foxes_opt.callbacks¶

Contains backend-independent callbacks for wind farm optimization.

WriteLayoutCallback requires a FarmOptProblem and accepts iteration or evaluation events from callback-enabled iwopy optimizers. PyGMO IPOPT does not support callbacks. Pipeline users pass an ordered callback list through LayoutOptimizerStage.callbacks.

Layout snapshots¶

Pass --write_layouts to examples/layout_wind_rose/run_pymoo.py or run_slsqp.py to enable intermediate images; output is disabled by default. These examples disable CSV output and write one image per generation or accepted iteration under examples/layout_wind_rose/results/, regardless of the working directory. --layout_image_type selects the format (default: jpg). Running an example again overwrites matching snapshot indices.

The callback selects the best first-objective individual. Each title names its objective and value; constraint-valid turbines are blue and those associated with violated constraints are red, with an upper-left legend. valid_color and invalid_color accept Matplotlib colors. Early snapshots may be infeasible and are not final results.

write_csv and write_image independently control output and both default to True. CSV files go directly into the output directory; images go into a format-named child directory such as jpg/. Images omit turbine annotations and show physical turbine radii with contrasting outlines. The examples disable per-file log messages, and generated output is excluded by .gitignore.

The step-zero layout snapshot evaluates the optimizer’s initial variables so its objective title and constraint-validity colors are available before the first optimizer iteration. Set write_initial=False to skip this evaluation and snapshot.

Initial history rows¶

WriteOptimizationHistoryCallback records the selected objective value and violated constraint-component count for each iteration, persisting each row immediately. It resets the CSV when a solve starts unless append mode is enabled.

Set WriteOptimizationHistoryCallback(write_initial=True) to also record the starting objective and violated constraint-component count in the history CSV before the first optimizer iteration. This evaluates the optimizer’s initial variables during callback initialization. The history option defaults to False and is independent of the initial layout snapshot.

from foxes_opt.callbacks import (
    WriteLayoutCallback,
    WriteOptimizationHistoryCallback,
)

callbacks = [
    WriteLayoutCallback("iterations", "layout", write_initial=True),
    WriteOptimizationHistoryCallback(
        "iterations/optimization_history.csv", write_initial=True
    ),
]

When these callbacks are passed to an optimizer, initialization writes layout index zero and a history row with iteration=0, objective, and n_violated_constraints. The normal iteration rows then begin at one. The configured iteration_offset also applies to the initial history row.

Continuing restart output¶

Use WriteLayoutCallback.step_offset to continue layout snapshot names after a persisted index. Disable write_initial when the selected snapshot already exists. Use WriteOptimizationHistoryCallback.iteration_offset with append=True to continue history rows. Append mode validates the existing CSV header before preserving its rows; without append mode, initialization replaces the history file. Leave the history callback’s write_initial=False when the restart’s starting row is already present, avoiding a duplicate entry at that index.

from foxes_opt.callbacks import (
    WriteLayoutCallback,
    WriteOptimizationHistoryCallback,
)

callbacks = [
    WriteLayoutCallback(
        "iterations",
        "layout",
        write_initial=False,
        step_offset=188,
    ),
    WriteOptimizationHistoryCallback(
        "iterations/optimization_history.csv",
        iteration_offset=188,
        append=True,
        write_initial=False,
    ),
]

The first optimizer iteration then writes layout and history index 189.