Dataset for Public verification of electricity reliability performance: Historical benchmarks and targeted disclosure in the United States
Description
This dataset supports research on public verification of electricity reliability performance, historical benchmarking and targeted disclosure. The main data were extracted and processed from official US Energy Information Administration Form EIA-861 annual reliability releases for 2013–2024. Auxiliary observations cover 2012–2021 for PacifiCorp California and 2018–2022 for Orangeville Hydro and were transcribed from their official reports. The files contain annual reliability measures and accounting bounds, 2,585 historical performance comparisons, disclosure allocations for 780 initially directionally indeterminate comparisons, reporting-stability checks, and results on record sharing, joint disclosure requirements, continuous reporting reform and historical backfilling. They support the study's assessment of how missing cross-classified reliability information affects the identification of historical performance changes. The data are annual aggregates and research-derived comparisons. No unknown joint reliability value is imputed. Calendar-based window analyses extend to 2032; empirical outage data end in 2024. Units, variable definitions, source boundaries and interpretation limits are documented in the accompanying README. The archive contains CSV data, JSON summaries, an English data guide and file checksums. Official source links and source-version checksums are included. Original annual archives, source PDFs and analysis code are not included.
Files
Steps to reproduce
1. Extract the archive and read README.md for file descriptions, variable definitions, units and source boundaries. Import CSV files as UTF-8, retain identifier columns as text, and preserve missing values. JSON files contain nested summaries. 2. Use annual_eia_measures_and_bounds.csv to inspect the annual measures. For records where joint_data_valid is True, verify the bounds L=max(0,B+C-A) and U=min(B,C), using the A/B/C definitions in the README. 3. Use historical_comparisons.csv and annual_record_contributions.csv to reproduce historical bounds. Combine coefficients for shared years before summing annual interval contributions. Compare the resulting lower and upper differences with zero to recover the directional classification. Do not impute missing joint values. 4. Use the allocation summary, selected-record, comparison-outcome, completion-frontier and random-benchmark files to check disclosure results by method and budget. The optimization covers only the 780 initially directionally indeterminate comparisons and assumes equal cost per annual record. The algorithms are described in the manuscript. 5. Use the continuous-disclosure and historical-backfill files for transition results. For Fig. 1, select disclosure_start=2020 and evaluation_year=2023 or 2024 in disclosure_window_structure_2018_2032.csv. For Fig. 2 and Table G.1, use the PacifiCorp annual measures and historical comparisons. Apply each source's own units and exclusion definitions. 6. Use CONTENTS_SHA256.csv to verify the files within this archive. official_source_checksums.csv records the original source versions and links. Original source archives and PDFs are not included. The deposit contains data and documentation; analysis code is not included.
Institutions
- Liaoning UniversityLiaoning, Shenyang