal cycle and the daily rhythm of wind at the site in a single compact picture.
5. Converting Wind Speed to Power
Each wind speed is converted into electrical output using the turbine's power curve — the manufacturer's published figures for how much power that specific model generates at each wind speed. Dividing by the turbine's rated output turns this into a capacity factor.
6. Restoring Day-to-Day Variation
A month-by-hour table holds only one value per combination, so applying it directly to a calendar would make every day in a month look identical — an obviously unrealistic flat pattern on an hourly chart.
Instead, each day of the typical year keeps its own individual readings, so a stormy week and a calm week look genuinely different.
This extra detail introduces a slight upward bias: because turbines respond very steeply to wind at lower speeds, a windy hour gains more output than a calm hour gives up. The day-by-day profile is therefore scaled so that its annual total matches the validated figure from the averaged table. That scaling factor is calculated from each site and turbine's own data, is never carried over between projects, and can only reduce the result, never inflate it. At one reference site, an uncorrected annual average of 33.06% was brought back to the validated 27.13% this way.
The annual energy total is unchanged by this step — only the shape across the year.
7. Final Output
The profile is mapped onto a standard 365-day year, producing exactly 8,760 hourly values, and converted to Central European local time. This is the same format expected for an uploaded profile.
Validation
The method has been checked against two real wind farms with publicly documented production records — one onshore, one offshore.
| Onshore | Offshore | |
|---|---|---|
| Site | Registered onshore unit, Hessen, Germany | Alpha Ventus, North Sea |
| Turbine | Vestas V90/2000, hub height 105 m | Adwen/Areva AD116-5000, hub height 90 m |
| Record used | 6 years (2019–2024) of metered production | 3 years of publicly reported capacity factors |
| Actual capacity factor | 15.75% (yearly range 11.03%–16.63%) | 48.1% (yearly range 42.7%–55%) |
| Modeled | 16.00% (+1.6%) | 47.37% (−1.5%) |
Both estimates fall inside the year-to-year range each site actually recorded — about as close as a model can get without on-site measurement. No general loss allowance or fitted correction is applied; the only adjustment anywhere in the process is the site-specific scaling described in step 6, which exists to preserve these validated levels.
Turbine Matching
Your turbine's manufacturer and model are matched against Re-Twin's library of 67 real turbines, each built from published manufacturer performance data. Naming in the German energy asset register (Marktstammdatenregister) lines up with this library, so most registered units match exactly.
Where there is no exact match, the closest available turbine by rotor diameter and rated output is used, preferring the same manufacturer.
Which Figures Are Used
A modeled estimate is only used when nothing more direct is available:
- Measured production, where the unit already has a reported generation history — its multi-year average is used directly.
- Your uploaded profile, where you supply one from a resource assessment.
- The modeled estimate described here, for planned or not-yet-built projects.
Known Limitations
- Two validated reference sites. Both fall within their recorded ranges, but two sites is a limited basis for accuracy across all European wind conditions.
- The offshore check is approximate. Alpha Ventus uses two different turbine models and only one is in the library, so a single-model estimate is compared against a whole-farm figure.
- Losses are not modeled. Wake effects between turbines, downtime, curtailment and electrical losses are excluded. A project-specific resource assessment accounts for these, which is why an uploaded profile always takes priority.
- Turbine matching may be approximate where your exact model is not in the library.