Methodology

Home energy audits provide valuable home energy performance insights and actionable recommendations. Until recently, residential energy performance information has been available only to the less than three percent of single-family homes which have undergone on-site assessments like U.S. Department of Energy (DOE)’s Home Energy Score (HES) and RESNET’S Home Energy Rating System (HERS) Index. Even then, this information is typically not readily available to prospective buyers of those homes. In-home audits provide valuable home energy performance insights and actionable recommendations, but by nature, they are difficult to scale up across the residential sector given the time, cost, and initiative required.

Virtual and automated Energy Models, like the one used in the Vermont Home Energy Profile (VHEP), can increase energy affordability and understanding by leveraging public data to provide fast and free energy estimates for all homes. Instead of the select few who undergo on-site assessments, which cost about $400 in Vermont, all homeowners have the opportunity to more holistically understand their home’s full operating costs and then access to resources and recommendations to lower their energy bills and improve the healthiness and comfortability of their homes.

Model Assumptions

The energy model used in the VHEP was developed by ClearlyEnergy. The model independently evaluates each component of a home’s energy consumption. To balance accuracy with ease of use, the project team developed the VHEP energy model with significant input from the Vermont Energy Labeling Working Group and the Montpelier Energy Efficiency Working Group. To streamline the process and to generate comparative results, the model makes some assumptions:

• Space and Water Heating: Water heating modeling uses historic water heater efficiency data in an approach similar to that used in U.S. DOE’s Home Energy Score. Space heating consumption depends on the type of home, weather, home insulation and air leakage, and the age and type of heating system.
• Occupancy Standardization: Some elements of consumption, such as water heating for showers and laundry, are dependent on the number of occupants. These elements use home size-to-people ratio assumptions as used in U.S. DOE’s DOE2 model originally used in the Home Energy Score.
• Weather and Thermostat Standardization: The model works with the assumption that the average user sets thermostats to 72 degrees Fahrenheit in the winter and 78 degrees Fahrenheit in summer. The model uses ZIP code level heating and cooling degree day data.
• Appliances: The tool asks for the age of appliances (refrigerator, freezer, dishwasher, washing machine, dryer) using time blocks correlating with changes in federal appliance standards.
• Lighting and Plug Load: Other consumption depends on home size, type of lighting, and intensity of electric appliance use. The lighting assumptions used in the model are based on U.S. DOE’s Residential End-Use Consumption Study. Plug load is calibrated using data from the U.S. Energy Information Administration’s Residential Energy Consumption Survey.

ClearlyEnergy Model Testing Results

Home Energy Test Benchmark

The U.S. DOE’s Home Energy Score, an industry standard for in-home energy assessments which uses 50+ parameters, also generates estimated cost and energy savings and uses standardized models for certain parts of the assessment. The Home Energy Score is the basis of a 1-10 point rating “miles-per-gallon” equivalent rating assigned to the home. Using 1,300 homes, the National Renewable Energy Laboratory (NREL) conducted an assessment of U.S. DOE’s Home Energy Scoring Tool, found that HES-predicted energy usage for electricity and natural gas differed about 24 percent from measured usage.

Comparison Study by ClearlyEnergy

At DOE’s request, ClearlyEnergy compared estimated home energy consumption and costs from its model with DOE HES results using a sample of approximately 7500 records provided by DOE. Results of the model-to-model comparison indicate that average estimates are very close for homes across fuel types and the correlation is highest for the most prevalent fuels: electricity and gas. With few exceptions, the fit is within 5% and the comparisons show no obvious biases in the treatment of heating and cooling load by geography, home age or size.

ClearlyEnergy’s AEM model uses a standardized information set available from public records to derive an estimated home energy consumption and cost. This study found the following correlation rates by fuel type: heating oil (58%), propane (68%), natural gas (70%), and electricity (91%).

ClearlyEnergy’s AEM+ model supplements public data with information on home systems, appliances and the building envelope which can be provided by homeowners or their realtor or inferred from the listing. This is the model used in VHEP. This study found the following correlation rates by fuel type: heating oil (79%), natural gas (85%), propane (90%), and electricity (91%).

Indepdent Assessment by the Rocky Mountain Institute

The Problem

The report notes that until recently, residential energy performance information has only been available to the less than 3% of single-family homes which have undergone on-site assessments like DOE’s Home Energy Score (HES) and RESNET’S Home Energy Rating System (HERS). Even then, this information is typically not readily available to prospective buyers of those homes.

The Solution

By leveraging public data to provide fast and free energy estimates for all homes, algorithm-based models are making home energy transparency more the rule than the exception. Instead of the select few who undergo on-site assessments (which cost about $400 in Vermont), all homeowners have the opportunity to more holistically understand their home’s full operating costs and the access to resources and recommendations to lower their energy bills and improve the healthiness and comfortability of their homes. RMI conducted an accuracy assessment of such models, namely that of VHEP’s ClearlyEnergy, in comparison to HES results.

The Results

RMI’s analysis of almost 8,000 homes across 27 states shows that remotely generated home energy estimates are just that—estimates—but they are accurate enough for a variety of useful applications. RMI concludes that automated home energy estimates may be sufficiently accurate for several use cases in light of accuracy standards acceptable in other industries, especially given that many focus on energy costs, where algorithm-based estimates can readily outperform other estimates by leveraging more granular utility rate data. RMI believes the vendor accuracy ranges are sufficient for use cases including resident home energy education and informed budgeting for homeowners and buyers. RMI expects that further market penetration, more claiming by homeowners, and scalable new data sources can all support a positive reinforcing loop that improves algorithm accuracy over time—just as MPG accuracy for cars improved over time.

Total Energy Use

Looking at individual homes, estimates showed a 20–30% average difference from HES estimates:

• Nearly three-quarters of all homes analyzed were less than 30% different
• Nearly half of all homes analyzed were less than 20% different
• More than one-quarter of all homes analyzed were less than 10% different

In aggregate, estimates showed an overall average difference within +/- 10% of HES estimates.

Total Energy Costs

Algorithm-based models can outperform on-site audits on energy cost estimates given their ability to use significantly more granular utility rate data (HES currently uses statewide average utility rates). Using the vendors’ implied utility rates, converted HES energy cost estimates were 9–22% different from reported HES cost estimates, sometimes closer to actual costs than those estimated by HES. In sum, these studies show that algorithm-based home energy assessment models are on par with industry standard.

Vermont Energy Investment Corp

In the preliminary phases of developing the Vermont Home Energy Profile, the Vermont Energy Investment Corporation (VEIC) analyzed a sample of 300 Vermont homes with historic bill data using VHEP. VEIC’s results were found to be comparable to the NREL and RMI studies, determining that the VHEP AEM is capable of producing reasonable high-level asset-based energy cost estimates.