Modeling an Existing Building in DesignBuilder/

EnergyPlus: Custom vs. Default Inputs,”

Citation
Wasilowski, Holly A. and Christoph F. Reinhart.2009. Modeling an Existing Building in
DesignBuilder/EnergyPlus: custom vs. default inputs. Proceedings of (Building Simulation) the
International Building Performance Simulation Association International Conference, Glasgow,
Scotland, July 2009: 1252-1259.

Permanent link
http://nrs.harvard.edu/urn-3:HUL.InstRepos:29663437

Terms of Use
This article was downloaded from Harvard University’s DASH repository, WARNING: No
applicable access license found.

Share Your Story

The Harvard community has made this article openly available.
Please share how this access benefits you. Submit a story .

Accessibility
Accepted at Proceedings of (Building Simulation) the International Building Performance Simulation
Association International Conference, Glasgow, Scotland, July 2009. Awarded Arup Engineering Best-
Paper Prize for Simulation in the Design Process

Energy Simulation of an Existing Building Using Customized Internal Load

Schedules and Weather Data as Opposed to Default Assumptions
Holly A. Wasilowski and Christoph F. Reinhart
Harvard University, Graduate School of Design
Cambridge, MA, USA
[email protected], [email protected]

makers and – as a direct result – an exponentially

ABSTRACT rising number of so-called ‘green’ new construction
This paper describes an effort to build and partially and renovation projects. In North America, the US
validate an energy model of an existing educational Green Building Council’s LEED (Leadership in
building located in Cambridge, MA, USA. This work Energy and Environmental Design) rating system
was carried out as part of a research seminar for (USGBC 2009) has established itself as a de-facto
graduate architecture/design students and included industry standard to demonstrate the ‘greenness’ of a
four related tasks: Modelling the building's geometry building. Most projects rated under LEED require an
and thermal properties in DesignBuilder/EnergyPlus, energy model to establish the number of energy-
generating a site-specific weather file based on near- efficiency credits for which the project is eligible
site measured data, assessing internal load schedules under LEED.
based on a detailed building survey, and collecting One outcome of the above-described rising demand
monthly metered data for heating, lighting, and for building performance simulation is an acute
cooling over a whole year. The purpose of the shortage of qualified consulting engineers who can
seminar was (a) to evaluate how effectively design provide their services to design teams. Until now,
students can use a state-of-the-art graphical user energy modelling has generally been performed by
interface (GUI) such as DesignBuilder and (b) to mechanical engineers or specialized consultants. This
quantify the value of using customized internal load practice has lead to the dilemma that integrated
schedules and weather data as opposed to default design practice, i.e. early involvement of the energy
GUI inputs. The authors found that the students consultants in the design process, leads to increased
quickly learned how to navigate the DesignBuilder up-front costs. In this climate, design teams might
GUI but were frustrated by the model data reconsider the notion of ‘who’ should actually carry
hierarchy/inheritance and that customized schedules out an energy simulation. While much of the more
cannot be assigned more efficiently. The benefit of advanced modelling tasks involving complex HVAC
using customized weather data as opposed to a local systems and advanced system integration are likely to
TMY3 file turned out to be small whereas using stay in the domain of engineer and energy
customized as opposed to default internal load consultants (Augenbroe, 2002), architectural firms
schedules reduced the relative error of predicted might actual want to start building up at least some
versus metered annual electricity use from 18% to energy modeling capabilities in-house. The most
0.2%. Each category of the customized internal load recent generation of commercial, high-end graphical
schedules including: occupancy & plug-loads, user interface seems to cater to the needs of
lighting, and air handling unit schedules contributed architectural firms with the developers suggesting
significantly to the increased accuracy of the annual that these tools have become so intuitive that they
energy load predictions. can be used by "everyone, even architects”. The
potential benefits for architectural firms to add
Keywords: measurement & verification, teaching
energy simulations to their portfolio are shorter
energy simulation, occupant behaviour,
communication paths and more effective design
benchmarking, weather files
feedback loops leading to shorter design times. The
INTRODUCTION disadvantages are equally clear as this could turn into
one more task that the architect must take on without
Across the North American building design industry,
necessarily increasing project budgets.
there is a growing interest in computer-based
building energy simulations. One of the drivers for These potential advantages and shortcomings
this change is a rising awareness of sustainable notwithstanding, an initial question worth
design practices among building owners and policy investigating is whether the current generation of
energy modelling software can actually be easily simpler than modeling Gund Hall. These
picked up by architectural students. Schmid recently assignments involved comparing architectural design
reported his experiences of teaching energy decisions, such as building massing, window-to-wall
simulation to architectural students in Brazil ratio, and envelope specifications in houses and
(Schmid, 2008). He found that students could single-zoned office buildings.
ultimately learn how to use energy modelling Finally, for the Gund Hall project, the students
software, but they expressed difficulty with the divided into four groups: a weather group, modelling
“user-friendliness” of the interface. One of the group, HVAC & metered energy group, and survey
limitations of the Schmid study was that the students group. Each group spent approximately six weeks on
used a simulation model written by one of the study their individual tasks, and at the end, combined the
authors. This model had a basic GUI but nothing information into one model of Gund Hall. The
comparable to some of the advanced current sections below describe this work in more detail.
commercial tools such as Sketchup/OpenStudio
The Case Study
(Google, 2009 and NREL, 2009), Ecotect (Autodesk,
2009) and DesignBuilder (DesignBuilder, 2009). All Gund Hall is a 16,350 gross m2 iconic modernist
three tools allow the user to build relatively complex building built in 1972 that exhibits a number of
building geometries and to export them into features that make it an interesting modelling object.
EnergyPlus for an energy simulation (US-DOE, Not only does Gund Hall have the expected academic
2009). For this study, the authors used the functions involving classrooms and offices, but also
DesignBuilder software since it also comes with a library, workshops, a small cafeteria, and a large
extensive data templates for a variety of building 24-hour studio space with seemingly erratic occupant
simulation inputs such as typical envelope schedules. The studio consists of a large open space
construction assemblies, lighting systems, and with multiple levels of balconies and a stepped,
occupancy schedules. These templates can be ziggurat-like roof. The building has over 100
especially enticing to beginners who may not have a separate roof surfaces and its envelope is
sense for when more accurate inputs specific to their characterized by large areas of glazing and exposed
building may be desirable. This invites the question, concrete (see Figures 1 and 2). In addition, Gund
"in which situations are these templates an acceptable Hall is a multi-faceted building serving a multitude
shortcut, and in which situations is it worth the effort of programs which vary throughout the year.
to generate custom inputs?" However, given that the building is connected to the
This paper investigates two questions: How campus’s district steam and chilled water system, the
successfully can a group of architectural students building does not include an on-site heating and
learn how to build an energy model of a complex cooling plant, somewhat simplifying the HVAC
commercial building over the course of a 13-week modelling and reducing the impact of part-load
term and how much accuracy can be gained by using performance curves.
customized weather data and internal load schedules
as opposed to default DesignBuilder inputs. These
two questions were addressed as part of a research
seminar on ‘Building Performance Simulation –
Energy’ that was offered during the Fall 2008 term at
Harvard University’s Graduate School of Design
(GSD). The case study building was the GSD’s own
building, Gund Hall, located in Cambridge, MA,
USA.
Figure 1 Photo of Gund Hall
METHODOLOGY
The Research Seminar
As stated above, the goal of the research seminar was
to teach building energy simulation to a group of
students who would subsequently build and evaluate
a model of Gund Hall using the DesignBuilder
interface for the EnergyPlus simulation engine.
Eleven graduate students, each working toward a
Figure 2 DesignBuilder Model of Gund Hall
master degree in architecture, urban planning,
sustainable design, or design technology worked on Weather Group
this task. The students had various professional Gund Hall is located in Cambridge, Massachusetts,
backgrounds including multiple practicing architects USA, which is part of the Boston Metropolitan Area.
but no mechanical engineers. None had any prior A TMY3 weather file for Boston-Logan Airport is
experience using energy simulation software. First, available from the U.S. Department of Energy
the students completed four assignments that were EnergyPlus climate file database (US-DOE 2009).
TMY3 data sets are derived from the 1991-2005 Geometry Group
National Solar Radiation Data Base (NSRDB) The responsibility of the geometry group was to
archives, i.e. they represent typical meteorological model Gund Hall in DesignBuilder using drawings
conditions over several years. In order to compare from multiple renovations (see e.g. Figure 4). The
energy model predictions to metered energy use for resulting DesignBuilder model contains over 100
Gund Hall the Weather Group compiled two custom different zones, 8 different exterior wall types, and 5
EnergyPlus Weather (EPW) files. The first file different window types. Envelope properties were
(EPW1) included data from November 1, 2007 to taken from an earlier Gund Hall analysis report
October 31, 2008. The data necessary was not readily written by Transsolar Inc. (Voit et al., 2007). The
available from a single source; therefore, data was overall mean U-factor for Gund Hall is 2.45 W/m2K,
acquired from three Boston & Cambridge area a high value for this climate by today’s standards,
weather stations, and then aggregated into an EPW due to large expanses of single-paned glazing and
format. Dry bulb temperature, dew point uninsulated fibreglass roof panels (Voit et al., 2007).
temperature, relative humidity, atmospheric station
pressure, wind speed, and wind direction were
acquired from data collected on the roof of the Green
Building on the Massachusetts Institute of
Technology’s (MIT’s) Cambridge campus.1
Radiation data was acquired from the University of
Massachusetts-Boston’s Center for Coastal
Environmental Sensing Networks (UMass-Boston).2
All of the other data for EPW1 was drawn from the
default Boston-Logan TMY3 file. For some
individual hours or days, MIT or UMass-Boston data
was unavailable; therefore, data was inserted from
the Boston-Logan TMY3 file. This likely causes
some inconsistencies and abrupt jumps in the weather
data. Figure 4 First floor of model with imported floor
The second EPW file, “EPW2,” included weather plan below.
data collected on top of Gund Hall from October 29 HVAC & Metered Energy Group
until December 4, 2008; the balance of the weather The HVAC and Metered Energy Group collected
file was the Boston – Logan TMY3 file. The Gund information about Gund Hall’s HVAC system design
Hall data was collected using a temporary HOBO3 and operation for input into the model. They also
weather station, shown in Figure 3. The weather gathered utility records for use in evaluating the
station collected wind speed, wind gust speed, wind model results.
direction, dry bulb temperature, relative humidity, Electricity is provided by Cambridge Electrical
dew point, and solar radiation in hourly time steps. Distribution. Building heat and domestic hot water
derives from high-pressure steam from dual-fuel
boilers at a central plant. Similarly, cooling for air
conditioning comes from chilled water, produced by
a central plant. All energy consumption is metered at
the building, rather than plant level.
The Gund Hall facilities manager provided an
overview of the Gund Hall HVAC system and the
operation schedules for the nine Air Handling Units
(AHUs). Due to limitations in the DesignBuilder
GUI at the onset of the project, purchased steam and
purchased chilled water were not available options
Figure 3 Gund Hall Weather Station for heating and cooling. Instead, the system was
modelled as fan coil units. In addition to the central
AHUs Gund Hall utilizes both radiant heat and VAV
1
Green Building location: 2.6 km from Gund Hall. systems with steam reheats. These systems were
Weather station hardware: Davis Vantage Pro 2 Software: excluded from the model, a known shortcoming that
VWS V12.08 will be discussed later.
2
UMass-Boston location: 9.4 km from Gund Hall. Weather The only spaces in Gund with operable windows are
station hardware: Davis Vantage Pro Plus. Software:
some of the offices. In these spaces, natural
unavailable.
3
HOBO weather station, Onset Computer Corporation,
ventilation was turned on in the model and set to be
Bourne, MA. www.onsetcomp.com Includes: weather automatically controlled by DesignBuilder based on
station starter kit, HOBO software, solar radiation sensor, occupancy schedule and indoor/outdoor temperature
light sensor level, and tripod kit. differences.
Annually, Gund Hall uses 160 kWh/m2 for heating, equipment schedule with the occupancy schedule.
180 kWh/m2 for cooling, and 146 kWh/m2 for Equipment was divided into those appliances which
electricity for a total of 486 kWh/m2. The national would be active as a percentage of the students’
average for a university building is 378 kWh/m2 and presence in the studio (i.e. lamps, coffee machines
361 kWh/m2 for a regional office building (regional etc.), and those appliances which would have a
university data is unavailable.) www.eia.doe.gov/ constant wattage regardless (i.e. refrigerators). For
emeu/cbecs/cbecs2003/detailed_tables_2003/detailed the “active consumption”, the quantity of each type
_tables_2003.html (CBECS, 2003). Therefore, Gund of appliance was converted into a ratio per student.
Hall uses approximately 32% more energy than The wattage per student was then multiplied by the
comparable buildings, presumably due in part to its occupancy fraction for each time slot and added to
poor envelope performance. the base wattage consumption.
Survey Group Due to the high wattage used by most of the
The objective of the Survey Group was to collect electronic equipment in the wood shop, the Survey
information on Gund Hall’s internal load schedules Group could not monitor their use over time with the
from building occupants, plug-loads and electric available watt meters. In order to sidestep this
lighting as well as window shade operation for input problem the band saw was metered over a week, and
into the Design Builder model. The information was its usage pattern identified. The same usage pattern
collected during October/November 2008 and was then assumed for the rest of the wood shop
estimated for the rest of the year. The Survey Group equipment.
separated the spaces of Gund Hall into twenty-three For equipment in the cafeteria kitchen, peak loads
categories, each category having similar use were assessed based on equipment labels or internet
characteristics. product searches. The average operating power for
The Survey Group conducted twenty walk-through that equipment was then estimated based on typical
observations of Gund Hall. In addition to these usage information provided by the cafeteria manager.
observations, an online questionnaire regarding For equipment that cycles on and off such as
occupant schedules and appliance usage was sent to refrigerators, annual energy consumption estimates
building occupants. Approximately 22% of 600 were obtained from the Energy Star Restaurant Guide
occupants responded to the questionnaire. The self- (U.S. EPA, 2007). Since the plug-loads in the offices
reported occupancy was higher in each time-slot than seemed fairly typical, the equipment power density,
the observed occupancy by an average of 5.4 W/m2, suggested by professional energy
approximately 20%. Since both methods of data consultants, Simpson Gumpertz & Heger was used
collection have their own limitations, the average (Waite, 2008).
occupancy schedule derived from the two methods The Survey Group also calculated lighting power
was used for the DesignBuilder model. densities for each of the 23 space categories based on
The survey data was collected over a four-week observation of each space, a list of lamp types
period during the normal academic session. provided by the facilities manager, and wattage
Therefore, to account for the summer term and information from the internet. The group then created
various holiday breaks, the Survey Group created four unique lighting schedules, based on information
schedules for 5 additional calendar periods based on from the facility’s manager, for spaces in which the
estimates provided by the facilities manager. For lighting operation is independent from the occupancy
example, during spring break, the occupancy of the schedules. In addition, four different window shade
studios was multiplied by 0.4 for each time slot. The schedules were created based on questionnaire
classroom occupancy schedules were created by responses, walk-though observations, and an
manually linking classroom-booking appointments interview with the facility manager.
from October to November 2008 with a class Simulations
enrolment list. A total of 19 different occupancy Finally, the work of the four groups was combined
schedules were generated for the model. into a DesignBuilder model. One set of three
The Survey Group also calculated plug-load densities simulations was run using each of the weather files:
for each space and created seven different plug-load the Boston – Logan Airport TMY3 file, “EPW1” the
schedules. As part of this effort, the group metered composite UMass/MIT file, and “EPW2” the Boston-
ten commonly used pieces of equipment in Gund Logan TMY3 file with one month of Gund Hall
Hall to determine their actual wattage using a weather station data inserted.
wattmeter4. Since equipment use is directly In another set of simulations, the natural ventilation
connected to occupancy, the researchers linked the and window shading were turned-off, one at a time,
to isolate the impact of these features on the
4 building’s energy consumption.
Watt meters used: watts up? Pro ES by Electronic
Education Devices, www.wattsupmeters.com and Kill A To investigate the energy impact of occupant
Watt EZ P4460 by P3 International Corporation, behaviour and the importance of surveying existing
www.p3international.com. conditions, another series of simulations was run.
This time several of the custom inputs were replaced
with default inputs from the DesignBuilder database.
First, the custom occupant densities and occupancy
schedules were replaced with default densities and
schedules. For each space, the most appropriate
DesignBuilder template was chosen. For example,
for the studio space, the DesignBuilder “University
Open Office Occupancy” was used. The cafe kitchen
became the default “University Food Prep
Occupancy” and so on.
Second, starting with the revised model described
above, the custom plug-load densities and schedules
were replaced with default DesignBuilder values in
Figure 5: Heating Load Comparing Weather Files
the same manner. Third, starting with this model, the
custom lighting schedules were similarly replaced. Natural Ventilation & Shading
In this model, the lighting densities were also The researchers ran simulations with and without
changed to default values by first setting each space both natural ventilation and window shading in the
to the IECC-2000 template standard of 3.4 W/m2 per model. The addition of natural ventilation caused a
100 lux and then changing the target illuminance in 7.3% increase in annual heating load, and a 2.6%
each space per the DesignBuilder templates. Finally, decrease in annual cooling load. The significant
the custom heating & cooling schedules were increase in heating load is surprising and warrants
replaced with default values for “University Open further investigation into the DesignBuilder
Office.” definition of the natural ventilation control set points
chosen.
RESULTS
The window shading had little impact on the heating
Weather Files and cooling loads, which is not surprising given that
As one might expect, the data in all three weather the shades are internal and present on less than half
files is similar except for wind speed, which is a very of the model’s glazing. The annual heating load
site-specific phenomenon. Comparing EPW1 (the increases by 0.3% and the annual cooling load
MIT/UMass-Boston file) and the default TMY3 file decreases by 0.14% with the addition of shades.
for the year reveals a daily mean difference in outside Electricity
dry-bulb temperature of 0.12 oC and 2.85 m/s in wind Figure 6 compares measured monthly electricity use
speed. Comparing EPW2 (the Gund file) and the for Gund Hall from August 2007 through July 2008
default TMY3 file, for the 37 days for which Gund to simulations using different combinations of
Hall data was recorded, reveals a daily mean custom and default internal loads. As one would
difference in outside dry-bulb temperature of 0.61 oC expect, the fully customized loads followed the
and 4.73 m/s in wind speed. metered electricity use more closely than simulations
The heating load results of annual simulations using based on default assumptions. For the fully
EPW 1 and EPW 2 are shown in Figure 5 along with customized loads the mean bias error5 (MBE) and
actual measured heating consumption. The root mean square errors6 (RMSE) were 0.2% and
November data in EPW 2 was obtained on-site. The 23% respectively compared to -17% and 64% for the
rest of EPW2 is an amalgamation of multiple years default loads. In order to better illustrate the
meant to represent a “typical year.” Therefore, one significance of the different internal load sources, the
would expect EPW 1 to produce results that are more middle dotted line in Figure 6 shows simulation
accurate in these 11 months. However, one can see results using default occupancy and equipment but
that, on a monthly or annual scale, both weather files customized electric lighting loads. For this case, the
produce similar accuracy. The same is true for results fell about halfway between the fully
cooling consumption. The remainder of the customized and all-default results, highlighting the
simulations discussed below used the EPW 1 weather significance of modelling internal loads for electric
file. lighting adequately. Since the electric loads do not
include heating and cooling, changing their schedules
had no effect on the electric loads.

1  n simulated  measured 

MBE   
5

n i1  measured 


simulated  measured
n 2

RMSE  
6
2
i1 measured



 DISCUSSION
Known and Suspected Shortcomings
While simulated and measured energy loads were
reasonably close for the Gund Hall model, the
authors do not rule out that hidden “lucky” mistakes
may be cancelling out each other. Some of the known
shortcomings of the model are:
 The building’s HVAC system was simplified,
therefore, the VAV system with steam reheats was
not modelled, meaning simultaneous heating and
cooling would be underestimated.
 It is suspected that the building systems cannot
Figure 6: Electricity - Measured vs. Simulations actually meet the peak cooling loads. Therefore,
the modelled cooling system, with its unlimited
Heating and Cooling capacity, exceeds the actual utility load in July.
The monthly metered and simulated heating and  In addition, manual heating setbacks during the
cooling loads are shown in Figures 7 and 8. The extended holiday and exam periods in late
students found abnormalities in the metered chilled
December and January were likely underestimated
water data from 2007/2008, so data from the previous
by the students in the model.
school-year are shown here.
 Finally, the chilled water meter was suspected of
From these graphs, one can see that the use of custom
malfunctioning and was replaced a few weeks prior
versus default settings, including HVAC operating
to this writing. Therefore, the authors are unsure of
schedules, again improved the simulation results,
the accuracy of the measured data shown in Fig. 8.
although the impact is not as consistent throughout
Weather Files
the year as with electricity. This is due in part to
variables cancelling out each other. For example, the Given the significant influence of weather conditions
default occupancy, plug-load and lighting loads on building performance, it is essential to use reliable
resulted in a 17% over estimation in annual heating climate data for energy modelling. Given that each
load, but then the addition of the default air handling of the three weather files tested produced similarly
unit schedule resulted in a 9% under-estimation. accurate simulation results, any of the three would
have been acceptable for this project. This is not
surprising given that all three files were collected
within a relatively small radius of several kilometres.
The different files had some significant discrepancies
in local wind conditions, but since the energy use of
Gund Hall is not very susceptible to wind, these
differences have little impact on the building’s
simulated energy use. The positive news for a
designer is that using a prepared climate file – by far
the easiest solution from a simulation standpoint –
does not seem to compromise significantly the
simulation accuracy. The caveat is that the climate
file must of course be representative of the particular
Figure 7: Heating – Measured vs. Simulations
building site.
In the absence of an already prepared local climate
file, the options for a design team are either to build a
climate file from scratch using local data from one or
several local sources (EPW1) or to collect one’s own
data (EPW2). Surprisingly, the latter option turned
out to be the more attractive one: The total cost for
the Gund Hall weather station is under $2500, the
equipment can be built up and run standalone at even
the remotest locations, and the measurements are
very close to the measured data from the MIT and
UMASS weather stations. During several months of
operation, the data logger produced a very reliable,
Figure 8: Cooling – Measured vs. Simulation synchronized data series that could be converted into
EPW format with little effort.
This finding strongly suggests that design teams
operating in locations for which climate data is not
available should collect their own weather data over largest error margin given that walk-through
at least several months in order to develop a more observations and occupant questionnaires lead to
accurate knowledge of the local climatic conditions different results. However, some occupancy analysis
of their building site. It should be noted, however, seems unavoidable for the creation of an accurate
that weather data might be of limited use if collected model of an existing building, especially when the
in an atypical year. plug-loads are so closely linked to occupancy. The
Conversely, the approach used for EPW1, combining number of site visits conducted in this experiment
multiple incomplete weather files, is not advised as it may be impractical for most project budgets, but
generated by far the most amount of work. certainly multiple site visits, and off-hours site visits,
Compiling data from several sources turned out to be as recommended by others (Waltz, 2000) seem
extremely time consuming and required a lot of necessary.
manual ‘cut and paste’ since data time steps were not It is interesting to note that the occupants consistently
always synchronized or constant and some time over-estimated the time spent at their desks, or
periods were missing altogether. perhaps the 22% of occupants who responded to the
Occupancy and Other Custom Inputs questionnaire tended to be an unrepresentative
sample. Waltz also reported that occupants tend to
Given the results shown in Figure 6, it seems that the
overestimate the amount of time they are spending at
detailed analysis of the building’s internal loads was
their workplace (Waltz, 2000). Therefore, although
worthwhile. The simulation with default occupancy,
more time-consuming for the researchers, walk-
plug-load, and lighting settings predicted an annual
through observations seemed to be a good
electrical consumption that was 18% lower than
supplement to, or replacement for, questionnaires.
metered data. The addition of custom lighting inputs
reduced this error to 12%, and custom occupancy and Little effort was invested in documenting window
plug-loads, reduced it further to 0.2%. This finding shade usage; however, the addition or subtraction of
underlines the benefits of carefully surveying a the shading in this simulation resulted in less than a
building during retrofitting projects. 1% change in annual energy consumption. One
should note that this number could be significantly
However, in the design phase of a project, the
larger for spaces with external shading. In Gund
modeller may have no additional information
Hall, the window shading is entirely internal, and it
available. In the Gund Hall project, deviation
exists on less than ½ of the glazing.
between “expected” and actual occupant behaviour
and plug-loads resulted in an additional 11.8% error In retrospect, the students’ detailed modelling
in electricity consumption. Yet an owner faced with strategy resulted in exponentially increasing
a nearly 12% delta between a design-phase energy complexity. Breaking the building into 23 different
simulation and the first year’s utility bills may be activity types resulted in exponentially more
tempted to blame the modeller’s ineptitude. schedules (occupancy, equipment, lighting, and
shading.) In addition, the 100+ model zones made
For buildings like Gund Hall, the analysis
inputs tedious and debugging difficult. Breaking the
methodology used in this project may be beneficial
schedules into short (two-hour) intervals made
but may not be feasible in projects outside of
adding calendar divisions time consuming, since they
academia with limited budgets. Therefore, the
were calculated as a percentage of the original
following discussion aims to pinpoint the most
occupancy. The added difficulty of trouble-shooting
effective analysis tasks employed on Gund Hall.
the model may outweigh the additional accuracy
First, the lighting was easy to document and
acquired through this level of detail. Therefore, the
decreased the error in electricity consumption by 6%
students would reduce the complexity of the model
of annual load, so a lighting analysis, both installed
next time with fewer zones, fewer schedules, and
density and operation schedule, seems advisable on
fewer time steps in the schedules.
every project. Inputting HVAC schedules is
similarly advisable (Waltz, 2000). The custom Energy Simulations for Architecture Students
HVAC schedules were easily obtained from the Following the experience of a semester long-course
facilities manager, and they significantly influenced on building energy simulation, four individual
the heating & cooling loads. modelling exercises and the group project of
Next, the detailed plug-load analysis was more modelling Gund Hall, the students were asked how
challenging. The three watt meters were a good comfortable they now felt with their modelling skills
investment, since they were a quick and easy way to and whether they would use the software again. As
gather accurate information. Given that smart watt novice energy modellers and non-engineers, they
meters that monitor energy use over an extended seemed to be reasonably satisfied with the simulation
period are becoming increasingly more affordable, results from the course project and there was a
using a higher number including watt meters for general expectation that with minor tweaking of the
larger equipment seems advisable. settings, the simulation results could be brought into
even better alignment with the measured data. Most
Finally, the detailed occupancy analysis was the most
students found that the exercise of modelling Gund
challenging piece and probably the one with the
Hall helped them understand the software better but models showing that there certainly remains a need
at the same time they felt that outside of academia, for modelling specialists, especially at the later
this type of benchmarking project would be better design stages.
left to simulation experts. A key lesson learned was that collecting one’s own
The students felt more comfortable using the weather data has become an affordable and easy-to-
software for smaller projects, where there is less implement option for design teams that leads to
room for modelling mistakes, and earlier in the reliable data sets. At the same time, it became
design process, when more basic decisions regarding apparent that this effort is only justifiable if no
programming and massing can be made based on the nearby climate file is available. Given the availability
software. The students also believed that using the of reliable low-cost weather station sets it actually
software to compare design decisions was a useful seems more effective for a design team to collect a
and interesting method for augmenting their building new climate file ‘from scratch’ than to assemble a
science curriculum. file from multiple local sources.
DesignBuilder/EnergyPlus was chosen for this course This experiment also attempted to quantify the
because, in comparison to most other building benefit of various building analysis tasks and custom
simulation software, DesignBuilder is a state-of-the- modelling inputs. The results are limited to one
art GUI. However, it still did not meet the building; therefore, the numbers cannot be readily
expectations of architecture students accustomed to extracted to another project, but rather offer insight
using sophisticated CAD and 3D modelling tools. In into the magnitude of the differences one might
particular, in DesignBuilder, the ability to select and expect between a non-standard building and the
organize model objects seemed limiting, and the illusive “typical building.” Collecting reliable
system of templates and attribute inheritance seemed internal load schedules is a very useful exercise for
both inflexible and unintuitive compared to other retrofitting projects. For new design projects, these
software for architects. Conversely, the students simulation assumptions should be carefully reviewed
needed to learn to abstract their models better. The with the building owner. Summing up, the students
students approached the geometries, schedules, and viewed the seminar and the course project as the
construction types with a minute level of detail much beginning rather than the culmination of their
more appropriate for an architectural model than a education in building systems and energy simulation.
building simulation. Nevertheless, in a survey, 10
out of 11 students said they would definitely ACKNOWLEDGEMENTS
recommend using DesignBuilder/EnergyPlus for The authors would like to thank the following
comparing architectural design decisions. 7 students and the teaching assistant for research
seminar GSD-6417 for all of their help with and
CONCLUSION dedication to this project: Diego Ibarra (TA), James
The paper documents the results from modelling a Kallaos, Anthony Kane, Cynthia Kwan, David
large educational building by simulation novices Lewis, Elli Lobach, Jeff Laboskey, Sydney Mainster,
using a state-of-the art graphical user interface. Rohit Manudhane, Natalie Pohlman, and Jennifer
Overall, it was found that over the course of a Sze. We further express our gratitude to the Harvard
semester design students are capable of learning how Graduate School of Design as well as the Real Estate
to set up a model of a larger complex building. As Academic Initiative at Harvard University for
suggest by others, students not only learned about supporting this effort.
energy simulation, but also learned about building
physics in the process (Schmid, 2008 and Batty & REFERENCES
Swann, 1997). Collecting their own weather data and Augenbroe, G. 2002. Trends in Building
carefully surveying the internal loads of a building Simulation, Georgia Institute of Technology,
helped the students to develop sensitivity for the College of Architecture, Atlanta, Georgia USA.
effect of these model inputs on simulation results.
Autodesk. last accessed in February 2009. Ecotect
Based on the students comments the authors believe
2009. http://ecotect.com/products/ecotect
that current state-of-the-art GUIs such as
DesignBuilder allow architectural students to build Batty, W. J., Swann, B. 1997. Integration of
meaningful energy models that can be used for initial Computer Based Modelling and an Inter-
design explorations. Learning how to set up an Disciplinary Based Approach to Building Design
energy model might further help architects to engage in Post-Graduate Education, Department of
in a more informed dialogue with their consultants. Applied Energy, Cranfield University,
At the same time, the students expressed their Bedfordshire, England.
discomfort with working on too complex building DesignBuilder version 1.9.0.003BETA. Last
accessed February 2009.
7 www.designbuildersoftware.com
The dissenting student believed the modelling and data
input process was too arduous to use realistically during the Energy Information Administration, Office of Energy
quickly evolving early design process. Markets and End Use. 2003. Commercial
 Buildings Energy Consumption Survey US Department of Energy (US-DOE). last accessed
(CBECS) – Forms EIA-871A,C,&E, February 2009. EnergyPlus Climate File
Washington, D.C. USA. Database.
Google. last accessed in February 2009. SketchUp http://apps1.eere.energy.gov/buildings/energyplu
Pro 7. http://sketchup.google.com/ s/cfm/weather_data.cfm
National Renewable Energy Laboratory (NREL) for U.S. Green Building Council (USGBC). last accesed
the U.S. Department of Energy. last accessed in in February 2009. LEED Rating Systems.
February 2009. OpenStudio Version 1.0.2 Build www.usgbc.org/leed/
37. United States Environmental Protection Agency.
http://apps1.eere.energy.gov/buildings/energyplu 2007. Putting Energy into Profits: Energy Star
s/openstudio.cfm Guide for Restaurants, 3, Washington, D.C.
Schmid A.L. 2008. The Introduction of Building USA.
Simulation into an Architectural Faculty: Voit, P., White, D., & Bummele, A. 2007. Gund Hall
Preliminary Findings, Departamento de – Analysis of Envelope and Thermal Comfort,
Arquitetura e Urbanismo, Universidade Federal Transsolar Inc., New York, New York USA.
do Parana, Curitiba Brazil. Waite, M. 2008. personal communication. Simpson
US Department of Energy (US-DOE). Last accessed Gumpertz & Heger, New York, New York USA.
February 2009. EnergyPlus Version 2.2.0.025. Waltz J.P. 2000. Computerized Building Energy
DLL default version embedded in Simulation Handbook, Fairmont Press, Lilburn,
DesignBuilder, Georgia USA
http://apps1.eere.energy.gov/buildings/energyplu
s/