End-Use Load Profiles for the U.S. Building Stock

NREL and its research partners have developed a database of end-use load profiles (EULP) representing all major end uses, building types, and climate regions in the U.S. commercial and residential building stock.

End-use load profiles are critically important to understanding the time-sensitive value of energy efficiency, demand response, and other distributed energy resources. This foundational dataset can help electric utilities, grid operators, manufacturers, government entities, and research organizations make critical decisions about prioritizing research and development, utility resource and distribution system planning, and state and local energy planning and regulation.

The team, which includes Lawrence Berkeley National Laboratory (LBNL) and Argonne National Laboratory, used an innovative approach that benefited from the reach, cost-effectiveness, and granularity of physics-based building stock modeling using the ResStock™ and ComStock™ capabilities developed by NREL for the U.S. Department of Energy. These stock models have been informed by and validated against the best-available ground-truth data—from whole-building interval meter data, submetering studies, and other data sources. The resulting public database was generated using the stock models and does not contain the ground-truth data used to develop and calibrate the models.

Foundational Dataset of ~1 Million End-Use Load Profiles for the U.S. Residential and Commercial Building Stock

ComStock logo
ResStock logo

Building stock models calibrated through 70+ model updates, supported by data:

  • Electric load data from 11 utilities and 2.3 million meters
  • 15 end-use metering datasets

Example: Texas Residential Load (modeled end-uses)

Side-by-side graphs showcasing a dataset example of electricity use by hour of the day in a Texas residential building.

The project centered around NREL's building stock modeling tools, resulting in a foundational dataset of approximately 1 million end-use load profiles for the U.S. building stock as well as calibrated models for evaluating the impact of future scenarios and technologies.

Dataset Access

At the most fundamental level, the end-use load profile dataset is the output of approximately 900,000 (550,000 ResStock plus 350,000 ComStock) building energy models. The output of each building energy model is 1 year of energy consumption in 15-minute intervals, separated into end-use categories. The dataset has also been formatted to be accessible in four ways—via files of individual model characteristics together with annual results, pre-aggregated load profiles in downloadable spreadsheets, a web viewer, and a detailed format that can be queried with big data tools—to meet the needs of many different users and use cases. There are separate lists of public datasets available for residential and commercial building stocks. 

Pre-aggregated load profiles are available at data.openei.org. See the README.md file for details.

Many users seek end-use load profiles that reflect the sum or average of all buildings of a given type in a given geographic area. These are referred to as "aggregate'' load profiles. To support many use cases, aggregates for the following geographic resolutions were created:

  • 16 ASHRAE/International Energy Conservation Code climate zones
  • 5 U.S. Department of Energy Building America climate zones
  • 8 Electric System independent system operator and regional transmission organization regions
  • 2,400+ U.S. Census Public Use Microdata Areas
  • 3,000+ U.S. counties.

Each aggregate load profile is represented by a single comma-separated value file, which can be opened using common data analysis tools, such as Microsoft Excel. There is a column for each end-use category and fuel type and a row for each 15-minute time period. In addition to the timeseries data, a file contains a list of identifiers for all the building energy models included in each aggregate. These identifiers can be cross-referenced against a separate tab-separated value file, which contains the characteristics (age, type, height, etc.) of each building energy model. In this manner, it is possible to understand the characteristics of the building stock for each aggregate.

County-level aggregates are primarily provided to facilitate further aggregation to larger geographic areas, such as utility service territories. ResStock and ComStock model buildings in proportion to how real buildings are distributed. Heavily populated urban counties contain many buildings, so ResStock and ComStock include many models in these areas, and the aggregate load profiles for these areas are robust. However, rural counties in sparsely populated areas do not contain many buildings. Thus, ResStock and ComStock do not include many building energy models in these areas, and the aggregate load profiles for these areas are less robust. The team highly recommends rolling up aggregates to include a minimum of 1,000 models to increase the robustness of the load profiles.

Other users want to be able to quickly filter, slice, combine, visualize, and download the results in custom ways. Because of the size of the datasets, this is only possible using big data technologies and skill sets that put this out of reach for many users. To address this challenge, the team developed online data viewers to handle the big data behind the scenes. These data viewers are available at comstock.nrel.gov and resstock.nrel.gov.

A full dataset of individual building/dwelling unit load profiles is available at data.openei.org. See the README.md file for details.

The raw dataset is a group of several hundred thousand files, each containing the outputs of an individual building energy model, totaling 17 terabytes. Although processing these results using conventional desktop computing is impractical, several cloud service providers make the required computing power and querying technology available to those with the technical skill set. Additionally, some users may have in-house access to advanced computing resources or want to download a small subset of individual building load profiles for their own custom use cases. To facilitate these use cases, the raw individual building results, along with the corresponding building characteristics, have been published to a public website. They may be downloaded directly from this website or queried in place using big data technologies.

The building energy models are available at data.openei.org. See the README.md file for details.

The large number of building energy models used by ResStock and ComStock to represent the building stock makes it impractical for most organizations to run the full set themselves. However, some users expressed the desire to have access to the models for other use cases, such as modeling a smaller geographic area using a subset of the models or serving as a starting point for other modeling efforts. For these reasons, the individual building energy models have been published online. These models are available in the OpenStudio® format, to allow for simulations with EnergyPlus™. A file of building characteristics allows users to identify buildings with a target set of properties to download and use.

Technical Reports

Led by NREL's end-use load profiles research team, End-Use Load Profiles for the U.S. Building Stock: Applications Methodology and Results of Model Calibration, Validation, and Uncertainty Quantification, is accessible through the NREL library, free and open to the public.  

Stay tuned for End-Use Load Profiles for the U.S. Building Stock: Applications and Opportunities, which will be published by LBNL to detail deployment opportunities for general users.

Technical Report

This technical report documents the EULP dataset, including detailed description of model improvements made for calibration, along with an explanation of validation and uncertainty of results.

Executive Summary

An executive summary of End-Use Load Profiles for the U.S. Building Stock: Applications Methodology and Results of Model Calibration, Validation, and Uncertainty Quantification is available as a separate document.

Report Archive

End-Use Load Profiles for the U.S. Building Stock: Market Needs, Use Cases, and Data Gaps highlights first-year project accomplishments including the formation of a technical advisory group, identification of market needs and data gaps, and next steps to verify the accuracy of calibrated model outputs.


End-Use Savings Shapes: Residential Round 1 Dataset Release Webinar, presented by Elaina Present from NREL, provides information on the residential End-Use Savings Shapes work approach, details on the measure packages included, sample results, and data access tips. Slides from this presentation can be viewed separately. (September 2022) 

End-Use Load Profiles Dataset Access Demonstration provides step-by-step instructions, showing several ways to access the EULP dataset. (October 2021)

Three Years in the Making: Calibrated, Validated, and Publicly Available End Use Load Profiles for the U.S. Building Stock, presented by Eric Wilson and Andrew Parker from NREL and Natalie Mims Frick from LBNL, provides an overview of the project, details options to access the end-use load profiles, and shares information on two forthcoming reports. (Oct. 28, 2021)

End-Use Load Profiles for the U.S. Building Stock: Market Needs, Use Cases, and Data Gaps, presented by LBNL and NREL, details the project's first year. (Dec. 10, 2019)

Continued Work

Continued research building on the success of the EULP project is creating national datasets of measure impact profiles (also known as savings shapes) to empower analysts to tackle a range of questions concerning the potential of building electrification measures and more. The first round of 10 residential energy efficiency and electrification packages is available on the residential datasets page. Commercial savings shapes and additional residential savings shapes will be available in future years.

We are currently soliciting input on residential measure packages for the second round of savings shapes work. Anyone interested in providing suggestions on measure packages or package prioritization is encouraged to fill out the input form, which will be available until at least Nov. 15, 2022. 

Technical Advisory Group

A technical advisory group guided the dataset development to ensure that the direction and outcomes of the work were aligned with market needs. The technical advisory group included roughly 100 representatives from organizations that were likely to use the resulting load profiles in their work, including utility companies, utility program implementers, grid operators, consultancies, research centers, state regulatory agencies, and regional energy efficiency organizations.

Technical advisory group presentations are available for the dates below:

Frequently Asked Questions

The profiles are simulated using the ResStock and ComStock models, which have been calibrated and validated against an array of empirical datasets. The validation results and uncertainty for quantities of interest are presented in a forthcoming report, End-Use Load Profiles for the U.S. Building Stock: Methodology and Results of Model Calibration, Validation, and Uncertainty Quantification.
Definitions of the end-use categories can be found in the data dictionary.
The building types are aligned with definitions used by the Energy Information Administration Residential Energy Consumption Survey and Commercial Building Energy Consumption Survey.
There are two years of data available: one with 2018 weather and one with typical meteorological weather. The building stock characteristics for both data years are intended to represent the U.S. building stock circa 2018.
Both the 2018 and typical meteorological weather data years start with Monday, Jan. 1.
Weather data used for the modeling has been provided for regression modeling or other forecasting. The OpenStudio models used to generate the load profiles have been provided for those who want to run the models using weather data from different years, which can be purchased from various providers .
For some measures in which the shape of savings is expected to be similar to the shape of consumption (e.g., lighting efficiency, air conditioner efficiency), using consumption load profiles is appropriate. For measures where the shape is expected to be different (e.g., controls, heat pumps), then it may not be appropriate.
We recommend using tools such as OpenStudio or BEopt™ to analyze building electrification with heat pumps. Users can identify profiles with electric space heating or water heating using the metadata file, but because the end-use load profiles dataset represents the existing building stock circa 2018, it does not include very many heat pumps in cold climates. We expect to develop a subsequent dataset that will include heat pump load profiles for all climates and building types.
No, the project collected validation data from pre-COVID 19 and did not attempt to reflect its impacts because the permanence of these impacts has yet to be determined.
The weather station name, latitude, and longitude for each profile is included in the accompanying metadata file.
For applications needing individual building or dwelling unit profiles, such as solar plus storage analysis or utility rate design, it is not appropriate to use location aggregate load profiles (which will lack realistic spikes in load) or single individual load profiles or OpenStudio models (which will not be representative of the diversity of the building stock). We recommend using multiple building or dwelling unit load profiles or OpenStudio models for these types of evaluations, to see the diversity in results. Depending on the analysis question, it can take 1,000 or more samples to see convergence in output quantities of interest (see report for further details).
The terms "load shapes" and "load profiles" are often used interchangeably. We use "shapes" to mean the shape of the hourly or subhourly load (i.e., normalized) and use "profiles" to mean both shape and magnitude (i.e., kilowatts) of load.
No, the anonymous utility data provided by partners for use in validation cannot be shared.
Broadly, the ResStock and ComStock models used to generate these profiles attempt to reflect the diversity of the characteristics found in the building stock. An overview of the data sources for ComStock can be found on the ComStock website. Data sources and assumptions for each of the ResStock input probability distributions can be found in the comments at the bottom of each housing characteristic .tsv file on the ResStock Github repository.

Suggested citation:

Wilson et al. 2021. End-Use Load Profiles for the U.S. Building Stock: Methodology and Results of Model Calibration, Validation, and Uncertainty Quantification. NREL/TP-5500-80889. https://www.nrel.gov/docs/fy22osti/80889.pdf.

Data Sharing Partner Acknowledgments

We are grateful to the data sharing partners who shared anonymized data with the project team or helped the team find data. Future users of the load profiles created by this project owe these partners a debt of gratitude; without their significant effort and contributions, the project would not have been possible.

Northwest Energy Efficiency Alliance, Pecan Street, Florida Solar Energy Center, Southern Company, and Massachusetts Energy Efficiency Advisory Council deserve special mention for their leadership and foresight into end-use load research, as do ComEd and ecobee for making anonymized data broadly available to researchers.

  • Adams 12 Five Star Schools
  • AES Indiana
  • Alliance Center
  • Ameren Missouri
  • Bert Brains
  • Bonneville Power Administration
  • Center for Energy and Environment
  • Center for the Built Environment
  • Cherryland Electric Cooperative
  • City of Fort Collins Utilities/Colorado State University
  • City of Tallahassee Utilities
  • Clarkson University
  • DNV
  • Ecotope
  • Efficiency Maine
  • Elevate
  • EPB
  • Horry Electric Cooperative
  • Hot Water Research
  • kW Engineering
  • Los Angeles Department of Water and Power
  • Massachusetts Program Administrators
  • National Rural Electric Cooperative Association
  • New City Energy
  • New York State Energy Research and Development Authority
  • Northeast Energy Efficiency Partnerships
  • NV Energy
  • PacifiCorp
  • PEPCO/Exelon
  • Portland General Electric
  • Powerhouse Dynamics
  • Resource Central
  • Seattle City Light
  • Southern Company
  • Vermont Energy Investment Corporation/Green Mountain Power
  • Xcel Energy.


We want to hear from you! How are you using EULP datasets?

To share your use case or for more information, contact load.profiles@nrel.gov.

Subscribe to Our Mailing List

Opt in below to receive study-related updates.

NREL will use your information to send timely news or event information to your inbox for as long as you are subscribed. Change your mind at any time by clicking the unsubscribe link in the footer of any email you receive from us. We will treat your information with respect. Learn more about NREL's security and privacy policies.