TEF is the top-level orchestration framework that couples:
- An internal system scaling block (
tef.system_scaling) that generates physically consistent awesIO system files from design variables (wing area, wing loading, tether stress, generator sizing). - The AWESPA inertia-free QSM performance model (power curves).
- The AWESPA AEP calculation.
- A storage sizing block that sizes the ground-station storage from the computed power curves (cycle energy imbalance method).
- The EcoMo economic model in AWESPA-connected mode (LCoE, ICC, OMC, ...).
The framework targets conceptual design and trend identification, not accurate cost prediction. awesIO-style YAML files are the central data exchange layer, mirroring AWESPA's architecture one level up.
AWESPA, EcoMo and awesIO are external dependencies installed in editable mode from their existing locations (no clones inside TEF):
pip install -e <path-to>/projects/AWESPA
pip install -e <path-to>/projects/ecomo
pip install -e <path-to>/projects/TEF
awesIO is installed from PyPI/its own repo (pip install awesio or an
editable clone).
config/base/ base system, scaling settings, QSM settings, wind
resource, cost inputs, economic settings template
config/studies/ study definitions (area sweep, generator power sweep)
src/tef/ the framework package: one module per framework block
scripts/ entry points
results/ one directory per study, one subdirectory per case
The package mirrors AWESPA's style — one clear module per model block
with a small stable interface, re-exported from tef:
tef/io.py YAML helpers and project paths
tef/system_scaling.py design variables -> scaled awesIO system file
(build_scaled_system, DesignVariables,
ScalingSettings)
tef/storage_sizing.py storage capacity from computed power curves
tef/wrappers.py AWESPA power / AEP / EcoMo interfaces
tef/pipeline.py full model chain for one case (TefCaseRunner)
tef/studies.py sweeps and combined summaries (run_area_sweep,
run_generator_power_sweep)
tef/design_space.py declarative optimised variables + bounds
(DesignSpace, DesignVariableSpec)
tef/optimisation.py model-agnostic optimisers (grid, optional scipy
refine) and the 1-D rated-power solver
tef/design_studies.py LCoE optimisation over the design space at each
fixed rated power (run_design_optimisation)
Each block writes exactly one output file (AWESPA style). The inputs
shared by all cases of a study are written once to
results/<study>/inputs/; a case directory holds only the
case-specific files:
results/<study>/
inputs/ base system, scaling settings, QSM settings,
wind resource, cost inputs, economics template
<case>/
design.yml the design variables of this case
system.yml scaling block output (updated by storage sizing)
power_curves.yml AWESPA output (+ .npz)
aep_results.yml AEP block output
economic_settings.yml generated; references ../inputs/ relatively
ecomo_results.yml EcoMo output
tef_summary.yml consolidated summary (design, scaling mass
split, storage sizing, performance, economics)
study_summary.yml / .csv
# 1. Scaled system files only (no AWESPA/EcoMo):
python scripts/run_generate_scaled_systems.py
# 2. One full case (default 25 m2):
python scripts/run_single_case.py --area 25
# Optimisation entry point (quiet, no validation, reusable work dir):
# from tef import evaluate_design, DesignVariables
# result = evaluate_design(DesignVariables(flat_area_m2=25.0),
# work_dir=Path('results/opt'))
# result.lcoe_eur_per_mwh # objective
# 3. The five-point area sweep (25..400 m2):
python scripts/run_area_sweep.py
# 4. The area x generator specific power grid:
python scripts/run_generator_power_sweep.py
# 5. Design optimisation: minimise LCoE over {wing area, tether stress}
# at each fixed rated power (25..300 kW). The generator is sized so
# the emergent rated power matches each target.
python scripts/run_design_optimisation.py
Run the tests with:
python -m pytest tests/ -v
wing_mass_model.type:tabulated_log_interpolation(thesis anchor table, default) orpower_law(smooth, for optimisation).kcu_mass_model.calibration_mode:raw_thesis(default) ormatch_v3_at_25(reproduces Bredael's 8.4 kg KCU exactly).sensor_mass_model.include_in_control_system_mass:truefor final TEF runs; setfalseonly to reproduce Bredael's 19 kg airborne mass.storage_sizing.mode/method: sizing from power curves via the cycle energy imbalance (smoothing to cycle-average power), with a generator-power-scaled fallback.
Reproducing Bredael's V3 inputs exactly: set calibration_mode: match_v3_at_25 and include_in_control_system_mass: false — this is
covered by tests/test_scaling_v3_reference.py.
TEF follows the AWESPA developer guide: snake_case functions/methods and parameters, mixedCase local variables and attributes, CamelCase classes, UPPER_SNAKE_CASE constants, Google-style docstrings. Dataclass fields that mirror YAML keys or CSV columns stay snake_case (they are the serialization schema, like the awesIO files).