Victor Okhoya November 2014
CARNEGIE MELLON UNIVERSITY
DOCTOR OF PROFESSIONAL PRACTICE
AREAS OF PRACTICE ASSIGNMENT
HOW RAPID ENERGY MODELING CAN HELP DESIGNERS IMPROVE BUILDING PERFORMANCE
IN EXISTING BUILDINGS
ABSTRACT
This paper introduces the concept of Rapid Energy Modeling as a possible solution to the
perceived difficulty of performing energy analysis studies on existing buildings. It begins by
describing the problem and suggests that traditional approaches like energy audits are time and
resource intensive. It then defines the concept of Rapid Energy Modeling. Three approaches to
energy analysis of existing buildings, including Rapid Energy Modeling, are discussed and
compared. Finally a comparative return on investment analysis is made by comparing Rapid
Energy Modeling to traditional energy analysis software tools.
INTRODUCTION
There is a perception in the design community that building performance analysis (BPA) for
existing buildings is complex, time consuming and difficult to master.1 For example, according to
the Canadian Industry Program for Energy Conservation (CIPEC)2, a traditional method of
energy analysis of existing buildings, the energy audit, consists of the following ten steps:
1. Conduct a condition survey
2. Establish the audit mandate
3. Establish the audit scope
4. Analyse energy consumption and costs
5. Compare energy performance
6. Profile energy use patterns
7. Inventory energy use
8. Identify Energy Management Opportunities
9. Assess the benefits
10. Report for action
These processes can take months to complete, are expensive and require a high level of
expertise. This problem of perceived difficulty of BPA in existing buildings has begun being
addressed by the digital design technology community. Solutions are being sought that bring the
benefits of BPA without excessive overhead in terms of learning and using the tools. One such
promising solution is Rapid Energy Modeling (REM).
1 Stumpf et al., 2011, p2.
2 See Natural Resources Canada, 2011.
1
Victor Okhoya November 2014
RAPID ENERGY MODELING
According to an Autodesk white paper Streamlining Energy Analysis of Existing Buildings with
Rapid Energy Modeling3 Rapid Energy Modeling refers to a streamlined and scalable approach
to performing energy assessments of existing buildings. Autodesk presents a three step process
for performing such assessments:
● Capturing existing building conditions
● Developing a 3D model of the building
● Performing analysis on the building
Existing conditions can be captured using digital photographs, aerial and satellite images, laser
scanning or, as we will see, thermal imaging technology.
The 3D model can be prepared using 3D modeling software that leverages captured digital data
as a background for model development. Reality capture tools like Insight3D, Agisoft Photoscan
and Autodesk Imagemodeler can also be used to help convert photo images into 3D models.
Once a 3D model has been prepared energy analysis simulations can be run using an
appropriate BPA tool. Outputs from such analyses include energy use intensities, annual energy
consumption by fuel type, heating and cooling loads, and carbon emissions among others.
The key benefit of REM is time and, therefore, cost savings. Users report that the REM exercise
for an average sized building (such as a typical three story office building) can be performed in
hours and days rather than the weeks and months of current processes.
The accuracy of analysis results from REM correlates favorably with real measurements as
discussed in the Department of Defence study below. Accuracy of building geometry generated
from reality capture methods is also quite high with modeled areas being within 7% of the actual
area according to the Autodesk study4.
In this paper we will look at three approaches to energy analysis of existing buildings:
● Energy Audits
● Traditional Software Approaches
● Rapid Energy Modeling by using Thermal Imaging as well as by using Autodesk
Software Workflows
We will compare these approaches in terms of cost, training time, time to perform an analysis
and accuracy of results. Finally, we will perform a return on investment comparison between
traditional software approaches and REM methods.
3 See Autodesk, 2011.
4 See Autodesk, 2011.
2
Victor Okhoya November 2014
We will use the following studies as our case study references5:
● Rapid Energy Modeling Workflow Demonstration, a US Department of Defence study.
● Rapid 3D Energy Performance Modeling of Existing Buildings using Thermal and Digital
Imagery, a study done at Virginia Polytechnic Institute and State University.
● Streamlining Energy Analysis of Existing Buildings with Rapid Energy Modeling, an
Autodesk study.
APPROACHES TO ENERGY ANALYSIS OF EXISTING BUILDINGS
Energy Audits
An energy audit is an inspection, survey and analysis of energy flows for energy conservation in
a building, process or system to reduce the amount of energy input into the system without
negatively affecting the output(s)6. According to the US DoD study, three levels of energy audits
are typically used based on the American Society of Heating, Refrigerating and Air-Conditioning
Engineers (ASHRAE) standard: walk through (ASHRAE Level 1), general (ASHRAE Level 2),
and investment grade (ASHRAE Level 3)7.
● Level 1, which is a rapid assessment of building energy systems done using a
walkthrough as well as energy benchmarking. This takes 1 -2 days and costs 500.00 -
700.00 per day.
● Level 2, which is a more detailed building survey of systems and operations. It includes
a breakdown of energy uses and sources, identification of energy conservation
measures and savings and identification of operational discrepancies. This takes 3 - 10
days and costs 500.00 - 700.00 per day or 1500.00 - 7000.00 per building.
● Level 3, which focuses on a whole building computer simulation and models the way the
building would respond to proposed energy saving measures. It requires longer term
data processing, computer models calibrated with field data and bid-level construction
cost estimating. This takes 10 - 50 days at a cost of 500.00 - 700.00 per day.
The DoD does not feel that REM workflows correspond directly to any of the levels of audit.
However, its outputs are a closer match to a level 2 audit and we will use this as a basis of
comparison.
Traditional Software Processes
Traditional Software approaches involve modeling the existing building conditions in a piece of
software and then running energy analysis exercises on this model. Several software solutions
exist for performing these analyses. More established tools include DOE-2, Energyplus and
Equest. These are typically free but have older interfaces and some are even MS-DOS based.
5 See the References section at the end of the paper.
6 Energy Audit. In Wikipedia. Retrieved in December 2014 from http://en.wikipedia.org/wiki/Energy_Audit.
7 Rupnow, J. & Sullivan, J., 2013, p80.
3
Victor Okhoya November 2014
Newer commercial tools include Designbuilder, Aecosim and IES VE. Traditional software
approaches can also be used as part of the energy auditing process.
Traditional software approaches, when used on existing buildings involve gathering existing
building condition data and then inputting this data into the energy analysis tool. This usually
takes the form of as-built or record drawings in CAD or PDF formats being redrawn in the
energy analysis tools. Sometimes, building measurements need to be undertaken first. Once
the existing conditions have been input, analysis simulations can be run and results reported.
Traditional software approaches provide the following challenges to the analysis of existing
buildings. First, the commercial versions carry a cost as high as thousands of dollars per
license. Second, these tools take time to learn with the learning curve being in weeks and
months for the more complex interfaces. Third, modeling existing conditions can take a long
time. Depending on the quality of existing conditions data this can range from days to weeks.
Finally, the time to perform analysis can also be lengthy depending on the complexity of the
energy simulation being run and the specifications of the hardware running it.
Rapid Energy Modeling using Thermal Imagery
We will refer to the Virginia Tech study conducted by Ham and Golparvar-Fard8 as an example
of REM using thermal imagery. The study uses digital and thermal imagery to rapidly create a
thermal model. The study notes that there are several challenges with existing approaches to
the modeling process for analyzing energy performance of existing buildings. These challenges
include:
1. Current energy modeling practices are time consuming and labor intensive. The process
of constructing models often takes weeks or months making it suitable mainly for high-
budget projects.
2. These energy models make assumptions that do not capture the variances of reality.
Sometimes these variances can be significant.
3. Creating these models requires skill and expertise that many designers simply do not
possess.
In order to address these challenges the study proposes to use image-based 3D modelling
techniques in conjunction with thermal imagery to rapidly create an existing conditions thermal
model.
This study is interesting for two reasons. First, it uses REM techniques to generate both
geometric as well as a thermal imagery which are then superimposed to create a composite
thermal model. Secondly, the composite model is itself a thermal analytical model. This means
that no further analysis of the model is required to produce results that can be used for design
decision making.
8 Ham, Y. & Golparvar-Fard, M., 2012.
4
Victor Okhoya November 2014
In this way, the authors’ note, the process is rapid both in its approach to generating the
analytical model but also in the fact that it eliminates the need for a separate analysis process.
Data capture for the study was performed in an office room of an existing instructional building
at Virginia Tech. The visual data was collected in the morning under natural daylighting
conditions.
The digital and thermal images were captured using an E60 thermal camera from FLIR Systems
which has a built in digital camera. The process began by capturing 429 unordered and non-
calibrated digital and thermal images. A streamlined image-based 3D reconstruction algorithm
was then used to generate a dense 3D point cloud of the scene. The density of the thermal
point cloud was 2,064,662 points while the density of the building geometry cloud was
8,488,888 points.
In order to generate a superimposed model it was necessary to co-register the thermal and
digital point clouds. Essentially it was necessary to correlate pairs of digital and thermal points.
However, because thermal images use gradient color coding which smooth over surface
intensities no distinctive features could be found for correlation. A novel approach had to be
devised.
Ordinary point cloud laser scanners are calibrated using calibration rigs like a checkerboard
located within the scene. However, low resolution thermal cameras cannot detect such a rig.
Therefore a thermal calibration rig was created using 42 small LED lights.
Using this thermal rig together with known parameters like the camera location and orientation
the thermal image was calibrated and registered to the digital model and a superimposed point
cloud created.
Finally, an augmented reality model viewer was used to enable visualization of the digital and
thermal models. A pair of the resulting images is shown below in Fig. 1.
Fig 1. REM using thermal images case study results.
5
Victor Okhoya November 2014
Rapid Energy Modeling using Autodesk Workflows9
We will refer to the US DoD study as an example of REM using Autodesk Workflows. This study
was undertaken by the Environmental Security Technology Certification Program (ESTCP) of
the US Department of Defense. Broadly speaking, the goal of the study was to evaluate REM in
order to determine if the workflow is capable of producing useful, rapid and cost effective
estimates of DoD buildings.
The study was conducted over a one year period using a population of 35 buildings and an
analyzed sample of 23 buildings. The buildings were spread across 8 locations and represented
7 different building types. Below is a summary of the test facility locations and types:
● 3 office buildings at the US Army Construction Engineering Research Laboratory in
Champaign, Illinois
● 1 office, 3 barracks and 1 gym at Fort Leonard Wood Army Base in Fort Leonard Wood,
Missouri
● 2 offices and 2 barracks at Joint Base Lewis McCord in Tacoma, Washington
● 2 offices and 1 barracks at the Naval Surface Warfare Centre in Panama City, Florida
● 4 offices at Peterson Air Force Base in Colorado Springs, Colorado
● 4 offices at Port Hueneme in Los Angeles, California
● 1 Barracks at Portsmouth Naval Shipyard in Kittery, Maine
● 1 Office, I Cafeteria, 1 School, 1 Fire Station and 1 Automotive Facility at Seymour
Johnson Air Force Base in Goldsboro, North Carolina
● 1 Office, 1 Automotive Facility and 1 Cafeteria at Naval Weapons Station Earle in Colts
Neck, New Jersey and Middleton, New York
● 2 barracks and 1 drill hall at Naval Station Great Lakes in North Chicago, Illinois
The study was done in a series of test phases namely Reality Capture phase, Modelling phase,
Analysis phase, Technology Transfer phase and the Reporting phase.
In the Reality Capture phase test sites and buildings were identified and building background
information was gathered using an installation energy questionnaire. In the Modeling phase,
Building Information Models (BIMs) were created. First, conceptual energy models were created
in Autodesk FormIt and Autodesk Vasari and then energy models were generated from the
conceptual models and run using Autodesk’s Green Building Studio web service.
During the Analysis phase modeled result data was compared to actual metered utility data. The
time and cost of the REM process was also compared to traditional energy audits. In the
Technology Transfer phase workshops, webinars and curriculum development took place and in
the Reporting phase the final report was developed.
The technology used for the study was mainly Autodesk software. Autodesk FormIt was used
for onsite modeling, Autodesk Vasari was used for modeling and analysis, Autodesk Revit was
9 Rupnow, J. & Sullivan, J., 2013
6
Victor Okhoya November 2014
used for additional model refinements and additional analysis tasks and Autodesk Green
Building Studio was the web based analysis engine and interface that did the actual calculation
and reporting.
The results of the study were both quantitative as well as qualitative.
Quantitative Performance Results
● Correlation of REM with annual energy electricity and fuel intensity
Results were within 10% error on 7 out of 25 buildings for electricity. Two buildings were
within +/-10% for natural gas. Overall there was 81.88% average accuracy for electricity
and 58.2% average accuracy for natural gas. This likely because electricity typically runs
on a schedule based on building use while natural gas varies more based on user
preferences.
The DoD felt that the results for electricity and natural gas were within good to
reasonable prediction levels as defined in the literature.
● Correlation of REM with overall annual energy use intensity
14 out of 25 buildings were within +/- 25% compared to baseline historical utility data.
Average accuracy was 77.56%. DoD felt that the REM EUI predictions were within good
to reasonable levels defined in the literature.
● Variance in monthly consumption
Results were within 15% of the target for 3 buildings using billing history and cost as
metrics. An additional 2 buildings were within 20% of the target. It was not realistic to
expect initial models to be within a 15% variance target since this is a target that
calibrated models aspire to.
● Testing the REM process for design alternatives to model potential energy savings
Energy Conservation Measures explored energy saving strategies for 5 buildings.
Savings greater than 30% were achieved in 3 out of the 5 buildings. The 2 buildings that
did not get to the 30% target had already undergone energy retrofits.
Qualitative Performance Results:
● Ease of learning technology and expertise required
Training completed at the time of publication indicated that DoD participants could learn
the REM workflow and begin creating and analyzing models in less than one day. This
was way below the target of 6 days.
● User satisfaction with REM technology
Participants indicated a high level of satisfaction with workflow as measured by survey
results.
● Ease of use creating REM models
7
Victor Okhoya November 2014
Preliminary results indicate that energy models can be completed in less than 3 hours
after the process is learned which is way superior to the 2 days per building target.
● Ability to scale the process across the DoD
3 individuals had been trained at the time of reporting compared to a target of 5
individual within the first year.
COMPARISON OF THE APPROACHES
In this section we provide a comparison of the approaches discussed above. We will compare
these approaches based on cost, training time, and time to perform analysis as well as
accuracy.
Cost Comparison
According to the DoD study a Level 2 Energy Audit of the 23 buildings they studied could cost
$179,673.0010 at an audit cost of $0.12/ft2. This is $7811.87 per building.
According to the DoD study an REM analysis of the 23 buildings studied could cost $6,900.00 or
approximately $0.005/ft2. This is $300.00 per building.
According to industry commentators11 a cost of $0.2/ft2 is a reasonable cost for traditional
energy modeling. Since the average size of the DoD buildings can be calculated as 60,000 ft2
the cost of energy modeling using a traditional approach would be $12,000.00 per building.
Fig 2. Cost per Building in dollars comparison.
10 Rupnow, J. & Sullivan, J., 2013, p80.
11 For example see http://energy-models.com/leed-and-energy-modeling retrieved in December 2014.
8
Victor Okhoya November 2014
Time to Train
According to the Canadian Institute of Energy Training (CIET) website a Certified Energy
Auditor course takes 3 days (24hrs) to complete.12
According to the DoD study REM workflow operator training takes 1 day (8hrs)13.
According to the IES VE website basic training on the product takes 3 days (24hrs)14.
Fig 3. Training Time in Hrs comparison.
Time to Perform Analysis
According to the DoD study a Level 2 Energy Audit takes 3 - 10 days. Let us take the median
value of 6.5 days (52 hrs)15.
According to the DoD study an REM analysis takes an estimated 3 hrs to complete16.
According to industry commentators17 energy modeling takes at least 40 hrs on average.
12 See http://cietcanada.com/events/certified-energy-auditor-cea/ retrieved in December 2014.
13 Rupnow, J. & Sullivan, J., 2013, p87.
14 See http://www.iesve.com/training/events retrieved in December 2014.
15 Rupnow, J. & Sullivan, J., 2013, p80.
16 Ibid.
17 For example see http://energy-models.com/leed-and-energy-modeling retrieved in December 2014.
9
Victor Okhoya November 2014
Fig 4. Time to Perform Analysis in Hrs comparison.
Accuracy of Analysis
According to the Ausgrid, an Australian electrical utility, Level 2 Energy Audits are typically
about 80% accurate18.
According to the DoD study an REM analysis is 77% accurate for Energy Use Intensity19.
According to Reeves et al. in their study, IES VE had accuracy, on average, for overall energy
usage of (86.45 + 51.90)/2 = 69.18.20
Fig 5. Accuracy Percent comparison.
18 See http://businessservices.ausgrid.com.au/~/media/Microsites/BusinessServices/Files/Metering/
Product%20and%20services/Energy_Audits.pdf retrieved in December 2014.
19 Rupnow, J. & Sullivan, J., 2013, p2.
20 Reeves et al., 2012, pp 584 - 586.
10
Victor Okhoya November 2014
COMPARATIVE RETURN ON INVESTMENT OF REM
We end with a return on investment (ROI) analysis that compares the ROI on an REM workflow
to the ROI on a more traditional energy analysis software workflow. As we have mentioned, part
of the deterrence to BPA is the perceived cost and difficulty of performing such analyses. If this
can be demonstrably reduced then there will be a higher rate of participation among designers.
Here we will use ROI as an objective measure of cost and ease of use.
In order to compare the ROI of using an REM workflow to the ROI of using a non-REM workflow
for BPA we will make a few assumptions. First we will use an ROI calculation method due to
Autodesk in their white paper BIM’s Return on Investment.21 According to this the ROI on design
software investment can be computed as:
(B - (B / (1 + E)) x (12 - C) / A + (B x C x D)
where:
A = cost of hardware and software
B = monthly labor costs
C = training time
D = productivity lost during training
E = productivity gain after training
We will compare an REM workflow using Autodesk Vasari to a non-REM workflow using
Integrated Environmental Solutions, Virtual Environment (IES VE).
Training time will refer to the time in months that it takes to complete formal training. We will
assume that the productivity lost during training stays in the same ratio as the training time. This
means if it takes three times as long to train on software A as B then the productivity lost for A is
three times as much as the productivity lost for B.
We will use an Effective Productivity Gain computed as:
Effective Productivity Gain = Productivity Gain x Accuracy Factor x Features Factor
This will penalise software that produces inaccurate results or software that lacks features that
are important for comprehensive BPA studies. The Accuracy Factor will be determined from
Vidmar’s paper22 where IES VE and Vasari are compared. Both pieces of software score 8/10
on this measure. The Features Factor will be determined from the following features which we
will consider as a minimal requirement for BPA tools:
21 See Autodesk, 2007.
22 Vidmar, 2013, pp 7-8.
11
Victor Okhoya November 2014
● solar analysis
● shadow analysis
● thermal analysis
● daylight analysis
● wind analysis
Features IES VE VASARI
Calculation speed/completion time 4 10
Visual environment & feedback 8 10
User interface simplicity & intuitiveness 5 8
Ease of use & learning curve 5 8
Software documentation 6 10
Community & technical support 10 7
Total out of 60 38 53
Figure 6. Comparative productivity gain for IES VE against Vasari
These features will be given equal weight in our analysis. Accordingly, IES has the full
complement of features and scores 1.0 while Vasari lacks daylight analysis features and scores
0.8.
Productivity gain after training will be computed based on Vidmar’s23 analysis (see Figure 6).
We will consider productivity to be a function of the following factors:
● calculation speed
● visual environment & feedback
● UI simplicity and intuitiveness
● ease of use and learning curve
● software documentation
● community and technical support
We will take the Productivity Gain to be the aggregate points awarded for these factors against
the total number of points possible.
According to both their websites Autodesk Vasari and IES VE both require at least dual core
CPU with 4GB RAM. The retail price for such a workstation at hardware resellers is about
23 Vidmar. 2013.
12
Victor Okhoya November 2014
$500.00. Also, according to the IES website, basic training on IES VE takes three days24 while
the US DoD reported that Vasari training could be taken in one day25. Monthly labor costs will be
taken uniformly at $16,000.00.
Accordingly first year ROI is as follows:
Vasari: (16000 - (16000 / (1 + 0.57)) x (12 - 0.05) / 1100 + (16000 x 0.05 x 0.05) = 60.89%
IES VE: (16000 - (16000 / (1 + 0.51)) x (12 - 0.15) / 5700 + (16000 x 0.15 x 0.15) = 10.57%
CONCLUSION
From the comparison of methods we saw that REM analysis performs better than either
traditional software processes or Level 2 energy audits for cost, training time and time to
perform analysis. We also saw that while REM analysis was not as accurate as Level 2 energy
audits for overall energy use, it was more accurate than traditional software methods.
IES VE VASARI
Cost of Hardware 500.00 500.00
Cost of Software* 5200.00 600.00
Monthly Labor Cost 16000.00 16000.00
Training Time (months) 0.15 0.05
Productivity Lost During Training (%) 0.15 0.05
Accuracy Factor 0.80 0.80
Feature Factor 1.00 0.80
Productivity Gain After Training**(%) 0.51 0.57
First Year ROI 10.57% 60.89%
*Vasari is free but attracts a cost for
cloud based analysis services
**Effective Productivity Gain
Figure 7. Comparative first year ROI for IES VE against Vasari
From the ROI analysis above we see that an REM tool has as much as a six times advantage in
ROI over traditional software approaches. This is because of the immense rapidity with which
24 See http://www.iesve.com/training/events.
25 Rupnow and Sullivan, 2013, p13.
13
Victor Okhoya November 2014
the software can be learned and deployed on projects. With such a return on investment it is
clearly a path to BPA that designers should consider.
We therefore believe that REM has an important role to play in helping designers perform BPA
studies on existing buildings and thereby achieving high performance solutions.
REFERENCES
Autodesk. (2007). BIM’s Return on Investment. Retrieved November 2014 from
http://images.autodesk.com/emea_s_main/files/gb_revit_bim_roi_jan07.pdf.
Autodesk. (2009). Rapid Energy Modeling for Existing Buildings. Retrieved November 2014 from
http://images.autodesk.com/adsk/files/rem_executive_summary.pdf.
Autodesk. (2011). Streamlining Energy Analysis of Existing Buildings with Rapid Energy
Modeling. Retrieved November 2014 from
http://images.autodesk.com/adsk/files/rem_white_paper_2011.pdf.
Ham, Y. & Golparvar-Fard, M. (2012). Rapid 3D Energy Performance Modeling of Existing
Buildings using Thermal and Digital Imagery. Advanced Engineering Informatics, 27(3), pp 395 -
409.
Loftness, V. (2014). Areas of Practice. Course Notes, Carnegie Mellon University, Autumn 2014.
Natural Resources Canada. (2011). Energy Saving Toolbox: An Energy Audit Manual and Tool.
Ottawa, Ontario. St Joseph Communication.
Reeves, T., Olbina, S., Issa, R. (2012). Proceedings of the 2012 Winter Simulation Conference:
Validation of Building Energy Modeling Tools: Ecotect, Green Building Studio and IES VE (pp
582-593). Presented at the IEEE 2012 Winter Simulation Conference.
Rupnow, J. & Sullivan, J. (2013). Rapid Energy Modeling Workflow Demonstration. Alexandria,
VA. ESTCP Program Office.
Stumpf, A., Kim. H., Jenicek, E. (2011). Early Design Energy Analysis Using Building
Information Modeling Technology. Champaign, IL. Construction Engineering Research
Laboratory.
Vidmar, J. (2013). Evaluation of simulation tools for assessment of urban form based on physical
performance. Retrieved November 2014 from
https://www.academia.edu/4820747/Evaluation_of_simulation_tools_for_assessment_of_urban_f
orm_based_on_physical_performance.
14
