Public Utility Data Liberation Project (PUDL) Data Release

v2026.7.2 (2026-07-14)

This is a monthly PUDL data release, primarily motivated by updating the EIA-860M monthly data through May 2026. As usual, it also includes all of the other changes that have accumulated on main since our last release.

This month, we have new EIA-176 tables, the EIA-860 early release, Parquet outputs for DBF assets, improved units handling, GeoParquet bugfixes, and better signal:noise ratio in unit test logging outputs.

Enhancements

• Added experimental Parquet outputs derived from the FERC DBF databases, and basic datpackage.json metadata describing their schemas to support querying and preview through the PUDL Data Viewer. See PR #5339.

• Standardized all unit strings in pudl.metadata.fields to Pint expression syntax, replacing ad-hoc abbreviations (gpm, min, F, cfm), underscore-separated compound units (lb_per_MMBTU, USD_per_MWh), and inconsistent capitalization. A new pudl.metadata.units module defines PUDL_UNIT_REGISTRY, a pint.UnitRegistry extended with energy-industry units (MMBtu, Mcf, MMcf, TBtu, VAr, USD). Added extensive new per-column units annotations. These changes should facilitate programmatic unit parsing, display, and conversion, and they are now surfaced in the PUDL metadata and datapackage outputs as machine-readable Pint-compatible unit definitions. See #5078 and #5361.

New Data

EIA-176

• Added detailed core EIA-176 continuation-line tables for natural gas imports, supplemental gaseous fuel supplies, gas exports, and other gas disposition. See #5240 and #5245.

Expanded Data Coverage

EIA-191

• Added EIA-191 data through end of March 2026. See PR #5396.

EIA-930

• Added EIA-930 data through end of June 2026. See PR #5396

EIA Electricity API

• Updated the bulk EIA Electricity API data used to fill in redacted fuel prices. See PR #5396.

EPA CEMS

• Added EPA CEMS data through end of March 2026. See PR #5396

EIA-860

• Added early release data for EIA-860 2025. See issue #5322 and PR #5324.

EIA-860M

• Added EIA-860M data through May 2026. See issue #5369 and PR #5371.

FERC Forms 2 and 6

• Updated the raw FERC Form 2 and 6 archives to include additional 2025 data. This data is converted to SQLite, but not deeply integrated into PUDL. See PR #5396.

FERC CID

• Updated the FERC company identifiers with data through end of June 2026. See PR #5396.

Documentation

• Expanded the developer docs around metadata naming, typing, and updates to explain how unit annotations, field namespaces, and namespace/table-specific metadata overrides should be defined and maintained. See #5361.

• Set up the sphinx_llm Sphinx extension to generate a Markdown version of the PUDL documentation, suitable for consumption by LLMs, based on the llms.txt convention. See PR #5381.

New Data Tests & Validations

• A new valid_datapackage_unit_strings_check() asset check factory validates all unit strings in the PUDL datapackage descriptor against the registry after each ETL run. About a dozen fields in PHMSA gas and EIA-860 FGD data that were typed as number but contain integer counts have been corrected to "type": "integer". A bug where convert_cols_dtypes and get_parquet_table overrides in FIELD_METADATA_BY_RESOURCE, were sometimes ignored has been fixed.

• Added dbt expect_column_values_to_be_between tests to codify range expectations for columns stated as percentages (0, 100) vs those that represent fractional values (0, 1). Column naming still needs to be standardized. FERC Form 1 fraction tests use error_if thresholds to accommodate a known small number of out-of-range values. See #5361.

Bug Fixes & Data Cleaning

• Three EIA-860 columns that EIA reports as percentages but PUDL describes as fractions have been corrected. standard_so2_percent_scrubbed (boilers) was already stored as a fraction but misnamed; it is now renamed standard_so2_fraction_scrubbed. max_oil_heat_input (multi-fuel generators) and dry_cooling_pct (cooling equipment) were extracted as percentages; both are now divided by 100 in the transform step and the cooling column is renamed dry_cooling_fraction. Field descriptions for the FERC1 *_fraction_cost columns have been updated to say “fraction (0-1)” instead of “percentage”. All true _pct columns now carry an explicit "unit": "percent" annotation. See #5361.

• Fixed several Click-based console scripts so shell callers now receive correct non-zero exit codes on failure. The script-entry conventions in pudl.scripts now use Click-native exits and call main() directly in the module launcher, which fixes automation that branches on pudl_check_for_build success or failure. See PR #5374.

• Fixed a DuckDB >= 1.5 incompatibility with PUDL’s GeoParquet outputs. DuckDB 1.5 requires CRS metadata in PROJJSON format, but the old PudlGeoParquetIOManager wrote a WKT string, causing "Geoparquet column 'geometry' has invalid CRS" when the spatial extension was loaded. Switching to native geopandas.GeoDataFrame.to_parquet() produces spec-compliant GeoParquet 1.0.0 metadata. See issues #4061, #5074 and PR #5347.

• Dropped the uninformative is_total column from core_rus7__yearly_distribution_services and out_rus7__yearly_distribution_services, renamed the service_status values to the self-explanatory connected_this_year, retired_this_year, total_in_place and idle_in_place, and documented why RUS Form 7 Part B subcomponents do not sum to the reported total within a year. See issue #5262 and PR #5323.

• Fixed several nightly/stable/branch deployment flow-control bugs: nightly builds were silently landing in staging instead of production because DEPLOYMENT_ENVIRONMENT was never wired into the build container’s Batch job, PUDL Viewer redeploys and Zenodo releases fired for every deploy type instead of being restricted to nightly builds and non-branch builds respectively, stale objects from earlier deployments were never cleared from the public GCS/S3 paths before new outputs were written, and a staging copy of a stable release tag was incorrectly treated as an immutable path that could never be cleared. See issue #5382 and PR #5384.

• Added two new optional arguments to get_pudl_dtypes and apply_pudl_dtypes. resource (aka table) name will now return all of the authoritative resource-specific dtypes instead of the generic or source-specific types. dtype_backend will now return the types for the specific file type, which extends the previous behavior of returning pandas dtypes. The dtype management now lives in pudl.metadata.dtypes. See #5361.

• Fixed pudl_deploy returning a plain integer exit code from its Click command, which Click’s standalone mode silently discards, so shell callers previously saw exit code 0 even when a deployment stage failed. It now uses ctx.exit() like the other scripts fixed in #5374. See #5382 and PR #5384.

• Fixed the longstanding issue with zenodo_data_release sandbox release failures happening even when the publication actually succeeded, because a client-side timeout on the publish request triggered a retry that legitimately 404s once a deposit is already published, and that raw 404 body was then parsed as if it were a real deposition. Also fixed the build-ID provenance marker file being a zero-byte upload, which Zenodo rejects outright. See PR #5384.

Performance Improvements

• Sped up PUDL deploys by compressing SQLite databases concurrently in a thread pool and uploading to all four GCS/S3 targets concurrently instead of one at a time. Also reduced the SQLite compresslevel from 9 to 6, trading a little archive size for much faster compression step. See #5382 and PR #5384.

Developer Experience

• Reduced spurious logging and error output from our unit tests. See PR #5362.

• Added two new optional arguments to get_pudl_dtypes and apply_pudl_dtypes: resource (aka table) name will now return all of the authoritative resource-specific dtypes instead of the generic or source-specific types. dtype_backend will now return the types for the specific file type, which extends the previous behavior of returning pandas dtypes. The dtype management now lives in pudl.metadata.dtypes. See #5361.

• Reworked the nightly PUDL build and deployment automation to send start and status notifications to the pudl-deployments Zulip stream directly from GitHub Actions and the batch build script, with per-stage timing summaries and direct links to build logs and outputs. The nightly build container also no longer installs or bootstraps a local PostgreSQL cluster solely to initialize Dagster, and the batch script now refuses to trigger deployment unless all required stages, including output upload, completed successfully. See PR #5374.

• Merged PudlGeoParquetIOManager into PudlParquetIOManager and retired the geoparquet_io_manager Dagster resource key. The four geo assets (out_censusdp1tract__states/counties/tracts and out_ferc714__georeferenced_respondents) now use parquet_io_manager. Updated the DuckDB dependency to >=1.5,<1.6. See issues #4061, #5074 and PR #5347.

• Extended the nightly build’s Zulip reporting and stage-tracking machinery (added in #5374) to pudl_deploy. Deployments now save logs to builds.catalyst.coop and send a per-stage duration + outcome report to Zulip. Centralized deployment logic and validation in a DeploymentPlan Pydantic model in pudl.deploy.pudl. Replaced some ad-hoc curl/JSON GitHub Actions dispatch with the gh CLI and a shared dispatch_github_workflow() helper. Added required-argument & duplicate-key validation to devtools/generate_batch_config.py Batch job config generator used by both the PUDL and FERC EQR build workflows. See issue #5382 and PR #5384.

• Added a guard so that deploying to a permanent, version-tagged stable-release path that already has content raises immediately instead of silently uploading over it, and updated the Zenodo data release Zulip notification to state plainly whether a release was sandbox or production and publish or draft, with a link to the resulting record. See issue #5382 and PR #5384.

Other PUDL v2026.7.2 Resources

• PUDL v2026.7.2 Data Dictionary
• PUDL v2026.7.2 Documentation
• PUDL in the AWS Open Data Registry
• PUDL v2026.7.2 in a free, public AWS S3 bucket: s3://pudl.catalyst.coop/v2026.7.2/
• PUDL v2026.7.2 in a requester-pays GCS bucket: gs://pudl.catalyst.coop/v2026.7.2/
• Zenodo archive of the PUDL GitHub repo for this release
• PUDL v2026.7.2 release on GitHub

Contact Us

If you're using PUDL, we would love to hear from you! Even if it's just a note to let us know that you exist, and how you're using the software or data. Here's a bunch of different ways to get in touch:

• Follow us on GitHub
• Use the PUDL Github issue tracker to let us know about any bugs or data issues you encounter
• GitHub Discussions is where we provide user support.
• Watch our GitHub Project to see what we're working on.
• Email us at hello@catalyst.coop for private communications.
• On Mastodon: @CatalystCoop@mastodon.energy
• On BlueSky: @catalyst.coop
• Connect with us on LinkedIn
• Play with our data and notebooks on Kaggle
• Combine our data with ML models on HuggingFace
• Learn more about us on our website: https://catalyst.coop
• Subscribe to our announcements list for email updates.