System map / Offline pipelines / Load profiles · NREL

Offline pipeline · load profiles

Load profiles: when in the day each appliance uses energy

On a time-of-use plan, a kWh at 6 pm can cost 40% more than one at noon, and solar only helps while the sun is up. So the model needs to know when each appliance draws power, not just how much per month. Those hourly shapes come from NREL ResStock: detailed simulations of real California housing stock, summarized per climate zone, end use and month.

Source
NREL ResStock 2025 Release 1 (AMY2018, OEDI)
Sample
~150 single-family homes per zone per run
Coverage
16 zones × 9 end uses × 12 months × 24 h
Built
2026-09-24 · replaces the old one-shape table

The pipeline

Offline, the script samples homes from NREL's public data lake, reads only the end-use columns it needs, and reduces each one to a 12-month × 24-hour table. Summing over the sample gives a stock-average shape, and each month is scaled so its 24 hours add to 1. At run time the model looks up your zone's shapes and spreads each electric appliance's monthly kWh across the day.

Why "upgrade runs"? Today's housing stock has few heat pumps, heat-pump water heaters and EVs, so a baseline sample can't say how they behave. ResStock also simulates the same homes with the upgrade installed. HVAC shapes come from run 4 (ducted heat pump), water heating from run 9 (HPWH), EV charging from runs 20/22. Lights, cooking, dryer and dishwasher use the baseline run.

Explore the shapes

Share of the day's energy in each clock hour. Solid = NREL for this zone (January, July). Dashed grey = the old single shape used for every zone and month. Shaded = PG&E peak window, 4–9 pm.

All nine end uses · · January

Red figure = share of the day's energy that lands in the 4–9 pm peak. Click one to show it above.

What it feeds

For each month, the hourly energy balance builds the home's electric load for one representative day: every electric appliance's monthly kWh ÷ days × its end-use shape, summed. Against that load it places solar output, runs the battery, and prices peak and off-peak purchases on your tariff.

L[m,h] = Σ_appliances kWh[m] / days[m] × shape[zone, end_use, m, h]
ApplianceNREL end use
Heat pump (heating / cooling)hvac_heating · hvac_cooling (each part spread separately)
Central AChvac_cooling
Heat-pump water heaterwater_heating
Heat-pump dryerclothes_dryer
Induction cooktop · ovencooking
EV chargingev_managed (charging avoids the peak)
Lights & plugs · anything elselights_plugs

What's modeled, what's simplified

  • Zone- and month-specific: winter heating peaks in the morning; summer cooling in late afternoon.
  • Physics-based simulations of real California homes, not survey guesses.
  • One clock with solar and the tariff: NREL's Eastern-time stamps are shifted to Pacific standard time, then daylight saving is applied.
  • A stock average, not your household: the shape is smooth; a family that runs the dryer at 7 pm peaks harder.
  • Managed EV charging assumed: EVs use the demand-flexibility run, which keeps charging out of the peak. Plug-in-and-charge would cost more.
  • Single-family detached only; weather year 2018 (AMY), while the energy totals use the TMYx typical year. A month with almost no use (heating in August) borrows that end use's annual shape.

Sources

NREL ResStock 2025 Release 1 · AMY2018Open Energy Data Initiative (OEDI) data lakeParker et al. 2025

Data includes information from the ResStock™ dataset developed by the National Laboratory of the Rockies (NLR) with funding from the U.S. Department of Energy.

Code
scripts/build_load_profiles.pydata/loads/end_use_profiles.jsondata/loads/sources/manifest.json src/load_profiles.py · DEVICE_END_USEdata/rates/device_load_shapes.json (old)docs/NREL_LoadProfiles_Plan.md