Applied Energy 66 (2000) 135±159

www.elsevier.com/locate/apenergy

Developing cost ecient control strategies

to ensure optimal energy use and sucient
indoor comfort
E.H. Mathews1, D.C. Arndt*,1, C.B. Piani1, E. van Heerden1
Department of Mechanical Engineering, University of Pretoria, South Africa

Received 5 October 1998; received in revised form 3 June 1999; accepted 4 June 1999

Abstract
Good HVAC control is often the most cost-e€ective option to improve the energy eciency
of a building. However the e€ect of changing the control strategy (i.e. e€ect on indoor com-
fort and energy consumption) is usually the most dicult to predict. To achieve this more
easily, a new simulation tool, QUICKcontrol, was developed. In this paper, the new tool is
used to investigate the energy savings potential in the Engineering Tower Building (ETB) of
the University of Pretoria. The in¯uence of reset control, economizer cycle combined with
CO2 control on the outside air ventilation rate and better system start-stop times were inves-
tigated. The simulation models were ®rstly veri®ed with actual measurements obtained from
the existing system to con®rm their accuracy for realistic control retro®t simulations. With the
aid of the integrated simulation tool it was possible to predict savings of 491 MWh per year
(34% building energy saving) by implementing these control strategies. These control strate-
gies can be implemented in the building with a direct payback period of less than 9 months.
# 2000 Elsevier Science Ltd. All rights reserved.

1. Introduction

Heating, ventilation and air-conditioning (HVAC) system energy eciency in

buildings means monetary savings for the owner and less greenhouse gases being
released into the atmosphere. Although very important, energy saving measures

* Corresponding author. Tel.: +27-012-420-9111; fax: +27-012-432-816.

E-mail address: 9192308@postino.up.ac.za (D.C. Arndt).
1
Also consultants of TEMM International (Pty) Ltd.

0306-2619/00/$ - see front matter # 2000 Elsevier Science Ltd. All rights reserved.
PII: S0306-2619(99)00035-5
136 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

Nomenclature

AHU air-handling unit

ETB engineering tower building
FCU fan coil unit
HVAC heating, ventilation and air-conditioning
IAQ indoor air quality
PID proportional, integral and derivative

must never compromise indoor air quality (IAQ). The reason is that IAQ has a
direct e€ect on the productivity of the occupants [1]. The cost associated with poor
IAQ far outweigh savings due to reduced energy consumption [2]. Popular belief in
the past was that good IAQ and energy eciency were in direct con¯ict [3]: it is now
agreed that energy savings of around 30% may be realised through retro®t projects
of existing buildings without compromising indoor comfort [4]. In South Africa,
studies have shown that, in the commercial sector, approximately 50% of energy is
used for air-conditioning [5]. This statistic clearly indicates that the area which o€ers
the most potential for energy saving is the HVAC system.
A cost e€ective way to improve the energy eciency of a HVAC system, without
compromising indoor comfort, is by implementing better control. However when
changing the control strategy of a system it is often dicult to predict the resulting
changes in system energy consumption and indoor comfort. To achieve these pre-
dictions, a simulation tool which can eciently and accurately simulate the building
with its HVAC system and controls in an integrated fashion is required. There are
many system simulation programs available. However they do not satisfy the
requirements of integrated, ecient and accurate simulation by the typical consult-
ing engineer [6±8]. A new simulation tool, QUICKcontrol was therefore developed
to satisfy these requirements. The models on which the simulation tool is based are
brie¯y discussed in Section 2. A case study was conducted to illustrate the value and
the capabilities of this new simulation tool. The potential for energy savings and
enhancement of indoor comfort by implementing new control strategies in the
Engineering Tower Building (ETB) of the University of Pretoria was successfully
investigated.

2. Simulation models

An electrical analogy [9] is used to model the heat transfer processes in the building
for the accurate prediction of the thermal performance of the building zones. The
HVAC component models link inputs and outputs of the basic thermodynamic vari-
ables in the system. They are based on simpli®ed fundamental principles combined with
correlation coecients derived from discrete empirical data [6]. The models are fully
component-based and allow simulation of a wide range of operating conditions. The
 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 137

calculation of the energy consumption of each component is included in each model.

The correlation coecients for a speci®c make and model of equipment can be
derived from data obtained from measurements or manufacturers' data sheets. To
simulate dynamic e€ects, a simple time constant approach is used. The user is
responsible for supplying the time constants. The approach is as follows:

d
 ˆ function …model input parameters† …1†
dt

with  the time constant of the model, and one of the output parameters of the
model. The relevant psychrometric relationships are employed in all models dealing
with moist air properties.
At present the simulation tool makes provision for proportional, integral, deriva-
tive (PID), on/o€ and step controllers. These controllers are used in most HVAC
applications. With these controllers, any measurable condition can be controlled
from a sensor. Water ¯ow rates, air ¯ow rates, steam ¯ow rates and the load capa-
cities of the system components are controllable variables. Controller output at each
step is only dependent on the previous time step values. This considerably reduces
the complexity of the solution algorithm. From a system point of view, this implies
that the controller acts like a controller that has a sampling rate corresponding to
the system integration time step size.
There are also energy management systems included in the simulation tool. With
these, the system's energy consumption can be reduced by more energy-ecient
control. The simulation tool provides the following energy management strategies:
. Setpoint-related energy management systems (temperature reset, zero energy
band control, enthalpy economizer cycle and adaptive comfort control).
. Advanced energy management strategies (demand limiting, duty cycling, load
resetting and CO2 control).
These energy management strategies are discussed in Ref. [10]. An integrated
simulation is accomplished by solving the combined performance of the building
model and the HVAC component models including the in¯uence of the controllers.

3. Building and HVAC system description

The building consists of oces above ground level and lecture halls underground.
The building has 15 ¯oors with levels 1 and 2 underground. All the oces have
adjustable external louvres for shading on the oce windows. The building's oce
windows face either east or west. Each oce is air-conditioned by its own fan coil
unit (FCU) and the lecture halls by air-handling units (AHUs) located in a plant
room. The total air-conditioned ¯oor area is 4098 m2. The building houses an average
of 600 people on weekdays. Lighting is done by ¯uorescent bulbs in the oces as well
as in the lecture halls. Most of the oces are equipped with computers and the lecture
halls with projectors.
138 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

The HVAC system consists of a central cooling plant, AHUs for the lecture halls,
an AHU for pre-conditioned air for the oces and FCUs located in the oces. The
system operates 24 h a day from Monday mornings through to Saturday afternoons.
The HVAC system layout is displayed in Figs. 1 and 2. Three AHUs supply a con-
stant volume of air to the lecture halls as illustrated in the schematic layout of the
lecture halls in Fig. 2. The air is conditioned by cooling coils located inside the AHUs
and an electric heater in the supply air duct of each lecture hall. Air is returned from
the lecture halls via grills and doors back to a common duct in the plant room,
where it mixes individually with outside air in each AHU (Fig. 2).

Fig. 1. Schematic layout of the HVAC system plant room.

 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 139

Fig. 2. Schematic layout of the lecture halls and one oce.

AHU 1 serves only one lecture hall and the indoor air temperature is controlled
from a temperature sensor located in the lecture hall. AHUs 2 and 3 serve more than
one lecture hall. The supply air temperatures of these two units are controlled from
temperature sensors located inside the supply air ducts downstream of the cooling
coils. These units supply air at constant temperature throughout the year. Individual
temperature control in the lecture halls is thus done by reheating. The output of the
electric heaters is controlled from temperature sensors in the lecture halls.
AHU 4 supplies pre-conditioned air via ducting to the fan coil units of the oces.
This unit makes use of 100% outside air and supplies air at constant temperature
throughout the year. Return air from each oce is mixed with pre-conditioned air in
its FCU. This air is then conditioned by a cooling coil and an electric heater before
supplying it back to the oce. The cooling and heating capacity of the FCUs are
140 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

only manually adjustable for indoor air temperature control. Two parallel water-
cooled chillers supply chilled water to all the cooling coils in the building, including
the coils of the FCUs. The condenser water of the chillers is cooled by two cooling
towers on the roof of the building. PID controllers are used to control cooling and
step controllers for heating.

4. Comfort and energy audit

The aim of the comfort audit was to evaluate the current indoor air conditions in
the ETB. Measurements were necessary to determine whether the current indoor air
conditions were up to standard and also to verify the simulation model. These
indoor conditions, if satisfactory, could then be used as standards for the evaluation
of retro®t options. The audit therefore consisted of measurements of actual comfort
levels, as well as a walk through audit. A walk through audit was conducted of the
ETB to determine any particular problem areas. The occupants opinions were also
obtained regarding comfort in the lecture halls.
To decrease energy consumption without a€ecting the comfort of the building
occupants, it was necessary to determine an end-user breakdown of the energy con-
sumption in the ETB. This could then be used to determine the largest energy con-
sumers, which are usually the areas with the largest energy savings potential. The main
aim therefore of the energy audit was to determine the end-user breakdown of the
energy consumption in the ETB. Measuring the energy consumption of the ETB was a
very labour intensive process. It was begun with a walk through audit, followed by
detailed measurements. The walk through audit was conducted to identify the usage of
lighting and other diverse equipment such as lifts, computers, etc. The consumption of
these energy consumers is important to determine if they can be improved.
The categories used in the energy breakdown are HVAC, lighting, computers and
diverse equipment (Fig. 3). Diverse equipment includes the four lifts operating in the
building, overhead projectors, etc. HVAC energy consumption was then broken
down into the various components of the HVAC system (Fig. 4). The HVAC system
in the ETB has the greatest energy saving potential as it is the major energy consumer.
It consumes approximately 64% of the total building energy. The reason for this is

Fig. 3. Breakdown of building's total electrical energy consumption.

 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 141

Fig. 4. HVAC system's electrical energy consumption breakdown.

the large internal load in the form of people in the lecture halls, which can accom-
modate as many as 714 people.
The control strategy of the ETB will be investigated to determine its energy sav-
ings potential. The energy audit identi®ed the major energy consumers as well as
areas where retro®ts could be made to obtain energy savings.

5. Veri®cation study

A veri®cation study was performed to the accuracy of the simulation models

(including control parameters) to calculate realistic energy savings with the retro®t
simulation options. The main purpose of the veri®cation study was not so much to
verify the building model and the HVAC models individually but to verify the ``real
life'' dynamic interaction between the building model and the HVAC system models
with the in¯uence of the controllers in an integrated fashion. The building model has
already been empirically veri®ed over a wide range of di€erent building structures
and climates in 140 case studies with great accuracy [9,10]. The building model can
therefore been used with con®dence without verifying it again, if it is properly
constructed.
For the veri®cation study, it was necessary to take measurements, set up the
simulation tool, execute the simulations and compare the measurements with the
predicted values. Measurements were taken over a period of two weeks during
March 1997. The measurements include temperatures, relative humidities, air ¯ow
rates, water ¯ow rates and electrical power required by each HVAC system compo-
nent. From these data, a single representative day was extracted to be compared to
the 24-h day predictions provided by the simulation tool.
The integrated simulation tool requires a large number of inputs. These inputs
comprise building structural data, outdoor climatic data, internal loads, occupancy
data, all the HVAC component data and the control parameters. Some of these
inputs vary with time, namely in our case, outdoor climate data, occupancy, internal
loads and the FCU usage by the occupants. In the case of the ETB, it was impossible
to monitor the occupancy and the FCU usage, which have a large in¯uence on the
HVAC system load. Most of the input data or if possible all of it should be available
to perform a veri®cation study with integrity.
142 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

A public holiday was therefore used for the 24-h veri®cation day to eliminate all
the inputs, which were impossible to monitor. During the public holiday, the building
was unoccupied, lights o€ and the FCUs not in use. The ideal would obviously be to
use a more typical day with more variation in the building load. This option how-
ever could have jeopardised the integrity of the study since all the inputs were not
available. This speci®c control strategy and the swing of the outside climate tem-
perature still allowed for enough variation to verify the interaction between the dif-
ferent components and the performance of the system.
The building was divided into ®ve zones. Each lecture hall zone represents the
lecture halls served by the same AHU. Lecture hall 1±12 is represented by zone 1.
Lecture halls 1±3, 1±4, 1±15 and 1±16 are simulated as zone 2. The rest of the lecture
halls 1±1, 1±2, 1±7, 1±8 and 1±9 are simulated as zone 3. Fig. 5 depicts the simula-
tion model layout. The oces make up the other two zones. The eastern side of the
building was simulated as one zone and the western side as the other zone. The walls
and ¯oors inside the zones are treated as partitions. The building structure data with
the zone dimensions were obtained from building drawings and a building investi-
gation. All of this was read into the simulation program. There were no internal
loads in the lecture halls since the day used for veri®cation was a public holiday as
mentioned before. The building space loads were therefore only caused by the out-
side air and the building envelope.
The input data needed for the mathematical models of the HVAC components
were obtained from the suppliers' performance data sheets. The measured chiller

Fig. 5. Computer layout of the simulation model.

 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 143

capacities did not match the supplier's performance data. This could be due to
fouling resistance since the chillers are more than 20 years old. The chiller model was
therefore obtained from measured data. The two chillers were simulated as one with
double the capacity with the necessary capacity steps. The two cooling towers on the
roof were also simulated as one with double the capacity. The control strategy,
including operating times and control parameters for most system components, were
obtained from the system's operating manuals. The parameters not available in the
manuals were obtained from measurements. The veri®cation results of the indoor air
temperatures, AHUs supply air temperatures, the chiller power, and the electric
heater power are presented graphically in this section (Figs. 6±14). The measured
indoor air temperature of each zone is the average temperature of all the lecture
halls in that zone. The indoor air temperatures of the lecture halls in the same zone
did not di€er much due to the same temperature set-point. The indoor air tempera-
tures of the oces were not used to verify the simulation model as it was not possible
to measure all 200 oces' temperatures at the same time. During the veri®cation
study, the oces did not have any in¯uence on the HVAC system as discussed
previously.
Figs. 6±8 display the simulated and actual measured indoor air temperatures of
each lecture hall zone. It is clear, from the ®gures, that satisfactory results were
obtained from all three zones. The indoor air temperature results are summarised in
Table 1. Only zone 3's result was not within 1 C for 100% of the time. The tem-
perature pro®le of the supply air of AHU 1 (Fig. 9) into building zone 1 must be
re¯ected in the return air (Fig. 6) when the space load is very small. It could be more
damped and out of phase due to the thermal mass of the building structure, but it
cannot be a straight line. Thus the only explanation for the measured temperature

Fig. 6. Indoor air temperature veri®cation results of zone 1.

144 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

pro®le in Fig. 6 is the possibility of strati®cation of the indoor air at the measuring
point. The traces in Figs. 7 and 8 do not follow one another exactly due to the
comparison of the average measured temperature of more than one lecture hall to
the simulated temperature of one representative zone. The re-heat dynamic respon-
ses of the individual halls are therefore not completely correctly modelled due to this
assumption.

Fig. 7. Indoor air temperature veri®cation results of zone 2.

Fig. 8. Indoor air temperature veri®cation results of zone 3.

 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 145

In Figs. 9±12, the measured and simulated AHUs supply air temperatures are
compared. The results are satisfactory. The dynamics of the coils and controllers can
be observed in Fig. 11, AHU 3. The bypass control valve and the electric heater
banks of AHU 4 did not respond to any supply air temperature changes to maintain
the constant set supply air temperature over the two week measuring period. The
valve position stayed at 10% open, which resulted in a constant water ¯ow rate
through the coil and the heaters were never switched on throughout this period.

Fig. 9. AHU 1 supply air temperature veri®cation results.

Fig. 10. AHU 2 supply air temperature veri®cation results.

146 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

These ®ndings were used as inputs during veri®cation. The good comparison between
the simulated and measured values displayed in Fig. 12 con®rms these ®ndings.
The simulated and measured chiller power are compared in Fig. 13. The simula-
tion tool successfully simulated the loading and unloading of the chiller capacity
steps. There is not a large phase di€erence between the predicted and simulated
values. This veri®es the time constants of the building structure and the HVAC

Fig. 11. AHU 3 supply air temperature veri®cation results.

Fig. 12. AHU 4 supply air temperature veri®cation results.

 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 147

system components. The di€erence between the measured and simulated energy
consumption over 24 h is only 3%. The total measured heater power of the system
was compared with the total simulated heater power. It was not possible to measure
the heater banks separately due to the number of heaters, so the total power was
measured in the plant room. Fig. 14 displays the heater power veri®cation results.
The simulated values do not follow the measured values exactly but the total energy

Fig. 13. Chiller power veri®cation results.

Fig. 14. Heater power veri®cation results.

148 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

Table 1
Summary of indoor air temperature veri®cation results

Zone Maximum error ( C) Within 1 C (%) Within 2 C (%)

1 0.82 100 100

2 0.66 100 100
3 1.3 98 100

consumed over 24 h compares well. There is a di€erence of 1% between the mea-

sured and simulated heater energy consumption over the 24 h day.
Twenty-four time steps per hour were needed to simulate the dynamics of the
building zones and the HVAC system with its controls to this degree of accuracy.
One may look at the simulation results and feel that the accuracy of the integrated
modelling is perhaps not that good. In some instances, the predicted and measured
values are out of phase and the graphs do not always have exactly the same pro®les.
However the objective was not to simulate the dynamics of the system 100% cor-
rectly but to predict accurate realistic daily (and yearly) system energy consumption
and accurate average indoor air temperatures to ensure indoor comfort.
The major contributing factor for most of the discrepancies is, that the system
conditions do not stay constant during each time interval due to the quasi-steady
state approach used by the simulation model. This implies that the simulation results
(dynamic response) are very sensitive to the time step size. For a system with a very
high dynamic response like this one, the time step size must be very small to simulate
the response correctly. This will however extend the simulation times, which one
would like to avoid in the case of year long simulations if this type of accuracy is not
necessary. Another factor to remember, which contributes to discrepancies is the
accuracy of the measuring equipment and of the sensors of the HVAC system. The
building has a very old pneumatic control system, which allows for the possibility
that the system does not operate any longer according to speci®cation. The results
obtained from the veri®cation study were suciently good for con®dent use of the
model in the retro®t simulations.

6. Control retro®t simulations

6.1. Preamble

An investigation was performed to establish the potential for energy savings in the
building. The savings were calculated by executing numerous energy saving control
retro®t simulations. These retro®t options include reset control on the AHUs,
economizer cycle combined with carbon dioxide control on the lecture halls' out-
door air ventilation rates, and better HVAC system start-stop times. All the retro®t
options investigated were done on the HVAC system of the building, since 64% of
 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 149

the total building energy consumption is consumed by the HVAC system. This
means that the biggest potential for the building's energy saving lies in HVAC sys-
tem retro®t options.

6.2. Reset control

The ®rst option investigated to establish the potential for energy savings in the
building was a reset control strategy on the AHUs. The existing system's AHUs
supply a constant temperature throughout the year. This temperature is set to
accommodate the maximum space load when the lecture halls are fully occupied
during a hot day. When the lecture halls are not fully occupied, the supply air is
unnecessarily cooled down to the AHU setpoint just to be heated up again to satisfy
the indoor air setpoint. By doing this, energy is wasted on cooling as well as on
heating.
To conserve energy on a system which makes use of reheating for temperature
zone control, heating should not be used unless absolutely necessary. Reset control
should be provided to maintain the cold air from the AHUs at the highest possible
temperature to satisfy the maximum space cooling requirement [11]. The above
control strategy was simulated in this building to establish the potential for energy
savings. The simulation model was derived from the model, which had been used in
the veri®cation study. The lecture halls were divided into two more zones to illus-
trate the control strategy and to calculate realistic energy savings. To make this
control strategy even more e€ective, it is necessary to balance the air ¯ows for each
lecture hall to the ratio of the design occupancy or ¯oor area of the lecture halls.
Year long simulations were executed to calculate the potential for energy savings
per year. The simulations were executed under the assumption that the lecture halls

Fig. 15. Chiller's electrical energy consumption for each month of the year.
150 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

Fig. 16. Heater's electrical energy consumption for each month of the year.

are only 60% occupied from 07:00 to 18:00 for 70% of the days of the year when the
system is running. This means the lecture halls are empty for the other 30% of
the days of the year. At present the system is o€ during weekends from 15:00 on
Saturday afternoon until 04:00 on Monday morning. Only 80% of the oces' FCUs
were running 24 h a day throughout the year. These assumptions are based on an
investigation into the occupancy of the building.
The year long simulation results are displayed in Figs. 15 and 16. These results
showed an energy saving of 21% (only reset control) on the existing system's total
consumption. This is a saving of 200 000 kWh per year. A worst case scenario with
50% of the lecture halls 60% occupied and the other 50% empty was simulated with
a saving of 180 000 kWh per year. This is, however, not at present an average
weekday since all the big lecture halls are more or less occupied during the same part
of the day.

6.3. Air economiser combined with CO2 control (including reset control)

To save more energy in the building, an investigation into the implementation of

an air economizer combined with carbon dioxide (CO2) control was performed. This
new control strategy was investigated with the inclusion of reset control on the
AHUs, as discussed in the previous section.
Studies have concluded [12] that 7 l/s of outdoor air ventilation per person will
control human body odour, such that roughly 80% of the unadapted persons will
®nd the odour at an acceptable level. The same level of odour acceptability was
found to occur at a CO2 concentration that is about 700 ppm (parts per million)
above the outdoor concentration, which at a typical outdoor level of 350 ppm yields
 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 151

an indoor CO2 concentration of 1050 ppm. The di€erence between indoor and out-
door levels of 700 ppm has become a measure of acceptability with respect to body
odour, irrespective of the outdoor CO2 concentration.
The CO2 control strategy used here is based on the above discussion. The follow-
ing CO2 control strategy was implemented to control the outside air dampers of the
lecture halls' AHUs (Fig. 17):

. If the CO2 concentration is below 700 ppm in a lecture hall, the outside air
damper should be 3% open.
. If the CO2 concentration exceeds 700 ppm in a lecture hall, the outside air
damper will open proportionally until it is fully open at 1050 ppm.

Fig. 17. CO2 control strategy.

Fig. 18. Complete retro®t control strategy.

152 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

Due to only one control damper for more than one lecture hall, the damper is
controlled from the lecture hall which needs the most outside air to maintain an
acceptable indoor CO2 concentration. This CO2 control strategy was combined with
an air temperature-controlled economizer which controls the same outside air damper.
The following economiser control logic was implemented (Fig. 18):

. The outside air damper should be 3% open if the outside air temperature
exceeds the lecture hall indoor air temperature setpoint or if the indoor air tem-
perature is below (no cooling needed) the indoor air setpoint (set point 22 C).
. If the indoor air temperature of the lecture hall exceeds the set point (cooling
needed), the damper will open proportionally until 100% open at 1 C above
the set point (23 C).
. Due to only one control damper for more than one lecture hall, the damper is
controlled from the lecture hall which requires the most cooling to maintain
the indoor air temperature.
. The cooling valve will now open only when the indoor lecture hall temperature
exceeds the set point by 1 C (23 C).
. The heaters will come on when the indoor air temperature is 0.5 C below the
set point.

Due to the combination of the two control strategies on the same damper, the
damper should always respond to the maximum outside air demand.
Fig. 19 displays the simulated indoor air temperature of one of the zones on a
summer day. The temperature drifts between 21.5 and 23 C in relation to the outside
air temperature. The acceptable temperature comfort band is between 21 and 24 C.

Fig. 19. Indoor air temperature, zone 2-2.

 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 153

The graph illustrates that this speci®c control strategy automatically implements a
strategy similar to adaptive comfort control. Adaptive comfort is based on the
assumption that the comfort perception of people will be dependent on the outdoor
climatic conditions [13]. For low outside air temperatures, people will be comfor-
table if the indoor temperature is lower. The opposite is true for high outside air
temperatures. This a€ords us the opportunity to potentially save even more energy.
If the indoor temperature can be cooler when the outside air temperature is low, less
heating would be required. During periods of high outside air temperatures, a higher
temperature would imply that less cooling would be required.
A year long simulation was executed with the same control and assumptions as
used previously to establish the potential for energy savings per year. The yearly
chiller and heater energy consumptions of the system with reset control are com-
pared to the reset control, CO2 and economizer control in Figs. 20 and 21. With this
combination of control strategies it was possible to save an additional 10% by
reducing cooling and heating energy consumption which makes the total saving
31% on the existing system's total consumption. The total simulated energy saving
per year is now 286 000 kWh (Table 2, option 1). The CO2 levels of all the lecture
halls were equal to, or less than 850 ppm for the entire year.

6.4. Better system start-stop times (including reset, CO2 and economizer control)

The last option to save building energy investigated in this study is the HVAC
system component start-stop times. This option can be applied where the indoor
temperatures have to be at the correct temperature only at speci®ed times. By
switching o€ the system at other times large amounts of energy can be saved.

Fig. 20. Chiller's electrical energy consumption for each month of the year.
154 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

Two options were investigated: switching o€ the FCUs of the oces and the cen-
tral system, including the cooling plant, during unoccupied building times. The ®rst
option mentioned was simulated with the FCUs o€ from 17:00 in the evening to
08:00 the next morning. Each FCU can be switched on manually when air-con-
ditioning is needed at any time during the day. The second option was investigated
with the entire HVAC system o€ from 22:00 at night to 06:00 in the morning with
the FCUs still o€ as mentioned in the ®rst option. It was necessary to start the sys-
tem at 06:00 in the morning to get the indoor temperatures to the set point tem-
peratures before 07:00.
The energy consumed by cooling did increase by 3% with both options included,
but the heater energy decreased by 45%. Figs. 22 and 23 display the indoor air
temperatures of a western oce and a lecture hall on a hot day, before and after the
retro®ts (worst case scenarios). The indoor air temperature of the oce stays below
24 C for the entire day with the retro®t, even when the cooling plant is o€ from
22:00 until 06:00 the next morning. The fan of the oce's FCU was running for the
entire 24 h. The indoor temperature of the lecture hall stays below 23.5 C from
06:00 until 22:00 in the evening after the retro®ts. The retro®t options still satisfy the
conditions for acceptable indoor comfort on a hot day.
Fig. 24 displays the indoor air temperature of the same lecture hall examined
previously on a cold day before and after the retro®ts. The indoor air conditions of
the lecture hall between 06:00 and 22:00 have improved with the retro®t. The indoor
air temperature of the oce, which is not displayed here, on a cold day stayed
exactly the same after the retro®t, due to individual control of each oce fan and
heater by the occupant. The main HVAC system has no in¯uence on the electric
heaters of the oces.

Fig. 21. Heater's electrical energy consumption for each month of the year.
 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159
Table 2
Economic analysis results

Retro®t option combinations Energy Saving fraction Saving fraction Straight Internal rate of return
savings of HVAC energy of total building energy payback period (%)
(kWh/year) (%) (%) (months)
3 years 4 years 5 years

1 Reset, CO2 and economiser control 286,000 31 20 15 52 59 62

2 Reset, CO2, economiser control and 328,000 35 22 13 63 70 73
FCU start±stop
3 Reset, CO2, economiser control and 491,000 53 34 9 109 115 116
total system start±stop
4 Reset, economiser control and total 435,000 47 30 5 223 226 227
system start±stop
5 Total system start±stop 192,000 21 13 1 1214 1214 1214

155
156 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

These results illustrate that the energy saving retro®t options will not reduce the
indoor comfort of the building at any time. Fig. 25 displays the total yearly HVAC
system energy consumed by the existing system and the system with the retro®t
options. Large energy savings were possible by switching the entire system o€ during
certain times as mentioned before (Table 2, options 3±5). The reason for this is the
high percentage of energy consumed by the pumps of the cooling plant and the fans
of the AHUs (48% of total system energy consumption).

Fig. 22. Oce's indoor air temperatures for a hot day.

Fig. 23. Lecture-hall's indoor air temperature for a hot day.

 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 157

6.5. Economic analysis

The capital cost of a CO2 sensor is very high and makes the implementation of
this retro®t option expensive. In this case, a minimum of seven sensors will be nee-
ded for proper control. The amount of savings per CO2 unit, due to the combination
of the number of lecture halls and their size, will not be sucient to persuade the

Fig. 24. Lecture-hall's indoor air temperatures for a cold day.

Fig. 25. Annual electrical energy consumption of the HVAC system.

158 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159

building owner to implement this option. Without the CO2 control, the imple-
mentation cost can come down by 40% while still realising a saving of 47% on the
HVAC system's energy consumption (Table 2, option 4). An economic analysis of
all the retro®t options is summarised in Table 2.

7. Conclusions

With the aid of the integrated simulation tool, it was possible to predict savings up
to 491,000 kWh per year by simulating various control retro®t options. Thirty per
cent of this saving was due to reset and CO2 control combined with economizer
control. This type of control strategy is strongly recommended for large auditoriums.
These retro®ts can be implemented in the building under consideration with a direct
payback period of less than 9 months. It is also possible to save up to 435,000 kWh
per year by implementing only reset control on the AHUs, economizer control on
the outside air ventilation rates, and HVAC system start-stop. This retro®t option
can be implemented in the building with a direct payback period of less than 5
months. Most building owners will be tempted to go for an option with such a short
payback period and large savings potential.
Without the aid of the integrated simulation tool, it would not have been possible
to evaluate these building energy savings and still ensure acceptable indoor comfort.
The building owner of the ETB did agree to implement some of the options. This
will provide us with the opportunity to verify the predicted energy savings in the
near future.

Acknowledgements

This work was made possible through funding by the International Institute for
Energy Conservation (IIEC)-Africa and the U.S. Environmental Protection Agency.
The authors wish to acknowledge the contributions made by Lloyd Wright, Je€
Wells, Pieter Strachan and Jason Hamilton.

References

[1] Woods JE. Cost avoidance and productivity in owning and operating buildings. Occupational
Medicine, State of the Art Reviews 1989;4(4).
[2] Sterling E, Collett C, Turner S, Downing C. Commissioning to avoid indoor air quality problems.
Proc. ASHRAE Transactions: Symposia 1994;January:867.
[3] Mathews EH, Piani CB. Establishing the energy savings potential in South African oce buildings.
Refrigeration and Air-conditioning 1996;12(4):59±65.
[4] Bevington R, Rosen®eld, AH. Energy for buildings and homes. Scienti®c American 1990;September:
207±15.
[5] TEMM International (Pty) Ltd. Energy savings potential and guidelines for e€ective energy use in
buildings. Project report compiled for the Department of Mineral and Energy A€airs: Energy
branch, Pretoria, March 1997.
 E.H. Mathews et al. / Applied Energy 66 (2000) 135±159 159

[6] Rousseau PG, Mathews EH. Needs and trends in integrated building and HVAC thermal design
tools. Building and Environment 1993;28(4):439±52.
[7] Hirsch JJ, Winkelmann FC, Buhl WF, Ellington KL. POWERDOE, a windows-based visually
oriented analysis tool. In: Mitchell JW, Beckman WA, editors. Building Simulation '95 Fourth
International Conference Proceedings. Madison (WI): IBPSA, 1995.
[8] Lebrun L. Simulation of HVAC system. Renewable Energy 1994;5(5):1151±8.
[9] van Heerden E, Mathews EH. A new simulation model for passive and low energy. In: Proceedings
of the 1996 PLEA Building and Urban Renewal conference, Louvain-la-Neuve, Belgium, 1996.
[10] van Heerden E. Integrated simulation of building thermal performance, HVAC system and control.
Thesis presented in partial ful®lment of the requirements for PhD degree, Faculty of engineering,
University of Pretoria, Pretoria, 1997.
[11] McQuiston FC, Parker JD. Heating, ventilation, and air conditioning analysis and design. New
York: John Wiley & Sons, Inc, 1994.
[12] Anon. ASHRAE, BSR/ASHRAE Standard 62-1989R, ventilation for acceptable indoor air quality
(Public Review Draft), Atlanta (GA): ASHRAE, 1996.
[13] Humphreys MA. Thermal comfort temperatures and the habits of hobbits. In: Nicol F, Humphreys
MA, Sykes O, Roaf S, editors. Standards for thermal standards. London: Chapman and Hall, 1995.