Performance Comparison of U.K. Low-Energy
Cooling Systems by Energy Simulation
by
Erik L. Olsen
B.S., Mechanical Engineering
Purdue University, 2000
Submitted to the Department of Architecture
in partial fulfillment of the requirements for the degree of
Master of Science in Building Technology
at the
Massachusetts Institute of Technology
June 2002
@2002 Massachusetts Institute of Technology
All rights reserved
Signature of Author
'-7
Department of Architecture
May 10, 2002
.
-
Certified by
Qingyan (Yan) Chen
Associate Professor of Building Technology
L Thesis Supervisor
Accepted by
V'V
Stanford Anderson
on
Graduate Students
Chairman, Department Committee
MASSACHUSETTS INSTITUTE
OF TECHNOLOGY
JUN 2 4 2002
LIBRARIES
ROTCH
blbrries
Mff
Document Services
Room 14-0551
77 Massachusetts Avenue
Cambridge, MA 02139
Ph: 617.253.2800
Email: docs@mit.edu
http://libraries.mit.edu/docs
DISCLAIMER NOTICE
The accompanying media item for this thesis is available in the
MIT Libraries or Institute Archives.
Thank you.
2
Performance Comparison of U.K. Low-Energy
Cooling Systems by Energy Simulation
by
Erik L. Olsen
Submitted to the Department of Architecture
on May 10, 2002, in partial fulfillment of the requirements
for the degree of Master of Science in Building Technology
Abstract
Building energy simulation is an important tool for evaluating the energy consumption of
a building and can provide guidance in the design of a building and its mechanical
systems. EnergyPlus is a new energy simulation program meant to be a major advance
over existing energy simulation programs. This study uses EnergyPlus to compare
several alternative low-energy cooling systems for an office building in suburban London
and compare them to the chilled ceiling system installed in the actual building.
Prior to modeling the full-scale building, several validation studies demonstrate the
accuracy of EnergyPlus and the author's competency as an EnergyPlus user.
Systems considered include displacement ventilation, traditional mixing variable air
volume ventilation, night cooling, and natural ventilation. Several changes were made to
the EnergyPlus source code to model these systems appropriately. Most notable are a
displacement ventilation three-node vertical temperature gradient model and a simple
model for prediction of the natural ventilation rate.
A detailed building model is created from inputs gathered from both building design
documents and measured data. An excellent comparison between simulated and
measured space temperatures over a one-month period demonstrates the accuracy of the
model. System comparisons show that systems using free cooling from outside air and
night cooling use the least energy and have the smallest equipment. Natural ventilation
alone is insufficient to maintain summer comfort within the building, but could be used
within a hybrid ventilation system.
Conclusions are that EnergyPlus should be adopted for general use, because it represents
a major improvement over previous energy simulation programs and is capable of
modeling real-world buildings. A hybrid ventilation system would have the lowest
building system energy use, but displacement ventilation is also a good choice, and could
be implemented with few changes to the existing building.
Thesis Supervisor: Qingyan Chen
Title: Associate Professor of Building Technology
Acknowledgements
Many thanks to Yan Chen for his guidance, and for knowing when it was needed and
when it wasn't. Grateful thanks to Steven Wisby, David O'Sullivan, Ian Howe, and
everyone else at BP Sunbury for their assistance in gathering building design data and
instrumenting the building. Many thanks to Christine Walker and Roger Chang for their
contributions during the site visit. Thanks to John Zhai and Brent Griffith for
programming and CFD assistance and many thought-provoking conversations. Thank
you to the National Science Foundation for providing the support that made this work
possible, and to my parents and community for providing the foundation on which it
rests.
Finally, thank you to Erin for her support and trust in me.
Dedication
For our future children, may this be the first step in many to make your world a better
place.
Table of Contents
7
CH APTER 1: IN TR O DU CTION ....................................................................................
7
1.1 PROBLEM STATEMENT ............................................................................................
1.2 PROBLEM DETAILS..................................................................................................
1.3 OVERVIEW ...............................................................................................................
8
10
CHA PTER 2: BA CKG RO UN D ....................................................................................
11
2.1 LOw -ENERGY COOLING SYSTEM S ........................................................................
2.2 ENERGY SIMULATION ...........................................................................................
11
CH APTER 3: VALIDA TION .....................................................................................
18
24
3.1 INTRODUCTION .....................................................................................................
3.2 IEA COMMERCIAL BENCHMARK .............................................................................
3.3 IEA EMPIRICAL VALIDATION ..................................................................................
3.4 D ISPLACEMENT VENTILATION ..............................................................................
24
24
29
42
3.5 MIT TEST CHAMBER ............................................................................................
46
CHAPTER 4: ENERGYPLUS MODIFICATIONS....................................................
54
4.1 INTRODUCTION......................................................................................................
4.2 D ISPLACEMENT VENTILATION ..............................................................................
54
54
4.3 A IRFLOW .................................................................................................................
4.4 PLANT LOOPS...........................................................................................................
4.5 BASEBOARD HEATER ............................................................................................
4.6 A IR SYSTEM .............................................................................................................
4.7 N EW ENERGYPLUS CODE ........................................................................................
61
62
CHAPTER 5: BUILDING
...................................................................
5.1 BASIC BUILDING MODEL ......................................................................................
5.2 EXISTING BUILDING SYSTEM S ..............................................................................
5.3 A LTERNATIVE BUILDING SYSTEMS.......................................................................
5.4 ENERGY CONSUMPTION ........................................................................................
5.5 SIM ULATION CASES SUMMARY.............................................................................
CHAPTER 6: RESU LTS ...........................................................................................
6.1
6.2
6.3
6.4
EXISTING BUILDING VALIDATION.........................................................................
ANNUAL ENERGY CONSUMPTION ........................................................................
EQUIPMENT SIZING ................................................................................................
THERMAL COMFORT ..............................................................................................
6.5 HYBRID SYSTEM ....................................................................................................
6.6 D ISCUSSION ...........................................................................................................
65
65
66
67
67
77
86
91
92
94
94
97
103
106
108
109
CHAPTER 7: RECOMMENDATIONS AND CONCLUSIONS.............111
7.1 BUILDING M ODELING IN ENERGYPLUS..................................................................
7.2 BUILDING A LOw -ENERGY COOLING SYSTEM S ....................................................
I11
112
REFERE N CE S ..............................................................................................................
115
APPENDIX A: CHANGES TO ENERGYPLUS CODE .................
120
A. 1
A .2
A .3
A .4
A .5
D ISPLACEMENT V ENTILATION ..............................................................................
A IRFLOW ...............................................................................................................
PLANT LOOPS........................................................................................................
BASEBOARD H EATER ............................................................................................
A IR SYSTEM ..........................................................................................................
120
122
125
127
128
Chapter 1: Introduction
1.1 Problem Statement
The "green building" movement has gained momentum over the last decade as energy
costs have risen (Hawken et al. 1999) and awareness of global warming and other fossil
fuel related issues has increased worldwide. Although public energy policy debate is
often focused on the development of alternative, renewable energy resources, demandside reduction can also significantly reduce dependence on unsustainable and
environmentally harmful energy sources. Buildings use one-third of total U.S. energy
and two-thirds of U.S. electricity (Hawken et al. 1999) and similar amounts in other
developed countries, making them an obvious target for demand reduction.
The single largest energy consumer in a building is the heating, ventilating, and airconditioning (HVAC) system, and much of the effort in the design of a low-energy
building is focused on reducing the energy this system uses to provide a comfortable
indoor environment. Many low-energy HVAC systems have been successfully
developed and demonstrated in all types of climates; because cooling is required for the
interior spaces of most office buildings year-round, these systems are generally referred
to as low-energy cooling systems. These systems often provide additional benefits such
as improved indoor air quality, which may enhance worker productivity. Many factors,
such as initial costs, lifetime energy costs, maintenance costs, and occupant comfort and
well being affect the selection of a specific system for a building.
However, due to the complexity of the physical processes within a building, it can be
very difficult to predict a building system's annual energy costs and ability to maintain
occupant comfort. Many simplified models have been developed to aid in the design of
low-energy cooling systems. However, most models are only useful for a specific
system, and their simplified nature can give incorrect results.
Energy simulation programs, however, are intended to model all energy flows within a
building on a more detailed level and predict annual building energy usage and indoor
environmental conditions. Their detailed nature makes them more adaptable to modeling
nearly any building design. Although currently primarily used to evaluate a building
design when it is complete, energy simulation can be used as a powerful tool for both the
architectural and mechanical design of a building.
This study demonstrates the use of an energy simulation program to evaluate various
mechanical system designs for a U.K. office building. Many U.K. office buildings use
traditional all-air heating and cooling systems. However, the mild climate throughout the
U.K. makes many alternative, low-energy cooling systems both technologically and
economically feasible. In this study, EnergyPlus, an energy simulation program released
in April 2001 by the U.S. Department of Energy, is used to predict annual energy use,
size equipment, and evaluate occupant comfort for several different systems. These
results provide much more information than is often available to a designer and allow for
a more informed building system selection.
U
1.2 Problem Details
1.2.1 Building
Many low-energy buildings have been designed and built in the U.K. over the last
decade. The mild climate makes many low-energy cooling technologies feasible, and
high energy costs provide motivation for using these technologies. The building selected
for this study is Building A, located on the BP (British Petroleum) Sunbury campus in
suburban London. Figure 1.1 shows an overall view of Building A. This building was
selected because the HVAC system is very sophisticated and is meant to have low energy
consumption. It therefore provides a good opportunity for the evaluation of an actual,
installed low-energy cooling system and the comparison of this system to other potential
systems. In addition, partial sponsorship of this research by BP allowed access to the
information necessary to model the building.
Building A uses underfloor air displacement ventilation with chilled ceilings, perimeter
trench heaters, and perimeter chilled beams. Figure 1.2 shows a schematic of this system.
Fresh air is supplied through many small diffusers located in the floor, above a supply
plenum. The air is extracted through the ceiling light fixtures into a return plenum. The
ceiling is made up of chilled ceiling panels that provide additional cooling. Chilled
beams provide additional cooling near perimeter windows to prevent overheating due to
solar gains, particularly during summer months. During winter months, trench heaters
near the perimeter are used to overcome heat losses due to conduction through windows
and infiltration of outside air.
1.2.2 Alternative Systems
Although this system can potentially save energy over many conventional systems, it is
very complex and is expensive to install. It is therefore desirable to compare this system
to other, less complex low-energy systems that could be used in this building. Several
systems have been selected based on their appropriateness to the U.K. climate and to this
building. The systems were also selected such that they could actually be installed in the
building without major architectural changes. Although in an ideal situation the
ventilation, heating, and cooling strategies are taken into account in a building's
Figure 1.1 BP Sunbury Building A
Woar
heat
transmission
Figure 1.2 Building A ventilation, cooling, and heating scheme
architectural design, this is often not the case, and the mechanical design is performed
after the architectural design is complete. The goal of this study is to compare systems
that could be installed in the building with its basic design intact, such that the building
could even be retrofitted with one of these systems if desired.
The systems selected are: mixing variable-air volume (VAV), displacement ventilation,
natural ventilation, and night cooling. The VAV system is the all-air system most
commonly used in modem office buildings. The displacement ventilation system is
essentially the same as the existing building without chilled ceilings. Natural ventilation
draws unconditioned outside air through building openings such as windows without
mechanical assistance. Finally, night cooling, where the system operates at night to
precool the building, can be used with any of these systems. Further reasoning behind
the selection of these systems, and more detailed descriptions, is given in section 2.1.
1.2.3 Energy Simulation Program
Many energy simulation programs are available; each has its strengths and weaknesses.
EnergyPlus was selected for this study for many reasons. First, it is the newest and most
complete energy simulation program available; EnergyPlus includes many features that
have not been available in previous programs. Further discussion of the features of
EnergyPlus is given in section 2.2.5. In addition, the U.S. Department of Energy now
only supports EnergyPlus and all major energy simulation development in the U.S. is
being performed within EnergyPlus. Finally, EnergyPlus has been purposefully
developed so that it may be easily modified. This is a necessary feature because standard
energy simulation models must be changed slightly to model portions of the systems in
this study.
1.3 Overview
The details of this study are discussed in the following chapters. Chapter 2 provides
background and literature review for low-energy cooling systems and energy simulation.
Chapter 3 discusses several small EnergyPlus validation studies performed by the author.
Chapter 4 presents several changes made to the EnergyPlus source code in order to
correctly model the building systems. Chapter 5 discusses the modeling strategy for the
building as a whole and each mechanical system. Chapter 6 presents the results of the
energy simulations and compares the performance of the various systems, and Chapter 7
recommends the best system for Building A and gives conclusions on the use of
EnergyPlus for modeling low-energy cooling systems.
Chapter 2: Background
2.1 Low-Energy Cooling Systems
Several low-energy cooling systems have been selected for comparison in this study.
This section discusses the reasoning behind their selection, and then provides a
description of each system. The basic operation of each system, how it saves energy, and
what complexities the system introduces into energy simulation are presented.
2.1.1 Selection
The first step in the comparison process is selecting which systems should be compared.
The field of potential systems to choose from is enormous; case studies of buildings using
many different technologies are given in Zimmermann and Andersson (1998). Table 2.1
lists some common low-energy cooling technologies. Some method must be used to
reduce this field to a manageable number of systems for comparison. The systems
considered here were selected for their suitability to the U.K. climate, to being modeled
in EnergyPlus, and to being installed in Building A.
Some systems can easily be eliminated due to their lack of suitability for the location
being studied. Evaporative cooling, for example, is. clearly inappropriate for the humid,
maritime U.K. climate, and sea/lake/river water cooling is not practical for Building A
due to the lack of a nearby body of water. Guides such as the International Energy
Agency "Selection Guidance for Low Energy Cooling Technologies" (IEA 1997) have
been developed to aid in the selection of suitable systems based on local conditions. This
guide provides brief summary sheets of several low-energy cooling technologies and a
selection sheet used to rate their suitability for a given site. Suitability considerations
alone, however, leave numerous systems that show potential for this site, as shown in
Table 2.1.
Table 2.1 Low-energy cooling technology selection chart
Suitable for
System Climate/Site
/
Night cooling
Natural ventilation
Slab cooling
Evaporative cooling
Dessicant cooling
Chilled ceilings
Chilled beams
Displacement ventilation
Ground cooling
Aquifer
Sea/river/lake water cooling
V
Suitable for
Building
V
The systems considered should also be appropriate to the building being modeled without
requiring major architectural changes. Although in an ideal situation the ventilation,
heating, and cooling strategies are taken into account in a building's architectural design,
the purpose of this study is to compare the performance of systems without making major
architectural changes, such that the building could actually be retrofitted with one of the
selected systems. This criteria eliminates slab cooling, which requires water or air
channels to be incorporated into the structural concrete slab, and ground cooling, which
requires water or air channels in the ground beneath the building. Finally, desiccant
cooling, although technically applicable to this building and site, was eliminated because
it is most effective when a waste heat source is available, which is not the case for this
building.
The remaining technologies were combined into five basic systems for comparison. The
actual building system, chilled ceilings with constant volume underfloor air ventilation
and perimeter trench heaters and chilled beams, was obviously selected. Underfloor air
displacement ventilation was selected because this is essentially the same as the existing
building without the chilled ceilings. A variable-air-volume mixing ventilation system
was selected to act as a baseline case. Natural ventilation was selected because the
U.K.'s extremely mild climate and the building's large glazed area make this approach
potentially feasible. Finally, any of these systems can be modeled with night ventilation.
Detailed descriptions and associated issue for each of these systems will now be
presented.
2.1.2 VAV Mixing Ventilation
The variable air volume mixing ventilation system, henceforth known as VAV, serves as
the baseline case because it is the most common system in modem office buildings. Air
is introduced at or near the ceiling level at relatively high velocities. The resulting jet is
intended to mix fully with the room air to maintain the desired space temperature. Air at
room temperature is exhausted near the floor from a sidewall. In the cooling mode, air is
supplied at a constant temperature, generally around 15*C, and the air flowrate is varied
to account for the variable space cooling load, hence the name variable air volume. In the
heating mode, the air flowrate is maintained at a fixed minimum and is heated to varying
temperatures by a terminal reheat unit to account for the variable space heating load.
2.1.3 Displacement Ventilation
A displacement ventilation system supplies air at or near floor level at very low velocities
(less than 0.5 m/s) and temperatures slightly below room temperature, typically 18*C.
The air spreads across the floor and rises as it is heated by sources such as people and
computers, creating a vertical temperature gradient, as shown in Figure 2.1. Exhausts are
located at or near the ceiling. The heat sources create thermal plumes that increase in
volume as they rise due to entrainment of ambient air. At the height where the plume
airflow rate equals the supply airflow rate, a stationary front exists, creating two zones
within the room. The lower zone has little recirculation flow (hence the term
displacement ventilation) while the upper zone has recirculation (Yuan et. al. 1998).
Figure 2.1 Displacement ventilation schematic
In a correctly designed system, the occupied space is entirely within the lower zone and
the vertical temperature gradient in the occupied zone is small enough to maintain
thermal comfort; ankle to head temperature differences less than 3C are generally
considered acceptable. Yuan et al. (1999a) provide design guidelines for displacement
ventilation. Most office buildings require cooling in core spaces year round; an auxiliary
heat source must be provided for perimeter spaces that require heating.
This system has two primary advantages. The first is improved indoor air quality
because 100% of the supply air reaches people and other contaminant sources, whereas in
mixing ventilation a portion of the supply air remains in the upper portion of the room
without reaching the majority of contaminant sources. The second is potentially reduced
energy consumption due to the vertical temperature gradient within the space. Only the
occupied zone must be maintained at the room setpoint temperature, while the remainder
of the space may be warmer. Hence, the temperature difference between supply and
exhaust air can be larger than for mixing ventilation. The cooling load, given by:
qi= i
cp, (Texhaust
-
T
(2.1)
can therefore be achieved with a lower airflow rate for displacement ventilation than
mixing ventilation, resulting in reduced fan energy. Additionally, the higher supply air
temperature can result in decreased chiller energy.
Several researchers have demonstrated energy savings with displacement ventilation (Hu
et. al. 1999; Chen et al. 1990). However, the complexity added by the non-uniform room
air temperature distribution makes displacement ventilation systems more difficult to
model and design than mixing ventilation systems. Yuan et al. (1998) provides a review
of many issues and models associated with displacement ventilation.
For energy simulation, it is essential that the temperature gradient be modeled correctly.
Energy simulation programs generally assume that the room air is well mixed and at a
uniform temperature. However, this is not the case for many actual systems, particularly
displacement ventilation, where the system performance is greatly influenced by the
vertical temperature gradient. Nodal models of varying complexity have been
demonstrated for modeling the temperature gradient in an energy simulation (Rees
1995;Van der Kooi and Bedeke 1983). Yuan et al. (1999b) review the development of
simplified models for the room temperature gradient. This study incorporates two of
these models into EnergyPlus in order to simulate displacement ventilation.
Although most previous studies focus on displacement ventilation where air is delivered
via large, floor level diffusers (as in Fig. 2.1), underfloor air supply (as in Fig. 1.2) can
also be thought of as a form of displacement ventilation. Here, the air is introduced to a
plenum beneath the floor of the occupied space, and the air then reaches the space
through numerous small diffusers located throughout the floor. Given low enough air
velocities, this system can be considered reasonably similar to displacement ventilation.
Because the building being considered has a raised floor with a supply plenum, this is the
system considered in this study.
Raised floor systems are advantageous because they reduce maintenance costs and
provide for easy reconfiguration of office space. If desks are rearranged, the raised floor
tiles can also be arranged to provide the best diffuser arrangement for the new
configuration. Ducts and pipes can be located in the plenum and are easily accessed for
maintenance simply by removing floor tiles.
2.1.4 Chilled Ceilings (with Displacement Ventilation)
In a chilled ceiling system, chilled water is circulated through tubes bonded to a thin
metal panel that forms some portion of the ceiling of the space. The panel absorbs
radiant heat directly from the occupants and other heat sources, and it absorbs heat
convectively from the room air. Ventilation is still necessary in order to provide fresh
air, and it can be provided by either a mixing or displacement ventilation system. Figure
2.2 shows a schematic of chilled ceilings with displacement ventilation.
Figure 2.2 Chilled ceilings with displacement ventilation
With either form of ventilation, the introduction of chilled ceilings introduces new
complexities which must be considered in the design and modeling of a system. The
radiant asymmetry created by the panels can be uncomfortable if too large. Humidity
control is essential in a space with chilled ceilings, because if the ceiling temperature
were to fall below the dewpoint temperature, condensation would occur on the ceiling
surface. For both of these reasons, ceiling temperatures are generally between 15 and
18'C. Feustel and Stetiu (1995) provide a review of issues associated with chilled
ceilings. Conroy and Mumma (2001) present a design methodology for chilled ceilings.
Because the volumetric heat capacity of water is 4000 times greater than that of air, it is
much more efficient to remove heat from a space using water instead of air because less
fluid must moved around the building; this is the primary reason chilled ceilings save
energy. Chilled ceilings are often used in conjunction with displacement ventilation
because displacement ventilation can only handle cooling loads up to 40 W/m 2 before
using excessively high air flowrates. A displacement ventilation system with chilled
ceilings can handle cooling loads greater than 100 W/m 2 .
The energy savings from a chilled ceiling system depend greatly on the cooling load and
how the system is installed. Niu et al. (1995) found that for the Dutch climate, which is
similar to the U.K. climate, the energy use of a chilled ceiling system is similar to that of
a VAV system, but that chilled ceilings would have much better performance in hot,
humid climates. In addition, the chilled ceiling has considerably better performance
when the water for the ceilings is directly chilled via cooling towers, which provide free
evaporative cooling. Novoselac and Srebric (2002) also found that a chilled ceiling
system becomes more efficient than a VAV system with increasing peak cooling loads.
Although the chilled ceiling system reduces fan energy, it uses more cooling energy than
a VAV system because the season for free cooling from outdoor air is longer for a VAV
system than a chilled ceiling system.
Chilled ceilings tend to counteract the vertical temperature gradient found with
displacement ventilation, because the air is cooled near the ceiling and flows downward
due to buoyancy, against the primary airflow direction. The extent of this effect is a
function of the amount of cooling load removed by displacement ventilation, making this
a key design parameter (Novoselac and Srebric 2002). A small displacement ventilation
load reduces the vertical temperature gradient, improving thermal comfort, but reduces
indoor air quality because of the increased mixing of room air. If the chilled ceiling load
is large enough, the displacement ventilation flow pattern might disappear completely
and the system characteristics are similar to mixing ventilation (Behne 1999).
Chilled ceilings introduce new complexities in energy simulation because they are
essentially room surfaces with embedded heat sinks. Both the transient conduction
characteristics of the ceiling panel and its convective and radiant effects on the room
must be modeled correctly. Strand and Pedersen (1997) and Niu (1994) discuss the
implementation of chilled ceiling models into energy simulation programs.
2.1.5 Natural Ventilation
Natural ventilation describes any system where unconditioned outside is drawn through
the building without mechanical assistance, usually through open windows. If the airflow
rate is large enough, the indoor air temperature can be maintained equal to the outdoor air
temperature, or even lower than the outdoor air temperature if appropriate thermal
storage techniques are used. Natural daytime ventilation is generally used in conjunction
with natural night ventilation in order to allow for such thermal storage. Slightly higher
air temperatures are also often acceptable in naturally ventilated buildings, because the
increased air velocities and individual control over opening of windows allow occupant
comfort at higher temperatures than with a mechanically ventilated building (Schiller
2000).
There are two primary forms of natural ventilation: wind-induced and stack-induced.
Wind-induced natural ventilation is caused by pressure differences between outside and
inside due to wind. Generally, the windward side of a building is under higher pressure
while the leeward side is under lower pressure. This causes a flow of outside air through
the building, especially for cross-ventilation, where open windows are available on both
sides of the building with an unrestricted airflow path between.
Stack-induced natural ventilation is caused by pressure differences created by the
buoyancy effect within a vertical space. The pressure at the bottom of the space is lower
than the outdoor pressure, while the pressure at the top of the space is higher than the
outdoor pressure. Hence, there is a flow of outside air from the bottom to the top of the
space.
In most buildings natural ventilation is cause by both wind and stack effects. A special
form of natural ventilation that generally relies on both stack and wind effects is singlesided ventilation, where a room only has openings on one side. Naturally ventilated
buildings must be carefully designed in order to provide sufficiently large ventilation
rates. Allard et al. (1998) discusses numerous design strategies for natural ventilation;
Alloca (2001) provides design guidelines for single-sided ventilation.
Accurate prediction of the ventilation rate is essential for the evaluation of a natural
ventilation scheme with energy simulation. Ventilation rates can be predicted using
correlations developed by experiment (Heiselberg et al. 2001) or by computational fluid
dynamics (CFD) simulation (Alloca 2001). More detailed predictions can be made by
coupling CFD models or multi-zone network airflow models such as COMIS directly to
the energy simulation. Several researchers have demonstrated the coupling of energy
simulation to both network airflow models (Geros et al. 1999; Dorer and Weber 1999;
Huang et al. 1999) and CFD models (Carrilho da Graga 2001).
An important element of predicting the ventilation rate is the modeling of the airflow
around a building. Although this is most accurately done with wind tunnel experiments,
such experiments are expensive and time-consuming, so CFD is often used to predict the
airflow patterns around a building. However, there are several issues associated with the
use of CFD to predict outdoor airflow patterns: the k-s turbulence model commonly used
has been shown to have limitations (Murakami et al. 1990), and the modeling of
buildings with very porous facades as solid objects can be inadequate (Straaten 1967).
Large eddy simulation can be used to overcome many of the limitations of the k-c model,
but requires lengthy computing times. Note that this study attempts only a first-cut
model of natural ventilation, and therefore uses the k-s model to model the building and
its surroundings as solid objects.
Natural ventilation need not always be used in isolation; if natural ventilation alone
cannot maintain comfort conditions throughout the year, a supplemental mechanical
system can be used during extreme weather conditions. The combination of natural and
mechanical ventilation is known as hybrid ventilation, and can be extremely useful in
locations where natural ventilation cannot always maintain comfort conditions. For
example, Levermore et al. (2000) found the a clinic building in the U.K. with single-sided
ventilation could maintain comfort conditions with pure natural ventilation in the north of
the country, but that mechanical assistance would be necessary during peak periods in the
area near London.
2.1.6 Night Ventilation
Night ventilation can be used with any of the systems discussed above. Natural or
mechanical ventilation is used to precool the mass of a building using cool night air. The
mass then absorbs heat throughout the day, reducing the amount of cooling which must
be provided by other means. For night ventilation to be effective, a significant amount of
mass, generally the structural concrete slab, must be exposed, and the diurnal temperature
variation must be sufficiently large.
Many factors affect the performance of night ventilation, such as the thermal properties
of the mass, its thickness and insulation level, and the ventilation rate. These factors
must be considered carefully when modeling and designing buildings with night
ventilation. If the thermal mass is too small, it will heat too quickly and the building will
overheat during the day. An excessively large mass, however, may have too large of a
time lag, resulting in a large heat release in the early morning hours just before the
building is occupied. Balaras (1996) reviews the factors affecting night ventilation and a
number of simplified models and design tools which have been developed for night
ventilation. Kolokotroni et al. (1998) developed a simple pre-design tool for night
ventilation.
Night ventilation has been demonstrated as an effective means of saving energy in the
U.K because of the region's relatively low peak summer air temperatures and medium to
large diurnal temperature swing. Kolokotroni (2001) found that for U.K. buildings with
daytime mechanical ventilation, mechanical night ventilation may not save energy
because of the extra energy necessary to run the fans at night. However, if natural
ventilation is used at night, annual cooling energy savings as large as 40% are possible
for a well-designed building. Braham (2000) showed that natural ventilation with
surface-cooled or core-cooled slabs and mechanical ventilation through hollow-core slabs
all provide energy savings.
Because energy simulation already accounts for the thermal mass of buildings surfaces, it
does not require any special modifications in order to simulate most night ventilation
cases other than making sure the night ventilation is appropriately scheduled. Several
researches have demonstrated energy simulation of both mechanical and natural night
ventilation (Ren and Dalenback 1995, Geros et al. 1999). Of course, some systems that
rely heavily on night ventilation, such as slab cooling, may still require special
modifications to an energy simulation program, but the modifications are necessary due
to the nature of the system and not the fact that it is running at night.
2.2 Energy Simulation
This section discusses the state of the art of energy simulation. The history of energy
simulation methodologies and their basic theory is reviewed, followed by a discussion of
validation techniques. Finally, the major elements of EnergyPlus are presented,
especially with reference to how they represent an improvement over previous energy
simulation programs.
2.2.1 History and Theory
Numerous methods have been used to size equipment and predict energy consumption of
HVAC systems. Strategies include very simple manual methods such as the degree-day
and bin methods (ASHRAE 2001), regression methods, and detailed hour-by-hour
computer simulation methods. The term energy simulation is generally reserved for true
simulation methods that use time as the independent variable rather than outdoor
temperature (Sowell and Hittle 1995).
Sowell and Hittle (1995) provide a thorough review of the evolution of building energy
simulation, which is briefly summarized here. The primary similarity between most
energy simulation programs is the sequence of simulation: load, system, plant, and
economics. These elements are generally simulated sequentially on an hourly basis for
an entire year.
Two primary approaches are used for load calculation: weighting factor and heat balance
methods. The weighting factor technique divides the instantaneous heat gain from
various sources over a time series, because some portion of the heat emitted from a
source is not immediately convected into the room air but radiates to room surfaces, from
which it is released over time. The first weighting factor program was developed in 1967
(GATC 1967); DOE-2, the most common commercially applied energy simulation
program in the U.S. today, still uses the weighting factor method.
The heat balance method uses more direct physical models by calculating instantaneous
convective gains for each time step and enforcing a heat balance on the air and each
surface. The first heat balance program was developed in 1978 (Kusuda. 1978). Two
programs currently using the heat balance method are BLAST (Hittle 1979) and ESP-r
(Clarke 1991). Both the weighting factor and heat balance method generally use
conduction transfer functions to calculate transient conduction through building surfaces.
Two primary approaches are also taken to simulating the secondary systems, consisting
of components such as coils, fans, and mixing boxes. Most programs, such as DOE-2,
provided the user with a fixed set of secondary systems. Other programs are componentbased, allowing users to build up custom systems from individual components. TRNSYS
(Klein et al. 1994) is the most popular component-based program. Component-based
simulations generally require more input effort and greater simulation times, but they
allow for complete studies of many specialized systems.
Plant models are generally simpler than system or load models. Manufacturer's data is
often used to generate curve fits for fan, chiller, and pump performance, and a simple
constant efficiency might be used for a boiler. Unit energy cost data may then be used to
estimate actual energy costs.
The sequential solution of loads, system, and plant clearly has its flaws, because these are
not linked only one direction. If the equipment is undersized, the space temperature may
actually "float" above the actual setpoint temperature, a condition that is not allowed in a
sequential solution. Simultaneous solution of the space, system, and plant introduces
difficulties with simulation stability, system control, and calculation of conduction
transfer functions on the smaller time steps used in this simulation method. Taylor et al.
(1991) demonstrated an early simultaneous solution method incorporated into BLAST,
which eventually became IBLAST, the predecessor to EnergyPlus.
An important consideration with energy simulation, as with any computer model, is the
use of correct inputs. No matter how well written and physically accurate a program is, it
will not give useful results unless the building is modeled correctly. The user should be
careful to use the best data available for the building being modeled. Mottillo (2001)
showed that the results of an energy simulation can be quite sensitive to some inputs,
although the relative importance of inputs certainly varies depending on building type.
Besides the building model, the annual weather data can significantly affect the results of
an energy simulation. Crawley (1998) concluded that typical weather year data, such as
TMY2 or WYEC2 data, is most appropriate for energy simulation of commercial
buildings.
2.2.2 Validation
An important part of the development of an energy simulation program is validation.
Several validation studies were performed as the first part of this work. Validation is
necessary to demonstrate both the accuracy of a program and the ability of a user to
correctly model buildings using a program. Bloomfield (1999) reviews various
validation techniques and projects. There are three primary means of validation:
analytical tests, inter-model comparisons, and empirical validation.
Analytical tests compare simulation results to analytical solutions for identical situations,
and are therefore only appropriate for very simple cases or subroutines of energy
simulation programs for which analytical solutions exists. These tests are generally
carried out by program developers.
Inter-model comparisons compare the results of various energy simulation programs.
Although the truth standard of such a comparison is clearly weak, it is still a valuable tool
for identifying major problems with a program; if the results of a program lie far out of
the range of results from other programs, it probably has an error. However, for a given
range of results from various programs, a program yielding results near the middle of the
range cannot be considered any better than programs yielding results at the edges of the
range (Judkoff and Neymark 1999). This is because of the lack of an absolute truth
standard; results near the edge of the range are equally likely to be closest to the actual
solution as results at the center of the range. The BESTEST benchmark tests are the most
commonly used set of inter-model comparisons in the U.S. (Judkoff and Neymark 1994).
Empirical validation compares simulation results to measured data for an actual building
or test cell. A carefully performed experiment provides an excellent truth standard for
comparison. However, much effort and skill is required to perform a good experiment:
the experimental setup must be sufficiently documented in order to accurately model the
building, and measured data must be reasonably accurate. Although all measurements
have some error, the amount of error must be characterized and should be smaller than
the amount of error expected from the simulation. Such experiments are difficult even
for small test cells and are essentially impossible for an actual building. Perhaps the best
publicly available dataset for empirical validation is the IEA Empirical Validation study
(Lomas et al. 1994).
2.2.3 EnergyPlus
Development of EnergyPlus began in 1995 as an effort to combine the best features of
BLAST and DOE-2 and incorporate some other new features(Strand et al. 2000). These
programs were believed to have reached maturity and their extremely old "spaghetti"
code made them increasingly difficult to modify. EnergyPlus solves this problem with a
modular program structure written in Fortran 90, which makes the code easy to
understand and modify. In addition, EnergyPlus was developed as only a simulation
engine, with no interface. Commercial developers can then develop proprietary
interfaces for different markets and users, using the EnergyPlus simulation engine.
The major improvement in EnergyPlus over previous energy simulation programs is the
integrated solution of loads, system, and plant, allowing accurate space temperature
predictions. The solution is based on the heat balance technique originally employed in
IBLAST and is referred to as the Predictor-Corrector Method (EnergyPlus 2001).
EnergyPlus adopts the standard energy simulation assumption that room air is well
stirred, providing a uniform temperature. The air heat balance is then based on the
equation:
dT
dt
dt
where:
h A (Ti -TZ)+
Q +
i
N.
Naces
Nsi
-
i=
Eflic(Ti -T)+
i1
rifc(Tif
T)+
sys
(2.2)
N~j
=
sum of internal convective loads from people, computers, etc.
i=1
Nsurfaces
hi A (T,5 - T,)
convective heat transfer from zone surfaces
ihic,(T,; - T,)
i=1
ri1
c p (Tif - Tz) = heat transfer due to infiltration
Q,,,s=
C
heat transfer due to interzone air mixing
=
iiyc ,(T, -Ti)
dT
2
dt
=
=
air system output
rate of energy storage in air
The zone temperature derivative is calculated with a third order finite difference
approximation:
dT z
~(St)~
dt
11
6
T
t
3I
3
+-T28
2Tt _1
-T
3Z
2z
t
-3
~
(2.3)
where:
Tz = zone mean air temperature
t = current time
St = system timestep
Equation 2.3 can be substituted in Eq. 2.2 and solved for the mean air temperature:
N., .,
EQ+
N~wf..
Nz_
hA Ti+
ic P T 5+ i fc T
6
8t
+r il
+ ZhA +
=i
c T
,,
-
J
8t
3 '128
S
3T t~ +- T.* - T ~.
c
+ rilc, + ri YSiic
i=I
This is the most fundamental equation in EnergyPlus. The introduction of the zone
capacitance term, Cz (dTz/dt), allows the zone air temperature to vary. In previous heat
balance programs, the left side of the heat balance equation (2.2) would be zero.
Two timesteps are used. The first is a fixed timestep input by the user: the simulation
timestep (default is 15 minutes), used for the surface heat balance, which has a relatively
stable time constant on the order of an hour. A variable timestep is used for the air heat
balance, which has a time constant dependent on the system load and can have an order
as small as one minute. At the beginning of each simulation timestep, the system
timestep is set equal to the simulation timestep; the system timestep is then reduced until
it is sufficiently small.
Using these two timesteps and Eq. 2.4, the predictor corrector technique then flows as
follows:
1) The required system load (Qys) necessary to maintain the setpoint temperature is
predicted by solving (2.4) for the system load with Tz equal to the setpoint
temperature
2) The system and plant are simulated using this Qys as the demand to determine the
actual system capacity.
3) This actual capacity is used in (2.4) to calculate the actual zone temperature.
4) The change in zone temperature from the previous timestep is calculated. If it is
greater than 0.3 K, the entire procedure is repeated with the system timestep
halved.
The iteration continues until the zone temperature change is small enough or the system
timestep is less than one minute. 0.3 K (1% of 300 K) is used as the criteria for timestep
halving to prevent the air temperature from changing too quickly; if the change is less
than 1 K, the system timestep is sufficiently smaller than the room air time constant. The
entire method is shown in flowchart form in Figure 2.3.
Other novel load modeling aspects of EnergyPlus are discussed thoroughly in the
program documentation, including a moisture balance, conduction calculations, and
window calculations (EnergyPlus 2001). Moisture balance calculations include optional
surface.mass transfer calculations. Surface conduction is calculated using conduction
transfer functions, but a new interpolation scheme has been used for calculating the
interzonal mixing and
surface T data for
current zone timestep
substitute setpoint temperature for
Tz in (1.4), calculate predicted
system load (predictor step)
.---
reet with2
neSt=t/
calculate actual system
supply capacity
use calculated system parameters in
(1.4) to calculate actual zone
temperature (corrector step)
ATz <0.3C
n
ye
move to next system timestep
Figure 2.2 Predictor-corrector method flowchart
temperature and heat flux histories. This prevents the storage of very large history arrays
when a shortened time step is used. Window calculations are based on the WINDOW 5
program (Winkelmann 2001).
The system solution method is iterative, unlike the single-pass methods used in DOE-2
and BLAST (Strand et al. 2000). The system input is component-based, allowing users
to create many types of systems. State variables are stored in nodes connecting various
air loop equipment (fans, coils, etc.) and zone equipment (reheat coils, VAV boxes,
plenums, etc.), and an iterative solution technique is used to solve for the state variables
and their associated controls. Considerable effort has been spent on implementing
appropriate system controls, although it should be noted that components are controlled
using the "known" zone load, rather than attempting to simulate actual control schemes
such as proportional-integral control.
Plant equipment is also input and simulated with the component- and node-based
approach. Curved-based models are currently used, but the modular code structure
makes it easy to implement other models as necessary (Strand et al. 2000).
Before EnergyPlus can be applied for practical simulations, it must be appropriately
validated. Witte et-al. (2001) have performed a variety of analytical and inter-model
comparative tests and found good agreement with published results. In addition, the
author has performed an additional set of validation tests in order to demonstrate both the
validity of EnergyPlus and the author's ability to model buildings correctly with
EnergyPlus.
Chapter 3: Validation
3.1 Introduction
Because of its recent release in April 2001, EnergyPlus is not yet in widespread use and
requires some experimental validation before it can confidently be used for building
energy simulation. The difficulty in obtaining complete whole-building data necessary to
experimentally validate an energy simulation required this validation to be undertaken
primarily as a series of four smaller, independent studies, which when taken as a whole
should provide reasonable confidence in a whole-building simulation. In addition,
limited comparisons between simulation results and measured data for Building A have
been performed; these are presented in Chapter 6.
Three of the validation studies involved comparison of EnergyPlus predictions to
previously published experimental data. The first is an inter-model comparison using
simulation results from six other energy simulation programs reported in the International
Energy Agency (IEA) Commercial Benchmarks study (Haapala et al. 1995). This study
served as a simple first step for evaluating general performance, especially annual heating
and cooling load predictions. The second study is an empirical validation using the
extensive data provided in the IEA Empirical Validation Package (Lomas et al. 1994).
This study provided a thorough validation of EnergyPlus models for building fabric heat
gains and losses. The third study is also an empirical validation, using transient cooling
load data for displacement ventilation (Chen 1988). This allowed validation of a
modification of EnergyPlus incorporating the use of CFD-derived models, and also
demonstrated the feasibility of this technique.
The fourth study compares predicted ventilation system performance to data specially
collected for this validation at the MIT Test Chamber facility. When taken together,
these studies validate the two essential portions of building energy simulation: fabric
gains and losses and ventilation system performance, and should provide confidence in
any simulation performed by a knowledgeable EnergyPlus user. The validation also
demonstrates that the author is a competent EnergyPlus user. The methods and results of
each study will now be presented individually.
3.2 IEA Commercial Benchmark
3.2.1 Case Description
The IEA Commercial Benchmark study serves an initial rough validation of EnergyPlus
by comparing the results of EnergyPlus simulations to the results of other energy
simulation programs. The study considers a module consisting of two identical office
rooms and a connecting corridor, as shown in Figure 3.1. This module is surrounded on
the four sides without windows by identical modules. Although the corridor is a
continuous space in a real building, the ends of the corridor within the module were
modeled as internal walls, as they were in the benchmark simulations. There is no
mixing of air between the zones. Table 3.1 shows thermal properties of the wall
Table 3.1 IEA benchmark wall properties (outside to inside)
Floor
Concrete Slab
Carpet
Ceiling
Carpet
Concrete Slab
Heavyweight
External Wall
Brick .
Foam Insulation
Concrete Block
Lightweight
Internal Wall
Plasterboard
layer
thermal
thickness
conductivity
m
W/mK
0.160
0.004
density
specific heat
thermal
km
J/kgK
resistance
m2K/W
1.130
0.300
1400
1600
1000
1380
-
0.004
0.160
0.300
1.130
1600
1400
1380
1000
-
0.102
0.061
0.100
0.950
0.040
0.510
1920
10
1400
920
1400
1000
-
0.010
0.160
950
840
-
Cavity
Plasterboard
0.050
0.010
0.160
950
840
0.180
-
Door
Chipboard
0.012
0.130
600
1380
-
Air Gap
Chipboard
0.020
0.012
0.130
600
1380
0.160
-
Interior wall for calculation
Room 2
(north)
corridor
Room 1
(south)
4n
4m
4.m
Figure 3.1 lEA Commercial Benchmark module
materials. All surfaces have a thermal emissivity of 0.9; solar absorptivity is 0.3 for
interior surfaces and 0.7 for exterior surfaces.
This study considered only one test case from the benchmark study, Case 3a, in which the
axis of the module is oriented north-south, the module has no exterior shading, and the
corridor is unheated. The other benchmark cases include several permutations of these
features and their opposites; the module can be oriented east-west, have exterior shading,
and the corridor can be heated. For this case, room 2 faces north and room 1 faces south.
The ventilation scheme is somewhat impractical; each space is ventilated by 100%
outside air at 3.0 ach from 07:00 to 17:00 and 0.5 ach from 17:00 and 07:00. Both rooms
have an internal 500 W load from 08:00 to 16:00, which is 50% convective and 50%
radiative. Both rooms are heated to a lower setpoint of 20*C and cooled to an upper
setpoint of 25*C from 07:00 to 17:00, and heated to 18*C from 17:00 to 07:00, with no
cooling during these hours.
In the benchmark study, this case was modeled in six different energy simulation
programs: BLAST, ESP, SERI-RES, S3PAS, TASE, and TRNSYS. The results of these
simulations are meant to be used for comparative validation. Complete results of these
simulations are available in Haapala et al. (1995).
3.2.2 EnergyPlus Model
The building fabric modeling was mostly straightforward using the geometry and
property specifications found in the benchmark study and discussed above. The glazing
optical properties provided did not include the properties needed in an EnergyPlus model.
The standard glass properties provided with EnergyPlus, shown in Table 3.2, were used
as a substitute when necessary, such that the only properties used from the IEA report
were a solar transmittance at normal incidence of 0.747 and a glass conductivity of 0.635.
The window was modeled as two 3.175 mm glazings separated by a 13 mm airgap with
no frame.
Rather than modeling a complete air system, ventilation with outside air was
accomplished by specifying a scheduled infiltration rate for each zone, which is
equivalent to ventilation with outside air. Heating and cooling in the rooms was
accomplished via purchased air, which also does not require specification of a complete
air system.
The most detailed models available in EnergyPlus were used in all cases. Both inside and
outside convection used the "detailed," or variable convection coefficient algorithm,
rather than the "simple," or fixed convection coefficient algorithm. For the inside, the
detailed algorithm uses flat plate correlations dependent on the temperature difference,
whereas the simple algorithm simply selects a coefficient dependent on the surface
Table 3.2 EnergyPlus standard window glass and airgap properties
Glass Properties
Value
Property
Solar transmittance at
0.9
normal incidence
Solar reflectance at
0.031
normal incidence
IR transmittance at
0
normal incidence
IR hemispherical
emissivity
Conductivity (W/mK)
0.84
Airg p Properties
Property
Units
Value
Density
kg/m3
1.29
Density temp.
derivative
kg/m3-K
-0.004
Conductivity
W/m-K
0.0241
Conductivity temp.
W/m-K2
7.6x10-5
kg/m-s
kg/m-s-K
1.73x10-5
1x10-7
Prandtl number
-
0.72
Prandtl number
temp. derivative
1/K
1
0.0018
derivative
0.9
Viscosity
Viscosity temp.
derivative
orientation. For the outside, the detailed algorithm uses correlations dependent on the
temperature difference, wind speed, and surface roughness, while the simple algorithm is
not dependent on temperature difference. Sky radiance was modeled as anisotropic,
which accounts for anisotropy of sky diffuse solar radiation incident on exterior walls. A
ten-minute timestep was used, which is the smallest timestep currently recommended for
EnergyPlus stability. Weather data was taken from the file drycold_ blast2.epw, which is
derived from the Denver, Colorado DRY-COLD.TMY file used in the benchmark
simulations. The EnergyPlus input file and weather file are included on the attached
compact disc.
3.2.3 Results
Table 3.3 shows EnergyPlus predictions and the mean of the predictions from the
benchmark simulations for various parameters. Minimum temperatures for rooms 1 and
2 are not shown because they are always heated, so the minimum temperature is fixed at
18*C. The predictions for annual heating energy and peak heating load are quite similar,
differing by less than 6 percent. EnergyPlus predictions for annual cooling energy and
peak cooling load are 10-20 percent less than the mean of the benchmark predictions.
This deviation is acceptable because the benchmark predictions have not been validated
with experimental data.
EnergyPlus slightly overpredicts the temperature extremes, especially for the corridor
minimum temperature, which is 2.6 *C less than the mean benchmark minimum.
However, the benchmark predictions for minimum temperature span a range of nearly 4
*C. Figure 3.2 shows temperature predictions for a winter day. EnergyPlus clearly
predicts lower temperatures than the other programs, but the temperature trends
throughout the day are very similar to the other predictions. The discrepancy can also be
explained by the fact that most of the other programs do not appear to account for
variations in outdoor air density due to temperature. The low temperature for this day is
-23*C. The resulting increased air density causes a higher mass flow due to infiltration,
resulting in a lower corridor air temperature. Figure 3.2 shows EnergyPlus temperature
predictions if the infiltration rate is reduced by 0.987/1.16, the ratio of the fixed air
density in Denver recommended in the benchmark study to the actual air density at 23*C. This adjustment increases the temperature by about 1*C. The temperature is
increased by another 1*C if the simple convection coefficient algorithm is used, which
Table 3.3 Comparison of EnergyPlus and mean IEA benchmark predictions
Parameter
Annual heating energy
Annual cooling energy
Peak heating load
Peak cooling load
Room I maximum temp.
Room 2 maximum temp.
Corridor maximum temp.
Corridor minimum temp.
Units
MWh
MWh
kW
kW
0
C
C
0
C
0
C
0
EnergyPlus
1.11
1.55
3.62
1.99
IEA Benchmark
1.18
1.71
3.72
2.48
29.4
29.4
29.8
6.7
28.8
0.6 *C
28.3
28.9
9.3
1.1 0 C
0.9 0 C
-2.6 *C
Difference
-5.9 %
-9.4 %
-2.7 %
-19.8 %
16
-E
- Ea - E+ low inf.
14
----- Eb - E+ low inf.,
simple conv.
-B
- BLAST
*0
12
I-
E
-E+
-P
10
-
-
S - SERI-RES
-
3 - S3PAS
- TASE
-T
8
ESP
R -TRNSYS
6
0
5
10
15
20
25
Time (h)
Figure 3.2 EnergyPlus and lEA Benchmark corridor Jan. 4 temperatures
tends to predict higher convection coefficients and hence more heat flux from the rooms
into the corridor. Some of the other programs used for simulation use fixed convection
coefficients.
Figure 3.3 shows temperature predictions for a summer day. EnergyPlus predictions are
at the upper range of the other predictions, but again the temperature trends are similar.
31
30
E+
-E-
E
-B-BLAST
(-29
E-P-ESP
2
1
3
28
3
-
S
-
SERI-RES
- 3- S3PAS
CD
0.
E
-
2-
-
T - TASE
R-TRNSYS
26
25
0
5
10
15
20
25
Time (h)
Figure 3.3 EnergyPlus and lEA Benchmark corridor July 27 temperatures
The discrepancy is smaller because difference in calculation of infiltration and
convection from the outer walls does not have as large an effect in the summer, when the
temperature difference between inside and outside is smaller.
Because the benchmark results do not include experimental data, they cannot provide a
robust validation. However, the comparison between EnergyPlus and benchmark results
is strong enough to show that EnergyPlus holds promise as an energy simulation program
and should be further validated.
3.3 lEA Empirical Validation
3.3.1 Case Description
The IEA Empirical Validation study provides a more robust validation of EnergyPlus's
simulation of fabric heat transfer by comparing simulation results to experimental data.
This study uses experimental data collected in the U.K. under the direction of the U.K.
Building Research Establishment (Lomas et al. 1994). Hourly temperature data was
collected from several small test rooms on the edge of an airfield 70 km northwest of
London. The rooms, shown in Figure 3.4, are of lightweight, timber framed construction
with a concrete slab floor elevated above the ground. The rooms are tightly sealed to
prevent infiltration. The roofspace above each room is vented. Each building consists of
two identical rooms separated by a heavily insulated wall; they study does not state why
each building contains two rooms. Data was collected for one room in each building.
Data was collected for two different time periods for three different test rooms. A
removable panel allowed for various glazings to be placed in the southern wall of the test
rooms. The temperature within the rooms was uncontrolled for a week in May 1990
during which data was collected. An oil-filled electric panel radiator with a maximum
power of 680 W was used to heat the rooms to a setpoint of 30'C from 06:00 to 18:00 for
a week in October 1987. Table 3.4 summarizes the six different cases considered.
Several energy simulation programs were originally used to model the experimental
cases. The results of these simulations are discussed thoroughly in Lomas et al. (1994).
The most general conclusion was that although most of the programs were able to predict
some parameters within the range of uncertainty, none accurately predicted all
parameters.
1506
1506
Figure 3.4 IEA Empirical Validation test rooms
Table 3.4 lEA Empirical Validation experimental cases
Case
Date
Heated?
Glazing type
Glazing area (M2 )
1
May 1990
No
None
0.0
2
May 1990
No
Single
1.5
3
May 1990
No
Double
1.5
4
Oct. 1987
Yes
None
0.0
5
Oct. 1987
Yes
Double
0.75
6
Oct. 1987
Yes
Double
1.5
3.3.2 EnergyPlus Model
The Empirical Validation Package includes extensive room construction information so
that the building fabric may be modeled as accurately as possible. The validation
package recommends specific surface constructions, which were implemented in the
EnergyPlus model. Each wall is divided into several surfaces in order to account for the
variable composition of the wall cross section. The only deviation from the
specifications recommended in the validation package was the modeling of the west wall,
which divides the test room from its neighboring twin. Rather than modeling the wall
with an extra, highly insulated outer layer in order to create an essentially adiabatic wall,
it was modeled with a symmetric section, with the outer surface temperature (the wall
temperature in the other room) specified to be the same as this inner surface temperature.
This also creates an adiabatic wall and is believed to better represent the actual physical
configuration.
As with the benchmark case, standard glass properties supplied with EnergyPlus were
substituted when the properties provided were insufficient. Nonstandard properties used
were a solar transmittance at normal incidence of 0.849 and a glass conductivity of 1.05
W/mK. The glazings were 4 mm thick, and the double-glazing had a 6 mm airgap, again
with no frame. An infiltration rate of 1 ach was assumed for the roofspace, as
recommended in the validation package.
The radiator was modeled using the "High Temperature Radiant System" model with a
60/40 radiative/convective split. This split is specified in the validation package and was
calculated using standard empirical results for convective and radiative heat transfer from
a vertical heated plate. The controls on the radiator model required some tweaking in
order to achieve the desired performance. An operative temperature throttling range of
26 - 30*C, corresponding to full to zero power, was found to have good performance.
The actual radiator was controlled by a PID controller, but this type of control cannot be
implemented in EnergyPlus. Attempts were made to account for the thermal mass of the
radiator using the "Low Temperature Radiant System," because the validation package
specifies the heater as having a 22-minute time constant. However, the results from this
model did not differ significantly from the high temperature model.
Exterior shading from the neighboring test buildings was included with the use of
detached shading surfaces. The most detailed models available in EnergyPlus, described
in the previous section, were used, along with a ten-minute timestep. The validation
package includes experimental weather and data files. The weather files were used to
create EnergyPlus weather input files using the weather format information in the
EnergyPlus manual (EnergyPlus 2001). The EnergyPlus input files and weather files are
included on the attached compact disc.
3.3.3 EnergyPlus Results
Table 3.5 shows measured and predicted values of maximum and minimum temperatures,
total energy consumption, and total south wall irradiation (svfr). The error bounds used
were given were derived by Lomas et al (1994) from modeling parameter uncertainties
using the Monte Carlo technique and SERI-RES. Hence, if the predicted value lies
within the error bounds, then any discrepancy from the measured value may be due to
uncertainty in the inputs, and not modeling errors in EnergyPlus. More accurate error
bounds could be obtained by applying the same technique in EnergyPlus (rather than
SERI-RES). However, the error bounds are not likely to change drastically because the
energy simulation programs use similar underlying principles. The measurement
uncertainty was estimated as ±0.2'C for all temperatures, and + 2% for the solar
irradiance and energy consumption
For the unheated cases, all of the parameters fall within the error bounds, which shows
that EnergyPlus modeling of building fabric heat gains and losses is very good. Few of
the parameters fall within the error bounds for the heated cases, indicating that there may
be some problems with the high-temperature radiator model, which are discussed below.
Only the two radiator models described above are currently available in EnergyPlus. Also
note that no radiator models were necessary to model Building A.
Table 3.5 Measured and predicted values of primary parameters and their uncertainty
Period
Description
Unheated
October
Heated
Parameter
Measured
Value
Lower
bound
Predicted
Value
Upper
bound
Max Temp ['C]
16.8
15.7
16.2
17.5
Min Temp ['C]
9.2
8.6
8.8
10.0
Max Temp ['C]
32.6
31.2
33.0
35.0
Min Temp ['C]
Max Temp ['C]
12.1
31.0
11.6
29.6
11.5
32.4
13.6
33.4
Min Temp ['C]
12.2
11.6
12.8
13.6
2
Svfr
Irrad. [MJ/m ]
82.8
76.8
82.5
88.8
Energy [MJ]
117.1
105.3
99.3
122.3
Opaque
Max Temp ['C]
29.8
29.4
29.9
30.2
Min Temp ['C]
14.6
14.0
12.7
16.4
Energy [MJ]
99.1
not available
75.2
not available
Max Temp ['C]
Min Temp ['C]
31.5
12.9
not available
not available
33.0
not available
12.0
not available
Energy [MJ]
89.3
78.1
65.1
92.7
Max Temp ['C]
37.8
36.5
42.0
40.5
Min Temp ['C]
11.9
11.5
11.6
13.9
Irrad. [MJ/m 2 ]
81.1
76.7
80.2
85.5
Small dbl. glz.
Double glaz.
Svfr.
V
Table 3.6 RMS error of predicted parameters
Period
Description
Solar
Irradiance
W/m
2
Opaque
Unheated
October
Heated
Single glaz.
Double glaz.
Opaque
Small dbl. glz.
35.60
52.53
Double glaz.
Room
Temp
"C
N. Wall
Temp
Floor
Temp
Ceiling
Temp
C
C
C
_
Roofspace
Temp
kJ
_C
0.93
0.65
0.63
0.74
3.02
-
1.33
1.09
1.42
1.83
2.35
1.58
1.09
1.67
2.90
3.19
-
2.30
2.50
1.46
1.95
1.04
0.93
1.24
1.40
3.40
201
3.16
289
2.93
2.81
2.16
2.58
3.48
325
Table 3.6 shows the RMS error of each of the predicted parameters over the entire
measurement period. The RMS temperature error tends to be 0.5 - 1.00 C higher for the
heated cases than the unheated cases. More detailed, hourly results will now be presented
for each parameter measured.
Solar Irradiance
Figure 3.5 shows time variation of south wall solar irradiance for the October
measurement period. The values do not agree exactly because the prediction is based on
diffuse and global horizontal solar data, whereas the measurements are taken from
instruments mounted on the south-facing wall. The agreement between predicted and
measured south wall solar irradiance is excellent, with only slight variations in peak
values noticeable. These may result from unpredictable variations in ground-reflected
diffuse radiation. The error in total irradiation over each measurement period is less than
2 percent.
1000
900-800 -700
600
500 400300 200 100
010/19
10/20
Energy
10/21
10/22
10/23
10/24
Date
Figure 3.5 South wall solar irradiance for October 1987
10/25
-
Room Temperatures
Figures 3.6 through 3.9 show time variation of room temperatures for the unglazed and
double glazed cases. The predictions match the measured room temperatures fairly well.
The unglazed, unheated case prediction (Fig. 3.6) slightly leads the temperature in time
and slightly underpredicts the measured temperature. These errors are small enough that
they may result from input uncertainties, as the maximum and minimum temperatures are
within the error bounds.
The glazed cases tend to overpredict the maximum temperature on days with large solar
irradiation. For the unheated case (Fig. 3.7), these small overpredictions (on 5/25 and
5/26) may result from coincident overpredictions in the solar irradiance. For the heated
case (Fig. 3.9), this attribution is not possible, because the peak solar irradiance is
35
30
C.)
25
E 20
a15
E
10
5
0 45/23
5/24
5/25
5/26
5/27
5/28
5/29
5/30
Date
Figure 3.6 Unheated, unglazed room temperature
35
30
,25
2 20
0.15
E
10
5
05/23
5/24
5/25
5/26
5/27
5/28
5/29
Date
Figure 3.7 Unheated, double glazed room temperature
5/30
0
25
20
E 15
10/19
10/20
10/21
10/22
Date
10/23
10/24
10/25
Figure 3.8 Heated, unglazed room temperature
454035-0(30
25
CD20
E
0.
15
10
5
-
010/19
10/20
10/21
10/22
Date
10/23
10/24
10/25
Figure 3.9 Heated, double glazed room temperature
actually underpredicted. The error in the heated cases is larger, probably because the
solar irradiance is larger for these cases. This indicates that there may be a problem with
the solar gains through the window, perhaps because of the incomplete window glass
specifications in the validation guidebook (i.e., no reflectance given). However, note that
the temperature extremes are not more than 0.1 *C outside of the error bounds for the
unheated cases, and only two temperature extreme values are outside of the error bounds
for the heated cases. These errors at the extremes are as large as 5 K for the heated,
double glazed case, again indicating possible problems with the radiator model.
The predicted room temperature leads the measured value in the heated cases. This is
because the radiator was specified as having a 22 minute time constant, which is not
included in the radiator model. This "step-function" radiator model also allows the
temperature to begin dropping sooner when the radiator is shut off, yielding slightly
IF--'
-
.
--
W
lower minimum temperatures. As previously stated, attempts to use different radiator
models that include a thermal mass were unsuccessful in obtaining more accurate
predictions than those shown.
Surface Temperatures
Figures 3.10 through 3.15 show time variation of representative interior surface
temperatures. For the opaque cases (Figs. 3.10 and 3.11), the predicted surface
temperatures follow the measured temperatures very well. For the glazed cases,
EnergyPlus consistently overpredicts the maximum surface temperature. This was
initially believed to result from the uncertainty in the surface absorptance or some other
parameter. However, sensitivity studies using the maximum and minimum input values
for several suspected parameters did not show variation in the surface temperatures large
enough to account for the error.
35
-
30 25
-
2015
-
10..... Predicted
f
Measured
0 -L
.
5/23
5/24
5/25
5/26
5/27
5/28
5/29
5/30
Date
Figure 3.10 Floor temperature for unheated, unglazed room
40
-
35 30
0
-
25 -
20
-
0.
E 1510
-
----- Predicted
5
I
-
Measured
0 4
.
10/19
10/20
10/21
10/22
Date
10/23
10/24
Figure 3.11 Floor temperature for heated, unglazed room
10/25
25 4
20 -
15E
0
n.j
-I
-....
--
Predicted
- Measured
5/23
5/24
5/25
5/27
5/26
5/28
5/29
5/30
Date
Figure 3.12 Floor temperature for unheated, single glazed room
35 30 25
~20*) 1510-
- Predicted
5
Measured
0--
5/23
5/24
5/25
5/26
5/27
5/28
5/29
5/30
Date
Figure 3.13 Ceiling temperature for unheated, single glazed room
40
35 -30 -
0
E15
10 --
---Predicted
10
5-
-- Measured
10/19
10/20
10/21
10/22
Date
10/23
10/24
10/25
Figure 3.14 Floor temperature for heated, small double glazed room
i--,
77
-
35
30
0 25
-
10
7
0.
10
5
-
5-
10/19
10/20
10/21
10/22
Date
10/23
10/24
10/25
Figure 3.15 Ceiling temperature for heated, small double glazed room
The problem therefore was thought to lie in the convection coefficient calculation,
especially at higher surface temperatures, where higher buoyancy forces would highlight
problems with natural convection coefficients. Because the room air temperature is
relatively accurate, the correct heat flux from the wall is being achieved, but the predicted
convection coefficient could be too low, yielding excessively high surface temperatures.
The convection coefficient is especially suspect because the overpredictions are largest
for the floor (highest h) and smallest for the ceiling (lowest h). A CFD study was used to
evaluate convection coefficients and compare them to the EnergyPlus predicted values.
The results of this study show that EnergyPlus does underpredict convection coefficients
for the test rooms. The methods and results of this study will be discussed below.
Roofspace Temperature
Figures 3.16 and 3.17 show roofspace temperatures for the unheated and heated double
glazed rooms. The roofspace temperature predictions are less accurate than any other
prediction, but this is to be expected, as the validation guidebook makes no effort to
3530
25
20 !
15-
E
10-
-Predicted
5
0
5/23
5/24
5/25
- Measured
5/26
5/27
5/28
5/29
5/30
Date
Figure 3.16 Roofspace temperature for unheated, double glazed room
V
35
----- Predicted
30
-Measured
25-
20
.
15
CL10
-
-
'-5
0
-5
-
10/19
10/20
10/21
10/22
Date
10/23
10/24
10/25
Figure 3.17 Roofspace temperature for heated, double glazed room
model the roofspace extremely accurately. The infiltration rate is the greatest unknown,
as it was assumed to be constant at 1 ach, but would actually vary greatly. Given this
unknown, the predictions are surprisingly good, especially for the unheated case. The
predicted temperature tends to lead the actual temperature, and the heated case tends to
overpredict the temperature extremes.
Energy Consumption
Figure 3.18 shows radiator energy consumption for the heated, double glazed room.
EnergyPlus consistently underpredicts the radiator energy consumption. This is primarily
due to the lack of any time lag in the radiator model, which allows the air to heat up faster
than it actually does, and thus allows the radiator energy consumption to decrease more
rapidly. The validation guidebook also suggests comparing energy savings of the doubleglazed case over the opaque case. The predicted energy savings is 34.2 MJ, while the
3000 Predicted
Measured
2500
2000
1500----
30
10001
1
500
10/19
10/20
10/21
10/22
Date
10/23
10/24
10/25
Figure 3.18 Radiator energy consumption for heated, double glazed room
38
Table 3.7 Surface temperatures for CFD simulation
Surface
floor
Temperature (*C)
34.7
ceiling
east wall
west wall
north wall
south wall
window
34.2
34.6
34.1
33.6
33.8
27.0
actual energy savings is 27.8 MJ. This comparison is not overly favorable; the error is 23
percent. These comparisons show that the current "step-function" radiator model is not
appropriate for evaluating energy consumption.
3.3.4 CFD Study
As stated above, a CFD study was performed using PHOENICS 3.3 in order to evaluate
the room convection coefficients for comparison to the EnergyPlus predicted coefficients.
The low Reynolds number model was used in order to account for the low velocities
occurring in a pure natural convection case and thus determine the convection
coefficients as accurately as possible. Wall temperature boundary conditions, shown in
Table 3.7, were specified using the EnergyPlus predicted wall temperatures coincident
with the peak room temperature for the unheated, single glazed case. A 40 x 42 x 44 (i x
j x k) grid was used, with the first grid node a distance of 0.6 mm from the wall in order
to yield a y* value of approximately 1 in the node adjacent to all surfaces.
Convection coefficients were calculated by two different means for comparison. The
first method is based on the net heat flux from the surface, such that
h=
q
Tswface -
(3.1)
Troom
The second method is based on the first node convection coefficient calculated via pure
conduction, and then transformed to a room convection coefficient using the node
temperature:
h =
k
Ax
(Tsace
-
(Tface -
Tr.st node)
(3.2)
Troom )
Following correction of an air density error within PHOENICS, these two values were
found to match exactly.
Figure 3.19 shows the velocities at the room midplane. The buoyancy effects of the cold
window on the southern wall establish a fairly strong natural convection circulation
pattern within the room, with maximum velocities greater 0.1 m/s. The circulation
pattern is likely enhanced by the unusually small room geometry. This strong circulation
pattern would clearly increase convection coefficients beyond those predicted by standard
correlations. This could explain why the EnergyPlus surface temperature predictions are
high for the glazed cases. The surface temperature predictions for the unglazed cases are
Bl/l I I
2
I
5W///
I
I
--
1.5
11 \
/
-
/
-
/
/
\ \ %lift
\
Jll'
-1
I
ItlaI
0.5
--~
-
0
~~~
-
,,,,l
id
I 0114im
-
'M
-
ilf11M3
-
MINf
y
0.2 m/s
Figure 3.19 CFD predicted velocities at room midplane
accurate because this circulation pattern does not exist in those cases, because the cold
window is not present.
Table 3.8 shows EnergyPlus and CFD predicted convection coefficient values. As
expected, the CFD values are consistently higher than the EnergyPlus values. Also note
that the difference in coefficients is 171 percent for the floor and only 20 percent for the
ceiling, as expected because the error in temperature predictions is greater for the floor
than the ceiling. These results make it appear likely that the convection coefficient
correlations used in EnergyPlus are the source of error in surface temperature predictions.
However, the error may be especially large for this case because of the unusual room
geometry.
Table 3.8 Comparison of convection coefficient predictions
E+ h (W/m 2-K)
1.78
CFD h (W/m 2-K)
4.82
% difference
171%
ceiling
0.78
east wall
west wall
north wall
1.56
1.33
1.04
0.94
2.55
20%
64%
2.28
3.74
71%
259%
window
2.34
3.63
55%
Surface
floor
As a test of this theory, EnergyPlus was modified such that the newly calculated
convection coefficients were used. Two methods were employed. In the simplest
method, the convection coefficient for each surface was treated as fixed, constant at the
values shown in Table 3.8. In the other method, the convection coefficients are variable,
calculated using the EnergyPlus correlations, but increased by a constant factor
determined by the ratio of the CFD and EnergyPlus values shown in Table 3.8, so that
h CFD, peak T
h= hh
-
(3.3)
correlation
h E+,peak T
Figure 3.20 shows floor temperatures for the hottest day of the measurement period with
all four convection coefficient calculation methods: fixed EnergyPlus, variable
EnergyPlus, fixed CFD, and variable CFD. It is clear that even fairly significant changes
in the convection coefficient do not have a large effect on the predictions. The floor
temperatures for the fixed value methods are nearly identical, with a peak temperature of
only 1.0C less than the original, variable EnergyPlus method. Therefore, although the
EnergyPlus correlations may not be accurate in all cases, they are not the major cause of
the error in surface temperature predictions.
Additional sensitivity studies were performed in an attempt to isolate other sources of
error in the surface temperature prediction, focusing on the floor temperature. The floor
temperature was found to be most sensitive to changes in the concrete solar absorptivity,
which governs the amount of solar energy directly absorbed by the floor. This explains
why the floor temperature predictions are fairly accurate for the unglazed, opaque wall
case, in which there is no solar input to the floor. The temperature prediction was also
sensitive to the concrete specific heat, because higher specific heats yield a lower
temperature increase for the same heat input. Figure 3.21 shows the lowest floor
temperatures obtained, when the absorptivity is lowered from 0.5 to 0.4 (the lower end of
-
E+ variable h
-
CFD variable h
E+ fixed h
- CFD fixed h
- Measured
)2
E
-
24
0
0
M~ 19
5/26/00
14 1
0:00
1
4:00
8:00
12:00
Time
16:00
20:00
0:00
Figure 3.20 Unheated single glazing floor temperature with various h calculation methods
30
25 -I
Old
New
floor ac
0.5
0.4
floor c,
920 J/kg-K
1012 J/kg-k
h calc
method
E+ variable
CFD fixed
10-
5 --
Measured
--- Original prediction
Meured
New prediction
-+va-bl---fie
0 1
5/23
5/24
5/25
5/26
5/27
5/28
5/29
5/30
Date
Figure 3.21 Unheated single glazing floor temperature: new and original predictions
the uncertainty band), the specific heat is raised from 920 to 1012 J/kgK (the upper end
of the uncertainty band), and the fixed CFD convection coefficient method is used. The
peak temperature is reduced by 2.4'C. The remaining error, 2'C may be accounted for
by errors in other input properties, the measurement error (±0.2*C), and a slight
overprediction of the solar radiation incident on the window. In addition the error band
given for the floor absorptivity is somewhat questionable, so the actual absorptivity could
be even lower than 0.4.
3.3.5 Conclusions
This validation exercise has shown that EnergyPlus can accurately model building fabric
gains and losses, yielding accurate room air temperature predictions. However, surface
temperature predictions can be high, especially for periods with large solar gains and high
room air temperatures. These errors were thought to result from the inapplicability of the
convection coefficient correlations used in EnergyPlus to the unusual room geometry.
Although the convection coefficients correlations were found to be incorrect in this case,
they are probably not the source of this error, which probably results from errors in
several input properties, such as the floor absorptivity. The radiator model can predict
the room temperature fairly well, but energy consumption predictions are not accurate
because the model does not include the time lag apparent in an actual radiator.
3.4 Displacement Ventilation
3.4.1 Case Description
The previous cases all rely on the assumption that the room air temperature is uniform,
which is true in many cases. However, for some cases, such as displacement ventilation,
Table 3.9 Displacement ventilation room materials
Surface
ceiling
floor
rear wall
side walls
parapet
Thickness
m
0.175
0.175
0.140
0.140
0.100
Density
kg/m
2300
2300
700
700
30
Specific heat
J/kgK
840
840
840
840
1470
Thermal conductivity
W/mK
1.9
1.9
0.23
0.23
0.035
Figure 3.22 Displacement ventilation test room
this is not the case. The displacement ventilation study provides a validation of one
method of modifying EnergyPlus in order to account for non-uniform air temperature.
This case is based on experiments carried out in a full scale climate room at Delft
University (Chen 1988). The room, shown in Figure 3.22, has dimensions 5.6 x 3.0 x 3.2
(x,y,z) m, with a parapet height of 0.9 m. Room enclosure material properties are shown
in Table 3.9. The floor and ceiling exterior surface temperatures are controlled to be
equal to the ceiling and floor interior temperatures, respectively, as if identical rooms
were located above and below the room. The wall exteriors are electrically heated such
that they are adiabatic. The temperature of the space outside the window could not be
found in published data; the window exterior surface temperature was assumed to be 24.5
'C. The room is cooled by displacement ventilation at a rate of 7 ach.
The room temperature was initialized at 23.0 *C, after which a step heat input of 950 W
was uniformly applied to the venetian blinds. The inlet temperature was then controlled
such that the temperature in the middle of the occupied zone (x = 2.8 m, y = 1.5 m, z =
0.9 m) remained constant at 23.0 'C. The purpose of this case is to study the effect of
adding a non-uniform air temperature distribution to EnergyPlus, in order to better
represent the actual conditions of displacement ventilation. This case has previously
been simulated using ACCURACY and non-uniform temperature distributions with
excellent results; complete results are available in Chen (1988).
3.4.2 EnergyPlus Model
The room surfaces were input using the properties specified in Table 3.9. Once again,
the property data available for the window was incomplete, and EnergyPlus standard
glass properties were substituted as necessary. The window is modeled as two 6 mm
glazings separated by a 12.7 mm airgap. The outer surface of the ceiling was specified as
the floor, and vice versa, in order to achieve a vertically re eating room. The wall
exterior was specified as having a near-zero (0.000 1 W/m K) total (convective and
radiative) heat transfer coefficient in order to prevent heat transfer through the walls.
EnergyPlus does not allow a convection coefficient equal to zero. The window outer
surface temperature was specified as fixed at 24.5 *C. The most detailed models
available in EnergyPlus, described previously, were used, along with a ten-minute
timestep. The choice of weather file was inconsequential because this model has no
interaction with the outdoors.
Cooling was again accomplished using "purchased air" at a supply temperature of 15 *C.
Although this creates a variable air volume system, whereas the experiment used a
constant air volume system, the effect on the room energy balance being considered is
unchanged.
The venetian blinds were modeled as an electrical load. The radiative/convective split
was determined using experimentally determined heat transfer coefficients (Chen 1989).
The coefficients are based on measured temperatures and heat fluxes at steady state, and
yield a 70 percent convective load. In reality, the radiative/convective split varies with
time, with the radiative portion decreasing as the room walls heat up. However, no
transient heat transfer coefficient data was available, so the load was initially modeled as
70 percent convective for the entire experiment.
Rather than directly coupling CFD with EnergyPlus in order to account for the nonuniform room temperature, CFD results previously obtained by Chen were used to
account for the non-uniformity. Rather than calculating the convective heat flux from the
surfaces based on the room air temperature, it was calculated based on a near-surface air
temperature for each different surface. These temperatures were calculated from nonlinear functions for the difference in temperature between the air near the surface and the
center of the room (23.0 *C):
ATceiing
=
0.036 + 6.99x10 3 - Q - 2.72xl0-. Q2
ATfloo = 0.026 - 1.83x10 3 - Q + 5.55xl0-
Q2
ATwindow = 0.036 + 6.99x10 3 - Q - 2.72x10-. Q2
ATwus = 0.047 +2.10xlO
- Q - 1.07x10 -
ATparapet = 0
where Q is the room cooling load and is calculated as
Q=ric (Tue - Ts)
Q2
(3.4)
(3.5)
(3.6)
(3.7)
(3.8)
(3.9)
a.
These functions were determined by curve fitting to temperature distributions determined
using CFD (Chen 1988). The EnergyPlus input files are included on the attached
compact disc.
3.4.3 Results
Figure 3.23 shows the predicted and experimental cooling load versus time. The cooling
load increases to the steady-state value of 950 W as the walls heat up and the air
temperature stratification increases. However, the simulations that do not use the
nonlinear temperature difference expressions, and instead assume uniform temperature
throughout the room, significantly overpredict the cooling load.
The predictions that do account for air temperature stratification match the experimental
data well. However, the fixed 70 percent convective heat load still overpredicts the
cooling load for the first three hours of the experiment. This may be because the heat
load from the blinds initially has a larger radiative portion because the walls are cooler
than they are at steady-state. To account for this, a simulation was performed with an
arbitrary hourly schedule for the radiative/convective load split, with the convective load
increasing from 54 to 70 percent, such that the simulation results for the non-uniform
temperature simulation match the experimental data. Use of this arbitrary schedule
without accounting for air temperature stratification still significantly overpredicts the
cooling load.
This validation shows the utility of coupling CFD with EnergyPlus when appropriate.
Although this validation did not involve a direct coupling of CFD with EnergyPlus,
similar results are obtained if such a coupling is used (Chen 1988). Alternatively,
temperature difference expressions such as those used here can be used in some cases.
900
800 -
$600
1500
x Experiment
0) 400
0300-
-.-.--70% convective - original E+
200
200
----- variable convective - original E+
100
0x
0
5
10
Time (hr)
15
Figure 3.23 Transient cooling load for displacement ventilation
20
if
3.5 MIT Test Chamber
The previous three studies have been concerned with modeling heat losses and gains
(both through the building fabric and due to internal loads) and the resultant heating or
cooling loads. This study provides a validation of the air system models used to meet
those loads. Experimental data for this study was collected at the MIT Test Chamber.
3.5.1 Experimental Facility
The test facility, shown in Figure 3.24, consists of a well-insulated enclosure separated
into two rooms by a partition wall with a large, double-glazed window. Not shown in the
m
'' 2A43
4. pesaM0W
e
wase
I. Am *esT
4. Am 0's.YT
4. 6OWLAIMV.
Pla
T.
SLOT
1.
neLAM
Cow."
Figure 3.24 Sketch of the test chamber
SUPPLYAIR
75000
40.00
30.00
-20.00
TC
I
1P4
X AIR
10.00
124.e61 CFM
TCSAN
TOFLOS
36.93 PERCNT 125.00 CFM
1
TOS
TCSA5
TCCCRII
TCSRHS
TCFPM
5.53 PERCNT
90.00 PERCNT 0.78 FPM
TCR14C
0.00 PRCNT
OATEMP
17.26 DEG C
HUMID I
OFF
OARH
59.75 PERCN'T
TCHUM
0.00 PRCNT
TST21
50.00
2C
TCPHC4000
0.00 PRCNT
OCCUPIED
30.00
TCLTD
LOW TEMP DET
20.00
OFF
10.00
TCLTDR
LOW TMP RSET
OFF
0.00
TCMAS 12.50
TCMAT 20.22
E
Figure 3.25 Control interface and schematic of HVAC system
00
00
1250
CCFAN
figure are two doors at either end, and the furniture shown in the figure was not present
during the experiment. All walls have an insulating value of R-30, or 5.3 Km 2/W.
The larger room is used as the test chamber and the smaller room as the climate chamber.
The test chamber is supplied with air via two linear ceiling diffusers, and air is exhausted
through a grill ceiling exhaust. The climate chamber is supplied with air via a rear wall
diffuser and exhausted via a rear wall exhaust. Each chamber has a separate HVAC
system. The two systems are nearly identical; the primary difference is that the supply
and return fans for the test chamber have a variable-speed drive, whereas the fan speed
for the test chamber is fixed. Figure 3.25 shows the configuration and control interface
of the HVAC. system. This interface allows the operator to control various system
setpoints. The control system also allows data for all data points in the control system to
be recorded at a time interval specified by the operator.
The facility measurement equipment includes a multi-gas monitor and analyzer and a
thermocouple data logger. The multi-gas monitor was used to measure levels of a tracer
gas, SF6, used to measure the ventilation rate. The concentration measurement error is
10%. Thermocouples were used to measure temperatures not monitored within the
HVAC system. The data logger was used to record the temperatures, resulting in an
overall error of 0.4*C. All thermocouples junctions were coated with aluminum paint to
reduce the effect of radiation on air temperature measurements.
3.5.2 Experimental Setup
Many experiments were required in order to identify and eliminate error sources before
the final experimental setup was achieved. In the initial experimental setup, the climate
chamber was heated to 32*C, while the test chamber was controlled to 20*C and 50%
relative humidity. Air was supplied through a single circular opening in the test
chambers, and a computer and several fluorescent lights were turned in order to create
additional cooling load. Many changes were made to this setup before arriving at the
final configuration described below; Table 3.10 shows the problems identified and the
actions taken to solve them.
Before beginning experimentation, there was some concern about the accuracy of the
HVAC system temperature and relative humidity measurements. These measurements
were verified by measuring the dry and wet bulb temperature at the sensor locations
within the HVAC system. A 10" stainless steel temperature probe was inserted through a
small hole in the duct, drilled in the same location as the HVAC system sensor, to
measure the temperature for verification. Wet bulb temperature was measured by
wrapping the probe with moistened paper towel. All HVAC system and probe
measurements were found to agree within the accuracy of the probe (±0.5*C).
In addition to the HVAC system temperature measurements, test chamber, climate
chamber, and laboratory air temperatures were measured with thermocouple arrays.
Figure 3.26 shows approximate locations of thermocouples inside and outside of the test
chamber. Thermocouples in the laboratory were placed approximately 10 cm away from
Table 3.10 Changes made to arrive at final test chamber experimental setup
Problem
Heating and cooling coils both run even when not
dehumidifying
Difficulty predicting latent cooling
Difficulty measuring power consumption of
computer and fluorescent lights
Difficulty controlling and predicting supply air
temperature
Uncontrolled laboratory temperature creates varying
conductive load through chamber walls
Possible infiltration from laboratory
Return air temperature measurement may not be
accurate
Test chamber air not well mixed
Possible infiltration from climate chamber
Varying climate chamber temperature creates
varying conductive load through chamber window
Solution
Change control logic - if mixed air T < supply air
setpoint, use heating coil, opposite for cooling coil
Eliminate humidity control
Remove electrical loads and switch to mild heating
case - test chamber = 240, climate chamber = 16*C
Set supply air setpoint as fixed and let test chamber
temperature float
Measure and record laboratory temperature with
several thermocouples
Run test chamber under positive pressure by
reducing return fan speed
Measure test chamber temperature with
thermocouple array
Change inlet to dual linear diffusers
Close down climate chamber inlet dampers until
climate chamber pressure <test chamber pressure
Measure and record climate chamber temperature
near window with several thermocouples
classroom
Location
1
1
Height (m)
0.5
1.5
2
1.0
3.0 (10 cm
above test
test chamber
climate
chamber
3
chamber roof)
laboratory
4
4
5
5
6
6
7
7
8,9,10
0.3
1.0
0.3
1.0
0.3
1.0
0.3
1.0
1.2
Figure 3.26 Temperature measurement locations
the test chamber wall. The laboratory temperature was determined by averaging the four
laboratory temperatures, test chamber temperature was determined by averaging the eight
chamber temperatures, and the climate chamber temperature was determined by
averaging the three climate chamber temperatures. The climate chamber temperature is
only measured near the window because the measurement is used as an input for the
window conduction calculation.
Both chambers were ventilated at a constant rate with 100% outdoor air. The test
chamber supply air temperature was maintained as a constant, while the climate chamber
supply air temperature was allowed to vary in order to maintain a constant return air
temperature. Because the facility is not equipped to directly measure coil loads, the
heating or cooling coil load was simply calculated from an energy balance across the coil:
lq=V
p cp (Tppy
,r
-Toutdr air)
(3.10)
where V is the known, constant volumetric flowrate.
To reduce the coil temperature difference measurement error, the system was allowed to
run overnight with no heating or cooling. The difference in supply air and outdoor air
temperature in this adiabatic case (0.2*C) was then attributed to measurement error and
added to the measured supply air temperature in order to achieve an accurate coil
temperature difference measurement.
In order to prevent infiltration, the test chamber was run under positive pressure relative
to its surroundings. This was achieved by running the supply fan at a faster speed than
the return fan (80% of maximum speed vs. 20% of maximum speed). The climate
chamber was also found to be under.positive pressure, so the supply dampers for the
climate chamber were closed down until it was at a lower pressure than the test chamber.
In the final configuration, the test chamber pressure relative to the laboratory was 28 Pa,
while the climate chamber pressure was 23 Pa.
The test chamber ventilation rate was measured using the tracer gas system. The tracer
gas, SF6, was injected at both diffusers at a constant flowrate and the system was allowed
to run overnight to reach a steady-state concentration distribution. The SF6 concentration
was then measured at each test chamber temperature measurement location (Fig. 3.26), as
well as at two locations in the return duct, approximately 20 cm beyond the room outlet.
The SF6 concentration in the supply air was measured just inside of both supply diffusers.
The concentration at each location was found to vary by less than ±10% from the roomaverage value, indicating that the room air is fairly well mixed. Because most air actually
exits the chamber through exfiltration, the room-average concentration was used to
calculate the ventilation rate via a mass balance:
S=
SF6(3.11)
outC.C in
where Cin is the SF 6 concentration (m3 SF6/m 3 air) of the supply air, and Cout is the SF6
concentration of the air exiting the chamber, equivalent to the room-average
concentration.
The validation case is based on a single experiment. The ventilation rate was 0.145 m3/s,
or 11.4 ach. This high ventilation rate was necessary to maintain sufficient positive
pressure in the test chamber. The chambers were first run to steady state with a climate
chamber return air temperature of 32*C and a test chamber supply air temperature of
14*C, in order to simulate a summer cooling case. The setpoints were then changed to
Table 3.11 Test chamber experimental parameters
add climate chamber return air temperature of 12*C and a test chamber supply air
temperature of 32*C in order to simulate a heating case. These parameters are
summarized in Table 3.11. The experiment was run from October 17 to October 24,
2001; the mode was switched from cooling to heating at 11:00 a.m. on October 19.
3.5.3 EnergyPlus Model
All insulated test chamber surfaces were modeled as resistance-only surfaces (no thermal
mass) with a resistance of 5.3 Km 2/W. The thermal mass was not included because it
was not known. The window glass properties shown in Table 3.2 were used because no
other window glass properties were available. The window was modeled as two 3 mm
glazings separated by a 12 mm airgap. The most detailed models available in
EnergyPlus, described previously, were used, along with a ten-minute timestep.
Two additional zones, representing the laboratory and the climate chamber, were created
in order to apply the measured laboratory and climate chamber temperatures on the
opposite side of the test chamber surfaces. The only purpose of these zones was to
ensure that the air temperature on the test chamber outer surfaces was correct. Both
zones were ventilated with purchased air in the EnergyPlus model. Hourly zone setpoint
schedules were input from the measured average hourly climate chamber and laboratory
temperatures. The resulting zone temperatures in the model match the measured space
temperatures.
The test chamber was ventilated with air supplied directly from the ventilation system
("Direct Air"). The heating coil was modeled as an electric heating coil with an
efficiency of 1.0. The cooling coil was modeled as a simple water cooling coil. The only
performance-related inputs for this coil are the UA value (1200 W/K) and the relative
humidity leaving the coil (95%). The model assumes that the relative humidity leaving
the coil will never be higher than this value. This coil model does not include any more
detailed latent cooling predictions and latent cooling was therefore not considered in this
study.
The test chamber supply air temperature was set as constant, 14*C for the cooling mode
and 32*C for the heating mode. The outdoor air temperature and humidity were supplied
via a weather file created from experimental data. The input fields of concern in the
- weather file were the outdoor dry bulb temperature, relative humidity, and dewpoint. This
model does not interact with the outdoors other than through the outdoor air intake, so
any other inputs in the weather file are irrelevant. The dewpoint was calculated from the
measured relative humidity and dry bulb temperature using a spreadsheet with a
psychrometric function calculator. The EnergyPlus input file and weather file are
included on the attached compact disc.
3.5.4 Results
The results are presented for the predictions of four parameters: heating coil load,
sensible cooling coil load, room air humidity ratio, and room air temperature.
The comparison of coil load predictions essentially evaluates the coil controls in
EnergyPlus, because the load for both the simulation and the measurement is calculated
using the measured air flowrate. Differences in coil loads therefore result from
differences in supply air temperature due to imperfect control of the actual coils.
EnergyPlus assumes the coil is controlled to always provide a supply air temperature at
exactly the setpoint; such perfect control cannot be achieved with an actual coil.
Figures 3.27 and 3.28 show predicted and measured heating and cooling coil loads. The
heating coil load prediction is very good. This is because the coil is electric and therefore
responds very quickly, providing control very close to the perfect control assumed in
EnergyPlus. The total predicted heater energy has a 2 percent error from the actual heater
energy for the experimental period. The cooling coil sensible load prediction is
somewhat less accurate. This is because the coil uses chilled water. The chilled water
inlet temperature is not maintained exactly constant and a three-way valve controls its
flowrate, so it cannot be controlled as perfectly as the heating coil. In addition, the coil
control loop may need additional tuning. The total predicted sensible cooling energy has
a 13 percent error from the actual sensible cooling energy for the experimental period.
Figure 3.29 shows the predicted and measured room air humidity ratio. The prediction is
fairly good and demonstrates the mass balance in EnergyPlus. A large source of error in
this comparison is the measurement itself. The humidity measurements are taken from
the HVAC relative humidity probes. The corresponding room air temperature is then
4000
35003000-
l0
2500
'S 2000C)
0
.51500 -
1000-
..
Predicted
Measured
5000 __
10/17
10/18
10/19
10/20
Date
10/21
10/22
10/23
Figure 3.27 Predicted and measured heating coil load
10/24
600
0
0
0
500
400300-
C
200
C
1000-10/17
10/18
Date
10/19
Figure 3.28 Predicted and measured cooling coil sensible load
0+
10/17
10/18
10/19
10/20
10/21
10/22
10/23
10/24
Date
Figure 3.29 Predicted and measured room air humidity ratio
used to calculate the humidity ratio. However, the room relative humidity measurement
comes from a single point in the return duct, rather than a more accurate room average.
Due to this large source of error, the prediction and measurement are considered
generally in agreement.
Figure 3.30 shows the predicted and measured room air temperatures. After reaching
steady state, the temperatures are in good agreement for the cooling mode, with errors
less than 0.30 C between October 18 and 19. For the heating mode, however, the error is
much larger, usually 3-4'C. Several factors may contribute to this error. Part of the error
may be due to incomplete mixing; the air supply is in the ceiling, which encourages
mixing when the supply air is cold, but when then supply air hot, it may tend to stay in
31.0
29.0
27.0
-
0~ 25.0
23.0
0-21.0
E
...Predicted
19.0
-- Measured
17.0
15.0
10/17
10/18
10/19
10/20
10/21
10/22
10/23
10/24
Date
Figure 3.30 Predicted and measured room air temperatures
the upper portion of the room. In addition, the construction details of the chamber are not
known with great accuracy, which can affect the accuracy of the heat balance calculation.
Finally, the temperatures on the outside of the chamber floor and north wall could not be
controlled or measured, because these surfaces are outside of the laboratory. This
unknown could also affect the heat balance calculation.
3.5.5 Conclusions
The results of the MIT test chamber study are mixed. EnergyPlus coil load predictions
are fairly accurate, although they tend to be better for electric coils, which more closely
reach the perfect control assumed in EnergyPlus. The mass balance calculations are also
accurate. However, although the temperature predictions are good for the cooling mode,
in general, the results of this experiment cannot be used to provide a good evaluation of
EnergyPlus temperature predictions because of numerous experimental unknowns. In
particular, the room construction is not known in the detail necessary to provide an
accurate energy simulation validation, and not all variables could be controlled or
measured. Fortunately, a limited comparison between measured and predicted
temperatures has also been performed for Building A; these results are presented in
Chapter 6.
Chapter 4: EnergyPlus Modifications
4.1 Introduction
Several modifications to the EnergyPlus vi.0 program code were necessary in order to
appropriately model the systems considered in this study. Some of these changes were
necessary to implement new physical models, while others were simple bug fixes or
changes to information flow.
The most complex change was the addition of a model for the displacement ventilation
vertical temperature gradient. EnergyPlus assumes the room air is well mixed and at a
uniform temperature, so this modification must be made in order to model displacement
ventilation. New, improved methods for calculating rates of interzone air mixing and
outside air infiltration (used for natural ventilation) were also added. Changes were made
to the plant loop simulation to allow more complex loop configurations. The baseboard
heater model was modified to allow it to also be used to model a chilled beam. Finally,
the supply and return air path simulation was changed slightly to allow the supply and
return plenums to be modeled appropriately, and a bug in the economizer model was
fixed. Each of these changes will now be discussed in detail.
All changes were made to EnergyPlus v1.0b23. The structure and calling tree of
EnergyPlus are fairly complicated; an introduction is provided in the program
documentation (EnergyPlus 2001). Although an effort has been made to document and
present the most vital portions of the changes in Appendix A, the best record and
explanation of the program is the code itself. The complete program code for each
variation of EnergyPlus used in this study is included on the attached compact disc.
4.2 Displacement Ventilation
Modeling of the vertical temperature gradient is essential to a meaningful energy
simulation of a displacement ventilation system. The vertical temperature gradient
affects occupant comfort, convective heat transfer from zone surfaces, and the air
flowrate necessary to maintain comfort conditions. Many models of varying complexity
have been proposed to account for the temperature gradient. This study uses a relatively
simple model as a demonstration of the implementation of a nodal model into
EnergyPlus. If more detailed, exact models are available for the system being
considered, they could be used in place of this model.
The simple model implemented into EnergyPlus has three nodes: the air temperature near
the floor (Tf), at the head level (Th), and at the ceiling, or exhaust, level (Te). The supply
air temperature is assumed to rapidly rise as the air enters the space, due to heat gains
from the floor. The temperature gradient is assumed linear between each node, as shown
in Figure 4.1. The room height is H, the head level height is a, and the distance from the
head to the ceiling is b. The volume-weighted zone average temperature, Tz, is:
aTf+HTh+bT(
a Tf+Th b Th+T
TH
2
H
2
2H
Ts
Tf
T
Figure 4.1 Displacement ventilation three-node model
This model influences the energy simulation in several ways. Rather than using the mean
air temperature for convection calculations from each surface, the air temperature near
the surface is used: Tf for the floor, Tz for the walls, and Te for the ceiling. The air
system is no longer controlled to maintain the mean air temperature at the setpoint;
instead, the head temperature (Th) is maintained at the setpoint. Finally, the ankle-head
temperature difference, Th - Tf, is important to thermal comfort. Temperature differences
greater than 3*C will result in a large fraction of dissatisfied occupants.
An essential portion of the model is the prediction of Tf and Th. These temperatures are a
function of nearly every factor affecting the room air: the supply air temperature and
flowrate, convective gains from surfaces, and convective gains from internal loads.
Analytical formulas or empirical correlations can be used to predict these temperatures.
The nodal temperatures are often expressed as dimensionless temperatures, 9j, for each
node j = f, h, or z):
0 =j _(4.2)
jT, -T,
where T. is the supply air temperature.
Several researchers have shown that the dimensionless air temperature near the floor
decreases as the ventilation rate decreases (Yuan et al. 1999b). Mundt (1990) assumed
that convection from the floor raises the supply air temperature from T. to Tf and that
radiative heat transfer from the ceiling to the floor maintains the energy balance on the
floor surface. She then developed a simple analytical formula to calculate Or
f =
A
K-+-
(a,
ac,
(4.3)
+1
where:
A
ar
acf
= floor area
= radiative heat transfer coefficient from the ceiling to the floor
= convective heat transfer coefficient from the floor to the room air
This formula is used to calculate Of in this study. The heat transfer coefficients are
assumed constant, such that:
-+
Ccr
=
2.54
(4.4)
CCef
Yuan et al. (1999b) used CFD to create a database of displacement ventilation parameters
for a variety of spaces. Assuming some fraction of the convective loads is added to the
air between foot and heat level, they used the database to develop a correlation for the
temperature difference between head and foot:
(4.5)
Th -T, = 0.295Q,, + 0.132Q, +0.185Qex
Tb-sysCT
where
= heat from occupants, desks lamps, and equipment
= heat from overhead lighting
Qoe
Qi
= heat from the exterior wall and window surfaces and the transmitted solar
Qex
radiation
This expression is used to calculate the head-foot temperature difference in this study.
The nodal temperatures must be incorporated into the air heat balance equation. The
temperature at any node can easily be expressed in terms of Te, Ts, and Oj:
Tj = OjTe
+
(4.6)
(1- Ojfts
In addition, the exhaust temperature can be expressed in terms of Th, Ts, and Oh:
(4.7)
Te =T+ (Oh
Oh
These expressions are substituted into the heat balance equation in order to solve for the
exhaust air temperature or the predicted system load.
When heat transfer from surfaces to air at different temperatures is accounted for, the
heat balance equation becomes:
dT N
Nnl
uams
Q + E hAf(T,
-T,)+
Zd _=C 0
fC
dt
N,
Nwailsrace,
hw,,(Ts -Tz)+
h'eTA(T
-)
Asi-Te)
H(4.8)
+rhc(Ti
zTx)
+ rhffiifcp(Tff
-T)+
in p inf
z +
p z
sys
i=1
where:
Ns,
= sum of internal convective loads from people, computers, etc.
1
i=1
Nfloor surfaces
ShfAf,(T,,
i=1
-
Tf) = convective heat transfer from zone floor surfaces
Nwansurfaces
hA,(T, -T)
i=1
convective heat transfer from zone wall surfaces
-
Nceiingsufaces
JhcAc (Tj,- Te) = convective heat transfer from zone ceiling surfaces
i=1
Nzones
= heat transfer due to interzone air mixing
ijc,(Tz -T)
i=1
ri iafc,(Tinf
(
-T.)
= rihyc,(T
Cd T
dt
=
=
heat transfer due to infiltration
-T)
air system output
=
rate of energy storage in air
The zone temperature derivative is again calculated with a third order finite difference
approximation:
~
(11
~ (8t)- -T
dT
_
28t
2
6
dt ,
12__
8t3
-3T t +-T
3
T-
t
(4.9)
Z
where the terms are defined as in Eq. (1.4).
For use in the predictor-corrector method, the heat balance equation (4.8) must be put
into two forms: one form solves for the system output, the other for the exhaust
temperature. This is most easily done by separating the room and surface temperature
factors, so that:
Naoorsufaces
Nnoorsufaces
ZhfAf(T,
-Tf)
=
IhfAfTSi,
Nfloorstfaces
hfAfTf
-
(4.10)
i=1
i=1
i=1
The summation notation is now dropped for simplicity; all summations remain implied.
The heat balance equation can now be expressed as:
C, (11 T -3T2
8t
6 Z
T-38t =
_
2
3
+
hfAfTi -hfAfTf + hAWTi -hA.Tz +
hCAcT,j -hcATe + iicp(Ti
- Tz)+ rifc, (Ti., -Tz)+ Q
(4.11)
This can be rearranged to solve directly for the system load:
hfAfTi +hAT, +h.A.T,
Q,=s
68t+h.A.
+rhic,
+IrifcPTz
+hfAfTf + hcAcT, - +1iicTz +r hinf cTi.
+
3T- t - 3 TY28t
+Q.
+
ITI38tJ
(
Note that the last term in (4.12) does not depend on the zone air temperatures and
therefore need not be recalculated for each iteration of the predictor step.
To solve (4.11) for the exhaust temperature, dimensionless temperatures can used to
express all other temperatures as a function of the exhaust temperature. First, the system
output is expressed in terms of the supply and exhaust temperatures and (4.11) is
rearranged with simplified notation:
HAT+ hfAfTf + HAeT, =HAT
(4.13)
where
HA - 11C +hA,
6 8t
+riic, +ri15,,c,
HA , = hCA. + rii,,,c,
HAT = ii,,,cT,, + hAT
-
( 3T-ta _
+hAT,
-2 +
+ hcAcT + riticT. + rif.cTi., +
Q
1T 1
Tz can be expressed as the volume-weighted average of the nodal temperatures:
HAz(
2H
T + ITh+
2
2H
Te)+ hfAfTf + HA.Te =HAT
(4.14)
Finally, the nodal temperatures are expressed in terms of the exhaust temperature, supply
temperature, and dimensionless temperature, and the equation is solved for the exhaust
temperature:
HA +(,
-)T,
Te=
2
(Of
2
H
-+hfA) +(O,
f)
H
-1)T,
1
(4.15)
+O +-J
+ hfAfOf + HAc
H
Note that HAT, HAz, and HA, do not depend on the zone air temperatures and therefore
need not be recalculated for each iteration of the corrector step.
The predictor-correct method for coupling the air heat balance and the system output
becomes considerably more complex when this displacement ventilation model is used.
This is because of the linking of the dimensionless temperatures to the system output.
The required system output is predicted based on the dimensionless temperatures, but the
dimensionless temperatures partially depend on the system output. Therefore, some
iteration is required within both the predictor and corrector steps. The solution method
flows as follows:
1) Predict the needed system load, Q,,, necessary to maintain Th at the setpoint
temperature:
1a) Using Oh from the previous timestep or iteration, T, from the previous timestep,
and Th equal to the setpoint, calculate Te from (4.7)
lb) Use Te to calculate Tf from (4.6) using Of from the previous timestep or iteration,
then calculate Tz from (4.1)
Ic) Solve for
Q,,
using (4.12) and the temperatures found in la) and Ib)
1d) Solve for insyic, using Ts from the previous timestep
1e) Calculate new dimensionless temperatures Oh and Of from (4.3) and (4.5)
If) Evaluate each AOj from the previous iteration. If either is greater than 0.01,
iterate with Oj = 0.15 Oj,new + 0.85 Oj,old
2) Simulate the system and plant using this Q,,, as the demand to determine the system
capacity.
3) Calculate the actual zone temperature:
3a) Using Oh and Of from the previous timestep or iteration and the actual system
supply temperature and flowrate, calculate Te from (4.15)
3b) Calculate new dimensionless temperatures from (4.3) and (4.5)
3c) Evaluate each AO3 from the previous iteration. If either is greater than 0.001,
iterate with Oj = 0.5 Oj,new + 0.5 6j,old
4) Evaluate ATz and each AOj from the previous timestep. If ATz> 0.3 K or any AOj>
0.005, repeat the entire procedure with the system timestep halved.
The entire procedure is summarized in Figure 4.2.
Several notes should be made concerning this procedure. The system load is predicted
using the supply air temperature from the previous timestep. However, the supply air
temperature is nearly constant over time, especially if the system is never overloaded, so
the prediction is still valid. The relaxation factors and convergence criteria in step If and
3c were determined to provide a reasonably stable yet fast solution. Although the criteria
for change in dimensionless temperature in steps 3c and 4 appear very stringent, they
were found to be necessary to obtain a consistently stable solution. Finally, the addition.
of the test of the change in the dimensionless temperatures from the previous timestep
(step 4) helps to prevent the room conditions from changing too quickly. Again, the
criterion was chosen after some experimentation by the author.
This computational method has been successfully implemented in EnergyPlus. Nearly all
of the changes are made in the module ZoneTempPredictorCorrector.f90, which handles
the heart of the predictor-corrector calculation. Major features of the code are presented
in Appendix A. 1, and the complete code is included on the attached compact disc.
set Th = setpoint temperature
solve for T. using Oh from previous
iteration, T,, and T
0
Solve for Q,,
Use m,,c, T. and T, to
calculate new Oj values
each A0j from
rvious iteration
< 0.01?
no
<+yes
calculate actual system
supply capacity
use calculated system
parameters and O3values from
previous iteration to calculate
actual T.
A-
W
iterate with
O = 0.5 0j; a +
05
.- Oj,ne.
zonal models calculate actual 03
values from system parameters
and actual T,
0
U
each A03 from
previous iteration
< 0.001?
no
yes
from previous system
timestep, AT, <0.3*C,
A0; < 0.005?
no
yes
move to next system timestep
Figure 4.2 Displacement ventilation model predictor-corrector method flowchart
4.3 Airflow
Two changes were made to the indoor airflow models. The first allows for a simple
prediction of the air flowrate in the natural ventilation case, while the second allows for a
variable rate of interzone air mixing based on the temperature difference between zones.
4.3.1 Natural Ventilation
The natural ventilation model assumes pure cross ventilation from the windward to the
leeward side of the building. Figure 4.3 shows this model in schematic form. Outdoor
air enters the windward side of the building. Air from the zone on the windward side of
the building then flows into the zone on the leeward side of the building, and air from this
leeward zone is exhausted outside. The actual airflow can be modeled using EnergyPlus
infiltration and mixing models, but some method must be used to predict the airflow rate.
The simplest method of predicting the air flowrate through an open window is with a
large-opening correlation, where some fraction of the static pressure difference across the
opening is assumed to be converted to dynamic pressure (velocity):
Vair = CDA
2A
(4.16)
Pair
where:
CD
A
Ap
= volumetric air flowrate through window
= opening discharge coefficient, usually ~0.6
= effective opening area
= pressure difference between inside and outside
This expression has been used for the simple natural ventilation model in EnergyPlus.
The discharge coefficient and effective opening area are assumed constant; the
determination of these constants for Building A is discussed in Chapter 5. The pressure
on an exterior surface relative to atmospheric pressure at the same height can be
calculated using a pressure coefficient, Cp, where some fraction of the wind dynamic
pressure is assumed to be converted to static pressure:
Pextrelative
=C
2
pa
where Umet is the meteorological windspeed.
leeward side
windward side
Figure 4.3 Natural ventilation schematic
(4.17)
The total ventilation rate for each floor can be calculated based on the pressure difference
between the two primary facades, located on opposite sides of the building. The pressure
difference, Ap, used in Equation (4.16), can therefore be calculated using:
Ap =(C
1
2-C,Pair m
(4.18)
where C, and Cp,2 are the surface-averaged pressure coefficients on the two primary
facades. The pressure coefficients are not constant but vary with wind speed and
direction. Determination of pressure coefficients for many wind speeds and directions is
complex and time-consuming. For this study, the pressure coefficient has been
determined for eight wind directions and assumed constant with wind speed for each
direction. The pressure coefficient for any wind direction is determined by linear
interpolation between two known pressure coefficients. The determination of pressure
coefficients for Building A is discussed in Chapter 5.
All of the changes necessary to implement this natural ventilation model were made in
the module HeatBalanceAirManager.f90. Major features of the code are presented in
Appendix A.2, and the complete code is included on the attached compact disc.
4.3.2 Interzone Mixing
A simple change to the EnergyPlus "CrossMixing" model was made to allow for a
slightly more sophisticated interzone mixing model. This interzone mixing model is used
to model mixing between the perimeter and central zones for each half-floor plate. The
original CrossMixing object allows for interzone air mixing at a constant rate that can be
modified by a schedule value. A temperature difference can be specified so that the
crossmixing is only active if the temperature difference is greater than some value.
In the new model, the crossmixing rate is assume to vary linearly between zero at a
temperature difference of zero and the design level at some specified temperature
difference, and remain constant at the design level for larger temperature differences:
A
IAT
crossmixing
=
Tdesign
.,g
'design1
design , I
AT
(4.19)
> ATdesign
where AT is the temperature difference between the zones. The implementation of this
model is very straightforward and is presented in Appendix A.2.
4.4 Plant Loops
EnergyPlus vl.0 only allows very simple plant (water-side) loops. The primary
restriction on plant loops is that only one set of demand side components may be placed
in parallel, because only one splitter and one mixer are allowed on the demand side.
Figure 4.4 shows the basic plant loop configuration. As many components can be placed
in parallel as desired. However, if another set of parallel components needed to be placed
after the demand-side mixer, this could not be done. Two new components have been
created to handle this need: crossover pipes and controlled crossover pipes.
4.4.1 Crossover Pipes
The crossover pipe allows more than one set of components in parallel by creating
multiple plant loops that exchange information. A crossover pipe is both a demand and
supply side component. It performs a very simple function: the demand side inlet
information is passed to the supply side outlet information, and the supply side inlet
information is passed to the demand side outlet information. This is illustrated in Figure
4.5. The primary plant loop is essentially the same as that in Figure 4.3, except that the
crossover pipe is placed after the demand-side mixer. In the secondary plant loop, the
only supply component is the crossover pipe and there is no supply bypass. Hence, the
demand side of the secondary loop receives the fluid that leaves the demand side mixer of
the primary loop, but all of the flow then returns to the supply side of the primary loop.
Figure 4.4 EnergyPlus plant loop configuration
Primary Plant Secondary Plant
Loop
Loop
Figure 4.5 Plant loop configuration with crossover pipe
The crossover pipe is specified as a new object in EnergyPlus that the user inputs in the
same manner as any other plant component. The implementation of this object is
presented in Appendix A.3.1.
4.4.2 Controlled Crossover Pipes
If the basic crossover pipe is used, there is no control over the temperature of the fluid in
the secondary plant loop. In order to allow for secondary plant loops that are maintained
at some fixed temperature setpoint, controlled crossover pipes were developed. A
controlled crossover pipe is essentially the same as a crossover pipe, except that it
controls the amount of flow passing through itself such that the secondary loop is
maintained at its setpoint temperature. Hence, the controlled crossover pipe must be used
in conjunction with a bypass. Figure 4.6 shows the basic configuration of a controlled
crossover pipe. Note that the controlled crossover pipe may be placed in parallel with
other components, but it may not be placed in series after a mixer (as with the crossover
pipe), because this would require a second splitter and mixer in order to accommodate the
bypass.
The controlled crossover pipe operates by calculating the flowrate necessary to maintain
the secondary loop at setpoint temperature. The cooling or heating load is determined
from the secondary loop mass flowrate and the temperature difference between the
crossover pipe inlet and the setpoint:
Qrequired
iicondary c
Primary Plant
Loop
(4.20)
- Tn
(ins
T
Secondary Plant
Loop
a
0
r
ontrolledcrossover pipe
|
I
Figure 4.6 Plant loop configuration with controlled crossover pipe
The required crossover pipe mass flowrate is the determined from this load and the
temperature difference between the crossover pipe inlets:
m
-
crossover pipe
-
Qrequired
c T
-
Ti
(4.21)
The remainder of the flow will be routed through the bypass by the EnergyPlus flow
resolver. The implementation of the controlled crossover pipes is presented in Appendix
A.3.2.
4.5 Baseboard Heater
Both the chilled beams and trench heaters in Building A were modeled using the
BASEBOARD HEATER:WATER:CONVECTIVE object. This component is
essentially a natural convection driven water-air heat exchanger located within the zone,
and can actually perform heating or cooling even though it is labeled as a heater.
However, some small changes to the model were necessary in order to fix bugs in the
existing model and control the component correctly.
Two bug fixes were performed: in the original code, the water mass flowrate was not set
to zero if the component was not active, and the calculated flowrate was never stored in
the Baseboard data structure. In addition, the original code set the air mass flowrate
constant and equal to a constant convective airflow speed (0.5 m/s), which is clearly an
error, because the total flowrate depends on both the speed and the cross sectional area.
In the new code, the air flowrate is still assumed constant, but it has been multiplied by
the density of air and the approximate area of the trench heaters and chilled beams for
Building A to yield a flowrate (5.4 kg/s). Finally, new control conditions appropriate to
Building A have been implemented. The water mass flowrate is set to zero if the air
temperature is between 20 and 25*C, allowing for a deadband where no chilled beams or
trench heaters are active. The water flowrate is also set to zero if both the water and air
temperatures are less than 20*C; this prevents a chilled beam from attempting to heat cold
air. The implementation of these bug fixes is presented in Appendix A.4.
4.6 Air System
Several small changes were made to the air system simulation in order to fix bugs and
eliminate restrictions originally imposed in EnergyPlus. Each of these will be briefly
discussed.
The original EnergyPlus code only allows for up to a three-deck system. This means that
there can only be up to three supply air outlets and return air inlets that connect the air
handling system and the distribution system in the building. However, six inlets and
outlets were found to be necessary to simulate the system in Building A appropriately.
The air system input processing code was easily modified to allow for as many inlets and
outlet as desired.
The author also found that although the data structure for air system mixers was in place,
there was no code to actually simulate the mixer. The addition of this calculation was
straightforward. The mixer outlet properties are determined by summing the product of
the mass flowrate and each property over all mixer inlets, and then dividing by the total
mass flowrate.
Several bugs relating to the simulation of the supply plenum and the economizer were
also discovered but easily fixed. The implementation of all of the above changes to the
air system is discussed in Appendix A.5.
4.7 New EnergyPlus Code
All of the changes discussed above have been successfully implemented in EnergyPlus.
Important portions of the actual code are presented in Appendix A. The complete
program code is very large - 121 modules totaling 7.3 MB. However, most of the
changes are concentrated in small segments of code, so that less than 10 modules have
actually been modified.
Two primary types of changes have been made: changes to calculations and changes to
inputs. Calculation changes simply consist of changing equations and affect very few
lines of code, and in the case of the displacement ventilation model, involve the addition
of new iteration loops. Input changes are also fairly simple because a library of input
processing routines already exists within EnergyPlus. Some input changes involve the
creation of new input objects, which must also be defined in the EnergyPlus input data
dictionary file (EnergyPlus.idd).
Three final versions of EnergyPlus were used in this study. The first, Crossover.exe,
includes all changes discussed above except for the displacement ventilation and natural
ventilation models. This version was used to simulate the existing building systems and
the VAV system. The second, DispVent.exe, also includes the displacement ventilation
model and was used to simulate the displacement ventilation system. Finally,
NatVent.exe only has the changes necessary to simulate natural ventilation, and was use
to simulate the natural ventilation system. Complete source code, compiled executables,
and input data dictionaries for each of these EnergyPlus versions are included on the
attached compact disc.
Chapter 5: Building A Models
Several EnergyPlus models were created to represent Building A and the various systems
considered in this study. These models are all based on the same physical configuration
of Building A, with only the mechanical systems changed. This chapter first presents the
basic Building A model, followed by the models of the various mechanical systems
studied. All EnergyPlus input files (.idf) are included on the attached compact disc.
5.1 Basic Building Model
5.1.1 Building A Description
Building A is located on the BP Sunbury campus, Sunbury-on-Thames, England, about
20 miles west-southwest of central London. The buildin has three floors which all open
to a central atrium. The net internal floor area is 4800 m . Note that throughout this
study the floors are referred to by the European numbering convention: ground, first,
second (G, 1, 2). Figure 5.1 shows a general section of the building. The building fagade
is nearly 100 percent glazing. The clerestory level provides daylight to the atrium.
Central mechanical equipment is located on the upper roof level, above the atrium.
Figure 5.2 shows the first floor plan. Each floor has two major zones: north and south,
which are separated by the atrium. Each of these zones will be referred to as a half-floor
plate. The location and number of interior partitions varies from floor to floor. In
general, the floor plan is very open. Figure 5.3 shows a more detailed section, including
the location of supply and return plenums, trench heaters, and chilled beams.
Atrium
Figure 5.1 Building A section A-A
Figure 5.2 Building A first floor plan
Figure 5.3 Building A detailed section
In the EnergyPlus model, each floor has eight zones, four for each half-floor plate. The
atrium is not modeled because smoke air flow visualization tests showed little interaction
between the air in the atrium and the rest of the building, and the internal and external
gains and losses in the atrium are very small compared to the remainder of the building.
The four zones corresponding to each half-floor plate are the central zone, perimeter
zone, supply plenum, and return plenum.
The occupied space is divided into a central zone and a perimeter zone to account for the
perimeter heating and cooling system, and because direct solar gains would tend to be
concentrated on the floor of the perimeter zone, rather than spread across the floor of the
entire space. The perimeter zone was chosen to have depth of 3 m on all three sides of
the occupied space, as shown in Figure 5.4. This distance was chosen because it is the
distance from the windows to the point where the dropped ceiling becomes level, and
because perimeter effects are unlikely to penetrate more than 3 m into the space. There is
some interaction between the perimeter and central zones that will be discussed later in
this chapter.
S
central zone
3
:3 m
perimet--
ne--------------3
--
perimeter zone
--
}3 m
-in
-----------------------------
54m
Figure 5.4 Definition of central and perimeter zones for each half-floor plate
Figure 5.5 Overall model geometry and detail of supply and return plenums
For each half-floor plate, the supply plenum is a single zone (54 x 15 m) located beneath
both the central and perimeter zones. Similarly, the return plenum is a single zone
located above both the central and perimeter zones. The structural concrete floor slab
separates the return plenum for one floor from the supply plenum for the floor above.
Figure 5.5 shows the overall geometry of the EnergyPlus model and a cutaway detail of
the supply and return plenum geometry.
5.1.2 Simulation Parameters
There are a number of general simulation parameters that must be specified for any
EnergyPlus model. Most of these involve a choice between different available models.
In general, the most detailed models available in EnergyPlus have been chosen. Table
5.1 shows the general simulation parameters for all simulations in this study. The
detailed convection algorithms use correlations to determine the convection coefficient,
rather than assuming a simple constant convection coefficient. The CTF solution
algorithm uses only conduction transfer functions for the surface heat balance
calculations, rather than the moisture balance algorithms also available.
The full exterior solar distribution means that detailed shadowing calculations are
performed for the building exterior, but all radiant solar energy transmitted to a zone is
assumed to strike the floor. EnergyPlus does include a full interior and exterior solar
distribution model, which projects the appropriate amount of transmitted solar energy
onto each internal surface. This model was not used because the only internal surfaces in
the model are floors and ceilings, so all of the radiant energy would strike the floor, and
because this model does not work with concave zones such as the perimeter zone.
Table 5.1 EnergyPlus simulation parameters
Parameter
Solar Distribution
Timesteps in Hour
Inside Convection Algorithm
Outside Convection Algorithm
Sky Radiance Distribution
Solution Algorithm
Setting Used
FullExterior
4
Detailed
Detailed
Anisotropic
CTF
5.1.3 Materials
There are relatively few materials in the building model, considering the complexity of
the building geometry. Table 5.2 lists the different constructions used and the properties
of their component layers. Note that complete data was not available for some materials,
such as the spandrel panel insulation and the raised floor carpet, so they were modeled as
thermal resistances with no mass. However, these materials have little thermal mass so
this does not adversely affect the model. The floor slab, which is uninsulated, is used
between the floors G and 1 and between floors 1 and 2. The ground slab is in contact
with the ground, underneath floor G, while the roof slab is above floor 2. Note that the
thermal and solar absorptivity of the plywood in the raised floor is very low because the
raised floor is backed with a foil lining.
Table 5.2 Building A material properties (outside to inside)
thickness
specific
thermal
density
W/mK
kg/m 3
J/kgK
conductivity
heat
I
thermal
solar
a
a
thermal
Jresistance
m2K/W
Floor Slab
Lightweight Concrete
0.11
1.10
1950
840
0.9
0.5
--
Ground Floor Slab
Polyurethane
Concrete
0.08
0.275
0.024
1.40
24
2400
159
900
0.85
0.9
0.9
0.5
--
Roof Slab
Polyurethane
Lightweight Concrete
0.08
0.12
0.024
1.10
24
1950
159
840
0.85
0.9
0.9
0.5
--
Spandrel
Aluminum Spandrel
Panel
0.003
200
2700
900
0.82
0.14
--
Airgap
0.015
--
--
--
--
--
0.16
--
-
Rockwool
0.1
--
--
--
0.87
0.87
2.7
Raised Floor
Plywood
0.031
0.12
540
1210
0.07
0.15
--
Carpet w/Rubber Pad
--
--
--
--
0.85
0.85
0.22
Dropped (Chilled)
Ceiling
Ceiling Insulation
Ceiling Panel
0.025
0.003
0.038
150
45
2700
710
870
0.07
0.15
--
0.85
0.9
--
The exterior of building A is nearly 100% glazing. The aluminum spandrel panels cover
the perimeter of the supply and return plenums, but the remainder of the fagade is
completely glazed. EnergyPlus does not allow a surface to be specified as all glazing;
windows must be specified as a subsurface of an existing surface. To model the glazings,
an inert surface with zero absorptivity and very high resistance was created, and a
window subsurface was then created which covered nearly the entire area of this surface.
The type of glazing is different for each fagade. Glass properties were selected from the
EnergyPlus database according to the type of glass specified for each glazing. Two types
of glass are used: green and clear, low-emissivity. Different thicknesses of these glass
types are used on the various facades. Table 5.3 shows the properties of the glass and the
various glazing constructions. All glazings are filled with air. The standard EnergyPlus
air properties (Table 3.2) were used. Curtain wall mullions were not included in the
glazing model. Building A uses an thermally broken curtain wall system, and the spacing
between mullions is fairly large (~I m), so their effect on conduction through the glazing
is fairly small.
Table 5.3 Building A glazing properties and constructions (outside to inside)
Property
Solar transmittance at
normal incidence
Solar reflectance at normal
incidence: outer side
Solar reflectance at normal
incidence: inner side
IR transmittance at normal
incidence
IR hemispherical
emissivity: outer side
IR hemispherical
emissivity: inner side
Conductivity (W/mK)
Green
Low-e Clear
0.487
0.6
0.056
0.031
0.056
0.17
0
0
0.84
0.1
0.84
0.84
0.9
0.9
Thickness
(mm)
South Facade
Green Glass
Airgap
Clear Low-e Glass
North Faeade
Green Glass
Airgap
Clear Low-e Glass
East/West Facades
Green Glass
Airgap
-Clear Low-e Glass
6
16
10
6
20
6
6
16
6
5.1.4 Shading
Building A has both interior and exterior shading. The exterior shading system consists
of arrays of solar panels on the southern fagade, roof overhangs, and walkways at each
floor level around the perimeter of the building. The interior shading system consists of
manually operated venetian blinds. Figure 5.6 shows a section of the shading system at
the second floor.
The walkways, roof overhangs, and solar panels were modeled as EnergyPlus
Surface:Shading:Attached objects. The shading surfaces are visible as the pink surfaces
in Figure 5.5. The walkways are metal gratings that are not completely opaque.
However, the grating is fairly dense, so the walkways were modeled as opaque surfaces.
Figure 5.6 Section of shading system at second floor
Pir
Figure 5.7 Detached shading surfaces
In addition to the shading system, Building A is also shaded by surrounding buildings.
The facades of buildings directly south, west, or east of Building A were modeled as
EnergyPlus Surface:Shading:Detached objects. Figure 5.7 shows the geometry of the
detached shading surfaces. The only difference between attached and detached shading
surfaces is the method used to input their geometry. Both models shade all direct and
diffuse solar radiation and do not account for reflected radiation.
Internal blinds present a modeling challenge because the occupants can operate them.
Close observations of blind positions in each zone at various times were made for several
days during a visit to the building in January 2002. These observations provided a
general idea of how the current occupants operate the blinds. Some blinds were observed
always shut, others always opened, and others open or shut depending on the amount of
glare on the window.
EnergyPlus WINDOWSHADINGCONTROL objects were used to control the blinds.
For zones where the blinds were always shut or always open, they were scheduled as so.
For zones where the blind position depended on the amount of solar glare, the trigger
SolarOnWindow was used to specify the blinds as open unless the amount of solar
radiation striking the window exceeds some threshold value, in which case the blinds are
closed. Thresholds of 100, 150, or 250 W/m 2 were chosen based on the observed
occupant sensitivity to glare and experimentation with EnergyPlus. Open blinds were
assumed to have no effect on the window. Closed blind properties were chosen from
listings for light colored venetian blinds in the ASHRAE Fundamentals (ASHRAE 2001)
and are listed in Table 5.4.
Table 5.4 Closed venetian blind optical properties
Property
solar transmittance
solar reflectance
thermal hemispherical emissivity
thermal transmittance
Value
0.05
0.55
0.85
0.05
5.1.5 Exterior Environment
EnergyPlus receives nearly all exterior environment information on an hourly basis from
the input weather file. The weather file used for all simulations was
UKEnglandLondonGatwick.epw, an EnergyPlus weather file derived from the
ASHRAE International Weather for Energy Calculations (IWEC) dataset for Gatwick
airport. The EnergyPlus documentation (EnergyPlus 2001) has more information on the
weather file format and contents. This weather file is included on the attached compact
disc.
In addition to the weather file, the Building A model interacts with the environment via
ground temperatures. Typical U.K. ground temperatures used in all simulations were
obtained from energy simulation researchers in the U.K. (Hand 2002). Table 5.5 shows
the U.K. ground temperatures at a depth of 1.5 m.
Table 5.5 Ground temperatures at 1.5 m depth
Month
Temp. (*C)
Jan
7.3
Feb
6.5
Mar
6.6
Apr
7.3
May
9.1
Jun
11.4
Jul
13.5
Aug Sep
14.3 14.1
Oct Nov
11.7 10.7
Dec
8.7
5.1.6 Internal Gains
Internal gains for Building A have three sources: lighting, equipment, and people.
Characterizing these loads accurately for any energy simulation can be challenging
because they are dependent on occupant behavior and the energy use of many electrical
devices (computers, printers, copiers, etc.). However, Building A was instrumented for
long-term energy monitoring in January 2002, providing the unusual opportunity to
obtain lighting and equipment gains from measurements.
The electrical distribution system includes individual distribution boards for plugs
(busbars in the supply plenum) and lights for each half-floor plate. These circuits have
been instrumented with data loggers that record the electrical power at fifteen-minute
intervals. Data from these loggers for February 2002 (1/28/02 - 3/06/02) was averaged
to obtain hourly load profiles for a typical weekday and weekend day for each half-floor
plate.
Figures 5.8 - 5.11 show the hourly electrical equipment and lighting loads for the typical
weekday and weekend. All loads have been normalized based on an internal floor area
12 E 10-
.
-J
C
64
0
2
0:
0:00
3:00
6:00
9:00 12:00 15:00 18:00 21:00
Time
Figure 5.8 Typical weekday
electrical equipment load profile
0:00
0
0:00
3:00
6:00
9:00
I
12:00 15:00 18:00 21:00
Time
Figure 5.9 Typical weekday
overhead lighting load profile
0:00
4.0
----
3.5
3.0
SOUTH G
NORTH G
-SOUTH
1
-
1
-NORTH
-SOUTH
c
4.0
---
3,5
SOUTH G
NORTH
-SOUTH
2
30
NORTH 2
1
-NORTH 1
-SOUTH
2 5-E
25-.5
01.0
.5
11.5
0.0
2
-NORTH
2
0*
'
_
_
0.0
0.0
0:00
3:00
6:00
9:00 12:00 15:00 18:00 21:00 0:00
Time
Figure 5.10 Typical weekend
electrical equipment load profile
0.0
0:00
3:00
6:00
9:00 12:00 15:00 18:00 21:00
0:00
Time
Figure 5.11 Typical weekend
overhead lighting load profile
of 810 m2 for each half-floor plate. The equipment load varies greatly from zone to zone;
peak loads vary from less than 4 W/m 2 to nearly 8 W/m 2 . This is because the density of
people and their associated equipment varies greatly. The first floor is the least dense and
has the lowest equipment loads. The equipment load for every floor is considerably
lower than typical office equipment loads, which are often 10 - 15 W/m 2 (ASHRAE
2001). This is because many energy-saving types of equipment, such as flat-panel LCD
monitors, are used throughout the building. The baseline load, which is fairly constant
over the night and weekend, also varies from zone to zone. It is generally higher for the
northern zones because vending areas are on the northern side of the building.
The lighting load is more constant from zone to zone because the installed lighting
density is the same throughout the building. However, daylight-sensitive dimmers and
motion sensors control the lights and therefore create some load variation. The load
peaks at the end of the day as the sun sets and daylight disappears. The loads also
decrease slightly with increasing height because the upper floors receive more daylight.
This load profile was assumed constant throughout the year. The load profile would
actually vary over the year as the length of the day and amount of daylight change, but
annual measurements providing such a complete profile are not yet available.
The portion of equipment and lighting loads that is radiant versus convective must also be
specified. For the equipment load, this was determined based on the number of each type
of equipment and its approximate load and radiant/convective split given in the ASHRAE
Fundamentals (ASHRAE 2001). The estimated split was constant at 16% radiant, 84%
convective throughout the building. The overhead lights fixtures act as the inlet to the
return plenum, so that the lights are ventilated. The lights were modeled so that when the
mechanical system is active, the convective load from the lights only reaches the return
air and does not enter the space. The radiant/convective split used was the ASHRAE
value for recessed, vented fluorescent lights: 59% radiant/41% convective.
Finally, the equipment and lighting loads must be split between the central and perimeter
zones. The installed lighting is approximately 10% in the perimeter zone, with the
remainder in the central zone, so this was the division for the load in the model. The
fraction of the equipment load in each perimeter zone was estimated from a count of the
number of each type of equipment in the perimeter zone. The fraction varies from zone to
zone, from zero for the second floor to 21 percent for the first floor north.
The internal gains from people must also be modeled. A single person was assumed to
produced a constant total load (sensible and latent) of 115 W, based on ASHRAE
recommended values for light office work (ASHRAE 2001). The EnergyPlus PEOPLE
model automatically calculates the portion of this load that is latent. The sensible load
was assumed to be 58% radiant/42% convective. The maximum number of people in
each zone, shown in Table 5.6, was estimated by counting the number of desks in that
zone. Note that the density of people varies greatly; the second floor has far more people
than the other two floors. This maximum is multiplied by an hourly fractional office
occupancy schedule, shown in Figure 5.12. This schedule, which is assumed constant for
the entire building, was estimated from the electrical equipment load data by determining
the ratio of the current equipment load to the peak equipment load for each hour of the
day, where the baseline equipment load has been subtracted from both loads so that only
the portion affected by occupants is considered. This schedule is only used for
weekdays; the building is assumed to be unoccupied during weekends.
Table 5.6 Maximum number of people in each zone
Location
Centra Zone
South G
46
North G
39
South 1
42
North 1
43
South 2
North 2
75
79
9
17
12
7
10
6
Perimeter Zone
Max.
#
People
1.00
0.90
0.80
0.70
0.60
0.50
0.40
0.30
0.20
0.10
--
-
-
- .
-
0.00
0:00
3:00
6:00
9:00
12:00
15:00
18:00
21:00
Time
Figure 5.12 Fractional weekday office occupancy schedule
0:00
5.1.7 Interior Airflow
The only interior airflow included in the EnergyPlus model is the mixing between the
central and perimeter zones. Airflow visualizations performed with smoke pencils in
Building A showed that the interaction between the atrium and each half-floor plate is
minimal, so airflow between half-floor plates through the atrium was not modeled.
The mixing between the central and perimeter zones is difficult to characterize because
there is no physical barrier between these zones. However, the mixing would be minimal
when the zones are at the same temperature because there is no force to drive airflow.
Similarly, the mixing rate would become very large at large temperature gradients,
because strong buoyancy affects would drive the airflow. In this study, the mixing rate
has been assumed to vary linearly with the temperature difference between the zones, up
to some maximum mixing rate and temperature difference, beyond which the mixing rate
is constant. A maximum mixing rate of 3 m 3/s (14 ach based on the perimeter zone
volume) was assumed to occur at a temperature difference of 5 K. Beyond this
temperature difference, the mixing rate remains constant. The implementation of this
model in EnergyPlus is discussed in Chapter 4.
5.2 Existing Building Systems
The existing building system consists of an underfloor air distribution system with chilled
ceilings, perimeter trench heaters, and perimeter chilled beams. Fairly complex chilled
and hot water loops are used to add and remove heat to these components. Each part of
the system has been included in the EnergyPlus model and will be discussed in detail
below.
The building as it is currently operated has many obvious opportunities for energy
savings. Therefore, two basic models of the existing building were developed. The first,
referred to as the Existing Building model, attempts to model the building as it is actually
operated. The second, referred to as the Improved Operations model, changes the
operation of the existing building systems to take of these obvious opportunities for
energy savings.
5.2.1 Air System
The Building A air system consists of two air handling units that deliver fresh air through
two independent supply risers. The supply risers both deliver the air to all six supply
plenums serving the six half-floor plates. Air is exhausted through the return plenums to
two return risers that exhaust air through the air handling units. Because both air
handling units serve all six half-floor plates, they have been lumped together and
modeled as a single unit within EnergyPlus.
Air Handling Unit
Figure 5.13 shows a schematic of the air handling unit, which includes supply and return
fans, a preheat coil, economizer, cooling coil, and a reheat coil. The modeling of each of
these components will be discussed briefly.
exhaust
air
fan
fa<
return air
economizer
outdoor
air
[j
supplysupply
airi
fa0jspl
preheat
coil
cooling
coil
reheat
coil
Figure 5.13 Air handling system schematic
The preheat coil and economizer were both modeled within an Outside Air System. The
preheat coil was modeled as a Simple Water Heating Coil with a UA value of 10,000
W/K and a maximum water flowrate of 5.2 kg/s. Although the exact UA value of the coil
was not known, it only affects the model in extreme conditions when the coil cannot meet
the load. Because this condition is very unlikely, all coils were modeled as simple coils
with fixed UA values. UA values were chosen to be sufficiently large to prevent to coil
from being overloaded. All coil maximum water flowrates were taken from building
design drawings and were found to be sufficiently high to prevent any of the coils from
overloading during peak periods. The preheat coil was controlled to prevent the air
temperature after the coil from falling below 5'C.
The economizer was modeled as an Outside Air Mixer. In the actual building, it is
controlled such that 100% outdoor air is used for the entire day after 8 a.m., 50% outdoor
air from 6 - 8 a.m., and 100% return air before 6 a.m. This control scheme is used in the
Existing Building model. However, there are clearly cases where using 100% outside air
in the early morning would be beneficial, especially in the summer when cool morning
outside air can be used for free cooling. Therefore, in the Improved Operations model,
the economizer is allowed to vary the mix of outside and recirculated air to keep the
mixed air temperature equal to the supply air setpoint when possible. The conditions
above for minimum amounts of outside air are unchanged, but more outside air can be
used if necessary.
The supply fan was modeled as a Simple Variable Volume Fan with a pressure drop of
500 Pa and a constant efficiency of 0.6. These parameters were simply chosen as rough
estimates because no fan data was available. The only affect the fan has on the model is
a slight rise in temperature due to fan inefficiencies, so these estimates are acceptable.
This model was not used to estimate fan electrical consumption; this estimate is discussed
later in this chapter. The return fan has a very small effect on the system other than its
electrical consumption and was not included in the model.
The cooling coil was modeled as a Simple Water Cooling Coil with a UA value of 40,000
W/K, maximum water flowrate of 13.8 kg/s, and maximum relative humidity leaving the
coil 90%. This means that no detailed latent cooling capacity calculation are performed.
The reheat coil was modeled as a Simple Water Heating Coil with a UA value of 10,000
W/K and a maximum water flowrate of 5.6 kg/s. There are three control modes for these
two coils: heating, cooling, and dehumidification. If the mixed air humidity ratio is
above the humidity ratio setpoint, dehumidification is not needed. In this case either the
heating or cooling coil is controlled to maintain the supply air temperature at setpoint. If
dehumidification is needed, the cooling coil is controlled to provide dehumidification
such that the air leaving the coil has a humidity ratio equal to the setpoint. The reheat
coil is then controlled to heat the supply air to the setpoint temperature. The humidity
ratio setpoint is 9 g/kg, which corresponds to a dewpoint 12.4'C. This low dewpoint is
needed to prevent condensation on the chilled ceilings. The actual system is controlled
by the air dewpoint; in EnergyPlus this type of control must be based on the humidity
ratio. The supply air setpoint is 18.5'C after 8 a.m. and 21.5'C before 8 a.m.
Air Distribution
Upon leaving the air handling unit the air is distributed to six supply plenums, one for
each half-floor plate. The air in the supply plenum enters the space through numerous
circular floor diffusers. Figure 5.14 shows the type of diffuser used in Building A. The
air leaves the space and enters the return plenums through the ceiling-mounted ventilated
light fixtures. The air from the six return plenums then returns to the air handling unit.
This distribution is performed in two stages within EnergyPlus. The air leaving the air
handling unit is split into six different supply air paths, one for each supply plenum. The
air flows through the supply plenum, and then flows through a splitter that supplies both
the central and perimeter zones corresponding to that supply plenum. The same approach
is taken in reverse for the return air. Figure 5.15 shows this distribution system in
schematic form. Although the second splitter does not physically exist, it must be
included in EnergyPlus in order to supply air to both the central and perimeter zones from
the same plenum. The heat transfer processes within the supply plenum are modeled in
the same method as for any other zone. Note that EnergyPlus does not actually model
ducts and therefore does not account for heat gains within the duct system. Finally, note
that the current version of EnergyPlus requires a terminal distribution unit to be specified
after the second splitter. In this case, a VAV box was specified with a minimum airflow
rate equal to its maximum airflow rate and an electric heating coil with a capacity of 0 W.
Figure 5.14 Building A floor diffuser
it.
it
--
i
+returnplenum ,
perimeter'
zone
1
1k
central zone
J
4-
supply plenum
Figure 5.15 Air distribution schematic
The underfloor air system can potentially create a vertical temperature gradient that can
make the assumption of uniform room temperature inaccurate. However, this system
uses very large airflow rates that tend to make the temperature gradient very small.
When the chilled ceiling is active, it further destroys the temperature gradient.
Measurements made during January 2002 showed that the floor to ceiling temperature
difference is generally less than 2*C. This small gradient makes the assumption of
uniform room temperature reasonably accurate, especially compared to the much larger
gradients arising in a displacement ventilation system.
Flowrates
The supply flowrate for the entire system and each zone is constant. Supply rates for
each zone were determined from measurements taken during January 2002. Duct
traverses were performed within the supply and return risers at each floor level using a
hot-wire anemometer. The average velocity across the duct was used to determine the
supply air flowrate. This flowrate was then divided evenly between the perimeter and
central zones based on the ratio of their floor areas (1:2.5). Table 5.7 shows the supply
air rates used in the Existing Building model. The flowrates are very high for a system
Table 5.7 Air supply rates for Existing Building and Improved Operations models
Location
Air
S
tpiRaBui ln
Improved Operations
Air Supply Rate (m 3/S)
South G North G
South 1
North 1
South 2
North 2
2.23
2.23
2.41
2.49
3.67
2.77
0.56
0.49
0.52
0.53
0.85
0.89
that is intended to provide only fresh air. The total system flowrate is approximately 40
L/s/person based on the maximum number of occupants shown in Table 5.6. The
Improved Operations model uses more reasonable flowrates for each zone of 10
L/s/person, a widely recognized standard for fresh air supply rate. The perimeter air
supply rate is held constant for each zone at 0.1 m3/s in the Improved Operations model.
The entire system is off during the night and weekends. Electrical monitoring showed
that the building system was generally turning on at 4 a.m. and off at 7 p.m. on weekdays.
This operation schedule was used in the Existing Building model. In the Improved
Operations model, the system turns on at a more reasonable time of 6 a.m. and still turns
off at 7 p.m.
5.2.2 Radiant and Convective Units
Chilled Ceilings
The entire ceiling of the central zone of each half-floor plate is composed of chilled
ceiling panels. Figure 5.16 shows a cutaway view of a typical chilled ceiling panel. The
materials composing the chilled ceiling are given in Table 5.2. The chilled ceiling
consists of metal chilled water tubing bonded to a thin metal panel. The back side of the
ceiling panel is lined with insulation.
Each central zone was modeled with a single Hydronic Low Temperature Radiant System
to represent the chilled ceilings. This EnergyPlus model is intended to represent chilled
ceilings, chilled concrete slabs, radiant concrete slabs, or any other surface that has an
embedded heat source. Geometric inputs for this model were estimated from building
design drawings. The tubing inside diameter was estimated to be 15 mm and the total
tubing length per half-floor plate was 2300 m for southern zones and 2700 m for northern
zones. Maximum cold water flowrates were also taken building design drawings and
vary from 2.8 to 3.6 kg/s.
The chilled ceiling control is based on the zone mean air temperature. The control varies
the flowrate linearly from zero flow at 24.5*C to the maximum flowrate at 25.5*C. This
corresponds to a 1 C throttling range centered around 25*C. This control approximates
the actual building controls, in which the panel flow is not variable, but simply on or off.
Flow to the panel is turned on if the proportional-integral temperature error exceeds some
set value. EnergyPlus does not have provisions for this type of control.
Figure 5.16 Chilled ceiling panel cutaway view
ChilledBeams
In the perimeter zone, the chilled ceiling is replaced with chilled beams, which have a
higher cooling capacity. The chilled beam is essentially a long and narrow finned tube
heat exchanger. Figure 5.17 shows a typical chilled beam. The chilled beam is mounted
near the ceiling. Hot air rises and passes through the chilled beam, where it is cooled and
falls downward.
Each perimeter zone was modeled with a single Convective Water Baseboard Heater to
represent the chilled beams. This model was modified slightly to allow it to be used to
model a chilled beam, as discussed in Chapter 4. The chilled beams were estimated to
have a UA value of 600 W/K. As with the heating and cooling coils, this value does not
have a large affect on the model because it only affects the performance of the beam
when it is fully loaded, which requires extreme conditions. Maximum cold water
flowrates were taken from building design drawings and vary from 1.7 to 2.2 kg/s.
The chilled beams are controlled to maintain a setpoint temperature of 25'C. If the
temperature is below 25'C, there is no flow to the chilled beam. When the temperature
reaches 25'C, the flow through the chilled beam is varied to maintain this setpoint. As
with the chilled ceilings, this control approximates the actual building controls, in which
the flow is not variable, but simply on or off. Flow to the beam is turned on if the
proportional-integral temperature error exceeds some set value. EnergyPlus does not
have provisions for this type of control.
Figure 5.17 Chilled beam
Trench Heaters
The perimeter zones also have trench heaters to overcome heat losses due to window
conduction during the winter. Figure 5.18 shows a schematic of the operation of the
trench heaters. The heater is placed in a trench at the base of the window. Air is cooled
as it comes in contact with the window and flows downward into the trench. It then rises
upwards after it is heated.
Each perimeter zone was modeled with a single Convective Water Baseboard Heater to
represent the trench heaters. The trench heaters were estimated to have a UA value of
400 W/K. As with the chilled beams, this value does not have a strong affect on the
Figure 5.18 Trench heater schematic
model. Maximum hot water flowrates were taken from building design drawings and
vary from 0.7 to 1.4 kg/s.
The trench heaters are controlled to maintain a setpoint temperature of 20*C. If the
temperature is above 20'C, there is no flow to the trench heater. When the temperature
falls to 20'C, the flow through the trench heater is varied to maintain this setpoint. This
is very similar to the actual building controls, because unlike the chilled ceilings and
beams, the trench heaters do have three-way valves allowing variable flow.
5.2.3 Plant Loops
The Building A plant loops consist of the hot water loop, serving the heating coils and
trench heaters, and the cold water loop, serving the cooling coil, chilled ceilings, and
chilled beams. As with the air system, the plant loop simulation does not account for
pressure drops or heat losses and gains between components.
Hot Water Loop
Figure 5.19 shows a schematic of the hot water loop. The only heat source is the boiler.
From the boiler, water can flow through a bypass, either of the heating coils, or into the
secondary loop. The secondary loop is maintained at a lower temperature than the
primary loop and is fed with water from the primary loop as necessary to maintain this
temperature. The secondary loop supplies the trench heaters.
The primary and secondary loops are modeled as two independent loops connected by a
heating crossover pipe. The heating crossover pipe is a new EnergyPlus component that,
combined with a bypass, represents the area within the dashed box in Figure 5.19. Its
operation is discussed in Chapter 4. Both loops operate at constant flowrate taken from
the building design drawings. The primary loop flowrate is 18.5 L/s; the secondary loop
flowrate is 7.28 L/s. The primary loop is maintained at a setpoint temperature of 80'C;
the secondary loop is maintained at a setpoint temperature of 520 C.
The secondary loop in the actual building actually has a variable setpoint temperature
dependent on the outside air temperature. This type of setpoint control is not possible in
Primary Loop
Secondary Loop
U,,
o
boiler
U,
Cc
I
0
0
heating crossover pipe
Figure 5.19 Hot water loop schematic
EnergyPlus. Exclusion of this variable setpoint from the model does not have a large
affect on the energy consumption of the system and is therefore acceptable.
Detailed pump performance information was not available and pump energy consumption
was therefore excluded from this study. This is acceptable because the pumping energy
is small compared to the fan, chiller, and boiler energy, and because the pumping energy
would be similar for the various systems considered. Pumps were included in the
EnergyPlus models, but only because flow will not move through the loop without them.
The secondary loop flowrate varies in the actual building in order to save pumping
energy. However, because the pumping energy is not considered, the use of a constant
flowrate is acceptable for this study.
Finally, the boiler was modeled as a simple gas-fired boiler with a constant efficiency of
0.95. Detailed performance data for the boiler was unavailable. The actual building uses
two boilers in parallel; they have been lumped together for this study.
Cold Water Loop
Figure 5.20 shows a schematic of the cold water loop. The only cooling source is the
chiller. From the chiller, water can flow through a bypass, or the cooling coil. The
warmer water exiting the cooling coil and bypass then flows through either another
bypass or the secondary loop. The secondary loop is maintained at a higher temperature
than the primary loop and is fed with water from the primary loop as necessary to
maintain this temperature. The secondary loop supplies the chilled ceilings and beams.
The primary and secondary loops are modeled as two independent loops connected by a
cooling crossover pipe, similar to the hot water loop. However, because the cooling
crossover pipe and its bypass are in series with the cooling coil, a second mixer and
splitter are required. Because a second mixer and splitter are not allowed by EnergyPlus,
a crossover pipe is used to create a third intermediate loop which has no components
other than the cooling crossover pipe and its bypass. The operation of the crossover pipe
is described in Chapter 4.
Primary Loop
Secondary Loop
cooling crossover pipe
Intermediate
Loop
Figure 5.20 Cold water loop schematic
As with the hot water loops, both cold water loops operate at constant flowrate taken
from the building design drawings. The primary loop flowrate is 33.5 L/s; the secondary
loop flowrate is 31.2 L/s. The primary loop is maintained at a setpoint temperature of
6*C; the secondary loop is maintained at a setpoint temperature of 15'C.
The primary loop in the actual building has a variable setpoint temperature dependent on
the need for dehumidification. The setpoint is 6*C when dehumidification is needed and
10*C at other times. This type of setpoint control is not possible in EnergyPlus.
Exclusion of this variable setpoint from the model does not have a large affect on the
energy consumption of the system and is therefore acceptable. This setpoint variation
was included in the calculation of chiller energy consumption, discussed later in this
chapter.
As with the hot water loops, pumps were included in the EnergyPlus models only
because flow will not move through the loop without them. The secondary loop
flowrate varies in the actual building in order to save pumping energy. However, because
the pumping energy is not considered, the use of a constant flowrate is acceptable for this
study.
Building A uses two 500 kW air-cooled chillers which are lumped together as one for this
simulation. EnergyPlus does not include an air-cooled chiller model, so the cooling was
provided by Purchased Chilled Water with a 1 MW nominal capacity. Calculation of the
electrical energy consumption of the chillers is discussed later in this chapter.
5.3 Alternative Building Systems
Several variations of the existing building model presented above were developed to
model alternative building systems. These systems include the VAV system,
displacement ventilation system, natural ventilation, and any of these system modeled
with night cooling. Every system is a variation on the Improved Operations model
presented above. Only the portions of the model discussed below have been changed
from the basic Improved Operations model.
5.3.1 Displacement Ventilation System
The displacement ventilation model is very similar to the Improved Operations model.
Three major changes are made: the chilled ceilings are removed, the air system has a
variable airflow rate, and the displacement ventilation vertical temperature gradient
model is used.
The chilled ceilings are removed from the model because the displacement ventilation
system is used to remove 100% of the cooling load in the central zone. Although the
chilled ceiling model is removed, the insulated ceiling panels remain in place with the
same physical properties as the chilled ceiling. The model simply treats the chilled
ceilings as if they are always off.
Because the ventilation system is used to remove the entire cooling load in the central
zone, it must have a variable airflow rate in order to meet the additional load that was
removed by the chilled ceilings. The minimum airflow rate for each zone is the same as
in the Improved Operations model, but the airflow rate in the central zones can increase if
necessary to maintain the setpoint temperature. The supply air rate in the perimeter zones
remains at a fixed minimum; additional loads in these zones are still handled by chilled
beams and trench heaters. The minimum amount of outside air is also unchanged. When
the airflow rate is increased above the minimum rate, the economizer mixes recirculated
air and outside air to maintain the mixed air temperature at the supply air setpoint if
possible.
The supply air temperature for a displacement ventilation system is normally about 18*C.
However, this system uses a supply plenum in which there is considerable heat gain,
especially during summer months. The air temperature can rise as much as 3*C within
the supply plenum due to heat gains from the raised floor. Therefore, a lower supply air
setpoint (for the air leaving the air handling unit and entering the supply plenum) was
used during warmer months so that the air leaving the supply plenum would be near
18*C. The supply air setpoint was scheduled to reset annually: April through October,
the supply air setpoint was 15*C for the entire day; November through March, the supply
air setpoint was 18*C after 8 a.m. and 20*C before 8 a.m.
Finally, the displacement ventilation vertical temperature gradient model was for the
central zone with this system. This model is discussed in chapter 4. Note that when this
model is used, the airflow rate is varied in order to maintain the head level temperature at
the setpoint, rather than the mean air temperature. The head level setpoint was 25*C.
The temperature distribution in the perimeter zones is different because of the trench
heaters and chilled beams; these zones are assumed to remain well mixed with a uniform
temperature distribution.
Several variations on the displacement ventilation model were considered. Because the
chilled ceilings are removed, condensation is no longer as serious of a concern and the
humidity requirement can be relaxed. A higher humidity model was created with a
supply air humidity ratio setpoint of 10 g/kg. There was also some concern that the heat
gains in the supply plenum partially resulted from heat from the warm return plenum
conducting through the structural floor slab, which is uninsulated. An insulated floor slab
model was created with a 25 mm layer of polystyrene insulation on the underside of the
first and second floor slabs (k=0.035 W/mK, p = 24 kg/m3 , c,= 1210 J/kgK, atjm =
0.85, asolar = 0.9). Finally, a model without the chilled beams was considered. In the no
chilled beams model, the perimeter airflow rate can increase above the minimum airflow
rate and is varied in order to maintain the setpoint of 25*C. The perimeter is assumed to
be well mixed, so the vertical temperature gradient model was not used for this zone.
5.3.2 VAV System
The VAV system is identical to the displacement ventilation system except that the
supply plenum is not used. Instead, the air is introduced directly into the space, and
diffusers are assumed to be located such that the room air is well mixed. Note that
although the supply plenum is not used, the raised floor is still in place. The return air is
still drawn through the return plenum. Because the room air is well mixed, the air
flowrate is varied to maintain the mean room temperature at the setpoint of 25 0C.
The supply air setpoint follows the same seasonal reset schedule as the displacement
ventilation supply air setpoint. This provides the 15*C supply air temperature typical of
VAV systems when cooling is needed, and a warmer 180C supply air temperature when
extra cooling is not needed in the central zone and heating is needed in the perimeter
zone. The ventilation rate cannot be decreased further when heating is needed because
fresh air ventilation is still required.
The baseline VAV model is meant to represent a traditional all-air system, so it also
excludes the chilled beams. The air flowrate in the perimeter zones is increased in order
to provide cooling to 25 0C when necessary. The trench heaters remain in place, so that
when heating is needed, the air flowrate remains at the minimum level and the water
flowrate to the trench heaters is varied to maintain the 20*C setpoint.
Two variations to the VAV model were also considered. In the first, the chilled beams
are included, so that the perimeter air flowrate is always at the minimum and the chilled
beams provide additional cooling, as in the displacement ventilation system. The second
case relaxes the space humidity requirement, as was done with displacement ventilation
system. The supply air humidity ratio setpoint is raised from 9 to 10 g/kg, corresponding
to 50% relative humidity at 25*C.
5.3.3 Night Cooling
Each of the three systems presented above can also be operated in a night cooling mode,
in which the air system is operated at night in order to precool the building. For all three
systems, the daytime system operation is unchanged from the cases above. At night, the
air system supplies a high flowrate of unconditioned, cool outside air to the central zone.
Night cooling operation begins at 11 p.m. and continues until normal system operation
begins at 6 a.m. At night, the system runs at a fixed maximum flowrate until a setpoint of
18*C is reached. The system then varies the flowrate in order to maintain the 18*C
setpoint, but does not let the flowrate drop below the daytime minimum flowrate.
Although in an actual building the controls might be set to turn the system off again once
the setpoint is reached, this type of control cannot be implemented in EnergyPlus.
Finally, note that night cooling is not used during the coldest months. The night cooling
mode is only operable April through October; the normal operating modes described
previously are used November through March. This simplified seasonal reset control
must be used because EnergyPlus does not have provisions for logic-based controls,
which might be used to activate night cooling based on the outdoor temperature.
Several other changes to the existing models were included in the night cooling models to
increase the effectiveness of night cooling. For both the displacement ventilation and
existing building (improved operations) case, the underside of the floor slabs were
insulated, as described previously for the displacement ventilation case. This is to
prevent any coolth stored in the floor slab from being released to the air in the return
plenum beneath the slab. For the VAV system, which does not use the supply plenum,
the raised floor was removed. Without the raised floor removed, there would not be any
thermal mass exposed to the supply air and there would be little potential for effective
night cooling. Finally, for the VAV and displacement ventilation systems, the relaxed
humidity setpoint of 10 g/kg was used.
Two different nighttime maximum air flowrates were considered. A higher flowrate may
help to further precool the building, but it also requires greater fan power. The higher
rate was 5 ach for each system. The lower rate was 2.5 ach for the existing building
system and 3.5 ach for the displacement ventilation and VAV systems. A slightly higher
rate was used for the alternative systems because they require higher maximum daytime
air change rates. In EnergyPlus, the maximum nighttime air change rate is also the
maximum daytime air change rate. If the maximum daytime rate is too low, there are
periods when the system is running at maximum flowrate and still cannot meet the
cooling load.
5.3.4 Natural Ventilation
The natural ventilation system is very different from the systems presented above. All of
the building systems discussed above are removed. The only conditioning of the building
is provided by naturally driven airflow through windows. Pure cross ventilation from one
88
side of the building to the other is assumed. The cross ventilation model and its
implementation into EnergyPlus are discussed in Chapter 4.
Controls on the natural ventilation rate have not been implemented. The windows are
assumed to be completely open at all times, yielding the maximum possible ventilation
rate. This is acceptable for evaluating summer conditions because in the mild U.K.
climate, the outdoor temperature is nearly always lower than the indoor temperature, so
the maximum ventilation rate is desirable throughout the day. However, this model
cannot be used for winter case where the ventilation rate should be kept very low and
heat recovery should be used. Therefore, this model is only used to evaluate summer
comfort conditions and determine the feasibility of natural ventilation for this building.
To apply the cross ventilation model to a specific building, the effective opening area,
opening discharge coefficient, and outside pressure coefficients must be determined. The
effective opening area was estimated using an architectural worksheet for natural
ventilation design (Moore 1993). The windows were assumed to be bottom-hinged
casement type windows with a total area of 120 m 2 and an effective opening ratio of 0.75
on both the southern and northern facades, yielding an effective opening area of 90 M2
The opening discharge coefficient (CD) for each set of windows (north and south) was
assumed to be 0.6. To estimate a discharge coefficient for the internal resistance of the
building, a CFD model of a single floor was created in PHOENICS 3.3. Figure 5.21
shows the geometry of this model. The south fagade was specified as an inlet at a
pressure of 2 Pa and the north fagade was specified as an outlet at a pressure of 0 Pa.
Using the total flowrate through the space and the effective opening area of 90 M2 , the
interior discharge coefficient was estimated to be 0.72. The product of this discharge
coefficient and the discharge coefficient for the windows on each fagade yields an overall
discharge coefficient of 0.26.
Finally, the surface-averaged pressure coefficients on the southern and northern facades
at each floor level were determined for eight different wind directions. A CFD model of
Figure 5.21 Second floor PHOENICS model geometry
Building A and the BP Sunbury campus was created in PHOENICS 3.3. Figure 5.22
shows the model geometry. The wind speed profile was estimated using a power-law
boundary layer model for suburban areas (ASHRAE 2001):
UH
= U m et
niet
(5.1)
( H
met
where
UH = local approaching wind speed at height H
Umet= 4.5 m/s = mean summer meteorological wind speed
wind boundary layer thickness at meteorological station
8met =270 m
Hmet - 10 m height of meteorological station anemometer
amet= 0.14 wind boundary layer exponent at meteorological station
6= 370 m suburban area wind boundary layer thickness
a = 0.22 = suburban area wind boundary layer exponent
Figure 5.22 BP Sunbury campus PHOENICS model geometry
6.0
4.02.0
0.0
-2.0
-4.0
-6.0
-8.0
-10.0
0
60
120
180
240
300
360
Wind Direction (deg CW from N)
Figure 5.23 Variation of pressure coefficients with wind direction
The pressures on each facade were extracted from the PHOENICS results and used to
determine pressure coefficients (C,) from Eq. (4.17). Figure 5.23 shows the variation of
pressure coefficients with wind direction. Winds from the south or northwest results in
the highest pressure differences and will yield the highest ventilation rates.
Three natural ventilation cases were considered. In the first, the building is left
completely unchanged from its existing configuration. However, the raised floors and
lowered ceilings in this case prevent any appreciable heat storage in the floor slab and
therefore limit the effectiveness of natural ventilation. Therefore, two additional cases
were considered: one with the raised floors removed, and one with the lowered ceilings
removed. Either of these actions partially exposes the thermal mass of the floor slab and
increases the potential for thermal comfort with natural ventilation.
5.4 Energy Consumption
This study considers energy consumption by fans, boilers, and chillers. As previously
mentioned, detailed pump performance information was not available and pump energy
consumption was therefore excluded from this study. This is acceptable because the
pumping energy is small compared to the fan, chiller, and boiler energy, and because the
pumping energy would be similar for the various systems considered. Very limited
information was also available for the fans, boilers, and chillers, so the methods used to
estimate the electrical and gas energy consumption of each of these components will be
discussed briefly. The boiler model is very simple and has been discussed previously.
The boiler is assumed to operate at a constant efficiency of 0.95.
The chiller model is based on a curve fit of the manufacturer's performance data. The
only performance data available was maximum cooling capacity and power input for
various ambient temperatures and chilled water temperatures; part load performance data
was not available. A curve fit of this data was used to generate an expression for the
chiller coefficient of performance (COP) as a function of ambient temperature T:
COP = aT2+ bT + c
(5.2)
Two sets of coefficients a, b, and c were used: one corresponding to a chilled water
temperature of 60C and one corresponding to a chilled water temperature of 10*C. Table
5.8 shows these coefficients. The higher chilled water temperature was used to determine
the COP whenever dehumidification was not needed.
Table 5.8 Coefficients for COP expression
Water
Temperature
60C
0.0014
-0.1846
7.7638
100C
0.0015
-0.2005
8.3009
The hourly chiller electricity consumption was then determined by dividing the hourly
chiller cooling load by the predicted chiller COP for that hour. The annual chiller
electricity consumption is the sum of the hourly values for the entire year.
The fan model is based on the cube-law: fan power is proportional to the cube of the
flowrate. Electrical monitoring showed that the fan power consumption for the existing
building is approximately 35 kW, and duct traverses Sdiscussed in 5.2.1) showed that the
total building air supply rate is approximately 15.8 m Is. These two measurements were
used to determine a proportionality constant that predicts the fan power for any of the
building systems:
P (in W)=8.9-V (in m3 /s)
(5.3)
This simple fan model does not account for variations in fan efficiency or the fact that the
power law often does not hold exactly, especially when flowrates becomes low and the
flow is not completely turbulent. However, no other information was available for
estimating the fan power.
5.5 Simulation Cases Summary
Several basic building simulation cases and their variations have been described. Table
5.9 summarizes the differences between these cases. The EnergyPlus input data file (.idf)
for each case is included on the attached compact disc.
Table 5.9 Summary of Building A simulation cases
Night
Minimum
Central Perimeter
Supply Ventilation
Chilled Chilled Trench Plenum
Zone
Zone
Supply Air Economizer
Ceilings Beams Heaters Used Maximum
Rate
Supply Air Supply Air
Rate
3
m /s
Supply Air
Setpoint
(April -
October)
Supply Air Humidity Insulated Raised
Setpoint
Ratio
Floor
Floor
(November.
Slab
Present
Setpoint
March)
ach
*C
Cc
g/kg
9
Lowered
Ceiling
Present
Existing Building
Existing Building
15.8
no
constant
constant
yes
yes
Improved Operations
yes
yes
0
18.5
18.5
no
yes
yes
3.84
yes
constant
constant
yes
yes
yes
yes
0
18.5
18.5
9
no
yes
yes
Slab
3.84
yes
constant
constant
yes
yes
yes
yes
0
18.5
18.5
9
yes
yes
yes
Night Cooling High Rate
3.84
yes
constant
constant
yes
yes
yes
yes
5
18.5
18.5
9
yes
yes
yes
Night Cooling Low Rate
3.84
yes
constant
constant
yes
yes
yes
yes
25
18.5
18.5
9
yes
yes
yes
Displacement Ventilation
3.84
yes
variable
constant
no
yes
yes
yes
0
15
18
9
no
yes
yes
Insulated Slab
3.84
yes
variable
constant
no
yes
yes
yes
0
15
18
9
yes
yes
yes
Higher Humidity
3.84
yes
variable
constant
no
yes
yes
yes
0
15
18
10
yes
yes
yes
No Chilled Beams
Night Cooling High Rate
3.84
3.84
yes
yes
variable
variable
variable
constant
no
no
no
yes
yes
yes
yes
yes
0
5
15
15
18
18
9
10
no
yes
yes
yes
yes
yes
Night Cooling Low Rate
3.84
yes
variable
constant
no
yes
yes
yes
3.5
15
18
10
yes
yes
yes
VAV
3.84
yes
variable
variable
no
no
yes
no
0
15
18
9
no
yes
yes
Higher Humidity
3.84
yes
variable
variable
no
no
yes
no
0
15
18
10
no
yes
yes
._____
.Insulated
Displacement Ventilation
VAV
Chilled Beams
3.84
yes
variable
constant
no
yes
yes
no
0
15
18
9
no
yes
yes
Night Cooling High Rate
3.84
yes
variable
variable
no
no
yes
no
5
15
18
10
no
no
yes
Night Cooling Low Rate
3.84
yes
variable
variable
no
no
yes
no
3.5
15
18
10
no
no
yes
Natural Ventilation
-
--
No Raised Floor
-
-
-
-
--
No Lowered Ceiling
-
-
-
-
Natural Ventilation
-
-
-
no
yes
yes
-
-
-
no
no
yes
-
-
-
no
yes
no
Chapter 6: Results
The results of the Building A energy simulations are presented in this chapter. Results
presented include a simple validation case, annual energy consumption, equipment
sizing, and thermal comfort evaluations.
6.1 Existing Building Validation
Extensive instrumentation of Building A installed in January 2002 has provided
experimental data which can be used for a simple validation of the Building A model.
This validation consists of a comparison of predicted central zone temperature to
measured zone temperatures.
For a comparison with measured data to be meaningful, the simulation must be
performed with weather data corresponding to the actual conditions at the time the
measurements were taken. A weather station was installed on the roof of Building A to
collect this data. Outdoor temperature, relative humidity, barometric pressure, wind
speed, and total horizontal solar radiation were measured at 15 minute intervals. Data for
February 2002 was translated into an EnergyPlus hourly weather file, included on the
attached compact disc. The measured total horizontal solar radiation was split into direct
and diffuse components for the weather file using estimates derived from standard solar
irradiance models (ASHRAE 2001).
Space temperatures in Building A were measured using compact HOBO thermocouple
dataloggers mounted on desk dividers, approximately at seated head height. The
temperature measurements have an error of ±0.5*C and were taken at 15 minute intervals.
Three dataloggers were distributed evenly across the floor area of the central zone for
each half-floor plate, as shown in Figure 6.1. The average of the three measurements is
taken as the zone temperature.
Figures 6.2 - 6.4 show predicted and measured space temperatures for the southern zones
for a single week in February. The agreement is very good; the error is generally less
than 1*C and nearly always less than 2*C, except for some anomalous periods discussed
X
xx
X
x
X
=temperature
datalogger location
Figure 6.1 Temperature datalogger locations on typical floor plan
I
--
26
024
e 22
0.
E
- 20
18
Monday
16
Tuesday Wednesday Thursday
Friday
Saturday
Sunday
1
02/18
02/19
02/20
02/21
02/22
02/23
02/24
02/25
Date
Figure 6.2 Ground floor south measured and simulated zone air temperatures
28
--..-.
South 1 Simulation
26
_--
South 1 Measurement
O 24
-
-
m22
-
E
i20A
02/18
02/19
02/20
02/21
02/22
02/23
02/24
02/25
Date
Figure 6.3 First floor south measured and simulated zone air temperatures
U
28
----- South 2 Simulation
26
-South
2 Measurement
~
UO2422
0
-
18Monday
16
Tuesday Wednesday Thursday
1
1
1
1
02/18
02/19
02/20
02/21
02/22
Friday
Saturday
02/23
Sunday
02/24
02/25
Date
Figure 6.4 Second floor south measured and simulated zone air temperatures
below. For the central zones, the largest factor affecting the daytime temperature
variation is the ratio of internal loads to supply air rate. These zones have little
interaction with the outdoors and little thermal mass, so the temperature is controlled by
the internal heat gains and cooling provided by the supply air. The agreement between
simulation and measurement during the day indicates that these key parameters have been
modeled with reasonable accuracy. This is particularly remarkable because the
simulation internal loads are on a schedule that is the same for each weekday, whereas
the actual internal loads vary according to day-to-day building usage.
There are exceptions to the generally good comparison. These occur during periods
when the building system clearly remained active through the night or over the weekend.
For the week shown, the system clearly remained on Wednesday night and during the day
on Saturday, because the measured temperature remains very stable over these periods.
This indicates a malfunction in the building control system and does not need to be
included in the simulation.
The rate of temperature drop at night, after the system has shut off, is largely controlled
by the thermal mass of the building and conduction losses through the perimeter. The
agreement between the simulation and measurement at night shows that these parameters
have also been modeled accurately.
This simple validation study provides confidence in the accuracy of the Building A
model. It shows that the internal loads, air systems, and building fabric have been
modeled accurately. Although it does not provide a direct validation of energy
consumption predictions, the energy used by the mechanical systems depends very
strongly on these parameters. This model can therefore be used to evaluate the
performance of various mechanical systems with confidence.
6.2 Annual Energy Consumption
The annual energy use of the three largest system energy consumers: chiller, boiler, and
fans, has been evaluated for each mechanical system presented in Chapter 5. The results
for these components are presented individually, followed by the total system annual
energy cost.
6.2.1 Chiller Electricity Consumption
Figure 6.5 shows annual chiller electricity consumption for each mechanical system. The
existing building cases require the most chiller energy. The improved operations case
requires slightly more chiller energy than the original existing building case. This is
because when the airflow rate is reduced in the improved operations case, the opportunity
for free cooling from air is reduced and the chilled ceilings must be used instead. Use of
the chilled ceilings always requires chiller energy in order to provide cooling.
Insulating the concrete floor slab has no appreciable affect on the chiller energy. This
indicates that the heat gains to the supply air within the supply plenum are originating
primarily from the raised floor above the plenum, rather than the concrete slab below the
plenum. The use of night cooling significantly reduces the chiller energy. The higher
30 -
-
T
25
- 20
15
150
0
4A1
~
00
c4)
Exstn
Buldn
DipaemnA
Vetiato
Fgr 6.5 Anua chile elcrct consuptio
nighttime ventilation rate yields a 23 percent reduction in chiller energy, while the lower
nighttime ventilation rate yields a 16 percent reduction in chiller energy.
The displacement ventilation system uses less chiller energy than the existing building.
This is because the replacement of the chilled ceilings with an all-air system in the central
zone increases the opportunity for free cooling from outside air. When the chilled beams
are removed, the chiller energy decreases even further for the same reason. As with the
existing building, insulating the floor slab has no appreciable affect. Increasing the
humidity setpoint reduces the chiller energy very slightly, indicating that only a small
portion of the chiller energy is used for overcooling to provide dehumidification. As with
the existing building, night cooling significantly reduces the chiller energy for the
displacement ventilation system. However, unlike the existing building, both ventilation
rates are equally effective, reducing the chiller energy by 20 percent. This is possible
because in both case, the ventilation rate decreases to the minimum fresh air rate once an
18*C setpoint is reached, so they can have equivalent precooling effects on the building.
Finally, the VAV system uses less chiller energy than either the displacement ventilation
or existing building systems. This is primarily because the VAV system is an all-air
system and therefore maximizes the potential for free cooling. However, even when
chilled beams are used with the VAV system, it still uses less chiller energy than the
displacement ventilation system. This is because in the displacement-ventilation system,
the central zone mean air temperature is actually higher than 25*C, so that mixing of this
air into the perimeter zone causes increased loads on the chilled beams, which maintain
the perimeter zone at 25*C. In the VAV system, the load on the chilled beams is lower
because the mean air temperature of the perimeter and central zones is the same.
Although the displacement ventilation system has slightly lower air supply rates and
therefore requires less outside air cooling, this difference between the systems is smaller
than the chilled beam load difference, so that the VAV system uses slightly less chiller
energy.
Increasing the humidity setpoint has a slightly larger effect on the chiller energy for the
VAV system than the displacement ventilation system (9% vs. 4% reduction). This is
because the VAV system uses higher airflow rates and therefore requires more
dehumidification of outside air. Night cooling is extremely effective for the VAV
system. The chiller energy is reduced by 33 percent with either ventilation rate. Recall,
however, that the night cooling mode requires the floor plenum to be removed, because
otherwise no thermal mass is exposed to the supply air.
These results show that reductions in the existing building chiller energy of up to 64%
(for the night cooling VAV system) are possible. All of the systems considered still rely
upon mechanical chillers in order to provide chilled water. Use of cooling towers for free
cooling of chilled water could provide even greater energy savings. This study focuses
on alternative space conditioning systems, rather than alternative plant systems, so
cooling towers have not been considered here.
6.2.2 Boiler Gas Consumption
Figure 6.6 shows annual boiler gas consumption for each mechanical system. Much of
the heating energy is used to heat outside air. The gas use for the existing building case
is four times higher than that of any of the other cases because the minimum supply rate
of outdoor air is four times higher for this case. The gas consumption for the improved
operations case is much lower because it uses the lower outdoor air supply rate.
Insulating the floor slab has no appreciable affect on boiler energy. Night cooling
increases the required boiler energy because when the system returns to normal operation
at 6 a.m., the supply air must be heated to the setpoint. However, up to 50 percent
recirculated air may be used before 8 a.m. If night cooling is not used, this recirculated
air is significantly warmer and less heating is required to reach setpoint. If better night
cooling controls could be implemented in the energy simulation, this small increase in
boiler energy would not be present.
The displacement ventilation and VAV systems both use about 25 percent less heating
energy than the existing building systems. This is because the existing building uses a
year round setpoint of 18.5*C, while the other systems operate with a 15*C setpoint for
much of the year and therefore require less heating to reach setpoint, especially during
early morning hours.
510
_________________-
140
- 120
p
100
l
~
w 80
0
~
S60
I
2 40
20
0
Cj
Existing
Building
c:
c
0c%
~
-
'
-
~4
Displacement
Ventilation
VAV
Figure 6.6 Annual boiler natural gas consumption
~~1
VariatiQns on the VAV and displacement ventilation systems follow identical trends.
The change in heating energy for any of these variations is relatively small (less than 6
percent). Increasing the humidity setpoint slightly reduces the boiler energy because
lessreheat energy is needed. Systems without chilled beams use slightly less boiler
energy because more warm air is available for recirculation. This increases the mixed air
temperature, yielding longer periods when no heating is of outside air neccesary. Night
cooling slightly increases the boiler energy for the same reasons as with the existing
building.
These results show that enormous reductions in boiler energy are possible by reducing
the outside air minimum supply rate. Further savings can be achieved by using systems
with lower supply air setpoints. Much of the boiler energy is used to heat outside air.
Most of this energy could be eliminated with the use of a heat recovery loop or enthalpy
recovery wheel that extracts heat from the exhaust stream. Again, this study focuses on
alternative space conditioning systems, so this option has not been considered. Because
the minimum outside air supply rate for all systems (except the existing building) is the
same, the energy savings from a heat recovery loop would be similar for all systems, and
the relative energy usage comparison would not change drastically.
6.2.3 Fan Electricity Consumption
The fan energy use is the most variable element between systems, because the cube law
translates somewhat small changes in fan flowrate into very large changes in fan power.
Figure 6.7 shows annual fan electrical energy consumption for each system. The existing
building requires much more fan energy than the other systems because it uses a very
high constant supply air rate. The improved operations case, however, uses a constant
supply air rate four times lower than the existing building and removes the cooling load
with the chilled ceilings. This results in a fan energy that is 1/64* that of the existing
building. The fan energy for all other systems is higher than that for the improved
operations case because they use a variable supply air rate, with the minimum rate equal
to the improved operations supply air rate.
Insulating the floor slab has no effect on the fan energy for the existing building because
the same constant supply air rate is still used. The night cooling cases, however, do
increase the fan energy. In the night cooling cases, the fans run for longer periods and at
higher speeds than for the existing building. The higher maximum night ventilation rate
in the night cooling high case results in an annual fan energy more than three times
greater than that for the night cooling low case.
The displacement ventilation case fan energy is 2.4 times the improved operations fan
energy. This is because the removal of the chilled ceilings requires an increased supply
air rate in order to meet the cooling load. When the chilled beams are removed, the fan
energy further increases, by a factor of 4.7, again because of the increased supply air rate
needed to meet the cooling load. Insulating the floor slab reduces the fan energy very
slightly because the supply air leaving the floor plenum is slightly cooler, requiring a
lower supply air rate to meet the cooling load.
100
136
Existing
Building
Displacement
Ventilation
VAV
Figure 6.7 Annual fan electrical energy consumption
The night cooling systems use more fan energy than the basic displacement ventilation
system, again because the fans operate for longer periods and at higher flowrates at night.
However, the change is not as dramatic as for the existing building, because the use of
night cooling reduces the daytime fan energy, whereas in the existing building the
daytime fan energy remains the same.
Finally, the VAV systems use the most fan energy because large air supply rates must be
used to meet the cooling load. When chilled beams are used, the fan energy is reduced
by 65 percent, but is still slightly higher than the displacement ventilation fan energy.
This is because the vertical temperature gradient allows displacement ventilation to meet
the same cooling load with a lower supply air rate.
The increase in fan energy for the VAV night cooling high case is much smaller than
with the other systems, and for the night cooling low case, the fan energy actually
decreases 26 percent. This is because the night cooling is so effective that the increased
fan running time is offset by large reductions in daytime fan power.
These results demonstrate the sensitivity of fan electrical energy use to fan speed. Supply
air rates above the minimum supply air rate must be justified by a reduction in energy use
in another system component. However, careful operation of fans for longer periods of
time, such as in night cooling, can actually result in decreased fan energy if sufficient
thermal mass is available.
101
6.2.4 Total Energy Cost
The annual energy use of three individual components has been presented. However, the
best measure of a system's energy performance is its total annual energy cost. Energy
costs can vary widely between markets and even with the time of day because of on- and
off-peak pricing schemes. Such detailed energy cost information was not available for
this study. Relative energy costs have been calculated based on a constant 3:1 electrical
to gas energy cost ratio.
Figure 6.8 shows the total annual energy cost for each mechanical system, normalized
with respect to the basic VAV system annual energy cost. The existing building energy
cost is more than four times greater than that of any other system, and 4.7 times greater
than the improved operations energy cost. The night cooling high case results in an
increased energy cost, but in the night cooling low case, the increased fan energy is offset
by the decrease in chiller energy to yield a very small reduction in energy costs. The
contribution of the fan energy to the total cost is small for the improved operations cases
because of the low supply air rates in these cases.
The displacement ventilation system energy cost is 17 percent less than the improved
operations energy cost. Insulating the floor slab has no significant effect on the energy
costs, and the reduction in chiller energy for the higher humidity case yields a very small
=5.74
1.6
U Fan Energy Cost
1.4
III
1.2
I
LI
UN
o er
M
L
r
nergy
ost
N Chiler Energy Cost
0.6
0.2 0.0
Existing
Building
VAV
Displacement
Ventilation
Figure 6.8 Normalized total annual energy costs (3:1 electricity:gas cost ratio)
102
energy cost decrease. The decreased chiller energy with the removal of chilled beams is
more than offset by the increased fan energy, resulting in a total energy cost increase. If
the night cooling rate is low enough, the increased fan energy is offset by decreases in
chiller energy to yield a very small cost decrease. The contribution of the fan energy to
the total cost is small for most of the displacement ventilation cases because this system
operates at the minimum supply air rate for much of the year.
The VAV system energy cost is slightly larger than the displacement ventilation system
energy cost. Again, the higher humidity case has a slightly lower energy cost. The
energy cost is further reduced with the addition of chilled beams, in which the chiller
energy increases but the fan energy decreases dramatically. Both night cooling cases also
have lowered energy costs due to decreased fan and chiller energy. The night cooling
low case has the lowest energy cost of any system, with a 12 percent cost reduction from
the basic VAV system.
6.3 Equipment Sizing
In addition to annual energy use, equipment sizing is important in the selection of a
mechanical system. A system may have extremely low annual energy consumption but
require unusually large equipment. If this equipment is too expensive, the payback
period for choosing this system will be very long, and it is unlikely the system will be
chosen.
Equipment size and cost is related to peak load, although the relation is generally not
linear. Peak loads for the chiller, boiler, and fans for each mechanical system are
presented below.
6.3.1 Peak Chiller Load
Figure 6.9 shows the peak chiller load for each mechanical system studied. The existing
building requires the largest chiller size of nearly 500 kW. Note, however, that the actual
building has two 500 kW chillers, indicating the installed cooling capacity is more than
twice what is necessary.
The improved operations case reduces the chiller size by 37 percent to just above 300
kW, because the decreased supply air rate eliminates unnecessary cooling and
dehumidification of outside air. Variations on the improved operations case do not
significantly affect the chiller size. This is because the peak load is dominated by the
chilled ceiling load, which does not change dramatically with these variations.
The displacement ventilation peak chiller load is 11 percent less than the improved
operations peak load. The increased outside air load due to higher supply air rates in this
case is counteracted by the removal of the chilled ceilings load. Insulating the floor slab
has no significant effect on the peak load. Raising the humidity setpoint reduces the peak
load 8 percent. Removing the chilled beams increases the peak load by 36 percent
because in this case, the additional outside air cooling load becomes very large. Both
night cooling cases significantly reduce the peak chiller load. The reduction is 26 percent
for the high ventilation rate and 22 percent for the low ventilation rate.
103
500
450
400
350
300
-77--
250
200
150
100
50
0
Figure 6.9 Peak chiller loads
The VAV system has the highest peak chiller load, 22 percent greater than displacement
ventilation peak and 8 percent greater than the improved operations peak. This is
because of the large amount of outside air cooling needed for an all-air system. When
chilled beams are added, the peak load is reduced 10 percent, but is still greater than the
displacement ventilation peak load. This is because the VAV system requires higher
supply air rates to meet the same cooling load. Raising the humidity setpoint also
reduces the peak load 10 percent. As with the displacement ventilation system, both
night cooling cases significantly reduce the peak chiller load. The reduction is 23 percent
for the high ventilation rate and 20 percent for the low ventilation rate.
These results show that chiller size depends greatly on the need for cooling of outside air;
the chiller size is largest for all-air systems. Although the economizer allows air to be
recirculated, the return air temperature is nearly always higher than the outside air
temperature, meaning the mostly outside air is used. Use of night cooling tends to reduce
the peak chiller load by about 20 percent. The addition of cooling towers to the plant
would not affect the chiller size. This is because peak chiller loads correspond to high
outdoor dry- and wet-bulb temperatures, which eliminate the potential for free
evaporative cooling.
6.3.2 Peak Boiler Load
Figure 6.10 shows the peak boiler load for each mechanical system. It is nearly constant,
at 140-150 kW, for all systems except the existing building case. This is because the
peak boiler load is dominated by the heating of outside air, and the existing building case
104
M456
140
.......
-
120
100
-
80
60
40
-~
Building
t
-
-
T
---
r
Displacement
Ventilation
Figure 6.10 Peak boiler loads
uses a minimum outside air rate four times greater than the minimum outside air rate for
the other systems. The peak boiler load is slightly lower for the VAV night cooling cases
because the raised floor is removed in these cases. This allows for thermal storage of
direct solar gains in the concrete floor slab, which reduces the load on the trench heaters.
Because the boiler load is dominated by the heating of outside air, it could be
dramatically reduced for any of these systems with the use of a heat recovery loop or
enthalpy recovery wheel. As has been discussed, these modifications have been excluded
from this study. Because the outside air supply rate is the same for each system, the peak
boiler load reduction would be similar for all systems.
6.3.3 Peak Fan Flowrates
Figure 6.11 shows the peak fan volumetric flowrate for each mechanical system. The
peak fan flowrate varies widely between systems. For the existing building cases, the fan
flowrate is determined by the minimum outside air rate, because these cases use a
constant supply air rate. When night cooling is used, the peak fan flowrate is determined
by the maximum nighttime ventilation rate.
The displacement ventilation system requires a peak fan flowrate 2.7 times that of the
improved operations case. This is because the supply air rate must be increased to handle
the cooling load that is removed by the chilled ceilings in the existing building.
Removing the chilled beams further increases the fan flowrate for the same reason.
105
iflzL'IllIzzzzzI_
Existing
Building
Displacement
Ventilation
VAV
Figure 6.11 Peak fan volumetric flowrates
Insulating the slab slightly reduces the fan flowrate because it lowers the supply air
temperature to the space slightly.
The effect of night cooling depends on the maximum night ventilation rate. The night
cooling high ventilation rate is greater than the displacement ventilation peak fan flowrate
and therefore increases the peak fan flowrate. The night cooling low case, however,
decreases the peak fan flowrate because of the decreased daytime cooling load.
The VAV system has the highest peak fan flowrate, 1.7 times greater than the
displacement ventilation peak and 4.5 times greater than the minimum supply air rate.
This can be reduced to only 1.2 times the displacement ventilation peak with the addition
of chilled beams. Both night cooling cases also reduce the peak fan flowrate, again
because of the decreased daytime cooling load provided by night cooling.
These results show that the use of supply air for cooling, rather than for fresh air only,
greatly increases the peak fan flowrate. This effect is most pronounced for an all-air
system with no chilled beams or chilled ceilings. Use of night cooling can reduce the
peak fan flowrate by slightly more than 10 percent.
6.4 Thermal Comfort
Each mechanical system uses setpoints that maintain space temperatures between 20 and
25'C year round. The equipment has been sized such that these setpoints are always met,
106
U--.
-
_________________________________________________________
and these systems therefore always provide acceptable thermal comfort. The natural
ventilation system, however, has no mechanical systems to maintain thermal comfort, so
its ability to maintain acceptable conditions must evaluated. In addition, the
displacement ventilation system creates vertical temperature gradients that may be
uncomfortable if overly large. Thermal comfort evaluations for each of these systems are
presented.
6.4.1 Natural Ventilation
This study evaluates the feasibility of using pure natural ventilation for Building A. The
feasibility of natural ventilation is determined by the summer comfort conditions within
the building. Levermore et al. (2000) present two criteria for evaluating thermal comfort
within a naturally ventilated building in the U.K.:
1. The temperature shall not exceed 25'C for more than 5% of the occupied year.
2. The temperature shall not exceed 28'C for more than 1%of the occupied year.
These criteria are used to evaluate the effectiveness of natural ventilation for Building A.
The occupied hours for Building A are 8 a.m. to 6 p.m. Monday to Friday, corresponding
to 2600 working hours. 5% of the occupied year is 130 hours and 1%of the occupied
year is 26 hours.
Figure 6.12 shows the number of occupied hours that the average building temperature
exceeds a given temperature for each natural ventilation case. Removing either the raised
floor or lowered ceiling shifts the curve downward because the thermal mass of the floor
slab is exposed. This allows more heat gains to be absorbed by the floor slab during the
day and released at night. Eliminating the raised floor shifts the curve the most because
the direct sunlight strikes the floor and can be directly absorbed by the floor slab when
the raised floor is removed.
250
-Existing
00
Building
- -No Lowered Ceiling_
--
No Raised Floor
0
150 -- --
-
-)F 100-
0
25
26
27
28
29
30
31
Temperature (*C)
32
33
34
35
Figure 6.12 Thermal comfort evaluation for naturally ventilated building
107
None of these cases meet the comfort criteria given above. The best case exceeds 25'C
for 191 hours and 28'C for 40 hours. Therefore, thermal comfort cannot be achieved
with a pure natural ventilation system. However, the building is quite close to being
within the comfort criteria. To achieve acceptable comfort conditions, a small
mechanical system could be installed to provide cooling during peak periods, while
natural ventilation could be used for most of the year. This is known as a hybrid
ventilation system. An estimate of the energy use of such a system has been performed
and is discussed below.
6.4.2 Displacement Ventilation
The vertical temperature gradient created by displacement ventilation can be
uncomfortable if too large. Head to ankle temperature differences less than 3 K are
generally considered acceptable. For all of the displacement ventilation cases considered,
the head to ankle temperature difference is always less than 3 K except for on the ground
floor.
For the ground floor, the temperature difference given by the vertical temperature
gradient model is greater than 3 K for about 300 hours a year. This is because the
concrete slab below the ground floor supply plenum is in contact with the ground.
Because the ground temperatures are quite cool, the supply air temperature increases
within the supply plenum much less than it does for the first or second floors. The lower
supply air temperature to the space results in higher head to ankle temperature
differences. Increased insulation levels beneath the ground floor slab or a small ground
floor supply air reheat system might remedy this problem.
6.5 Hybrid System
Because the natural ventilation system cannot maintain summer comfort conditions, a
hybrid system has been proposed. This system would use natural ventilation for most of
the year and a mechanical system for peak periods where heating or cooling is required.
The energy use of such a system has been estimated by dividing the year into three
seasons: natural ventilation, cooling, and heating. During the natural ventilation season,
the mechanical system does not operate and no energy is used. During the cooling and
heating seasons, the VAV system operates, with night cooling during the cooling season.
The length of these seasons was determined with the no raised floor natural ventilation
model. Interior temperatures were found to exceed 25'C only between June 15 and
August 31, so this was specified as the cooling season. To determine the heating season,
the natural ventilation rate was assumed to be controllable (by cracking windows) to
provide the minimum fresh air rate for each zone during occupied hours. With this
ventilation scheme, interior temperatures fell below 20'C between November 1 and
March 31, so this was specified as the heating season.
The energy use of the hybrid system is the energy use of the VAV system during the
heating and cooling periods. Figure 6.13 shows the annual energy use this system
compared to the best cases of the other mechanical systems, normalized with the annual
108
1.4
-
MFan Energy Cost
1.2
1.2
MBoiler Energy Cost
N Chiller Energy Cost
0
~1.0-
0O.8
0.6
0.4
z
0.2
0.0
Existing
Building
Improved
Operations
Displacement VAV - Night VAV/Natural
Ventilation - Cooling Low Ventilation
Night Cooling
Hybrid
Low
Figure 6.13 Normalized total annual energy cost of best-case systems
energy use of the basic VAV system. The energy use of this system is 22 percent lower
than that of the best purely mechanical system (VAV night cooling) and 42 percent lower
than that of the existing building. Both the heating and fan energy are significantly
reduced from the year-round VAV night cooling case.
The estimate presented here represents the maximum possible energy use of the hybrid
ventilation system, where the system is switched into a mechanical mode for an entire
season. An actual hybrid system might switch between mechanical and natural modes on
a daily basis during the heating and cooling seasons, further reducing the system energy
use. This type of control has not been implemented in EnergyPlus and was therefore not
simulated.
6.6 Discussion
Several observations can be made when these results are taken as a whole. The first, and
most obvious, is the importance of using appropriate outdoor air supply rates. The 40
L/s/person rate currently used in Building A results in annual energy costs more than four
times greater than the costs for any of the other systems, which use a 10 L/s/person
outdoor air supply rate.
Second, use of free cooling from outside air should be used to fullest practical extent.
The displacement ventilation and VAV systems both save energy over the existing
building systems because they take advantage of this opportunity. In the existing
building, any cooling that is not provided by air, which is supplied at a fixed minimum
rate, must be performed by the chilled ceilings. The chilled ceilings operated on a chilled
water loop that cannot take advantage of free cooling. Besides reducing energy use, free
cooling also reduces the peak chiller load.
109
However, there is a limit to the advantages of free cooling. When chilled beams are not
used in the perimeter, the energy consumption of both the displacement ventilation and
VAV systems increases. The reduction in chiller energy comes at the cost of
dramatically higher fan energy. Very high supply air rates are necessary to cool the
perimeter. This also results in much larger fans when chilled beams are not used.
Two changes that have little impact on the system performance are raising the humidity
setpoint and insulating the floor slabs. Although raising the humidity setpoint even
higher (above 10 g/kg) might allow for more appreciable energy savings, there could be
some risk for condensation, so this action is not recommended.
Night cooling is an effective technique for the displacement ventilation and VAV
systems, reducing both energy use and equipment sizes. However, the maximum night
ventilation rate must be chosen carefully - the cost of the fan energy becomes too high
for this strategy to be beneficial if the ventilation rate is overly large. A maximum
ventilation rate of 3.5 ach was more effective than 5 ach. Even when the floor slab mass
is not directly exposed to the space, as with the displacement ventilation system, night
cooling works because the coolth stored in the floor slab helps to keep the air in the
supply plenum cool. Because the supply air rate is stays at a fixed minimum for the
existing building cases, they cannot take full advantage of the coolth stored in the floor
slab, and night cooling is not effective for these cases.
Of the purely mechanical systems studied, the VAV and displacement ventilation systems
with night cooling are the best choices for Building A. These systems have the lowest
energy costs and smallest equipment. The VAV system uses slightly less energy than the
displacement ventilation system because the impacts of night cooling are the strongest for
the VAV system. However, the VAV system requires larger fans and chillers than the
displacement ventilation system. In addition, displacement ventilation systems provide
better indoor air quality than VAV systems. Therefore, the displacement ventilation
system is the best choice for Building A if natural ventilation is not used.
Although it presents the possibility of a zero-energy system, pure natural ventilation is
not an effective strategy for Building A. Acceptable comfort levels cannot be maintained
throughout the summer. However, Building A is not far out of the envelope for
acceptable natural ventilation comfort conditions. A hybrid system using natural
ventilation for part of the year and the VAV system during peak periods can maintain
comfort conditions year-round. This system uses more than 20 percent less energy than
the best purely mechanical system and more than 40 percent less energy than the existing
building. This hybrid system is therefore the single best choice for Building A.
110
Chapter 7: Recommendations and Conclusions
This study had two primary goals:
1) Demonstrate the use of EnergyPlus and evaluate its ability to model a technically
sophisticated building.
2) Through this demonstration, compare the performance of several low-energy
cooling systems for an office building in the U.K.
Conclusions and recommendations for each of these goals are presented individually.
7.1 Building Modeling in EnergyPlus
This study has thoroughly tested the capabilities of EnergyPlus through the development
of several building models. Very simple models were created for the validation studies
and an extremely detailed model of Building A was created. Several conclusions have
been reached concerning EnergyPlus's building modeling capabilities.
First, EnergyPlus uses good physical modeling techniques that provide trustworthy
results. This is the most important benchmark for evaluating an energy simulation
program. Both small-scale and full-scale empirical validation studies demonstrated that
EnergyPlus results are well within the accuracy needed for building design. However,
these validation studies cover a limited class of buildings. The small-scale validation is
for a lightweight wood-frame structure, and the full-scale validation is for a lightweight
concrete, fully glazed, deep plan office building. Further validation of EnergyPlus for
more building classes should be performed to expand its range of applicability.
EnergyPlus is a very flexible program. An experienced user can develop simplified
building models quickly, but a building can be modeled in great detail if desired. This
shows its potential as both a building design tool and a final design assessment tool.
Important building elements can be considered in a simple model very early in the design
stage, and greater detail, such as exterior and interior shading and complex mechanical
systems, can be added as the design progresses. This also makes a detailed building
model easier to construct. The Building A model began with only a few basic elements,
and layers of detail were gradually added. The model can be tested repeatedly as it is
built up, reducing the chance for error and time needed to fix the model.
The EnergyPlus source code is well written and organized. This makes it fairly easy to
change the program, making it even more flexible. Many engineers' programming
experience should be sufficient to allow them to make simple changes without great
trouble. For this study, many changes were implemented very quickly, often in less than
a day. Even the displacement ventilation model, which is quite complex and intertwined
with the existing program, was implemented in only a few weeks. Such a complex
change is unnecessary for a typical building design, but the implementation of such a
complex model by someone other than an original program developer shows the code's
transparency.
111
The plant loops and system controls are the main components of EnergyPlus needing
improvement. Although the models used for plant loop components are acceptable,
connecting these components is difficult. To a beginner, it is the most confusing portion
of EnergyPlus input. The requirement of a single splitter and mixer limits the program's
ability to model real-world plant loops. To model the Building A plant loops correctly,
new components had to be created to allow multiple splitters and mixers.
The building system controls also need improvement. The number of control schemes is
relatively limited, and the control schemes can be inconsistent between components. For
example, the trench heater water flow is controlled to exactly maintain the zone air
setpoint temperature, while the chilled ceiling water flow is throttled from zero the
maximum flow across a given range of zone air temperatures. Logic-based control
schemes, such as operating night cooling only if the daytime maximum temperature was
above some minimum, cannot be used. Greater flexibility in the system controls should
be implemented, so that the user can define the control scheme for each component and
use logic-based controls if desired.
A graphical input interface for EnergyPlus is also needed. Although EnergyPlus was
purposefully developed as only a simulation engine with no interface, it will not be
adopted for general use without an interface. The text input files are easy to read and
modify, and a simple input file editor has been released, but the input process is still
fairly laborious and the learning curve is steep. Building geometry, air loops, and plant
loops would be especially easier to input with a graphical interface. Proprietary
interfaces will hopefully solve this problem.
In conclusion, EnergyPlus should be adopted as the new standard for energy simulation
in the U.S., especially after graphical interfaces are released that make the program easier
to use. It provides accurate results and greater flexibility than most programs currently in
use. Although it is not perfect, EnergyPlus represents an improvement over DOE-2, the
most widely used energy simulation program in the U.S., and holds tremendous potential
for use in the design and evaluation of low-energy buildings.
7.2 Building A Low-Energy Cooling Systems
Simulations performed using a detailed model of Building A have allowed the
comparison of several alternative building systems. Systems considered include the
existing building systems (chilled ceilings with underfloor air ventilation), underfloor
displacement ventilation, variable-air volume (VAV) ventilation, night cooling, and
natural ventilation. Several conclusions have been reached concerning the advantages of
the various systems.
First, the use of appropriate outside air supply rates is essential. The existing building
uses 40 L/s/person of outside air. When this is reduced to a more standard 10 L/s/person,
the annual energy consumption of the building is reduced more than fivefold, and peak
boiler and chiller loads are also dramatically reduced.
112
Free cooling by ventilation with outside air is very effective in the mild U.K. climate.
Both the displacement ventilation and VAV systems, which use a variable supply air rate
to condition most of the building, take advantage of free cooling. The annual energy cost
for these systems is about 20 percent less than for the existing building, which uses a
fixed minimum supply air rate that does not take advantage of free cooling. However,
this strategy is not effective for conditioning the perimeter of the building because of the
large fan energy required to meet the perimeter cooling load.
Night cooling is also effective in the mild U.K. climate, although the night ventilation
rate must be chosen carefully. If the ventilation rate is too large, large fans that use an
excessive amount of electricity are needed. With appropriate ventilation rates, the annual
energy consumption, peak chiller load, and peak fan flowrate are all reduced. The largest
impact is on the peak chiller load, which is reduced by 20 percent. Night cooling is most
effective for the VAV system, which does not use a supply plenum and has thermal mass
directly exposed to the occupied space. However, it also works with the displacement
ventilation system, where the thermal mass is only exposed to the supply plenum. Night
cooling is not effective for the existing building because of the low amount of cooling
performed by the air in this system.
The VAV and displacement ventilation systems with night cooling are the best
mechanical systems for Building A. These systems have the lowest energy costs and
smallest equipment of the systems studied. In addition, these systems eliminate the
chilled ceilings, which are expensive.
The VAV system uses slightly less energy than the displacement ventilation system
because the impacts of night cooling are the strongest for the VAV system. However, the
VAV system requires larger fans and chillers than the displacement ventilation system.
In addition, displacement ventilation systems provide better indoor air quality than VAV
systems. Therefore, if a purely mechanical system is used, the displacement ventilation
system is the best choice for Building A.
The natural ventilation simulation has been performed with inputs determined using
computational fluid dynamics (CFD). Natural ventilation alone cannot maintain
appropriate summer comfort conditions for Building A. Even if the raised floor is
removed to expose additional thermal mass, the comfort criteria fall slightly out of
acceptable limits. However, a VAV system can be used to maintain comfort during
extreme periods. This type of hybrid system is the best choice for Building A, using at
least 20 percent less energy than any mechanical system.
Finally, note that even greater energy savings would be possible with modifications to the
air and plant loops. Most importantly, a heat recovery loop or enthalpy recovery wheel
could reduce the boiler energy used to heat outside air, and cooling towers could provide
free cooling to the chilled water loop for much of the year.
In conclusion, a hybrid ventilation system would be the lowest energy solution for
Building A. VAV or displacement ventilation systems also provide significant energy
113
savings and use smaller equipment than the existing system. All of these systems use less
energy than the chilled ceiling system currently installed while also having lower first
costs, because the expensive chilled ceilings are eliminated. With the addition of
appropriate controls, the systems in the existing building could be modified to operate as
the displacement ventilation system studied here. The building owners should consider
taking this action.
114
References
Allard, F., et al. 1998. Natural Ventilation in Buildings:A Design Handbook. London:
James and James Ltd., 1998.
Alloca, C.2001. "Single-sided natural ventilation: design analysis and general
guidelines." Massachusetts Institute of Technology, Dept. of Mechanical Engineering.
ASHRAE, 2001. 2001 ASHRAE Handbook Fundamentals, ASHRAE, Atlanta. Ch. 2831.
Balaras, C.A. 1996. "The role of thermal mass on the cooling load of buildings. An
overview of computational methods." Energy andBuildings, 24, p. 1-10.
Behne, M. 1999. "Indoor air quality in rooms with cooled ceilings. Mixing ventilation
or rather displacement ventilation?" Energy andBuildings, 30, p. 155-166.
Bloomfield, D.P. 1999. "An Overview of Validation Methods for Energy and
Environmental Software." ASHRAE Transactions, 105 (2).
Braham, G. D. 2000. "Comparative Performance of U.K. Fabric Energy Storage
Systems." ASHRAE Transactions, 106 (1).
Carrilho da Graga, G., et al. 2001. "Simulation of wind-driven ventilative cooling
systems for an apartment building in Beijing and Shanghai." Energy and Buildings, 34,
p. 1- 1 1.
Clarke, J.A., et al. ESP-r,A buildingandplant energy simulation environment. Energy
Simulation Research Unit, ESRU Manual U91/2. Glasgow, UK: University of
Strathclyde.
Chen, Q., T.G. Hoornstra, and J. Van der Kooi. 1990. "Energy Analysis of Building with
Different Air Supply and Exhaust Systems." ASHRAE Transactions,96 (1).
Chen, Q. 1989. "Convective Heat Transfer in Rooms with Mixed Convection."
Proceedingof InternationalSeminar on Air Flow Patternsin Ventilated Spaces, February
1989, Liege, Belgium, p. 69-82.
Chen, Q. 1988. "Indoor Airflow, Air Quality, and Energy Consumption of Buildings."
Delft University of Technology, Faculty of Mechanical Engineering.
Crawley, D.B. 1998. "Which Weather Data Should You Use for Energy Simulation of
Commercial Buildings?" ASHRAE Transactions, 104 (2).
115
Conroy, C.L., and S. A. Mumma. 2001. "Ceiling Radiant Cooling Panels as a Viable
Distributed Parallel Sensible Cooling Technology Integrated with Dedicated Outdoor Air
Systems." ASHRAE Transactions, 107 (1).
Dorer, V., and A. Weber. 1999. "Air, contaminant and heat transport models:
integration and application." Energy and Buildings, 30, p. 97-104.
EnergyPlus 2001. "Engineering Reference." EnergyPlusManual, version 1.0. April
2001.
GATC. 1967. Computerprogramfor analysis of energy utilization in postalfacilities.
Vol. 1, User's manual. Niles, IL: General American Transportation Corporation.
Geros, V., et al. 1999. "Experimental evaluation of night ventilation phenomena."
Energy and Buildings, 29, p. 141-154.
Haapala, T., et al. 1995. "Energy Analysis Test for Commercial Buildings (Commercial
Benchmarks)." International Energy Agency, Annex 21/Task 12, Tampere, Finland.
Hand, J. 2002. Personal contact. Energy Simulation Research Group, University of
Strathclyde, Scotland.
Heiselberg, P., K. Svidt, and P.V. Nielsen. 2001. "Characteristics of airflow from open
windows." Building and Environment, 36, p. 859-869.
Hittle, D.C. 1979. Building loads analysis and system thermodynamics (BLAST) user's
manual. Technical Report E-153. Champaign, IL: U.S. Army Construction Engineering
Laboratory.
Huang, J., et al. 1999. "Linking the COMIS Multi-Zone Airflow Model with the
EnergyPlus Building Energy Simulation Program." Proceedingsof the International
Building PerformanceSimulation Association Conference 1999, p. 1065-1070.
International Energy Agency (IEA) Annex 28 - Low Energy Cooling. "Selection
Guidance for Low Energy Cooling Technologies." United Kingdom: Energy
Conservation in Buildings and Community Systems Programme, December 1997.
Judkoff, R. and J. Neymark. 1994. "Building Energy Simulation Test (BESTEST) and
Diagnostic Method". U.S. National Renewable Energy Laboratory Report NREL/TP472-6231.
Judkoff, R., and J. Neymark. 1999. "Adaptation of the BESTEST Intermodel
Comparison Method for Proposed ASHRAE Standard 140P: Method of Test for Building
Energy Simulation Programs." ASHRAE Transactions, 105 (2).
116
Kooi, J. van der and K. Bedeke. 1983. "Improvement of cooling load programs by
measurements in a climate room with mass." Proceedings of the XVIth International
Congress of Refrigerationn, vol. E, p. 54-60, Paris.
Kusuda, T. 1978. NBSLD, the computerprogramfor heating and cooling loads in
buildings. NBS Building Science Series No. 69-R, Washington, D.C.: NBS.
Lomas, K.J., et al. 1994. "Empirical validation of thermal building simulation programs
using test room data. Volume 2: Empirical validation package." International Energy
Agency, Annex 21/Task 12, September 1994.
Feustel, H.E., and C. Stetiu. 1995. "Hydronic radiant cooling - preliminary assessment."
Energy and Buildings, 22, p. 193-205.
Hawken, P., A. Lovins, and L.H. Lovins. NaturalCapitalism. Boston: Little, Brown and
Co., 1999, p. 85.
Hu, S., Q. Chen, and L.R. Glicksman. 1999. "Comparison of Energy Consumption
between Displacement and Mixing Ventilation Systems for Different U.S. Buildings and
Climates." ASHRAE Transactions,105 (2).
Klein. S.A., et al. 1994. "TRNSYS - A transient system simulation program." Madison:
Solar Energy Laboratory, University of Wisconsin.
Kolokotroni, M., B.C. Webb, and S.D. Hayes. 1998. "Summer cooling with night
ventilation for office buildings in moderate climates." Energy and Buildings, 27, p. 231237.
Kolokotroni, M. 2001. "Night Ventilation Cooling of Office Buildings: Parametric
Analyses of Conceptual Energy Impacts." ASHRAE Transactions, 107 (1).
Levermore, G.J., A.M. Jones, and A.J. Wright. 2000. "Simulation of a Naturally
Ventilated Building at Different Locations." ASHRAE Transactions,106 (2), p. 402-407.
Moore, F. 1993. EnvironmentalControlSystems: heating cooling lighting. New York:
McGraw-Hill, 1993.
Mottillo, M. 2001. "Sensitivity Analysis of Energy Simulation by Building Type."
ASHR AE Transactions,107 (2).
Mundt, E. 1990. "Convective flow above common heat source in rooms with
displacement ventilation." ProceedingsofROOM VENT '90, Oslo.
Murakami, S., et al. 1990. "Examining the k-s model by means of a wind tunnel test and
large eddy simulation of the turbulence structure around a cube." Journalof Wind
Engineeringand IndustrialAerodynamics, 41-44, p 87-100.
117
Niu, J. 1994. "Modelling of Cooled-Ceiling Air-Conditioning Systems: Influences on
Indoor Environment and Energy Consumption." Delft University of Technology, Faculty
of Mechanical Engineering.
Niu, J., J.v.d. Kooi, and H.v.d. Ree. 1995. "Energy saving possibilities with cooled
ceiling systems." Energy and Buildings, 23, p. 147-158.
Novoselec, A., and J. Srebric. 2002. "A critical review on the performance and design of
combined cooled ceiling and displacement ventilation systems." Energy and Buildings,
34, p. 497-509.
Rees, S. J., and P. Haves. 1995. "A Model of a Displacement Ventilation System
Suitable for System Simulation." ProceedingsofInternationalBuilding Performance
Simulation Association Conference 1995, p. 199-205.
Ren, J.X., and J.O. Dalenback. 1995. "Night Ventilation for Cooling Purposes. Part I Reference Building and Simulation Model." ProceedingsofInternationalBuilding
Performance SimulationAssociation Conference 1995, p. 15 8-165
Schiller, B.G., and R. de Rear. 2000. "A standard for natural ventilation." ASHRAE
Journal,42 (10), October 2000.
Sowell, E.F., and D.C. Hittle. 1995. "Evolution of Building Energy Simulation
Methodology." ASHRAE Transactions, 101 (1).
Straaten, V. 1967. Thermal Performance ofBuildings. Amsterdam: Elsevier, 1967.
Strand, R.K., and Pedersen, C.O. 1997. "Implementation of a Radiant Heating and
Cooling Model into an Integrated Building Energy Analysis Program." ASHAE
Transactions, 103 (1).
Strand, R.K., et al. 2000. "EnergyPlus: A New-Generation Energy Analysis and Load
Calculation Engine for Building Design." Proceedingsof the ACSA Technology
Conference, Cambridge, MA, July 2000.
Taylor, R.D., et al. 1991. "Impact of Simultaneous Simulation of Buildings and
Mechanical Systems in Heat Balance Based Energy Analysis Programs on System
Response and Control." ProceedingsofBuilding Simulation '91, August 1991, Nice,
France.
Winkelmann, F.C. 2001. "Modeling Windows in EnergyPlus." ProceedingsofSeventh
InternationalIBPSA Conference, August 2001, Rio de Janeiro, Brazil.
118
Witte, M.J., et al. 2001. "Testing and Validation of a New Building Energy Simulation
Program." Proceedingsof Seventh InternationalIBPSA Conference, August 2001, Rio
de Janeiro, Brazil.
Yuan, X., Q. Chen, and L.R. Glicksman. 1998. "A Critical Review of Displacement
Ventilation." ASHRAE Transactions, 104 (1).
Yuan, X., Q. Chen, and L.R. Glicksman. 1999a. "Performance Evaluation and Design
Guidelines for Displacement Ventilation." ASHRAE Transactions, 105 (1).
Yuan, X., Q. Chen, and L.R. Glicksman. 1999b. "Models for Prediction of Temperature
Difference and Ventilation Effectiveness with Displacement Ventilation." ASHRAE
Transactions, 105 (1).
Zimmermann, M., and Andersson, J. 1998. "Case Studies of Low Energy Cooling
Technologies." International Energy Agency, Energy Conservation in Buildings and
Community Systems Programme, Annex 28 - Low Energy Cooling, August 1998.
119
Appendix A: Changes to EnergyPlus Code
The theory behind the changes made to the EnergyPlus code is presented in Chapter 4.
This appendix presents highlights of the actual program code. All changes were made to
EnergyPlus vi.0b23. The structure and calling tree of EnergyPlus are fairly complicated;
an introduction is provided in the program documentation (EnergyPlus 2001). Although
an effort has been made to document and present the most vital portions of the changes,
the best record and explanation of the program is the code itself. The complete program
code for each variation of EnergyPlus used in this study is included on the attached
compact disc.
A.1 Displacement Ventilation
The background theory for the displacement ventilation model implemented in
EnergyPlus is presented in Chapter 4.2. Nearly all of the changes necessary to implement
this model were made in the module ZoneTempPredictorCorrector.f90, which handles the
heart of the predictor-corrector calculation. A new subroutine, CalcThetas, was created
in this module to perform the calculation of the dimensionless temperatures from Eqs.
(4.5) and (4.3), the nodal models. The calculation portion of this function is shown in
Figure A. 1. Note that logical conditions were used to prevent the model from
accidentally giving unreasonable results: if Th is less than Tf, then it is set equal to Tf, and
if Th is greater than Te, Th is set equal to Te.
The predictor step is performed in the CalcPredictedSystemLoad subroutine. This
subroutine has four different sections for calculating the system load based on the type of
setpoint: single heating, single cooling, single heating/cooling, or dual setpoint with
deadband. The displacement ventilation model is only implemented into the single
cooling setpoint calculation, because this is how a displacement ventilation system is
normally controlled.
!Mundt simple model
NewThetafloor(ZoneNum) = 1.0/(2.54*Coefhas/Zone(ZoneNum)%.FloorArea + 1.0)
IF (NewThetafloor(ZoneNum) .gt. 1.0) NevThetafloor(ZoneNum) = 1.0
Tf temp = Tsupply(ZoneNum) + NewThetafloor(ZoneNum)*(Tetemp -
Tsupply(ZoneNum))
!Yuan model
qinternal = 0.295*(SUMC(ZoneNum)-ZoneIntGain(ZoneNum)%QLTCON)
qlights
= 0 . 132*ZoneIntGain(ZoneNum)%QLTCON
qwallstop = 0. 185*( SUMHATf (ZoneNum)-SUMHAf (ZoneNum)*Tf temp
&
+SUMHATc(ZoneNum)-SUMHAc(ZoneNum)*Tetemp
&
+SUMHATw(ZoneNum)-SUMHAwv(ZoneNum)*0. 5*(Floorratio(ZoneNum)*Tf temp
&
+Exhaustratio(ZoneNum)*Tetemp)
qwallsbot = 1.0 + 0.185*0.5*SUMHAv(ZoneNum)/Coefhas
Thtemp = (Tftemp + (qinternal+qlights+qwallstop)/Coefhas) / qwallsbot
IF (Thtemp
It. Tftemp)
Thtemp = Tf temp
IF (Thtemp .gt. Tetemp) Thtemp = Tetemp
NewThetahead(ZoneNum) = (Thtemp -
Tsupply(ZoneNum))/(Tetemp - Tsupply(ZoneNum))
RETURN
Figure A.1 Calculation portion of CalcThetas subroutine
120
IF (DispVentActive(ActualZoneNum)
eq. 1.0) THEN !determine true zone T for disp. vent.
Thetachange = 1.0
Thtemp = TempZoneThermostatSetPoint (ActualZoneNum)
Oldthetahead = NewThetahead(ActualZoneNum)
Oldthetaf loor = NewThetaf loor(ActualZoneNum)
k=0
IF ( MAT(ActualZoneNum) .gt. (Thtemp-5.0) ) THEN
DO WHILE ((Thetachange .gt. 0.001) and. (k It. 100))
k-k+1
NewThetahead(ActualZoneNum) = 0.15*NewThetahead(ActualZoneNum) + 0.85*Oldthetahead
NewThetafloor(ActualZoneNum) = 0.15*NewThetafloor(ActualZoneNum) + 0.85*Oldthetafloor
Tetemp - (Thtemp + (NewThetahead(ActualZoneNum) - 1.0)*Tsupply(ActualZoneNum))/NewThetahead(ActualZoneNum)
Tftemp - Tsupply(ActualZoneNum) + NewThetafloor(ActualZoneNum)*(Tetemp - Tsupply(ActualZoneNum))
TZtemp - 0.5 * (Floorratio(ActualZoneNum)*Tftemp + Thtemp + Exhaustratio(ActualZoneNum)*Tetemp)
LoadToCoolingSetPoint - (TempDepZnLd(ActualZoneNum)*TZtemp + SUMHAf (ActualZoneNum)*Tf temp &
+ SUMHAc(ActualZoneNum)*Tetemp - TempIndZnLd(ActualZoneNum))
Oldthetahead = NevThetahead(ActualZoneNum)
Oldthetaf loor = NewThetaf loor(ActualZoneNum)
Coefhas = LoadToCoolingSetPoint
/
(Tsupply(ActualZoneNum) - Tetemp)
CALL CalcThetas(ActualZoneNumCoefhasTetempTftemplThtempl)
Thetachange = ABS(NewThetahead(ActualZoneNum) - Oldthetahead)
Thetachange = MAX(ThetachangeABS(NewThetafloor(ActualZoneNu)
END DO
- Oldthetafloor))
ELSE
LoadToCoolingSetPoint = 0 .0
ENDIF
ELSE
Figure A.2 Displacement ventilation portion of CalcPredictedSystemLoad subroutine
Figure A.2 shows the displacement ventilation portion of this subroutine. Several
features should be noted. A flag (DispVentActive) has been created to determine if
displacement ventilation is active; the setting of this flag will be discussed below. The
displacement ventilation calculation is performed if the flag is set, otherwise, the normal
EnergyPlus heat balance calculation is performed (not shown in Fig. 4.4). If the mean air
temperature is 5'C or more below the setpoint, the load prediction is not performed and
the required system load is assumed equal to zero. This speeds up the calculation by
avoiding iterations in the predictor step when extra cooling is clearly not required.
Finally, the number of iterations is limited to 100 in order to speed up the calculation and
prevent infinite loops.
The corrector step is performed in the CorrectZoneAirTemp subroutine. Figure A.3
shows the evaluation of the DispVentActive flag within this subroutine. If the zone has
been specified as being conditioned with displacement ventilation, and the system is
running (mass flowrate greater than zero), then the displacement ventilation model is
used. Note that this assumes that the temperature gradient immediately disappears and
the entire zone is at the mean air temperature when the system turns off, where in reality
it would dissipate over time.
Figure A.4 shows the iteration loop for the displacement ventilation calculation in the
CorrectZoneAirTemp subroutine. Again, the displacement ventilation calculation is only
performed if the DispVentActive flag is set.
121
Tsupply(ZoneNum) = Node(ZoneEquipConfig(ControlledZoneNum)%InletNode(1))%Temp
IF (ZoneEquipConfig(ControlledZoneNum)%DispVent and. &
Node(ZoneEquipConfig(ControlledZoneNum)%InletNode(1))%MassFlowRate .GT. 0.0) THEN
DispVentActive(ZoneNum) = 1.0
ELSE
DispVentActive(ZoneNum) = 0.0
ENDIF
Figure A.3 Evaluation of the DispVentActive flag
DO WHILE ( (Deltathetafloor
> 0.001)
or.
(Deltathetahead
> 0.001)
)
!loop to converg theta, T values
NevThetafloor(ZoneNum) = 0.5*NevThetafloor(ZoneNum) + 0.S*Thetafloortemp
NevThetahead (ZoneNum) = 0. 5-NevThetahead (ZoneNum) + 0. 5-Thetaheadtemp
!0.5 relaxation factor to speed
! convergence
Thetatop = ( (CoefSumhaz*Floorratio(ZoneNum) + SUKHAf(ZoneNum))*(1.0 - NewThetafloor(ZoneNum)) &
+ CoefSumhaz*(1.0 - NewThetahead(ZoneNum)) )-Tsupply(ZoneNum)
Thetabot = Floorratio(ZoneNua)-NewThetaf loor(ZoneNum) + NewThetahead(ZoneNum) + Exhaustratio(ZoneNum)
Texhaust(ZoneNum) = (CoefSumhat + Temphist - Thetatop)/ &
(CoefSumhaz-Thetabot + SUMHAf (ZoneNum)*NewThetaf loor(ZoneNum) + CoefSumhae)
!nodal models
!store old theta values for comparison
Thetaf loortemp = NewThetaf loor(ZoneNum)
Thetaheadtemp = NewThetahead(ZoneNum)
CALL CalcThetas(ZoneNum, Coef has ,Texhaust (ZoneNum), Tf loor(ZoneNum) ,Thead(ZoneNum))
Deltathetafloor = ABS(NewThetafloor(ZoneNum) - Thetafloortemp)
Deltathetahead = ABS(NevThetahead(ZoneNum) - Thetaheadtemp)
END DO
Figure A.4 Displacement ventilation iteration loop in CorrectZoneAirTemp subroutine
Several smaller changes were also made the ZoneTempPredictorCorrector.f90 module.
Most of these dealt with storing current and previous timestep dimensionless
temperatures so that they may be compared for the timestep-halving criterion. In
addition, the zone interaction with the air system is based on the zone node temperature,
which is stored in the air system data structure and represents the temperature of the air
leaving the zone. In the original EnergyPlus model the zone node temperature is the
same as the mean air temperature. In this model however, this temperature is changed to
be equal to the zone exhaust temperature, Te.
Small changes were also made to the ManageHVAC subroutine (in the module
HVACManager.f9O) in order to implement the new timestep-halving criterion based on
the dimensionless temperature change from the previous timestep. Finally, the
InitlnteriorConvectionCoeffs subroutine (in the module
HeatBalanceConvectionCoeffs.f90) was modified so that convection coefficients for
walls, floors, and ceilings are calculated based on their respective air temperatures. The
detailed convection coefficient correlations are unchanged; only the air temperatures used
in the correlations are different. The complete program code is included on the attached
compact disc.
A.2 Airflow
A.2.1 Natural Ventilation
The background theory for the natural ventilation model implemented in EnergyPlus is
presented in Chapter 4.3.1. All of the changes necessary to implement this natural
122
ventilation model were made in the module HeatBalanceAirManager.f9O. Several
changes were made to the subroutine GetSimpleAirModellnputs in order to input new
information relating to natural ventilation. A new input object, "Cp Values," was created
to input the pressure coefficients for each fagade for eight wind directions. The only
inputs for this object are the eight pressure coefficient values corresponding to wind
directions 0 - 3150 at 450 increments.
The "Mixing" and "Infiltration" objects were used to represent the natural ventilation
airflow through each zone. Each floor is modeled as two zones; Chapter 5 gives a more
complete description of the zones. The infiltration object is used to introduce outdoor air
into the zone on the windward side of the building. The mixing object is used to
introduce air from the windward zone into the leeward zone. Hence, the combination of
the infiltration and mixing represents cross-ventilation through the building.
A new value (%WhenActive) was introduced to the Mixing and Infiltration data types to
represent whether a zone is on the leeward or windward side when the pressure difference
is positive. In addition, a new value (%Level) was created to represent what floor of the
building the zone is on. These inputs are also processed in the GetSimpleAirModellnputs
subroutine.
Figure A.5 shows the calculation of the natural ventilation air flowrate in the subroutine
SetConvHeatGains. The DO loop index (Loop) corresponds to the each building level
(Building A has three levels). The pressure coefficient for each side of the building is
first determined by linear interpolation with the current wind direction. The pressure
difference is then calculated from Equation (4.18) using the current wind speed. Finally,
the air flowrate is calculated from Equation (4.16).
Figure A.6 shows the implementation of the natural ventilation model into the infiltration
calculation in the subroutine SetConvHeatGains. A logical statement is used to
determine if natural ventilation infiltration is active for the current zone, and if it is, the
infiltration rate is set equal to the natural ventilation rate corresponding the building level
in which the zone is located. If not, the infiltration rate is equal set to the design level
according to the standard EnergyPlus model. A similar operation is carried out for the
mixing model in the subroutine CalcHeatBalanceAir. The complete natural ventilation
program code is included on the attached compact disc.
!simple natural ventilation calculation
Cd = 0.26 !overall discharge coefficient
A = 90.0
leffective opening area
!cp value interpolation
WindDir = MOD(WindDir,360.)
J = 1 + INT( (WindDir-MOD(WindDir.45.))/45.
DO Loop = 1,3
!cp value interpolation
cpl(Loop) = Cpvalues(LoopJ) + ((Cpvalues(Loop,J+1)-Cpvalues(Loop,J))/45.)&
*(WindDir - 45.*REAL(J-1))
cp2(Loop) = Cpvalues(Loop+3,J) + ((Cpvalues(Loop+3,J+1)-Cpvalues(Loop+3,J))/45.)&
*(WindDir - 45.*REAL(J-1))
!calculations
deltaP(Loop) = (cpl(Loop)-cp2(Loop))*(0.5*AirDensity*VindSpeed-2.)/100. 'south - north F difference
Ivolumetric flowrate
Q(Loop)
= Cd(Loop)*A(Loop)*SQRT(2.0*ABS(deltaP(Loop))/AirDensity)
END DO
Figure A.5 Natural ventilation air flowrate calculation
123
! Process the scheduled Infiltration for air heat balance
DO Loop=lTotInfiltration
NZ=Inf iltration(Loop)%ZonePtr
natural ventilation calculation
IF ( ((Infiltration(Loop)%UhenActive EQ. 1) AND. &
(deltaP(Infiltration(Loop)%Level) .GT. 0.0))
OR. &
((Infiltration(Loop)%UhenActive EQ. -1) AND. &
(deltaP(Infiltration(Loop)%Level) .LE. 0.0)) ) THEN
MCPI(NZ)=Q(Infiltration(Loop)%Level)*AirDensity*CpAir
ELSE
IVF(NZ)=Infiltration(Loop)%Designevel*GetCurrentScheduleValue(Infiltration(Loop)%SchedPtr)
MCPI(NZ)=IVF(NZ)*AirDensity-CpAir*( Infiltration(Loop)%ConstantTeraCoef +
&
ABS(OutDryBulbTemp-MAT(NZ) )*Infiltration(Loop)%TemperatureTermCoef
&
+ UindSpeed*(Infiltration(Loop)%VelocityTeraCoef + &
WindSpeed*Inf iltration(Loop)%VelocitySQTermCoef)
ENDIF
OAMFL(NZ)=MCPI(NZ)/CpAir
MCPTI (NZ)=MCPI(NZ)*OutDryBulbTemp
ENDDO
Figure A.6 Natural ventilation infiltration calculation
A.2.2 Interzone Mixing
A simple change to the EnergyPlus CrossMixing model has been implemented to allow
the crossmixing rate to be calculated from Eq. (4.19) as discussed in Chapter 4.3.2. The
CrossMixing model differs from the Mixing model (used in the natural ventilation code
above) because it accounts for the interchange of air between two zones, whereas the
Mixing model accounts for air flowing from one zone into another; air does not flow in
both directions. Figure A.7 shows the new CrossMixing calculation, which is found in
the CalcHeatBalanceAir subroutine of the module HeatBalanceAirManager.f90. The
only change to this calculation from the original EnergyPlus code is in the determination
of the coefficient A.
COMPUTE CROSS ZONE
AIR MIXING
DO J=1,TotCrossMixing
N=CrossMixing(J)%ZonePtr
M-CrossMixing ( J)%FroaZone
TD=MTC(J)
IF (TD .GE. 0.0) THEN
TZN-MAT(N)
TZM-MAT(M)
IF ( (ABS(TZM-TZN) LT. TD)
A = ABS(TZM-TZN)/TD
ELSE
A =
AND. (TD
NE. 0.0) ) THEN
1.0
ENDIF
SET COEFFICIENTS
MCPxN-A*MVFC(J)*
&
cpairfn(ZoneAirHuxRat(n),REAL(tzn))*rhoairfn(OutBaroPress.REAL(tzn),ZoneAirHumRat(n))
MCPM(N)=MCPM(N)+MCPxN
MCPxM=A*MVFC(J)*
&
cpairfn(ZoneAirHuaRat(m),REAL(tzm))*rhoairfn(OutBaroPressREAL(tzm),ZoneAirHumRat(m))
MCPM(M)=MCPM(M)+MCPxM
MCPTM(N)=MCPTM(N)+MCPxM*TZM
MCPTM(H)=MCPTM(M)+MCPxN*TZN
ENDIF
ENDDO
Figure A.7 Linear CrossMixing model calculation
124
A.3 Plant Loops
A.3.1 Crossover Pipes
The function of the crossover pipe object is presented in Chapter 4.4.1. The crossover
pipe is specified as a new object in EnergyPlus (CROSSOVERPIPE) that the user inputs
in the same manner as any other plant component. Figure A.8 shows the input data
dictionary (idd) definition of the crossover pipe. The user inputs a component name and
four node names corresponding to the inlets and outlets. This processing of this input is
fairly straightforward; it is performed in the subroutine GetCrossoverPipelnput in the
module PlantPipes.f90. The crossover pipe is simulated in the subroutine
SimCrossoverPipes in the module PlantPipes.f90; the simulation is called by either the
plant loop supply side manager or the plant loop demand side manager in the same
manner any other component. Figure A.9 shows the simplicity of the actual simulation of
the crossover pipe; it simply passes the node information from each inlet node to each
outlet node.
CROSSOVERPIPE,
Al,
A2,
A3,
A4,
A5;
\memo Passes inlet node state variables to outlet node state variables
\field PipeName
\field In et1 Node Name
\field Outlet1 Node Name
\field Inlet2 Node Name
\field Outlet22 Node Name
Figure A.8 Crossover pipe idd definition
!PASS INFORMATION FROM INLET TO OUTLET NODE
EnergySource:
&
SELECT CASE (CompType)
CASE ('CROSSOVERPIPE')
PipeNum=FindItemInList(NameCrossoverPipe%Name,NumCrossoverPipes)
IF (PipeNum /= 0) THEN
Node(CrossoverPipe(PipeNum)%OutletNodeNum2) = Node(CrossoverPipe(PipeNum)%InletNodeNual)
Node(CrossoverPipe(PipeNum)%OutletNodeNum1) = Node(CrossoverPipe(PipeNux)%InletNodeNum2)
ELSE
CALL ShowFatalError('CrossoverPipe not found='//TRIM(Name))
ENDIF
CASE DEFAULT
CALL ShowFatalError('Invalid component, expected CROSSOVERPIPE='//TRIM(CompType))
END SELECT EnergySource
Figure A.9 Crossover pipe simulation
A.3.2 Controlled Crossover Pipes
The function of the controlled crossover pipes is presented in Chapter 4.4.2. The input
for a controlled crossover pipe is identical to that for a crossover pipe. Two additional
objects are used for controlled crossover pipes: HEATING CROSSOVERPIPE, and
COOLING CROSSOVERPIPE. The two are essentially identical, except the heating
pipe is used for hot water loops and the cooling pipe is used for chilled water loops. The
input statements are processed in the subroutines GetHeatingCrossoverPipelnput and
GetCoolingCrossoverPipelnput, in the modules PlantOutsideHeatingSources.f90 and
PlantOutsideCoolingSources.f90, respectively.
125
U.
The simulation method for a controlled crossover pipe is based on the existing simulation
for PURCHASED:CHILLED WATER in EnergyPlus. Figure A.10 shows the simulation
of the cooling crossover pipe in the subroutine SimOutsideCooling in the module
PlantOutsideCoolingSources.f90.
The component's maximum capacity is determined from the maximum available flowrate
and the difference in temperature between the demand and supply side inlets. The
subroutine CalcCompCapacity then determines the component load (MyLoad) necessary
to maintain the CalcCompCapacity then determines the component load (MyLoad)
necessary to maintain the secondary loop setpoint temperature.
SimCoolingCrossoverPipe then simulates the actual component by determining the
flowrate needed to meet this load, and finally, the inlet temperatures are passed to the
outlets and all mass flowrates are set equal to the calculated mass flowrate.
CASE ('COOLING CROSSOVERPIPE')
CompNum = FindItemInList(Name, CoolingCrossoverPipeName, NumCoolingCrossoverPipes)
!Calculate Load
! MinPlr, MaxPir, OptPlr are not defined. Hence assume min = 0, max=opt=Noxcap
InletNodel
= CoolingCrossoverPipe(CompNum)%InletNodeNum1
InletNode2
= CoolingCrossoverPipe(CompNum)%InletNodeNum2
OutletNodel
= CoolingCrossoverPipe(CompNum)%OutletNodeNum1
OutletNode2
= CoolingCrossoverPipe(CompNum)%OutletNodeNum2
IF (LoadFlag /=0) THEN
MinCap = 0
MaxCap = CPCW(Node(InletNode2)%Temp)*(Node(InletNode2)%Temp - &
Node(InletNodel)%Temp)*Node(InletNode2)%MassFlowRateMax
OptCap = MaxCap
CALL CalcCompCapacity(RemLoopDemand, MaxCap, MinCap, OptCap, LoadFlag, MyLoad)
RETURN
END IF
CALL SimCool ingCrossoverPipe( RunFlag, CompNux, MyLoad ,Flowlock)
Node(OutletNodel)%Temp = Node(InletNode2)%Temp
Node(OutletNode2)%Temp = Node(InletNodel)%Temp
Node(OutletNodel)%MassFlowRate = MassFlowRate
Node(OutletNode2)%MassFlowRate = MassFlowRate
Node(InletNodel)%MassFlowRate = MassFlowRate
Node(InletNode2)%MassFlowRate = MassFlowRate
Node(InletNodel)%MassFlowRateMaxAvail = MassFlowRate
Node(OutletNodel)%MassFlowRateMaxAvail = MassFlowRate
Node(OutletNode2)%MassFlovRateMaxAvail = Node(InletNode2)%MassFlowRateMaxAvail
Node(OutletNode2)%MassFlowRateMinAvail = Node(InletNode2)%MassFlowRateMinAvail
Figure A.10 Cooling crossover pipe simulation
MassFlowRate = 0
!set inlet and outlet temperature variables
InletTemp = Node(InletNode2)%Temp
DeltaTemp = Node(InletNode2)%Temp - Node(InletNode2)%TempSetPoint
IF (DeltaTemp .LE. 0) RETURN
IF (MyLoad == 0
OR.
NOT. RunFlag)RETURN
! If FlowLock is True, the new resolved mdot is used to update Purchased Cooling.
IF (FlowLock==0) THEN
IF (DeltaTemp /
0) THEN
DeltaTemp = ABS(Node(InletNode2)%Temp - Node(InletNodel)%Temp)
MassFlowRate = ABS(MyLoad/CPCW(Node(InletNode2)%Temp)/DeltaTemp)
ELSE
CALL ShoFatalError('DeltaTemp = 0 in SimPurchasedCooling mass flow calculation
END IF
ELSE
! If FlowLock is True
MassFlowRate = Node(InletNode2)%MassFlowRate
END IF
Figure A.11 Cooling crossover pipe calculation
126
)
Figure A. 11 shows the calculation portion of the subroutine SimCoolingCrossoverPipe.
Flow through the pipe is only allowed if the secondary side inlet temperature is greater
than the secondary side setpoint. If so, the mass flowrate is calculated from Eq. (4.21).
Any remaining flow will be routed through the bypass by the EnergyPlus flow resolver.
The heating crossover pipe is simulated similarly with appropriate signs reversed.
Complete source code including the crossover pipe and controlled crossover pipes is
included on the attached compact disc.
A.4 Baseboard Heater
Several bug fixes have been implemented in the BASEBOARD
HEATER:WATER:CONVECTIVE simulation routine. Figure A.12 shows the
simulation of the baseboard heater in the subroutine SimHWConvective in the module
BaseboardRadiator.f90; the actual calculation portion of the subroutine has been removed
for brevity. Two bug fixes are included: in the original code, the water mass flowrate
was not set to zero if the component was not active, and the calculated flowrate was
never stored in the Baseboard data structure. In addition, the original code set the air
mass flowrate constant and equal to a constant convective airflow speed
(SimpConvAirFlowSpeed, 0.5 m/s), which is clearly an error, because the total flowrate
depends on both the speed and the cross sectional area. Here, the air flowrate is still
assumed constant, but it has been multiplied by the density of air and the approximate
area of the trench heaters and chilled beams for Building A to yield a flowrate (5.4 kg/s).
Finally, new control conditions appropriate to Building A have been implemented. The
water mass flowrate is set to zero if the air temperature is between 20 and 25'C, allowing
for a deadband where no chilled beams or trench heaters are active. The water flowrate is
also set to zero if both the water and air temperatures are less than 20'C; this prevents a
chilled beam from attempting to heat cold air.
WaterInletTemp = Baseboard(BaseboardNum)%WaterInletTemp
AirInletTemp = Baseboard(BaseboardNum)%AirInletTemp
CpWater - CPHW(WaterInletTemp)
CpAir = CpAirFn(Baseboard(BaseboardNum)%AirInletHumRatAirInletTemp)
!AirMassFlovRate = SimpConvAirFlovSpeed
AirMassFlowRate - 5.4
WaterMassFlovRate = Node(Baseboard(BaseboardNu)%WaterInletNode)%MassFlovRate
CapacitanceAir = CpAir * AirMassFlovRate
IF ((AirInletTemp GE. 20) and. (AirInletTemp .IE. 25)) WaterMassFlovRate - 0 !ELO 2/8/02
IF ((WaterInletTemp I.T.20) and. (AirInletTemp LT. 20)) WaterMassFlowRate = 0 1ELO 2/15/02
IF (GetCurrentScheduleValue(Baseboard(BaseboardNum)%SchedPtr) .GT.0. &
.AND. WaterMassFlovRate.GT.0.0) THEN
'ALL CALCULATIONS PERFORMED HERE
ELSE
WaterMassFlovRate = 0. 'ELO 2/15/02
CapacitanceWater = 0
CapacitanceMax - CapacitanceAir
CapacitanceMin = 0.
NTU = 0.
Effectiveness = 0.
AirOutletTemp - AirInletTemp
WaterOutletTemp = WaterInletTemp
LoadMet = 0.
Baseboard(BaseboardNum)%WaterOutletEnthalpy
END IF
Baseboard(BaseboardNum)%WaterInletEnthalpy
Baseboard(BaseboardNu)%WaterOutletTemp = WaterOutletTemap
Baseboard(BaseboardNu)%AirOutletTemp = AirOutletTemp
Baseboard(BaseboardNux)%Pover = LoadMet
Baseboard(BaseboardNum)%WaterMassFlowRate = VaterMassFlowRate
ladded by ELO 2/7/02
Figure A.12 Baseboard heater simulation with calculations removed
127
K
A.5 Air System
Several small changes were necessary to the air system models as discussed in Chapter
4.6. The air system input processing code was easily modified to allow for as many inlets
and outlet as desired. Figure A. 13 shows the code for processing the input statements for
the air handling system outlets, found in the subroutine GetAirPathData in the module
SimAirServingZones.f9O. The air handling system inlets are handled similarly.
Code to simulate air system mixers was easily implemented in the subroutine
UpdateBranchConnections in the module SimAirServingZones.f9O. Figure A. 14 shows
the mixer simulation. The mixer outlet properties are determined by summing the
product of the mass flowrate and each property over all mixer inlets, and then dividing by
the total mass flowrate.
! Get the supply nodes
CALL GetNodeNums (Names ( B) ,NumNodes,NodeNums)
! Allow at most 3 supply nodes (for a 3 deck system)
'IF (NumNodes > 3) THEN
CALL ShowSevereError('Air System:Only 1st 3 Nodes will be used from:'//TRIM(Names(8)))
! ErrorsFound= true.
!ENDIF
IF (NumNodes.EQ.0) THEN
CALL ShowSevereError('Air System:there must be at least 1 supply node in system '//TRIM(Names(l)))
ErrorsFound= .true.
END IF
AirToZoneNodeInf o ( AirSysNum)%NunSupplyNodes - NumNodes
I Allocate the supply node arrays in AirToZoneNodeInfo
Allocate(AirToZoneHodeInfo(AirSysNun)%ZoneEquipSupplyNodeNum(AirToZoneNodeInfo(AirSysNum)%NumSupplyNodes))
Allocate(AirToZoneNodeInfo(AirSysNum)%AirLoopSupplyNodeNum(AirToZoneNodeInfo(AirSysNum)%NumSupplyNodes))
I Fill the supply node arrays with node numbers
DO I - 1, AirToZoneNodeInfo(AirSysNum)%NunSupplyNodes
AirToZoneNodeInfo(AirSysNu)%/ZoneEquipSupplyNodeNum(I) = NodeNums(I)
END DO
Figure A.13 Air loop outlet input processing
IF (PrimaryAirSystem(AirLoopNum)%MixerExists) THEN
! if we are at an outlet branch, pass data through the mixer
IF (PrimaryAirSystem(AirLoopNum)%MixerBranchNumOut . EQ. BranchNum) THEN
OutletNodeNun = PrimaryAirSystem(AirloopNum)%MixerNodeNumOut
Node(OutletHodeNux)%Temp = 0
Node(OutletNodeNum)%HumRat = 0
Node(OutletNodeNum)%Enthalpy = 0
Node(OutletNodeNum)%Press - 0
Node(OutletNodeNux)%MassFlowRate = 0
! set the outlet properties flows
DO InletNum=lPrimaryAirSystem(AiroopNum)%MixerTotalInletNodes
InletNodeNum = PrimaryAirSystem(AirLoopNum)%Mixer%NodeNumIn(InletHum)
MassFlowRate = Node(InletNodeNum)%MassFlowRate
Node(OutletNodeNum)%Temp = Node(OutletNodeNum)%Temp + MassFlowRate*Node(InletNodeNum)%Temp
Node(OutletNodeNum)%HumRat = Node(OutletNodeNum)%HumRat + MassFlowRate*Node(InletNodeNum)%HumRat
Node(OutletNodeNum)%Enthalpy = Node(OutletNodeNum)%Enthalpy + MassFlowRate*Node(InletNodeNun)%Enthalpy
Node(OutletNodeNum)%Press = Node(OutletNodeNum)%Press + MassFlowRate*Node(InletNodeNum)%Press
Node(OutletNodeNum)%MassFlowRate = Node(OutletNodeNum)%MassFlowRate + MassFlowRate
END DO
IF (Node(OutletNodeNum)%MassFlowRate NE. 0) THEN
Node(OutletNodeNum)%Temp = Node(OutletNodeNum)%Temp/Node(OutletNodeNum)%MassFlowRate
Node(OutletNodeNum)%HumRat = Node(OutletNodeNum)%HumRat/Node(OutletNodeNun)%MassFlowRate
Node(OutletNodeNum)%Enthalpy = Node(OutletNodeNum)%Enthalpy/Node(OutletNodeNum)%MassFlowRate
Node(OutletNodeNum)%Press = Node(OutletNodeNum)%Press/Node(OutletNodeNum)%MassFlowRate
ENDIF
END IF
END IF
Figure A.14 Air loop mixer simulation
128
A small but very influential bug was found in the simulation of the supply plenum. For a
variable volume system, after the zone has been simulated and the required flowrate
determined, this flowrate must be passed backwards up the supply air path, so the total
required flowrate is known by the air handling unit. However, the EnergyPlus authors
inadvertently placed the reversed indexing used to step backward through the supply air
path in the wrong DO loop, so that this information was never passed backwards, and the
total air flowrate would not vary appropriately. The corrected simulation, found in
subroutine SimZoneEquipment in module Zoneequipmentmanager.f90, is shown in
Figure A.15.
Finally, two bugs affecting the control of the economizer were found in the subroutine
CalcOAController in module MixedAir.f90. The minimum fraction of outside air,
defined as the ratio of the minimum allowable outside air mass flowrate to the total air
loop mass flowrate, was not originally calculated using the correct variables, but was
easily corrected. In addition, the outside air fraction was being set to the minimum
fraction if the outside air temperature was greater than the mixed air setpoint. This is not
a logical rule, because the outside air could still be cooler than the return air, in which
case 100% outside air should be used. This condition was removed from the economizer
control simulation. Again, complete program code including these changes is on the
attached compact disc.
! Process supply air path components in reverse order
DO SupplyAirPathNun = NuaSupplyAirPaths, 1, -1
DO CompNun = SupplyAirPath(SupplyAirPathNum)%NunOfComponents, 1. -1
SELECT CASE (SupplyAirPath(SupplyAirPathNum)%ComponentType(CompNum))
CASE ('ZONE SPLITTER')
CALL SihAirLoopSplitter(SupplyAirPath(SupplyAirPathNum)%ComponentName(CompNum),
FirstHVACIteration, FirstCall, SplitterInletChanged)
&
IF (SplitterInletChanged) THEN
If the Splitter inlet conditions have been changed, the Air Loop must be resimulated
SimAir =
TRUE.
END IF
CASE ('ZONE SUPPLY PLENUM')
CALL SinAirZonePlenux(SupplyAirPath(SupplyAirPathNum)%ComponentName(CompNum),
FirstHVACIteration, FirstCall)
CASE DEFAULT
CALL ShowFatalError('Invalid Supply Air Path Component='// &
TRIM(SupplyAirPath(SupplyAirPathNu)%ComponentType(CompNum)))
END SELECT
END DO
Figure A.15 Supply air path reverse simulation
129
&