# China Energy and Electricity Yearbook: Structured Data Release

This release brings together historical Chinese energy and electricity statistics extracted from Excel and PDF yearbooks. Source files were inventoried, deduplicated and standardized without modifying the originals.

## Release overview

- 1,797 original files, approximately 17.35 GiB.
- 32 archives unpacked, adding 4,625 unique spreadsheets.
- 5,990 unique Excel workbooks and 10,408 worksheets.
- 2,142,776 non-empty Excel cells and 1,491,468 automatically extracted observations.
- 33,183 duplicate observations arising from merged cells removed.
- Four electricity yearbook PDFs for 2022–2025: 758 pages and 78,315 extracted observations.
- 1,569,783 records in the unified observation layer.
- Public core dataset: 2,987 records, 23 indicators and 32 regions, covering 1979–2024.
- 114 core consistency checks; maximum relative difference of 0.515%.
- No workbook extraction failures or non-finite observations.
- No tables with more observations than numeric source cells.
- Source sampling: 16 Excel tables and 328 cells, with no mismatches.

## Public downloads

- [CSV + JSON package](key-indicators.zip): lightweight core indicators.
- [CSV](key-indicators.csv): for spreadsheets, R and Python.
- [Parquet](key-indicators.parquet): columnar data for batch analysis.
- [English data dictionary](DATA_DICTIONARY.en.md): fields, units, keys and quality labels.

The download files retain the original field names, Chinese indicator and region labels, source units and numeric values. The bilingual website translates presentation labels without changing the underlying dataset. `metric_id` and `region_code` provide stable identifiers for analysis.

National capacity and generation series extend back to 1979. Provincial coverage varies by indicator and generally begins in 2017. Missing years are not interpolated, and missing observations are not converted to zero.

## Provenance and review

Use `source_relative_path`, `table_id`, and source page or worksheet coordinates to locate the original evidence. `edition_year` records the publication edition, while `year` records the statistical year; they are distinct fields.

Core indicators pass key-uniqueness checks. Capacity and electricity consumption also have comparisons between national totals and the sum of 31 provinces; differences are retained in the consistency-check records. The latest available edition is retained when historical figures are revised.

The larger automatically extracted observation layer prioritizes coverage. Automatically composed indicators, missing years or units, and complex headers require review; use `quality_tier` and `quality_flags` when working with that layer. The website's core release is a separately selected subset.

## Release structure

The release workspace retains core CSV/JSON/Parquet files, standardized observations, indicator/region/unit dimensions, source and table catalogues, consistency checks, rejected candidates, and a manifest of file sizes and SHA-256 hashes. These processing records support reproducibility; the public download package contains the lightweight core dataset.

Source extraction and standardization do not themselves grant republication rights for the original yearbooks. Distribution of complete original tables or large source collections requires separate consideration of publication permissions and database rights.

## Citation

PlanetaryEnergy. (2026). *China Energy Yearbook Structured Dataset* (PE-DATA/001). https://planetaryenergy.ac.cn/