Navigate to resstock.nlr.gov/datasets (ResStock) and nlr.gov/buildings/comstock (ComStock) to browse available dataset releases; as of 2025, ResStock 2025.1 and ComStock 2025-2 are the latest releases — note the dataset release version for reproducibility.
Access the underlying data from AWS Open Data Registry: the datasets are hosted in the S3 bucket s3://oedi-data-lake/nrel-pds-building-stock/ under organized prefixes by dataset type and release year; use the AWS CLI or boto3 with anonymous access (no credentials required for this public bucket).
Use the BuildStockQuery Python library (available on PyPI and documented at NREL's GitHub) to query ResStock and ComStock results efficiently without downloading the full multi-terabyte dataset; BuildStockQuery supports filtering by geography, building type, vintage, upgrade scenario, and energy variables against Parquet files on S3.
Understand the dataset structure: each building is represented by a sampled 'dwelling unit' with characteristics from the ResStock or ComStock probability distributions; results include annual energy consumption by end use (HVAC, water heating, lighting, appliances) and optional timeseries load profiles at 15-minute or hourly resolution.
For load shape modeling, download the end-use load profile (EULP) timeseries files for the target climate zone and building type combination; files are in Parquet format and contain 8,760 hourly (or 35,040 fifteen-minute) rows per representative building.
Cite the dataset DOI and release version in any published analysis; NREL updates datasets annually and results can differ between releases due to model improvements — do not mix results from different release years.
Known gotchas
The full ResStock and ComStock datasets are estimated at ~10 TB each in their complete timeseries form; downloading the entire dataset is rarely necessary and is costly in both time and egress — use BuildStockQuery's S3 query interface to retrieve only the aggregations or samples you need.
ResStock models residential buildings and ComStock models commercial buildings; they use different building energy simulation tools (EnergyPlus with OpenStudio) and sampling methodologies — do not merge results from the two datasets without carefully aligning the geographic and temporal resolution.
Upgrade scenario results (e.g., heat pump installations, insulation packages) are included as separate result columns alongside the baseline; baseline and upgrade results reflect modeled outcomes, not measured data, and should be validated against measured building stock data before use in utility planning.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp