OOWL

Onshore and Offshore Wind Layouts (OOWL, pronounced /aʊl/ like owl) generates diverse, exact-count wind-farm layouts from composable geometry and placement methods. It provides boundaries, turbine layouts, wind resources, turbine models, and curated reference farms, and writes them as windIO.

OOWL is not a wake model or a wind-farm layout optimizer, and it does not try to reconstruct a particular real wind farm. Those are downstream applications that consume OOWL's output.

Installation

OOWL requires Python 3.12 or newer and is imported as oowl.

pip install oowl

To work from a checkout, or to develop OOWL.

git clone https://gitlab.windenergy.dtu.dk/sys/OOWL.git
cd OOWL
pip install -e .

# or the locked development environment
pixi install --locked --environment dev

See Installation for the optional extras and the supported dependency bounds.

Draw a farm

Describe the scenario you want as a sampler, then draw from it.

import oowl
from oowl import boundaries as B
from oowl import layouts as L

spec = oowl.Sampler(
    B.OffshorePolygon(area=45_000_000, seed=3),
    L.Grid(n_turbines=92, lattice_angle=60, seed=4),
)

farm = spec.sample()[0]
farm.plot()
farm.save("farm.yaml")

The sampler is validated when built and can be drawn from repeatedly with the same result. sample(n) returns a tuple of scenarios. Boundaries, strings, layouts, surfaces, wind resources and turbines are six composable layers. See Concepts.

Draw a dataset

Give any value a (low, high) range, an f(rng) -> value callable or an SP.Choice, and the same description becomes a population.

import oowl
from oowl import boundaries as B, layouts as L, sampling as SP

spec = oowl.Sampler(
    boundary=B.OffshorePolygon,
    layout=SP.Choice(
        [L.Grid, L.PoissonDisc, L.Delaunay],
        probabilities=[0.5, 0.3, 0.2],
    ),
    n_turbines=(40, 90),
    density={"minimum_spacing": 800.0},
    seed=7,
)

farms = spec.sample(200)
for index, farm in enumerate(farms):
    farm.save(f"dataset/farm_{index:04d}.yaml")

density names how tightly turbines are packed, here a minimum turbine distance in meters. Count, area and density constrain one geometry, so a sampler takes any two of them. A count and a spacing scale each boundary to fit, which keeps every farm buildable. n_farms= makes each scenario a Cluster of farms placed together, and domain= a region of development sites.

Seeds are derived per scenario, so spec.sample(3) repeats the first three scenarios of spec.sample(200), spec.sample(50, start=150) shards the population, and spec.stream() yields scenarios one at a time. [farm.summary() for farm in farms] gives one row of metrics per farm.

windIO

farm.save(path) and oowl.load(path) write and read one farm as a windIO wind_energy_system, with its wind resource and turbine types. Cluster.save and Cluster.load use a directory. OOWL writes windIO directly and does not depend on the windIO package. Substations and cable arrays are out of scope.

Documentation

The documentation is at https://oowl-4781ed.pages.windenergy.dtu.dk, with a Quickstart, Concepts, the API reference and executable notebooks, including integrations with PyWake, TOPFARM and OptiWindNet.

Citation

If you use OOWL, please cite it.

@software{OOWL_0_1_0_2026,
  author = {Rasmussen, Frederik Peder Weilmann and Schøler, Jens Peter and de Alencar, Mauricio Souza},
  title = {{OOWL: Onshore and Offshore Wind Layouts}},
  year = {2026},
  version = {0.1.0},
  url = {https://gitlab.windenergy.dtu.dk/sys/OOWL},
  license = {MIT},
}

Citation metadata and the author list are kept in CITATION.cff.

Contributing

Contributions are welcome. See the contribution guide. Run pixi run --environment dev lint, format-check, typecheck, test and docs before submitting, and record user-visible changes in the changelog.

License

MIT. See LICENSE. Copyright (c) 2026 DTU Wind and Energy Systems.