Idlelandinwindfarms.
How much of the licence area allocated to wind farms is actually used? I calculated the answer across six pilot sites.
6 pilot sites at a glance.
- 171km²
- 59–172
- 287–826MW
- 881–2,518thousand MWh/yr
73% of the 234 km² total site area
on top of 71 existing turbines
on top of ~288 MW existing capacity (by scenario)
on top of ~748 thousand MWh/yr existing generation (by scenario)
These 6 sites were chosen to develop and test the methodology across different site sizes and complexities.
According to the EPDK query of 15 August 2026, there are 304 wind farms in operation. These results cover only 6 of them (1.97%); the potential across the full portfolio is much larger.
The analysis was prepared with a methodology in which human expertise and AI agents work together.
73% of the licensed area is idle.
Project sites are drawn geometrically according to the number of turbines and the layout plan. While permitting takes years, turbine technology moves on. So by the time a project is commissioned, the same installed capacity can be built with a different layout and fewer turbines. Because the site boundaries are not updated, idle areas appear inside the sites.
To see how much area is left idle, I picked 6 pilot sites licensed by EPDK. Under the Regulation on the Technical Evaluation of Wind-Resource-Based Applications, a total of 234 km² was allocated to these sites. When the sites are redrawn under the regulation using the turbines in operation, the area needed for the same capacity is only 84 km².
Why does the sum exceed the licensed area?
The area actually needed (84 km²) plus the idle area (171 km²) comes to 255 km², which exceeds the licensed area (234 km²). That is because the site calculated by the rules from the turbines in operation extends beyond 21 km² of the old, irregularly shaped licence boundaries. This is a separate finding: the old licence boundaries and the current method do not fully overlap.
The approach: three metrics.
- A
Idle area
How much surplus area is being held?
- B
Placement potential
How many additional turbines fit?
- C
Generation potential
Does it make sense in terms of wind?
The end-to-end workflow.
- 01
Data intake
EPDK licence and turbine data is read.
- 02
Metric A
The calculated field and the idle area are found from the merged 12-gon buffers.
- 03
Metric B
With a slope and power-density filter, ILP places the most additional turbines in the idle area.
- 04
Threshold layer
Weak candidates are removed with p10 / min / min80, derived from the site's own power-density distribution.
- 05
Metric C
Capacity factor and additional MWh/yr are found with Weibull/GWA (reference) or the EPİAŞ ratio (current).
- 06
Independent audit
A separate agent goes down to the raw data and recalculates every step to verify it.
- 07
Output
Delivered to the decision maker as GeoPackage, Excel, QGIS and KMZ.
How is the idle area calculated?
- Official site boundary
- Calculated field
- Idle area
- Spill-over beyond the licence
- STEP 1
Official boundary and turbines
EPDK licence and turbine data is read: the official site boundary and the turbines in operation.
- STEP 2
A 12-sided polygon on each turbine
A 12-sided polygon (12-gon) with a 700 m radius is drawn around the centre of every active turbine.
- STEP 3
The calculated field
12-gon clusters closer than 1,000 m to each other are merged with a convex hull, without any extra buffer. Isolated clusters farther than 1,000 m get their own +300 m safety band. Once all parts are merged, the calculated field appears. The part of the field that spills over the licence boundary is shown in purple. The current regulation has no +300 m band. Because these sites date from the earlier regulatory period, the extra band was preferred.
- STEP 4
The idle area
The part of the official licence boundary that lies outside the calculated field is the idle area. The orange region is the gap that could open to new turbines; the purple region is where the calculated field spills over the licence boundary. This is a schematic example, not a real site.
Placing additional turbines.
- CRITERION 1
14D × 6D elliptical exclusion zone
An elliptical exclusion zone is defined for every turbine (D = rotor diameter). The long axis of the ellipse runs along the wind direction. No other turbine may enter this ellipse. It was chosen so that wake (shading) losses can be neglected, since no micrositing is performed.
- CRITERION 2
The most turbines, using ILP
Integer linear programming (ILP) places the largest number of candidate turbines without conflicts. The choice of method is the result of an iteration: with the GWA power-density (PD) map as the base layer and slope removing the areas where turbines cannot be placed, 12 different layout algorithms were tested to find the result closest to real micrositing. The combination of fixed grid + ILP + step-wise fine-tuning gave the highest turbine count and power density, so it was chosen as the official method.
- CRITERION 3
Site-specific 95th-percentile slope threshold
A site's own existing turbines already show at what slope they were built. So a site-specific 95th-percentile slope threshold is used. A universal, fixed limit would not reflect the terrain reality of the site.
- CRITERION 4
One blade length from the boundary
Candidate turbines are allowed to approach the site boundary down to at least one blade length (the rotor radius).
- DATA AND CONSTRAINTS
Data resolution and constraints
The wind resource data (GWA) has a resolution of about 250 m, the terrain and slope data (Copernicus DEM) about 30 m. In the real world, candidates may have been placed at points where turbines cannot be built because of topography or the wind resource. To prevent this, professional micrositing is needed for every site. To partly automate the screening, a power-density (PD) threshold is also applied in the Metric C stage. A separate module is being developed to add country-wide constraints (heritage sites, transmission lines and so on) to the sites.
Generation potential.
Current-turbine scenario
- The plant's actual capacity factor is read from real EPİAŞ generation data. Since tower and power-curve data for all operating turbines is not available, this is an approximate but directionally right estimate.
- The capacity factor and annual generation of a candidate point are scaled by the ratio of its own power density (PD) to the plant average.
Reference-turbine scenario
- GWA's multi-height (10/50/100/150/200 m) Weibull A/k layers are sheared to each candidate's own hub height by log-height interpolation.
- A modern reference turbine power curve (Nordex N175/6.X) that is sensitive to the Weibull wind distribution and air density is used.
- The A and k parameters at the candidate point are convolved with this curve to calculate the theoretical capacity factor and annual generation.
- An 11% generic loss, representing availability, electrical loss, soiling and curtailment, is applied to the gross capacity factor. This loss applies to the reference scenario only; the current scenario is untouched because it already comes from EPİAŞ's actual net generation data.
The wind resource threshold: why and how?
Metric B only finds candidates that fit geometrically. But if a candidate sits at a weak point for wind, placing a turbine there can make no economic sense. Running a micrositing study on every site is not feasible either. So a threshold derived from the power-density (PD) distribution of the site's own operating turbines is applied. It is not a global, fixed number; it is specific to each site. Three parallel scenarios are defined on purpose.
- P10
The lowest 10% slice of the operating turbines' PD. The most conservative.
- MIN
The lowest PD value of the operating turbines.
- MIN80
80% of that minimum. The most flexible.
Example: Site-1
The average PD of the site's 9 operating turbines, and the thresholds. If a candidate point's PD is below the threshold of a scenario, it is not included in the generation of that scenario.
The pilot scope: why these 6 sites?
- 01 · SITE-1
Typical-scale reference
- ~28
- 34%
- ~84,000
- 02 · SITE-2
Large site + anomaly
- ~47
- 33%
- ~135,000
- 03 · SITE-3
Large site + anomaly
- ~100
- 23%
- ~200,000
- 04 · SITE-4
Typical-scale reference
- ~20
- 21%
- ~37,000
- 05 · SITE-5
Small site
- ~1
- 38%
- ~2,700
- 06 · SITE-6
Reference site
- ~92
- 36%
- ~290,000
Pilot results: the overall table.
| SITE | TOTAL CAPACITY (MWm) | SITE AREA (km²) | IDLE AREA (km²) | ADDITIONAL TURBINES IN THE IDLE AREA (count / MWm) | |||||
|---|---|---|---|---|---|---|---|---|---|
| CURRENT TYPE | REFERENCE TYPE | ||||||||
| MIN80 | MIN | P10 | MIN80 | MIN | P10 | ||||
| SITE-1 | 28 | 6 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 |
| SITE-2 ★ | 47 | 74 | 67 | 101 / 485 | 49 / 235 | 47 / 226 | 73 / 511 | 56 / 392 | 54 / 378 |
| SITE-3 ★ | 100 | 114 | 91 | 56 / 224 | 22 / 88 | 4 / 16 | 32 / 224 | 17 / 119 | 3 / 21 |
| SITE-4 | 20 | 6 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 |
| SITE-5 | 1 | 1 | 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 | 0 / 0 |
| SITE-6 | 92 | 33 | 14 | 15 / 86 | 12 / 68 | 8 / 46 | 13 / 91 | 11 / 77 | 8 / 56 |
| TOTAL | 288 | 234 | 171 | 172 / 794 | 83 / 392 | 59 / 287 | 118 / 826 | 84 / 588 | 65 / 455 |
★ At Site-2 and Site-3 the distance between clusters is greater than 1 km, which does not comply with the regulation (see the multi-part anomaly). The figures cover only the candidates in the idle area that pass both the slope and the PD threshold. Candidates inside the calculated field are counted as belonging to the existing plant owner. Areas and capacities are rounded, so totals may differ because of rounding.
Generation and capacity potential.
Additional capacity in the idle area (MW)
Additional generation in the idle area (thousand MWh/yr)
A large area does not mean high potential.
1.09 turbines/km²
73 additional turbines on 67 km² of idle area (min80, reference): high yield.
0.35 turbines/km²
Despite a wide idle area of 91 km², only 32 additional turbines fit (min80, reference): low yield, probably because of terrain and wind constraints.
Schematic example: a multi-part site.
Schematic example: the blue areas (calculated field) form detached clusters more than 1 km apart. The continuity the regulation expects is not met at Site-2 and Site-3. The orange areas show the wide idle zones between the clusters. These are not real sites.
- Official site boundary
- Calculated field
- Idle area
SITE-5: limited potential
A small site (~1 km²) with almost no idle area. Net additional turbine potential: 0.
Schematic example: the 12-gon of a single turbine exceeds the site boundary, leaving no idle area.
- Official site boundary
- Calculated field
- Spill-over beyond the licence
The candidate turbine funnel.
- Unconstrained (raw, mixed)286
- After slope (in the idle area)209
- MIN80118
- MIN84
- P10 (final)65
The same funnel for the current type (in the idle area): 513 (raw, mixed) → 302 (after slope) → 59–172 (p10–min80).
Assumptions and limitations.
Reference turbine
Nordex N175/6.X (7.0 MW). The model is flexible enough to run with any reference turbine.
3 parallel threshold scenarios
p10 / min / min80 are deliberately not reduced to a single number; they are presented side by side.
Site-specific slope threshold
There is no fixed, universal slope limit. The 95th percentile of each site's own operating turbines is used.
Minimum data rule
A year of actual generation data is left out unless there are at least 3 months / 2,160 hours of data.
Data resolution gap
Wind data (GWA) ~250 m, terrain data (Copernicus DEM) ~30 m. The roughly 8× gap between them should be kept in mind when interpreting the results.
Wake loss neglected
CFD-based micrositing is not performed. The 14D×6D elliptical rule was chosen so that wake loss can be neglected.
Generic loss assumption
An 11% generic loss is applied to the gross capacity factor of the reference scenario. The current scenario is unaffected, because it already comes from EPİAŞ net data.
These are calibratable assumptions. They are not a final, unchangeable result.
Turbines in the calculated field.
| SITE | ADDITIONAL TURBINES IN THE CALCULATED FIELD (count / MWm) | |
|---|---|---|
| CURRENT TYPE | REFERENCE TYPE | |
| SITE-1 | 11 / 36 | 2 / 14 |
| SITE-2 | 4 / 19 | 1 / 7 |
| SITE-3 | 23 / 92 | 8 / 56 |
| SITE-4 | 10 / 20 | 1 / 7 |
| SITE-5 | 0 / 0 | 0 / 0 |
| SITE-6 | 4 / 23 | 2 / 14 |
| TOTAL | 52 / 190 | 14 / 98 |
Because these turbines already lie inside the existing site's calculated (required) area, they are not free area that could be allocated to a separate investor, unlike those in the idle area.
The team: human and AI together.
Project owner / decision maker
Chief architect / independent auditor
Model: Claude Sonnet 5
ANTIGRAVITYCode writer / executor
Writes and runs all the Python and interface code.
Model: Gemini 3.1 Pro
Parallel site processing, for example running the Metric C calculations across sites at the same time.
- bug-hunter
Tries to find bugs and hidden problems in the code.
- epias-cf-calibration
Checks the estimates against actual generation data.
- geodata-ingestion
Reads raw map files and puts them in order.
- graphify-maintenance
Keeps a guide showing how all parts of the project connect.
- metric-a-idle-area
Finds the area standing idle on a site.
- metric-b-repowering-potential
Calculates how many new turbines fit in that empty area.
- metric-c-wind-resource-potential
Estimates how much electricity the turbines placed there could produce.
- qgis-integration
Makes the results viewable on a map.
- web-dashboard-app
Builds a simple screen where you can see the results.