Changelog¶

Introducing command line application foxes_opt_yaml: Runs optimization from yaml parameter input file, no Python script needed.
Changes in FarmVarsObjective: Renaming contraction rule mean into mean_no_weights.
Adding run-time factories to problems, objectives, and constraints.
Introducing output sub package, and first outputs SingleOptResultsWriter and MultiOptResultsWriter
New example yaml_input, demonstrating how to run *foxes_out through a yaml parameter file

Full Changelog: https://github.com/FraunhoferIWES/foxes/commits/v0.3

v0.4¶

Support for foxes v1.3

Full Changelog: https://github.com/FraunhoferIWES/foxes/commits/v0.4

v0.5¶

Support for foxes v1.4

Full Changelog: https://github.com/FraunhoferIWES/foxes/commits/v0.5

v0.6¶

Compatibility with foxes v1.5
Dropping support for Python 3.8
Introducing minimal package versions of dependencies

Full Changelog: https://github.com/FraunhoferIWES/foxes/commits/v0.6

v0.7.0¶

Compatibility with foxes v1.7.0
Moving currently not running or deprecated opt problems to new scratch folder: GeomLayoutGridded, GeomLayout, RegGridsLayoutOptProblem
Bug fixes:
- Bugs fixed that caused error with RegularLayoutOptProblem
- Bug fixed with GG and layout problems

Full Changelog: https://github.com/FraunhoferIWES/foxes/commits/v0.7.0

v0.7.1¶

Compatibility with foxes v1.7.4
Compatibility with iwopy v0.5.0
New optimization pipeline: LayoutPipeline, combining different optimizations for generating an optimal wind farm layout

Full Changelog: https://github.com/FraunhoferIWES/foxes/commits/v0.7.1

v0.8.0¶

General:
- Compatibility with foxes v1.9.2
- Fixed vectorized wake-steering optimization with gridded rotor weights
- Dropping support for Python 3.9
- Introducing type annotations
- Raising overlapping dependency minimum versions in pyproject.toml to be no lower than the corresponding minimum versions in foxes
Documentation:
- Migrating Sphinx API docs from sphinx_immaterial.apidoc.python.apigen to AutoAPI
Examples:
- New example layout_field_data, demonstrating pymoo-based layout optimization with FieldData states loaded from multiple NetCDF files

Full Changelog: https://github.com/FraunhoferIWES/foxes/commits/v0.8.0

v0.9.0¶

General:
- Expanded type annotations and tightened code checks across the package
Dependencies:
- Raised the minimum supported versions to foxes v1.9.6 and iwopy v0.5.0, and aligned overlapping dependency floors with both projects
Callbacks:
- Added an optimizer-independent layout callback with optional intermediate CSV output, configurable image snapshots and objective titles, and red highlighting of constraint violations in the legend
Pipelines:
- LayoutPipeline now retains stage results and supports layout plot and CSV output
- Added LayoutOptimizerStage for optimizer-driven refinement of layouts from previous stages, with user-selectable optimization problem types
- Added RandomSubsetStage for successive layout optimizations over random turbine and state subsets, forwarding layout optimizer configuration through its base stage
Examples:
- Added pymoo and SLSQP wind-rose examples using the layout callback
- Added a pygmo-IPOPT wind-rose example without callbacks, since pygmo exposes neither exact live IPOPT iterations nor their decision vectors
- Added a pygmo-IPOPT single-state layout example with configurable, layout-scale finite differences; progress output reports native IPOPT iterations separately from objective function evaluations
Bug fixes:
- Weighted state contractions now normalize by the selected weight sum
- Fixed multi-state population optimization for population-major foxes state ordering

Full Changelog: https://github.com/FraunhoferIWES/foxes-opt/commits/v0.9.0

v0.9.1¶

Dependencies and iwopy integration:
- Require FOXES >=1.9.8 to inherit its corrected runtime security floor
- Require iwopy >=1.0.0 and consume CSR get_jacobian/calc_jacobian matrices, iwopy.create_optimizer, flat native options, shared max_iterations / max_evaluations controls, and pymoo pure-limit selectors throughout maintained workflows and documentation; YAML uses family/optimizer, while layout stages use optimizer_family, including boundary-repair overrides. Retained scalar-gradient terms, native convergence, and distinct population/batch units; documented sparse native boundaries and corrected spacing-selection ownership
General:
- Added support for Python 3.14
- Added coding-agent policies, including developer-owned uv setup and mandatory uv run --no-sync, plus FOXES Optimization-specific architecture, naming, development, docstring, ADR, data-classification, and visual-design guidance
- Recorded Fraunhofer corporate design for FOXES Optimization visual surfaces
Callbacks:
- Layout image snapshots now omit turbine annotations, use true turbine radii with color-matched outlines, use configurable validity colors that default to high-contrast blue and red, and live in a file-type subdirectory. Added restart offsets and optional initial step-zero snapshots, enabled by default; initial objectives and constraints are evaluated before plotting, including without integer variables. Fixed validity colors for small turbines and nearest-group constraint violations
- Added a CSV optimization-history callback that records objective values and violated constraint-component counts per iteration, with restart offsets and validated append mode. Optional initial evaluation (write_initial=False by default) records the starting values before the first optimizer iteration
Pipelines:
- LayoutOptimizerStage now forwards callbacks to the optimizer and accepts stage-local FOXES algorithm parameter overrides through algo_pars
- LayoutPipeline accepts finite initial coordinates without a preceding layout-generation stage and can restart a selected stage from a validated persisted layout index without resetting earlier pipeline results
- Automatic minimum-distance constraints accept their own parameter mapping, allowing boundary and spacing families to use independent tolerances
- Layout plots now match callback snapshots with true turbine radii, high-contrast colors, matched outlines, and validity legends
Constraints:
- MinDistConstraint now provides analytical position derivatives, vectorized variable-dependency masks for faster large-farm Jacobians, and bound-certified nearby-pair selection preserving full component identities and checks; callers enforce selection bounds, while current iwopy SLSQP evaluates all registered pairs
- Added NearestGroupConstraint, enforcing nearest-distance and minimum connected-group-size limits with piecewise analytical position derivatives
- Added analytical derivatives for area boundaries and geometrical layout constraints, including composed and internally excluded geometries, with cached bulk signed-distance gradients for area-geometry Jacobians
Objectives:
- Added analytical derivatives for turbine-count and geometrical layout objectives; FOXES-result objectives continue to require iwopy.LocalFD
Bug fixes:
- Layout candidates now update turbine coordinates before FOXES state initialization, while final individual layouts restore scalar positions
- Partial-turbine population objectives now select the turbine axis and the matching turbine-dependent weights
- Local-square layout bounds remain anchored to the initial layout during vectorized population evaluations

Full Changelog: https://github.com/FraunhoferIWES/foxes-opt/compare/v0.9.0…v0.9.1