U.S. Synthetic Residential Building Telemetry Dataset Generated with EnergyPlus

This dataset provides national-scale synthetic residential building telemetry generated using EnergyPlus. It includes detached and attached residential models developed from probability distributions derived from actual ecobee DYD homes and the U.S. residential building stock. The dataset covers 48 contiguous U.S. states and 13 ASHRAE climate zones. The dataset contains 540 unique EnergyPlus models and their corresponding full-year simulation-output files in Parquet format. Each model is assigned a statistical weight in case_description.csv; applying these weights to the 540 unique configurations produces a probability-weighted analytical population of 1,200 residential cases, including 1,000 single-family detached and 200 single-family attached cases. The released dataset contains three core case-level components: case metadata (case_description.csv), EnergyPlus input files (*.idf), and synthetic telemetry (*.gz.parquet), together with supporting documentation, weather mapping, and QA/QC files, as described in Table 1. Table 1 Overview of the dataset components, file formats, and contents. Component Format Contents Case description case_description.csv Building and simulation metadata for the 540 unique EnergyPlus models, including building, envelope, infiltration, and HVAC information, including source properties and corresponding EnergyPlus inputs.. The case_weight indicates how many analytical cases each unique configuration represents within the probability-weighted 1,200-case population. EnergyPlus input *.idf Complete EnergyPlus version 9.5 model definitions and simulation assumptions for each of the 540 unique cases. Synthetic telemetry *.gz.parquet One full year of EnergyPlus outputs at five-minute intervals for each case (105,120 records per complete case). Dictionary dictionary.csv Variable names, descriptions, data source, and units for the fields reported in case_description.csv. Weather data summary weather_data_summary.csv Mapping of the selected ecobee city, selected EnergyPlus weather station, ASHRAE climate zone, and exact weather file used for each state. Miscellaneous data *.csv Supporting QA/QC and validation files: HVAC COP updates (idfs_v2_cop_change_report.csv), telemetry quality checks (telemetry_data_quality_summary.csv, telemetry_data_quality_variables.csv), and CPM climate-zone summaries (cpm_climate_zone_summary.csv). In the revised dataset release, case_description.csv, the EnergyPlus input files (.idf), and the corresponding synthetic telemetry files (.gz.parquet) were updated to incorporate corrected HVAC efficiency parameters. The revised case_description.csv reports both source efficiency information and the corresponding EnergyPlus input values. The affected *.idf files were updated with the corrected heating and cooling COP values, and the corresponding *.gz.parquet telemetry files were regenerated using the revised models. Case-level changes to the HVAC COP inputs are documented in idfs_v2_cop_change_report.csv, which reports the updated values for the affected EnergyPlus models. Additional supporting files document variable definitions, weather-station and climate-zone mappings, HVAC efficiency updates, telemetry quality checks, and climate-zone-level change-point-model summary statistics. The synthetic telemetry was evaluated against ecobee field data using HVAC runtime–outdoor-temperature relationships derived from a four-parameter change-point model (CPM). The comparison evaluates heating and cooling slopes, balance temperatures, and broad climate-dependent HVAC-runtime patterns. The CPM results are summarized in cpm_climate_zone_summary.csv. The Python code used for the change-point-model validation is publicly available through a cited GitHub repository: https://github.com/yjkim9411/residential_hvac_cpm. The dataset is intended to support the development and benchmarking of surrogate and telemetry-based building energy models, evaluation of retrofit and HVAC-sizing scenarios, climate-zone and state-level analyses, and validation of modeling methods against known physical parameters before their application to field data.