WindKit 2.1.0 released

[Bjarke Tobias Olsen]

WindKit v2.1.0 has been released

Docs

Highlights of release

• LinRegMCP with residual noise — new LinRegMCP.predict_with_noise() samples Gaussian residuals per sector to recover realistic wind-speed distributions that preserve variance. MCPRegressor.predict(return_std=True) exposes the per-timestep residual standard deviation, and LinearRegression.residual_std is computed and stored during fit().
• New windkit.io.wasp low-level subpackage — read and write .tab, .lib, and .owc / .omwc files as plain dicts with numpy arrays, without xarray overhead. read_tab / write_tab / format_tab, read_lib / write_lib / format_lib, read_owc / parse_owc_element.
• Correlated synthetic TSWC data — create_tswc(not_empty=True) now generates temporally and spatially correlated wind data using an AR(1) / Weibull model (one year at 10-minute frequency by default). New create_tswc_pair creates correlated TSWC pairs with configurable speed bias, directional bias, and target correlation.
• windIO format support (experimental) — new windkit.io.windio subpackage adds readers and writers for the IEA windIO format, covering BWC / WWC / TSWC wind climates, wind turbines / WTGs, and full plant composition. Top-level readers and writers (read_wwc, read_bwc, read_tswc, read_wind_turbines, read_wtg, and their *_to_file counterparts) now accept YAML. See New features below for the full API.
• Dependency minimum bumps (breaking) — minimum supported versions raised: numpy 2.0, xarray 2024.9, pandas 2.3, scipy 1.13, geopandas 1.0, rasterio 1.4, pyproj 3.7, lxml 5.2, netcdf4 1.7. Users on older versions must upgrade.