The Home Energy Saver's heating and cooling calculations are based on DOE-2, an extensively validated HVAC modeling methodology.  Other algorithms and data are integrated in order to produce a whole-house analysis.  Many users report very good agreement with their actual energy bills.

In 2012, with leadership from the Florida Solar Energy Center, we completed an unprecedented accuracy assessment of the Home Energy Saver model. The tool was found to be extremely accurate (within 1% across groups of dissimilar homes), when given high-quality inputs including not only about the physical characteristics of the building but also how appliances and other equipment are operated by occupants. The following abstract will give you the essence of the findings, but please read the full report here:

Accuracy of the Home Energy Saver Energy Calculation Methodology

Danny Parker, Florida Solar Energy Center

Evan Mills, Leo Rainer, Norm Bourassa, and Greg Homan, Lawrence Berkeley National Lab


The Home Energy Saver (HES) suite offers popular online simulation tools that enable U.S. homeowners and energy professionals to rigorously evaluate home energy use and develop recommendations on how energy can be saved across all end uses. The underlying analytical system is also available as a web service to power third-party energy analysis tools. Given the system’s diverse uses, it is important that the simulation is robust and accurate. While the engineering methods are extensively documented and subjected to peer review, it is useful to evaluate how well HES predicts energy use in occupied homes. In this paper we compare measured to predicted energy use for 428 occupied homes in Oregon, Florida, and Wisconsin, representing a diversity building types, energy intensities, and occupant behaviors. We show how audit depth, knowledge of operational details, and sub-metered energy data can be valuable to the process of improving model accuracy—particularly for individual households, where energy use can vary three-fold for homes with virtually identical physical characteristics. Accuracy is strongly proportional to the quality and completeness of inputs, yet audit data are often deficient. Predictions are best—and the tendency of models to over-predict actual consumption is mitigated—when behavioral inputs match actual conditions. When averaged across groups of homes, HES predicts energy use within 1% of actual consumption when physical characteristics and occupant behavior are well accounted for. New research findings promise to confer even greater accuracy as they are incorporated into simulation tools.

You can read here about the scenarios we developed to assess the model in four modes of operation:

In performing the accuracy assessment, we had the benefit of extraordinarily high-fidelity submetered data for 10 identical occupied houses in Homestead, Florida.

A pre-release version Home Energy Scoring Tool (HEScore), which is an Asset Rating tool, has been separately assessed by the National Renewable Energy Laboratory (NREL).  The study is posted here. Note that this assessment was done on a BETA version of the Home Energy Scoring Tool.  An appendix includes some late assessment on a version much closer to the system as publicly launched in 2012, but those results are unfortunately not described in the Executive Summary or main body of the report.

Note: If you've come across the 2008 "Oregon EPS" study, please review our critique.