open-bps

Open, auditable screening of building performance standard (BPS) gaps and penalty exposure, built from the benchmarking data cities already publish. It starts with Washington, DC (BEPS) and Denver (Energize Denver).

Cities with a BPS law publish each building's energy use, and some publish its target. None of them publishes a ready answer to the question an owner, lender, buyer or reporter asks: how far is this building from its target, and what does missing it cost? open-bps answers that for every covered building:

• the law, encoded in dated rule files that cite the regulator's own documents;
• checked against the regulator's published verdicts wherever they exist;
• published as a refreshed open dataset, with a one-building Excel workbook anyone can audit.

Disclaimer. Every figure is a screening estimate from public data under the cited rules. It is not a compliance determination, and not legal, engineering or investment advice. Official figures come only from the jurisdiction (DC DOEE, Denver CASR).

What you get

Outputs:

• Data: dc and denver tables (Parquet and CSV), each with a .meta.json recording the rule version, the source snapshot (URL, retrieval time, SHA-256) and the attribution, plus a SUMMARY.md of headline figures. Released monthly: data-latest, with dated archives that include the raw snapshot.
• A workbook per building: inputs with their sources, every calculation a live formula, and a Checks tab that confirms the workbook reproduces the engine.
• A lookup app: search by building ID, name or address (app/streamlit_app.py).

How far you can trust it

• DC verdicts. DOEE publishes whether each building met its standard. The encoded rule reproduces that verdict for 1,797 of 1,806 buildings (99.5%). The 9 exceptions are boundary values and later corrections that DOEE decided outside the plain rule; they are listed in dc.meta.json, not fitted away.
• DC distances. DOEE's published distance from the standard is reproduced exactly for every EUI-metric building that has one.
• Denver penalties. The worked examples in Denver's own guidance are reproduced exactly by the tests. Denver publishes targets but no performance verdicts, so the gap and penalty figures are new information.
• Data quality. The monthly refresh stops, and publishes nothing, if the DC match rate falls below 97% or most Denver buildings stop scoring. Either would mean the upstream data changed shape.

Known limits, also stated in every output:

• Denver figures are upper bounds. The city subtracts renewable, transportation and electrification credits before comparing EUI to target, and those credits are not public.
• Denver schedule. Whether a building opted into the 2028/2032 extension is not public, so both schedules are shown.
• DC Performance Pathway. DOEE evaluates it on Adjusted Site EUI, which is published only on the owner portal. The estimate uses public site EUI (the 2018–2019 average against the latest year) and is labelled Estimated.
• DC Standard Target Pathway. Payments for buildings scored on ENERGY STAR need DOEE's own conversion models, so they are left blank, not guessed.
• Exposure assumes energy use stays at the latest reported value ("cost of inaction"). Retrofit costs are not estimated in v0.1.

Quick start

Python 3.11 or later.

pip install "open-bps @ git+https://github.com/Skippy-8/open-bps"   # PyPI release: pip install open-bps

open-bps fetch all          # dated raw snapshot of DOEE and Denver open data, with a manifest
open-bps score all          # data/scored/{dc,denver}.parquet|csv|meta.json and SUMMARY.md
open-bps lookup denver "union station"
open-bps excel denver 2428 --out one_union_station.xlsx
open-bps validate           # DC rule vs DOEE's published verdicts
open-bps rules              # rule versions and their primary sources

Use it from Python:

from open_bps.engine import denver, dc

denver.kbtu_not_achieved(eui=52.0, target_eui=48.3, gfa=150_000)    # 555,000 kBtu (Denver TG v3, Table 16)
dc.performance_payment(gfa=100_000, reduction_achieved=0.10)         # $500,000: half of the $1M maximum

To run the app: pip install "open-bps[app] @ git+https://github.com/Skippy-8/open-bps", then streamlit run app/streamlit_app.py. Without local data, it reads the latest release.

How it is built

open_bps/rules/*.yaml     the law, by jurisdiction and version; every value cites its primary source
open_bps/sources/         fetchers: DC GIS FeatureServer (DOEE), Denver open data hub (CASR)
open_bps/canonical.py     one schema per city; every building keeps a screen_status, none dropped silently
open_bps/engine/          pure functions per jurisdiction: verdicts, gaps, payments, penalties
open_bps/dataset.py       scored tables, metadata, quality gate, SUMMARY.md
open_bps/excel.py         one-building workbooks: live formulas and a Checks tab
app/                      Streamlit lookup
tests/                    offline; fixtures are samples of the real public data

See docs/methodology.md for the rules and assumptions, and docs/data_dictionary.md for every output column.

Data sources and licences

• DC: District of Columbia Department of Energy & Environment (DOEE), via Open Data DC, licensed CC BY 4.0.
• Denver: City and County of Denver, Climate Action, Sustainability & Resiliency (CASR), Energize Denver Master Dataset. Provided as is, without warranty; not for engineering purposes.
• open-bps code: MIT (LICENSE).
• open-bps scored datasets: CC BY 4.0, keeping the upstream attributions above (DATA_LICENSE.md).

Adding a jurisdiction

NYC Local Law 97 and Boston BERDO are next. Both cities publish building-level data. CONTRIBUTING.md describes the four pieces a new city needs:

1. a rule file citing primary sources;
2. a fetcher;
3. a canonical mapping;
4. an engine with the regulator's worked examples as tests.

Citing

See CITATION.cff. Please cite the data release you used (its tag carries the date).