Troubleshooting

Common Errors

"Could not find or infer 'rated_power'"

Cause: DatabaseGenerator cannot determine turbine rated power.

Solution: Add to your turbine definition:

turbines:
  performance:
    rated_power: 15000000  # Watts

Or include "XMW" in turbine name: name: "IEA 15MW Reference"

"Feature not found in dataset"

Cause: Requested feature doesn't exist after preprocessing.

Check: 1. Run with preprocessing.run: true 2. Include recalculate_params in steps 3. Verify input resource file has required variables

"UMBRA not installed"

Cause: Bayesian calibration requires the UMBRA package.

Solution:

pip install umbra
# or
pixi add umbra

Alternatively, use a different calibrator (MinBiasCalibrator).

LeaveOneGroupOut with single group

Cause: All farms mapped to one CV group.

Solution: Check groups config matches farm names exactly:

groups:
  GroupA:
    - Farm1  # Must match 'name' in farms list exactly
    - Farm2

Performance Issues

Database generation is slow

• Reduce n_samples for initial testing
• Use coarser grid_res in blockage metrics
• Consider running farms in parallel (future feature)

Memory errors with large databases

• Process farms individually, combine afterward
• Reduce number of flow cases
• Use chunked xarray operations