Whole-Building Electrification and Efficiency
ApprovedPublished 2025-02-07
Foreword
This document has been prepared by the OpenEAC Alliance, a collaborative group of energy engineering and measurement and verification professionals focused on creating transparent and standardized methods to measure the environmental impacts of demand side interventions. These methodologies are designed to align with best practices and reflect input from industry experts. The OpenEAC Alliance aims to enhance accountability and consistency in documenting energy and carbon savings from distributed energy resources (DERs).
About the OpenEAC Alliance
The OpenEAC Alliance is an organization of professionals dedicated to standardizing the measurement and verification of DERs. By focusing on clarity and rigor, the Alliance aims to overcome challenges that have hindered the proper valuation of DERs for their climate benefits. Central to this effort is the Energy Attribute Certificate (EAC), a standard that provides a detailed, auditable record of energy project impacts, including attributes such as type, timing, and location of energy production or reduction.
The OpenEAC Alliance collaborates to develop open-source methodologies, ensuring that DER projects can be accurately measured, verified, and valued in a way that promotes trust and drives market adoption.
Introduction
An electrification or energy efficiency Intervention changes how a building uses fuel and electricity. Where the building has a fuel meter and an electricity meter, those meters record the change directly, and the impact can be bound to the building rather than inferred from a population.
This document specifies how to quantify that impact from whole-building interval or billing data. The method that applies depends on the resolution of the data available.
Where the meter records hourly intervals for both periods, a regression model fitted at hourly resolution supplies the impact directly, and nothing is borrowed from outside the Asset. This is the preferred method, and every other method in this document is a fallback from it.
Where the meter is read monthly or daily, which is the common case for fuel, no model fitted to it can produce an hourly result on its own. A daily regression model then supplies the quantity, and a load profile from the NREL ResStock or ComStock dataset supplies only the hourly shape within each day. The meter decides how much energy the Intervention displaced. The load profile decides only when, within a day, that displacement occurred.
Hourly resolution is what allows an Energy Attribute Certificate to carry the timing of the impact rather than an annual total.
Scope
This document specifies the calculation of hourly avoided fuel consumption and hourly additional electricity consumption for an electrification or energy efficiency Intervention, from whole-building interval or billing data.
This methodology supports:
-
Electrification of end uses such as residential space heating, water heating, and full electrification.
-
Energy efficiency Interventions such as insulation, window replacement, or HVAC upgrades.
-
Metered analysis at hourly resolution, in which a model fitted to hourly interval data supplies the impact with no external load profile.
-
Metered analysis at daily or billing resolution, in which a daily model supplies the quantity and an external load profile supplies the hourly shape.
-
Commercial buildings, for space heating electrification only, modelled against the NREL ComStock dataset.
-
Impacts claimed after the first Reporting Period year, from a model of that first year.
-
Adjustment of the Counterfactual for the weather that occurred during the Reporting Period.
-
Interval data and billing data. Billing data is the common case for fuel.
This methodology does not currently support:
-
An Asset with no meter data in either the Baseline Period or the Reporting Period. The device-level deemed methodology applies instead, because nothing here binds the estimate to the building.
-
An Asset with Device-level temperature and flow rate sensors. The heat pump sensor methodology takes precedence, because a sensor measures the Device rather than the building.
-
Commercial water heating or cooking electrification. ComStock ships space-heating packages alone, so no upgrade matches those end uses and the Asset is refused rather than approximated against a residential package.
-
A fuel of propane or delivered wood, which the load profile datasets do not carry.
-
Conversion of energy impact to carbon impact. The Decarbonization Accounting methodology owns that conversion.
At the boundary, the calculation falls back through the defined sequence in Methodology 1.1, and the Implementer records every fallback taken. Where Reporting Period fuel data is absent, Methodology 1.2.4 scales a load profile to the Baseline model output. Where Baseline Period data is absent but Reporting Period data exists, Methodology 1.3.3 models the Reporting Period instead. Where neither period has sufficient data, the Asset is not eligible under this methodology, and the device-level attribute-based methodology applies.
A fallback is not automatic where the Implementer has declared what evidence the Asset carries. Methodology 1.1.5 states the rule: a declared evidence class is never silently downgraded, and an Asset that does not meet it is refused with the missing data named. Silent downgrading lets a claim describe weaker evidence than the one a Verifier was told to expect.
Normative References
The following documents are essential for the application of this specification. For references with no date, the latest edition applies:
-
OpenDSM EEMeter, which defines the daily model: https://lfenergy.org/projects/opendsm/
-
OpenDSM EEWeather: https://eeweather.readthedocs.io/en/latest/
-
NREL ResStock Documentation: https://resstock.nrel.gov/
-
NREL ResStock Data Dictionary: https://resstock.nrel.gov/data-dictionary
-
NREL End-Use Load Profiles Technical Documentation, 2022, which defines the measure upgrades: https://oedi-data-lake.s3.amazonaws.com/nrel-pds-building-stock/end-use-load-profiles-for-us-building-stock/2022/EUSS_ResRound1_Technical_Documentation.pdf
-
OpenEAC Alliance Shared Guidelines: https://methods.openeac.org/shared-guidelines
Terms and Definitions
-
Asset: The building in which the Intervention took place.
-
Intervention: A change made in a building that leads to a measurable difference in energy use patterns and carbon emissions.
-
Measure Type: A specific end use within a building that the Intervention changed, such as conditioned heating, water heating, or range oven.
-
Implementer: The party that submits the project data and applies this methodology.
-
Verifier: The party that checks conformance with this methodology.
-
Baseline Period: The year of meter data ending on the date the Intervention started.
-
Reporting Period: The period after the Intervention completed for which impact is claimed.
-
Counterfactual: The energy consumption of the Asset in the Reporting Period in the absence of the Intervention.
-
Daily Model: A regression of daily energy consumption against heating and cooling degree days, fitted by the OpenDSM EEMeter daily method.
-
Hourly Model: A regression fitted by the OpenDSM EEMeter hourly method, which segments the year and models each segment against weather at hourly resolution.
-
Load Profile: An external hourly profile, from the NREL ResStock dataset for a residential Asset or the NREL ComStock dataset for a commercial one, used only to distribute a daily quantity across the hours of that day.
-
Evidence Class: The Baseline and Reporting model resolutions that the Implementer declares an Asset to carry, which Methodology 1.1.5 forbids downgrading silently.
-
Loadshape: A time series of energy consumption at hourly resolution.
-
8760: A loadshape that covers one non-leap year at hourly resolution, being 8,760 hours.
-
Data Sufficiency: The fraction of a period covered by valid meter readings, as required under Data Requirements.
-
Scale Factor: The ratio that rescales a ResStock load profile to an annual quantity derived from meter data.
Data Requirements
The Implementer shall submit all of the following. Where a requirement is not met, the stated consequence applies.
1 Whole-building electricity meter data for the Baseline Period, as interval or billing data, covering at least 90% of the year ending on the date the Intervention started.
2 Whole-building fuel meter data for the Baseline Period, as interval or billing data, covering at least 90% of the same year.
3 Whole-building electricity meter data for the Reporting Period, covering at least 90% of the period claimed.
4 Whole-building fuel meter data for the Reporting Period, covering at least 90% of the period claimed, except where the Intervention removed the fuel service entirely. Where the fuel service was removed, the Implementer shall submit evidence of the removal, and the Reporting Period fuel consumption is zero.
5 The date the Intervention started and the date it completed. The Baseline Period ends on the start date and the Reporting Period begins on the completion date, so no period covers the installation itself.
6 The Measure Types that the Intervention changed, drawn from the Measure Types in Appendix 2.
7 The building attributes that Appendix 4.2 marks as required, which Methodology 2.2 consumes to select the load profile that supplies the hourly shape.
8 The location of the Asset, at sufficient resolution to resolve a weather station under Methodology 3.3.
9 Hourly outdoor air temperature covering the Baseline Period and the Reporting Period, from a source that satisfies Methodology 3.3.
10 The Evidence Class the Asset carries, where the Implementer declares one. This is optional. Where it is declared, Methodology 1.1.5 applies.
Where Baseline coverage under 1 or 2 is below 90%, the Asset is not eligible for the methods in 1.2.1 and 1.3.1, and the Implementer shall apply 1.3.3. Where Reporting Period coverage under 3 is below 90%, the Implementer shall shorten the Reporting Period to a span that meets the threshold, and shall report the excluded span. Where no period meets the threshold for either fuel or electricity, the Asset is not eligible under this methodology.
The Implementer shall not claim impact for any hour outside the Reporting Period as defined in 5. An impact claimed across the installation itself attributes the disturbance of the installation to the Intervention.
The Implementer shall report the uncertainty of the fitted model alongside the impact, as Methodology 2.1.5 requires. A savings figure carries no useful meaning to a Verifier without the uncertainty of the model that produced it.
Methodology
1 Impact calculation
1.1 Method selection
1.1.1 The Implementer shall calculate an hourly Counterfactual avoided fuel loadshape under 1.2, and an hourly additional electricity loadshape under 1.3, for the same Reporting Period.
1.1.2 The method that applies to an energy source is determined by the resolution of its Baseline data and of its Reporting Period data, not chosen freely. The Implementer shall resolve the resolution of each period first, then apply the method that 1.2 or 1.3 gives for that pair.
1.1.3 The order of preference is hourly, then billing with hourly reporting, then daily, then billing. The Implementer shall apply the first method whose data the Asset carries, and shall record the resolution found for each period.
1.1.4 A method that consumes a Load Profile is a fallback from the hourly method in 1.2.1 and 1.3.1. The Implementer shall record which fallback was taken and what data was missing, because the fallback introduces a shape the Asset did not produce.
1.1.5 Where the Implementer declared an Evidence Class, and the Asset does not carry the data that class describes, the Implementer shall refuse the measurement and shall report which energy source and which period fell short. The Implementer shall not fall back to a weaker method in that case. A declared class tells a Verifier what evidence to expect, so downgrading it silently misdescribes the claim.
1.1.6 Where the Asset is commercial and the Intervention electrified water heating or cooking, the Implementer shall refuse the measurement. The ComStock dataset carries space-heating packages alone, so no Load Profile represents those end uses.
1.2 Counterfactual avoided fuel consumption
1.2.1 Hourly Baseline and hourly reporting data
1.2.1.1 Fit an Hourly Model of Baseline Period fuel consumption under 2.1.
1.2.1.2 Predict hourly fuel consumption over the Reporting Period, which is the hourly Counterfactual. No Load Profile is used, because the model already resolves each hour.
1.2.1.3 Subtract the measured Reporting Period fuel consumption from the Counterfactual, hour by hour, to obtain the avoided fuel consumption.
1.2.2 Billing Baseline data and hourly reporting data
1.2.2.1 Fit a Daily Model of Baseline Period fuel consumption from the billing series under 2.1, and predict it over the Reporting Period.
1.2.2.2 Distribute each daily prediction across the hours of its day using the measured hourly reporting series as the shape, rather than a Load Profile. The Asset’s own reporting shape is closer to the Asset than any external profile.
1.2.2.3 Subtract the measured Reporting Period fuel consumption from the Counterfactual, hour by hour, to obtain the avoided fuel consumption.
1.2.3 Daily or billing data in both periods
1.2.3.1 Fit a Daily Model of Baseline Period fuel consumption under 2.1, and predict it over the Reporting Period.
1.2.3.2 Derive an hourly Load Profile under 2.3, summing the columns that Appendix 2 gives for the Baseline fuel.
1.2.3.3 Match the hours of that Load Profile to the hours of the Reporting Period under 3.1.
1.2.3.4 Rescale the matched loadshape so that each day sums to the daily prediction from 1.2.3.1.
1.2.3.5 Subtract the measured Reporting Period fuel consumption from the Counterfactual, hour by hour, to obtain the avoided fuel consumption. Fuel that the Asset still consumes is not avoided.
1.2.4 No Reporting Period fuel data
1.2.4.1 Derive an hourly Load Profile under 2.3 as an 8760, summing the columns that Appendix 2 gives for the Baseline fuel.
1.2.4.2 Fit a Daily Model of Baseline Period fuel consumption under 2.1, using weather data at the location of the Asset from 3.3.1.
1.2.4.3 Predict that model over the weather file that NREL published alongside the Load Profile, obtained under 3.3.2, to obtain an annual fuel quantity on the same weather basis as the profile.
1.2.4.4 Calculate the Scale Factor as the annual quantity from 1.2.4.3 divided by the sum of the 8760 from 1.2.4.1, and multiply every hour of the 8760 by it.
1.2.4.5 Apply the weather normalization in 3.2 to obtain the hourly Counterfactual for the Reporting Period.
1.2.4.6 Where the Implementer submitted evidence that the fuel service was removed at the Intervention, the Reporting Period fuel consumption is zero and the whole Counterfactual is avoided. Where no such evidence exists, the Implementer shall refuse the measurement rather than assume zero. Assumption 2 states why.
1.3 Additional electricity consumption
1.3.1 Hourly Baseline and hourly reporting data
1.3.1.1 Fit an Hourly Model of Baseline Period electricity consumption under 2.1, and predict it over the Reporting Period to obtain the hourly Counterfactual.
1.3.1.2 Subtract that Counterfactual from the measured Reporting Period electricity consumption, hour by hour, to obtain the additional electricity consumption.
1.3.2 Daily or billing data in either period
1.3.2.1 Fit a Daily Model of Baseline Period electricity consumption under 2.1, and predict it over the Reporting Period.
1.3.2.2 Derive an hourly Load Profile under 2.3, summing the columns that Appendix 2 gives for electricity, and match its hours to the Reporting Period under 3.1.
1.3.2.3 Rescale the matched loadshape so that each day sums to the daily prediction from 1.3.2.1.
1.3.2.4 Subtract the rescaled Counterfactual from the measured Reporting Period electricity consumption, hour by hour, to obtain the additional electricity consumption.
1.3.3 No Baseline Period data
1.3.3.1 Fit a Daily Model of Reporting Period electricity consumption under 2.1, and predict it over the Reporting Period to obtain a stable annual electricity quantity.
1.3.3.2 Derive an hourly Load Profile under 2.3 as an 8760, summing the columns that Appendix 2 gives for electricity.
1.3.3.3 Rescale the 8760 so that its annual sum equals the annual quantity from 1.3.3.1.
1.3.3.4 Apply the weather normalization in 3.2 to obtain the hourly additional electricity consumption for the Reporting Period.
1.3.3.5 This method produces no measured Baseline, so the Counterfactual for the fuel side comes from 1.2.4 or from the device-level attribute-based methodology. The Implementer shall record which.
1.3.4 Impacts claimed after the first Reporting Period year
1.3.4.1 Fit a model of Baseline Period consumption under 2.1, and a second model of the first Reporting Period year under 2.1, both at the resolution that 1.1 selected.
1.3.4.2 Predict both models over the period claimed. The Baseline model output is the Counterfactual. The first-year model output is the modeled observed consumption.
1.3.4.3 Where the selected resolution is daily or billing, derive a Load Profile under 2.3, match its hours under 3.1, and rescale each day of each series to the corresponding daily prediction from 1.3.4.2.
1.3.4.4 The difference between the two series is the additional electricity consumption for the period claimed.
1.3.4.5 The Implementer shall not apply this method beyond the point where the Device’s performance can be assumed stable. Assumption 5 states the effect.
1.4 Net impact
1.4.1 For each hour of the Reporting Period, the fuel impact is the negative of the avoided fuel consumption from 1.2, because the fuel is no longer consumed.
1.4.2 For each hour of the Reporting Period, the electricity impact is the additional electricity consumption from 1.3.
1.4.3 The Implementer shall report the fuel impact and the electricity impact as separate hourly series, in the base unit of each energy source. Summing them discards the distinction between the two energy sources, which the Energy Attribute Certificate carries.
1.4.4 The Implementer shall also report the Counterfactual projected onto the Reporting Period by calendar alignment alone, with no weather normalization applied. Its gap to the measured series is the raw year-over-year change, and a large difference between it and the normalized Counterfactual shows how much of the reported impact rests on the weather model.
1.4.5 Where an hour of the Reporting Period lacks meter data or weather data, the Implementer shall exclude that hour from both series, and shall report the excluded fraction of the Reporting Period.
2 Energy estimation methods
2.1 OpenDSM models
2.1.1 The OpenDSM EEMeter daily method fits a time-series regression of daily energy consumption against heating and cooling degree days. It accepts daily interval data and billing data.
2.1.2 The OpenDSM EEMeter hourly method segments the year and fits each segment against weather at hourly resolution. It requires hourly interval data, and it is the only method in this document that resolves an hour from the Asset’s own data.
2.1.3 The Implementer shall fit the model to the period named by the calling clause, and shall not fit one model across the Intervention. A model fitted across the Intervention absorbs the change it is meant to measure.
2.1.4 The Implementer shall submit the fitted model parameters and the model’s own goodness-of-fit output. A Counterfactual that a Verifier cannot refit carries no evidentiary value.
2.1.5 The Implementer shall calculate and report the uncertainty of the fitted model over the Reporting Period. Where the model has too few degrees of freedom to support an uncertainty estimate, the Implementer shall report that instead of omitting it.
2.1.6 The Implementer shall use the same weather source for the fit and for the prediction. Assumption 7 states the effect of using two sources.
2.2 Load Profile building selection
2.2.1 This method selects the single dataset building whose attributes best represent the Asset, to supply the hourly shape only. The quantity comes from the meter data.
2.2.2 Select the dataset by building type: ResStock for a residential Asset, ComStock for a commercial one. The upgrade numbering of the two datasets is not interchangeable, so an upgrade identifier from one dataset applied to the other names a different Intervention.
2.2.3 Match the Intervention to one of the predefined measure upgrades that the selected dataset publishes. Where no listed upgrade corresponds to the Intervention, and the Asset is residential, the Implementer shall select the whole-premises total columns in Appendix 2 and shall record the substitution. Where the Asset is commercial, 1.1.6 applies.
2.2.4 Apply every filter in Appendix 4.1, and the filter for every attribute that Appendix 4.2 marks as required. The result is the candidate population.
2.2.5 Where the candidate population is empty, the Implementer shall relax the floor area filter to the adjacent band, and shall record the relaxation. Selection supplies a shape rather than a quantity here, so a looser match costs less than it does in a deemed methodology.
2.2.6 Narrow the candidate population to one building using the attributes in Appendix 4.2 that are not marked as required. Where the value of such an attribute at the Asset is known, the Implementer shall filter on that value. Where it is unknown, the Implementer shall filter on the value held by the largest number of buildings in the candidate population, and shall record the substitution.
2.2.7 Where the narrowing leaves more than one building, the Implementer shall select the building whose floor area is closest to that of the Asset, and shall record the size of the remaining population. An unordered selection from a tied population is not reproducible by a Verifier.
2.3 Load Profile derivation
2.3.1 Obtain the hourly output file for the building selected in 2.2, for the measure upgrade matched in 2.2.3.
2.3.2 Sum, for each hour, the columns that Appendix 2 gives for the energy source being derived.
2.3.3 The result is an 8760 of energy intensity per unit floor area. Where the calling clause rescales the profile to a metered quantity, the units of the profile cancel and only its shape survives.
3 Helper methods
3.1 Loadshape time-matching
3.1.1 Given an hour in a Load Profile, match it to an hour of the Reporting Period that has the same hour of day, the same day of week, and the same ISO week number.
3.1.2 Where the Reporting Period contains an hour with no counterpart in the Load Profile, the Implementer shall match that hour to the nearest hour that shares its hour of day and day of week, and shall record the adjustment. An ISO week 53 has no counterpart in an 8760.
3.2 Weather normalization from an 8760 to the Reporting Period
3.2.1 Obtain hourly outdoor air temperature for the weather file that produced the Load Profile, under 3.3.2.
3.2.2 Obtain hourly outdoor air temperature at the location of the Asset, covering the Reporting Period, under 3.3.1.
3.2.3 Aggregate the 8760 to daily totals, and fit a Daily Model to those totals against the weather data from 3.2.1, using the method in 2.1.
3.2.4 Predict daily consumption over the Reporting Period, using the weather data from 3.2.2.
3.2.5 Match the hours of the 8760 to the hours of the Reporting Period under 3.1.
3.2.6 Rescale the matched hourly loadshape so that each day of the Reporting Period sums to the daily prediction from 3.2.4.
3.3 Weather data sources
3.3.1 The Implementer shall use an outdoor air temperature source that publishes its method and its station coverage. EEWeather satisfies this requirement within the United States.
3.3.2 The Implementer shall use, for a fit against a Load Profile, the weather file that NREL published alongside that profile. Fitting against a different year of weather than the one that produced the profile attributes weather response to the wrong conditions.
3.3.3 The Implementer shall use one source for the Baseline Period and the Reporting Period. A change of source between the periods introduces a step change that the model attributes to the Intervention.
Assumptions
The assumptions are ordered from the largest to the smallest expected effect on the reported impact. Assumption 3 applies only to the fallback methods that consume a Load Profile, and not to the hourly method in 1.2.1 and 1.3.1.
1 The whole-building meter attributes all of its change to the Intervention. Every model is fitted to whole-building consumption, so any non-weather change in the Asset during the Reporting Period lands in the reported impact. An occupancy change, an appliance replacement, or a second Intervention all bias the result, in whichever direction they act. This is the principal limitation of a whole-building method, and it is the price of binding the estimate to the building. Nothing in this methodology detects such a change, so the Implementer’s record of the site is the only control on it.
2 A removed fuel service means zero, and an absent meter does not. Methodology 1.2.4.6 treats the Reporting Period fuel consumption as zero only on evidence that the service was removed. Without that evidence, an Asset that still burns fuel for an end use the Intervention did not change has that residual consumption reported as avoided, and the error direction is not conservative. This is why the clause refuses rather than assumes, and why 1.2.3 takes precedence whenever Reporting Period fuel data exists.
3 A Load Profile supplies only the shape within a day. The fallback methods rescale each day of the matched profile to that day’s model prediction, so a shape error redistributes energy between the hours of a day and never changes a daily total. The effect is bounded within each day, and it affects the hourly attribution of an Energy Attribute Certificate rather than its quantity. This is why a looser building match is acceptable here, under 2.2.5, than in a deemed methodology. It is also why the hourly method is preferred: it borrows no shape at all, so this assumption does not arise.
4 Degree days capture the weather response of the Asset. The models regress consumption against weather. A building whose consumption responds to humidity, solar gain, or wind in a way the model does not track carries that error into the Counterfactual. The assumption weakens as Reporting Period weather moves outside the range present in the Baseline Period, and a prediction outside that range is an extrapolation. Reporting the non-normalized Counterfactual, as 1.4.4 requires, is what lets a Verifier see how much of the impact rests on this assumption.
5 A model of the first Reporting Period year represents later years. Methodology 1.3.4 treats the first-year model output as the observed consumption for a later period. Equipment performance drifts, and controls are adjusted, so the further the claim is from the first year the weaker the assumption. The direction of the error is not known in advance. This is why 1.3.4.5 bounds the method rather than extending it indefinitely.
6 A period at 90% coverage represents the whole period. The Data Sufficiency threshold admits a gap of up to 10%, which the model fills by prediction. The assumption is that the missing days resemble the covered days at the same weather. A gap concentrated in one season is worse than the same gap spread through the year, because it removes the weather range the model needs, and the threshold alone does not distinguish the two cases.
7 Two weather sources agree. Methodology 1.2.4 fits a model against weather at the Asset and predicts it over the Load Profile’s own weather file, so a bias in either source propagates into the Scale Factor and therefore into both sides of the impact. Because it enters both sides, the effect on the net impact is smaller than its effect on either series alone.
8 A commercial Asset resembles a ComStock building. ComStock represents commercial stock more coarsely than ResStock represents housing, and it ships space-heating packages alone. A commercial Asset therefore carries a weaker shape match than a residential one, which is a further reason to prefer the hourly method for commercial buildings, where interval data is more often available.
9 Occupancy patterns align by weekday and week number. The time-matching in 3.1 pairs hours by hour of day, day of week, and ISO week number rather than by calendar date. Holidays, and any occupancy pattern tied to a date, therefore land on the wrong hour. The effect redistributes consumption within the year without changing any daily total.
Appendix 1: Load Profile measure upgrades
This appendix is informative. The NREL End-Use Load Profiles Technical Documentation, cited in Normative References, defines each ResStock upgrade. The ComStock upgrades are documented separately and cover space heating alone.
1 Basic enclosure
2 Enhanced enclosure
3 Heat pumps, minimum efficiency, electric backup
4 Heat pumps, high efficiency, electric backup
5 Heat pumps, minimum efficiency, existing heating as backup
6 Heat pump water heaters
7 Whole-home electrification, minimum efficiency
8 Whole-home electrification, high efficiency
9 Whole-home electrification, high efficiency with basic enclosure package
10 Whole-home electrification, high efficiency with enhanced enclosure package
Appendix 2: Load Profile column mapping
This appendix is normative. Methodology 2.3.2 consumes it. It applies only to the fallback methods, because the hourly methods in 1.2.1 and 1.3.1 consume no Load Profile.
A column name carries the energy source as a substring, shown here as `<fuel>` for the Baseline fuel and `electricity` for electricity. Every name is prefixed `out_` and suffixed `_energy_consumption_intensity`, which the table omits for width.
| Replacement type | Baseline fuel columns | Electricity columns |
|---|---|---|
| Heating only | `<fuel>_heating`, `<fuel>_heating_hp_bkup`, and `<fuel>_fireplace` where the fuel is natural gas | `electricity_heating`, `electricity_heating_fans_pumps` |
| Water heater only | `<fuel>_hot_water` | `electricity_hot_water` |
| Whole building | `<fuel>_heating`, `<fuel>_heating_hp_bkup`, `<fuel>_hot_water`, and both `<fuel>_fireplace` and `<fuel>_range_oven` where the fuel is natural gas | `electricity_heating`, `electricity_heating_fans_pumps`, `electricity_hot_water`, `electricity_range_oven` |
Two rules apply to that table:
-
Where the replacement carries an electric resistance backup, the Implementer shall add `electricity_heating_hp_bkup` to the electricity columns.
-
The `range_oven` and `fireplace` columns appear for natural gas only, because the datasets carry no fuel oil, district hot water, or district steam consumption for those end uses.
Where Methodology 2.2.3 found no matching upgrade for a residential Asset, the whole-premises totals apply instead, for both energy sources:
| Case | Baseline fuel column | Electricity column |
|---|---|---|
| No matching upgrade, or an envelope-only Intervention | `<fuel>_total` | `electricity_total` |
Appendix 3: Worked illustration
This appendix is informative. It carries one set of realistic figures through an Asset whose two energy sources take different methods, which is the common case.
The Asset is a single-family detached house with a natural gas furnace and a natural gas storage water heater. The Intervention installed an air source heat pump and a heat pump water heater. The Intervention started on 2025-03-15 and completed on 2025-03-31, so the Baseline Period is 2024-03-15 to 2025-03-15 and the Reporting Period begins on 2025-03-31.
The gas meter is read monthly and the electricity meter records hourly intervals. Under Methodology 1.1.2 the two sources therefore take different methods: gas takes 1.2.3, which consumes a Load Profile, and electricity takes 1.3.1, which does not.
| Quantity | Value |
|---|---|
| Baseline gas billing coverage | 100% |
| Baseline electricity interval coverage | 98% |
| Method selected for gas | 1.2.3, Daily Model with a Load Profile |
| Method selected for electricity | 1.3.1, Hourly Model, no Load Profile |
| Baseline annual gas consumption | 13,244 kWh (452 therms) |
| Baseline annual electricity consumption | 7,180 kWh |
The models fitted under 2.1 and predicted over the first Reporting Period year return the Counterfactuals below. The Reporting Period was colder than the Baseline Period, so the gas Counterfactual exceeds the Baseline consumption.
| Quantity | Value |
|---|---|
| Counterfactual gas consumption | 14,010 kWh |
| Measured Reporting Period gas consumption | 1,320 kWh |
| Avoided gas consumption, under 1.2.3.5 | 12,690 kWh |
| Counterfactual electricity consumption | 7,390 kWh |
| Measured Reporting Period electricity consumption | 11,880 kWh |
| Additional electricity consumption, under 1.3.1.2 | 4,490 kWh |
The measured Reporting Period gas consumption of 1,320 kWh is the gas range and the fireplace, which the Intervention did not change. Clause 1.2.3.5 subtracts it, so it is not reported as avoided. Assumption 2 describes the alternative. Had the gas meter stopped reporting with no evidence of service removal, 1.2.4.6 refuses the measurement. The residual 1,320 kWh is 10.4% of the avoided figure, and reporting the full 14,010 kWh overstates the impact by that much.
The electricity figures used no Load Profile at all. The Hourly Model resolved each hour of the Counterfactual from the Asset’s own Baseline intervals, so Assumption 3 does not apply to that series.
The hourly rescaling in 1.2.3.4 applies to the gas series only. For 2026-01-15 the daily model predicts a gas Counterfactual of 78.4 kWh, while the matched day of the Load Profile sums to 61.2 kWh.
| Quantity | Value |
|---|---|
| Daily Counterfactual from the model | 78.4 kWh |
| Sum of the matched Load Profile day | 61.2 kWh |
| Rescaling ratio | 1.281 |
| Load Profile value at 07:00 | 4.10 kWh |
| Rescaled Counterfactual at 07:00 | 5.25 kWh |
The daily total after rescaling equals the model prediction exactly. This is the property that Assumption 3 relies on: the Load Profile moved energy between the hours of 2026-01-15 and changed nothing about the quantity claimed for that day.
Appendix 4: Load Profile selection filters
This appendix is normative. It applies only to the fallback methods that consume a Load Profile. The field names are those of the NREL ResStock Data Dictionary cited in Normative References; the ComStock dataset carries its own equivalents.
Appendix 4.1: Fixed filters
The Implementer shall apply both filters in this table.
| Field | Description | Value | Reason |
|---|---|---|---|
| in.vacancy_status | Whether the building is occupied or vacant | Occupied | An unoccupied building carries no occupancy-driven load, so its Load Profile does not represent an occupied Asset. |
| in.hvac_heating_efficiency | Presence and efficiency of the heating system | Starts with “Fuel” | An Asset that was already all-electric has no avoided fuel, so the Baseline heating system is fuel-fired. |
Appendix 4.2: Attribute filters
The Implementer shall filter on every attribute marked required, under Methodology 2.2.4. The remaining attributes narrow the candidate population under Methodology 2.2.6.
| Field | Description | Required |
|---|---|---|
| in.state | The state in which the building is located | Yes |
| in.geometry_building_type_recs | Building type classification based on RECS data | Yes |
| in.geometry_floor_area | Total floor area of the building | Yes |
| in.geometry_stories | Number of building storeys | Yes |
| in.vintage | Period in which the building was constructed | Yes |
| in.hvac_heating_type_and_fuel | Presence, type, and fuel of the heating system | Yes |
| in.insulation_wall | Wall construction type and insulation level | No |
| in.insulation_floor | Floor insulation level | No |
| in.insulation_ceiling | Ceiling insulation level | No |
| in.heating_setpoint | Baseline heating setpoint with no offset applied | No |
| in.heating_setpoint_offset_magnitude | Magnitude of the heating setpoint offset | No |
| in.infiltration | Air leakage rate of the living space and the garage | No |
| in.hvac_heating_efficiency | Presence and efficiency of the heating system | No |
| in.hvac_heating_type | Presence and type of the heating system | No |
| in.water_heater_fuel | Fuel type used for water heating | No |
The postal code of the Asset resolves the climate region used for selection, so no separate climate zone attribute is filtered. Because selection supplies a shape and not a quantity here, Methodology 2.2.5 permits the floor area filter to be relaxed, which it does not permit in a deemed methodology.