Weather-related power-outage frequency model

This repository contains the data-building, LightGBM leave-one-year-out modeling, evaluation, feature-comparison, and publication-figure scripts used for the Florida weather-related outage-frequency study.

Start with README_USAGE.txt for the complete run order, expected inputs, and outputs.

Main workflow

1. scripts/build_conus404_outage_dataset.py
2. scripts/threshold_justification.py
3. scripts/loyo_frequency_model.py
4. scripts/evaluate_frequency.py
5. scripts/compare_feature_sets.py
6. scripts/compare_spatial_temporal.py (optional diagnostic)
7. scripts/make_publication_figures.py
8. scripts/make_supp_figs.py (optional supplementary regeneration)

make_publication_figures.py is the final manuscript-style plotting workflow. The older plotting scripts are kept under legacy/ only for provenance.

The raw data and full processed modeling CSV are not bundled in this code package. See data/README.txt.