GOOD datasets

Builds input data for the GOOD model (GOOD 2.x format) for the United States from EPA and EIA sources: existing plants, candidate wind, solar and battery sites, inter-regional transmission, hourly load, wind, solar and hydro profiles, and state renewable portfolio standards, organized by EPA IPM region.

Build

pip install -r requirements.txt
python build.py

This writes the processed files to Data/US/Processed/ and GOOD graphs to Outputs/, taking about two minutes. Every graph is validated against GOOD's input schema before it is written.

Output	Contents
Data/US/Processed/assets.json	every asset, keyed by handle
Data/US/Processed/lines.json	transmission lines
Data/US/Processed/profiles/<key>.json	8,760 hourly per-unit values per profile
Data/US/Processed/policies.json	state RPS policies as GOOD attribute filters
Data/US/Processed/metadata.json	format version, units, and every assumed parameter with its source
Outputs/US.json.gz	the whole country as one GOOD graph
Outputs/<region>.json.gz	ERC, FRCC, MIS, NENG, NY, PJM, SPP, S and WEC subgraphs
Outputs/California.json.gz	the six California regions, aggregated with good.aggregate (ratio 0.1)

Outputs/ is not tracked; rebuild it with python build.py.

import good

graph = good.graph.graph_from_json("Outputs/WEC.json.gz")
policies = good.utilities.read_json("Outputs/policies.json")
network = good.Network(steps=(4608, 4776)).from_graph(graph, policies)

Units

GOOD 2.x units: MW, MWh, hours, $/MWh for variable costs, $/MW for overnight capital costs, $/MW-yr for fixed O&M, kg/MWh for emission rates and Btu/kWh for heat rates. Costs are in 2016 dollars, as in the EPA tables. Loads are positive MW with profiles that peak at 1.0.

Sources

File in Data/US/Raw/	Source	Used for
eGRID2021_data.csv, eGRID2020.xlsx	EPA eGRID 2021 and 2020	plant location, primary fuel, emission rates
needs_v617_parsed.csv	EPA NEEDS v6 (parsed)	units, capacity, IPM region, fuel and O&M costs
needs_v6_transmission.csv	EPA Platform v6	inter-regional transfer capability and wheeling tariffs
table_2-2.csv	EPA Platform v6 Table 2-2	hourly load by region
table_4-38.csv, table_4-41.csv	EPA Platform v6 Tables 4-38, 4-41	candidate wind and solar capacity by resource and cost class
table_4-39_onshore.csv, table_4-43.csv	EPA Platform v6 Tables 4-39, 4-43	wind and solar generation profiles
table_4-40.csv, table_4-44.csv	EPA Platform v6 Tables 4-40, 4-44	wind and solar capital cost adders ($/kW)
table_4-15.xlsx, table_4-16.xlsx	EPA Platform v6 Tables 4-15, 4-16	regional cost factors, base capital and fixed O&M costs
eia860_2021_energy_storage.csv	EIA-860 2021, Schedule 3-4 (operable units)	existing battery energy capacity, for durations
capacity_factor.csv	not recorded	monthly hydro capacity factors by region
rps_fraction.csv	not recorded (the layout matches NREL ReEDS inputs)	RPS shares by state and year

EPA Platform v6 is the November 2018 reference case. Please fill in the two unrecorded sources.

Assumptions to review

PARAMETERS in Data/US/process.py holds every value the build introduces rather than reads from a table, each with its source; metadata.json repeats them. These are placeholders or assumptions:

• Wind and solar capacity credits (0.15 and 0.10). EPA gives 0-90% ranges that fall with penetration.
• Pumped hydro duration and efficiency (10 hours, 80%); EIA-860 does not report pumped-storage energy.
• A 2-hour default duration for batteries without an EIA-860 match (the 2021 fleet median), and a 15-year life for new batteries.
• Geothermal runs as must-run at full capacity; its capacity factor is not in the source data.
• Mean dispatch costs of coal and oil plants are rescaled to $23 and $32/MWh, a calibration carried over from earlier versions whose source is not recorded.
• renewable is true for wind, solar, geothermal, biomass and landfill gas. RPS eligibility differs by state.

Tests

pytest -q                 # unit tests and checks on Data/US/Processed
python build.py --check   # a fresh build must match Data/US/Processed

CI runs both. The build is deterministic: missing costs are filled by sampling with a fixed seed, and hour columns are selected by name, so results do not depend on the pandas version. The check allows numbers to differ in their last written digit, because floating-point results vary slightly across platforms.

Changes in GOOD 2.x format (2026)

• Outputs use GOOD 2.x units, classes and declarative policy filters.
• Wind profiles had 25 values per day: table_4-39_onshore.csv carries an index column, so position-based slicing started each day at "Day Of Month". Columns are now selected by name. (Also addressed by data edit in PR #1.)
• Hydro profiles dropped the first hour and were never attached to plants ("REGION:hydro:" vs "REGION:hydro"). They now use the monthly capacity factors directly as a daily energy budget.
• New wind and solar capital costs were $/kW divided by 1e6; they are now $/MW, with fixed O&M and EPA's capital charge rate.
• Storage has durations (EIA-860 for batteries) and efficiencies; new batteries use EPA Table 4-35 costs.
• Transmission lines carry EPA inter-regional losses, and the two directions of a path share a corridor.
• Missouri's state code had a trailing space, so its RPS applied to no assets.
• Offshore wind, biomass and landfill gas are flagged renewable.
• Existing wind and solar plants use a capacity-weighted average of their region's resource-class profiles instead of whichever class came first.
• Nuclear dispatch cost no longer includes fixed O&M (mean $21.2 to $6.8/MWh).
• Fill-in costs are sampled with a fixed seed.
• The notebooks are replaced by build.py; aggregation uses good.aggregate.

Known limitations

Profile file names contain ":" (for example WEC_BANC:wind:3.json), which Windows does not allow, so the repository cannot be checked out on Windows.

License

GPL-3.0; see LICENSE.