This dataset is a subset of the Ecobee Donate Your Data (DYD) dataset. The Ecobee DYD data comprises user-reported metadata (of home and occupant characteristics), and data collected by Ecobee thermostats (reported in 5-minute intervals). Participant data are pulled from the Ecobee servers, and then anonymized to remove any personally identifiable information. This subset selects 1,000 single family homes in four states – California, Texas, New York, and Illinois, and span the entire year of 2017. In addition to the measurements, a metadata JSON file is included to illustrate the high-level contextual information of the dataset. The dataset can be analyzed to understand how a single-family heating, ventilation, and air-conditioning (HVAC) system operates, occupant behavior, and building thermal dynamics.
Use case:
Smart Home HVAC Analytics for Energy Efficiency and Occupant Behavior Insights
| Example Observation | indoorTemperature: 21.7 °C indoorHumidity: 38 % occupancy: 0/1 hvacMode: heat/cool/auto/off heatSetpoint / coolSetpoint: 20 °C / 24 °C hvacState: heating/cooling/idle HVAC runtime (s): 180 s outdoorTemperature: 10 °C time-of-day, day-of-week: 18:25 / Monday metadata (area, ZIP code, climate): |
| Example Action | ΔheatSetpoint ∈ {-2, -1, 0, +1, +2} °C ΔcoolSetpoint ∈ {-2, -1, 0, +1, +2} °C hvacMode change ∈ {heat, cool, auto, off} Fan on/off (binary, optional) |
| Example Reward | rt=−(c⋅Energy_kWht+λ⋅ComfortPenaltyt) Energy_kWh_t = (HVAC runtime_t / 300 s) × P_rated (≈ 3.5 kW heating, 2.8 kW cooling) ComfortPenalty_t = max(0, |indoorTemp_t – TargetComfort| – ε) Coefficients c, λ are adjustable |