2. gains. However, static decouplers may not always be able to provide
satisfactory control performance. In contrast, dynamic decouplers
require detailed process models, but they provide better perfor-
mance than static decouplers provide [14,15]. For practical opera-
tions, the emphasis is typically placed on suitability and causality
needs, which makes precise configurations difficult to achieve,
especially for high-dimensional MIMO processes. To settle these
difficulties, most of these methodologies focus on TITO (two input
and two output) systems [16,17]. The main shortcoming of the
dynamic methods lies in the complexities of the decoupler ele-
ments, which are obtained from the apparent process model. The
difficulty becomes greater for sophisticated plants because the
technique incorporates the determinant of the model transfer
function [18]. Additionally, the requirement for the decoupler is
that all of its elements must be proper, causal and stable [19]. A few
studies in the literature have focused on the inverted decoupling
methods that are used to reduce variable interactions in the process
[18e22]. Gagnon [10] demonstrated that the performance of
inverted decoupling depends on the scheme of implementation.
When inverted, decoupling is implemented with a lead-lag and
delay function process, and the control performance retreats.
Normalised decoupling control design methodology was used by
Shen [23]. For this type of decoupling, the ETF (equivalent transfer
function) of each element in the transfer function matrix was
required to derive the closed-loop of the plant model, including the
algorithm of the control system. Then, the decoupler was obtained
by multiplying the inverse of the ETF by a stable, proper and causal
ideal-diagonal transfer function.
This paper seeks to analyse and discover the paramount choice
of controlled parameters in the HVAC systems, which are reflected
in optimisation controller performances. However, the controller's
performance is related to buildings' energy efficiency, which is
most directly affected by the decoupling problem. Therefore, in this
study, the extensive and elaborate models of a building that has
HVAC system components are used to simulate a real system.
Deriving the matrices of decoupling, inverted decoupling or ETF
from such a complex model is challenging because all of its ele-
ments must be proper, causal and stable. In concision, the HVAC
control systems use both temperature and RH as references instead
of using temperature only, which is what the earlier mode did.
Because temperature and RH are coupled, it is difficult to control
them separately for a certain desired value [11].
Nomenclature
Symbols
A surface area, m2
C heat capacitance, J/C
dEs/dt rate of change in storage energy of the system, J/s
E;
in energy rate entering the system, J/s
E:
out energy rate leaving the system, J/s
M mass, kg
Cp specific heat, J/kgC
m: mass flow rate, kg/s
Mcp heat capacitance, J/C
T temperature, Co
u humidity ratio, kgw/kgda
h latent heat/heat transfer coefficient, J/kg, W/(m2C)
Q:
cooling load, WC
CF surface cooling factor, W/m2
U construction U-factor, W/(m2C)
DT cooling design temperature difference, C
OFt, OFb, OFr opaque-surface cooling factors
DR cooling daily range, C
CFfen surface cooling factor, W/m2
UNFRC fenestration U-factor, W/(m2C)
PXI peak exterior irradiance, W/m2
SHGC solar heat gain coefficient
IAC interior shading attenuation coefficient
FFs fenestration solar load factor
Et, Eb, ED peak total, diffuse, and direct irradiance, W/m2
Tx transmission of the exterior attachment
Fshd fraction of the fenestration shaded by overhangs or fins
L site latitude, N
SLF shade line factor
Doh depth of the overhang, m
Xoh vertical distance from the top of the fenestration to the
overhang, m
Fcl shade fraction closed (0e1)
j exposure (surface azimuth), measured as degrees from
south
V;
volumetric flow rate, L/s
DF infiltration driving force, L/(s cm2
)
thermal resistance, C/W
Noc number of occupants
Nbr number of bedrooms
aroof roof solar absorbance
t time constant, s
I infiltration coefficient
Du indooreoutdoor humidity ratio difference, kgw/kgda
Subscripts
m air in mixing box
r room/return
o outside
os outside supply
i inside
He heat exchanger
a air
w water
aHe air in the heat exchanger
L leakage
Win water input
Wout water output
Wl wall
room inside room
out outside room
g glass
fg heat of vaporization
Opq opaque
inf infiltration
fen fenestration
f indoor and outdoor
t at time t
flue flue effective
es exposed
ul unit leakage
ig internal gains
l latent
s sensible/supply
fur furniture
cl closed
R.Z. Homod / Energy 74 (2014) 762e774 763
3. It is possible to solve a problem in which the variables of tem-
perature and relative humidity are coupled. The first modification in
AHUs is the addition of a fresh air pre-cooling coil that is used to
alleviate the coupling intensity, which is particularly necessary in
humid climates. The second modification for control objectives is the
increase of the optimisation parameters of the output controller by
adding a model of the PMV index in order to evaluate indoor thermal
comfort. Next, decoupling and reduction in energy are simulated by
comparing three different systems under real weather conditions
within certain set point comfort limits. The first system is a con-
ventional system in which the objective is to achieve the tempera-
ture and relative humidity that are within the limits of the desired
conditions. The second system is similar to the first, with the only
difference being the addition of a reheating coil and a wet main
cooling coil in AHUs that are used to solve the coupling problem.
However, these additional reheating and wet main cooling coils
double the energy consumption of the unit due to the addition of
two processes: an implemented sub-cooling process that reduces the
RH and reheating the supplied air in order to meet the desired levels
of thermal comfort. The third system is the same as the first, but it
has an additional pre-cooling coil and controller objective where a
PMV model is added to facilitate the controller optimisation for four
outputs (i.e., the dry bulb temperature, the radiant temperature, the
relative air velocity and the relative humidity for an indoor condi-
tioned space). Controller (TSKFIS (TakagieSugenoeKang fuzzy
inference system)) optimisation is achieved by manipulating the five
AHU inputs (control outputs), which are in the form of the flow rate
of chilled water for the pre-cooling coil and the main cooling coil, the
flow rate of the supply air (fresh air and return air) and the fan speed
of the supply air. Additionally, the PMV model strategy does not
require the use of a reheating coil for decoupling purposes.
The main contribution of this paper is to address the coupling
problem, which arises in the hot and humid climatic region of
South Iraq, by modifying the AHU and applying the algorithm of the
adaptive multi-variable control TSKFF (the Takagi-SugenoeKang
fuzzy forward).
2. Control system design
The present paper attempts to address the shortfall on energy
savings and decoupling for buildings with HVAC control systems in
the hot and humid climatic region of Iraq. Careful assessments in
simulated environments are considered. The PMV model is added
to enable controller decoupling of temperature and RH. Increasing
manipulation parameters are used to compensate for any bounded
variations that may arise due to the limitation of the dampers
range. This is considered as a limitation because the HVAC control
systems have set upper and lower control limits for the dampers
range in order to maintain ventilation for acceptable indoor air
quality, according to the ANSI/ASHRAE 62 standard [24].
2.1. TSKFF controller
The industry standard PID controller exhibits the inability to
control the objectives of the HVAC system that have inherently
adverse characteristics, such as a nonlinear, large-scale system with
a large thermal inertia, a pure lag time, constraints and factors of
uncertain disturbances. Additionally, the indoor thermal comfort
must be decoupled from the temperature and relative humidity.
Hence, fuzzy logic controllers are used due to their flexibility and
intuitive use [25] in controlling the aforementioned characteristics.
2.1.1. Basic description of the control system
The most important motivation for adopting this type of
controller is due to it being able to treat multi-controlled variables
because it converts a TSKFIS (TakagieSugenoeKang fuzzy inference
system) model into a memory layers parameters (TKS) model. The
output routine of the classical TSKFIS model requires numerical and
logical operation tasks, and these tasks take a long time to be
completed. However, the TSK model uses the gradient algorithm,
which is a faster online tuning method that requires less mathe-
matical manipulations than other traditional methods, such as the
backpropagation method for neural networks. The most important
aspect of online tuning is that it can tune a multivariable controller
with multiple outputs; this tuner can improve the controller's
ability to deal with MIMO models that possess a large-scale
nonlinear aspect, are heavily coupled, have a pure lag time,
contain large thermal inertia, possess uncertain disturbance factors
and have constraints, which are common properties in HVAC sys-
tems. For the purpose of this study, each strategy of the control
structure is developed by upgraded layers of memory in order to
coordinate the modification of AHUs, which follows a change in the
online tuning system.
2.1.2. Model identification architecture
The main concept of the TSKFF (Takagi-SugenoeKang fuzzy
forward) structure is based on obtaining the consequent parame-
ters by mapping them from the antecedent space to the consequent
space. The obtained parameters of the consequent space are
organised as layers in the memory space. The parameters in these
layers function to the inputs of the model. These inputs calculate the
outputs' data set, which can be clustered into seven groups within a
time frame of 24 h, where each cluster for each output is repre-
sented by TSK rules. The outputs Yj(X) must fit the data set. This can
be achieved by modulating the nonlinear equation for each output
yi. The modulation can be attained by tuning the parameters ai and
bi. The offline tuning method is performed by using the GNMNR
(GausseNewton Method for the Nonlinear Regression) algorithm,
which has the capability to express the knowledge that is acquired
from inputeoutput data in the form of layers of parameters. The
Equation of the final model's outputs is characterised by aggre-
gating the clusters' outputs and obtaining the singleton fuzzy
model, which belongs to a general class of the universal model
output. Subsequently, the outputs Yj(X) can be obtained as follows:
Yj ðX Þ ¼
XN
i
uiai
1 À eÀbix
(1)
where X ¼ [x1, x2 … xm]T
is the input variables vector, i is a rule
number subscript, ai and bi are the Tagaki-SugenoeKang parame-
ters functions, ui is the basis function (weight), and j is the cluster
number subscript.
The TSK model can be structured in layers f (x; ai bi) and the
weights framework that is shown in Fig. 1 where f (x; ai bi) is a
nonlinear function of the TSK parameters and the independent
variable x.
The TSKFF is modelled by collecting training data from the
building and the HVAC system equipment. Learning of the pa-
rameters in the TSKFF model is accomplished by the offline GNMNR
algorithm. One of the advantages that the GNMNR algorithm offers
is the real-time implementation of computational cost reduction.
This is possible because the proposed method requires a lower
number of iterations to perform the learning/training procedure;
therefore, the tuning time will be reduced when it is implemented
in real-time [5]. The controller method is realised by the TSKFF feed
forward model to increase the response and time steady state
control for the HVAC system. Additionally, the feed forward model
is tuned online by using the gradient algorithm to enhance the
stability and to reject the disturbances and uncertain factors. By
using the gradient algorithm, a faster online tuning method is
R.Z. Homod / Energy 74 (2014) 762e774764
4. found that requires less mathematical manipulations than other
methods do, such as the backpropagation method for neural net-
works. The most important aspect of this online tuning is that it can
tune a multivariable controller with multiple outputs [11].
2.2. Decoupling problem and objectives' setting
The cooling coils in AHUs are categorised into dry and wet types.
The temperature and relative humidity of air that is introduced to
the AHU that has a dry cooling coil are characterised by coupling
loops due to the constant air humidity ratio. Once the temperature
is decreased, the relative humidity will be increased and vice versa.
The thermal comfort can be controlled through the PMV index by
using this type of AHU, with either air temperature or air relative
humidity being a control variable (but not with both being control
variables at the same time). The rest of the PMV variables are
considered to be disturbances. It is desirable to control temperature
and relative humidity independently and accurately in certain in-
door conditions. In these cases, the AHU with a wet cooling coil is
used; both temperature and RH are varied independently based on
the flow rates of air and chilled water. It is impossible to set one
variable without affecting the other when the design of the AHU
does not take into account the coupling dynamics between these
variables; therefore, the importance of decoupling techniques that
are used to implement an appropriate AHU is realised.
The proposed strategy is implementing a twin cooling coil AHU
and an advanced multi-variable control system. The pre-cooling
coil (wet) is equipped to cool and dehumidify the fresh air intake.
The main cooling coil (dry) is used to cool the supply air. The deeply
chilled water is only necessary for (pre-cooling coil) removing the
moisture from the fresh air. The main cooling coil requires
moderately cool water, according to the building load. This type of
order helps in save energy for buildings with HVAC systems
because higher chilled water temperatures indicate better COPs
(coefficients of performance). Furthermore, the use of the PMV
index (the indoor air temperature, the radiant temperature, the
relative air velocity and the relative humidity for an indoor condi-
tion space) as a desired objective enables the control system to
optimise the input plant by controlling air velocity and manipu-
lating the flow rate of fresh air in regard to thermal comfort levels.
The main difference between the proposed strategy and the
other two strategies is in their control objectives of the operating
system and AHU equipment. The AHU for the conventional strategy
is similar to what it is for proposed strategy, but there are two
differences: first, it does not contain a pre-cooling coil, and second,
the controlled variables include two variables that have restricted
values. These variables are temperature and relative humidity; both
of them are set at desired specific values. The objective of this
control strategy acts as a control reference of the online tuning that
reflected negatively on its performance due to stiff references and a
limited number of input plant variables that are used for optimi-
sation. The controlled variables for the third (adding the reheating
coil) strategy are similar to those of the conventional strategy, but
the difference is that the AHU is equipped with a wet main cooling
coil and a reheating coil to consolidate the controller for the
decoupling problem.
The objectives of this paper are to:
1. Assess the feasibility of using the proposed strategy in a South
Iraq climate
2. Characterise the energy savings and decoupling of the proposed
system
3. Test the potential of the controller for multi-objective optimi-
sation in the HVAC system.
These aims will be achieved by comparing three scenarios of the
AHU control system in order to assess the decoupling problem and
energy savings of the simulated HVAC system.
3. Analysis of energy and mass flows of a building
The purpose of the control strategy is to minimise the total
power consumption of the HVAC system by optimising the vari-
ables of the indoor thermal comfort (i.e., the indoor air tempera-
ture, the radiant temperature, the relative air velocity and the
relative humidity for the indoor condition space). Generally, the
electric power consumption of the HVAC system is a function of the
COP (coefficient of performance) of the chillers, the EER (energy
efficiency ratio) of the building and the cooling load of the building.
The EER and COP are constants for a specified building and chiller,
respectively, whereas the total cooling loads of the building vary,
depending on the disturbances and the controllable variables.
Therefore, the total electric power consumption can be summarised
by Equation (2) [26,27]:
EP ¼
XN
i
chli
copi
þ EPAHU ¼
TBCL
EER
(2)
where EP is the total electric power consumption, N is the number
of chillers, chl is the chiller power, EPAHU is the electric power that is
consumed by AHU, and TBCL is the total building's cooling load.
From Equation (2), it can clearly be observed that the EP can be
derived by using two different methods that are based on the
energy and mass balance equations of the building's fabric (the
right term of Equation (2)) and of the AHU subsystems' equipment
(the middle term of Equation (2)). Therefore, the theories
regarding the conservation of energy and mass are applied to
thermally analyse and model the overall behaviour of an HVAC
system. These theories are based on the fact that in the control
volume of any subsystem, energy is transferred from/to a sub-
system by two types of processes: mass transfer and conventional
Fig. 1. Schematic diagram of the TSK model as layers of memory.
R.Z. Homod / Energy 74 (2014) 762e774 765
5. heat transfer (conduction, convection and radiation). These pro-
cesses are dominant in HVAC systems. In this research study, the
system is subdivided into the building's and the AHU's control
volumes. The building's energy and mass transfer can be demon-
strated by Fig. 2. To evaluate the sensible heat gain of the building,
the following thermal balance equation is applied to the building's
control volume:
The term on the left side of Equation (3) denotes the output of
the AHU, which represents the heat and mass that is transferred to
the building's control volume. On the right side of Equation (3), the
first part (the accumulation or storage of energy) represents the
thermal mass that is stored in the inner wall, indoor air and
furniture, while the second part (the difference between the input
and output of energy) represents other inputs/outputs to the con-
trol volume of the building.
The latent heat gain of the building is related to the moisture
transfer, which can be evaluated by applying the conservation of
time-dependent mass law to the control volume of the building,
which is shown in Equation (4);
The term on the left side of Equation (4) is the rate of moisture
that is absorbed by the AHU. On the right side of Equation (4), the
first part (the rate of moisture change) is the change in the rate of
air moisture in the building at time interval dt, and the other terms
are related to the indoor input/output and the generated moisture.
To evaluate the sensible and latent heat gains of the building, it is
necessary to calculate the left-hand sides of Equations (3) and (4),
which can be obtained by applying the laws of conservation of energy
and mass to the control volume of the AHU. The AHU is subdivided
into three subsystems: the mixing air chamber, the pre-cooling coil
and the main cooling coil. Energy is only consumed in the pre-cooling
and main coolingcoils,socalculations for the energyand mass control
volumes are applied on these two subsystems, as follows:
The term “energy absorbed by the coil” in Equation (5) refers to
the sensible and latent heat load that is exerted by the pre-cooling
coil. On the right side of the equation, the first term (energy
accumulation in the metal mass of the coil) refers to the rate of
change for the heat storage in the coil mass, while the second term
(the sensible energy delivered by air) refers to the sensible cooling
load of the fresh air, and the third term (the latent energy delivered
by moisture withdrawal) refers to the latent energy that is absorbed
by the coil due to the condensation of moisture. The third term on
the right side of Equation (5) can be evaluated by applying the law
of mass conservation to the air flow stream that is used for the pre-
cooling coil. The following is obtained:
By using the same procedure as was used for the pre-cooling coil
to obtain the sensible and latent heating loads for the dynamic
subsystem equations, the main cooling coil can be written mathe-
matically by using the time-dependent equation of the control
volume, as follows:
_Qs
z}|{
Cooling load
¼ _Qair þ _Qfur
zfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflffl{Accumulation or storage of energy
þ _Qopq þ _Qfen þ _Qslab þ _Qinf þ _Qig;s
zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
Difference between input and output of energy
(3)
_ms
À
ur;t À us;t
Ázfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflffl{
rate of moisture withdrawal by AHU
¼
dMrur;t
dt
zfflfflfflffl}|fflfflfflffl{
rate of moisture change
þ
_Qig;l
hfg
zffl}|ffl{
rate of moisture generation
þ _minf uo;t À _mrur;t
zfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflffl{
rate of moisture transfer
(4)
_mw;tcpw
À
Two;t À Twin;t
Ázfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
energy absorbed by the coil
¼ MHecpHe
dTh;t
dt
zfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflffl{
energy accumulation in the metal mass of coil
þ _mo;tcpa
À
To;t À Tos;t
Ázfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
sensible energy delivered by air
þ _mo;t
uo;t À uos;t
hfg
zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
latent energy delivered by air dehumidification
(6)
_mw;tcpw
À
Two;t À Twin;t
Ázfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
energy absorbed by the coil
¼ MHecpHe
dTh;t
dt
zfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflffl{
energy accumulation in the metal mass of coil
þ _mo;tcpa
À
To;t À Tos;t
Ázfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
sensible energy delivered by air
þ _mcon:;thfg
zfflfflfflfflfflffl}|fflfflfflfflfflffl{
latent energy delivered by moisture withdrawal
(5)
R.Z. Homod / Energy 74 (2014) 762e774766
6. The rate of thermal energy transfer (the sensible cooling load)
from the building by the mechanical ventilation air flow (Qvent) is
calculated by using Equation (8).
_Qvent ¼ _ms;tcpa
Tr;t À Ts;t
(8)
The power of the air supply system in the mechanical ventila-
tion state (the transmission power) is mainly from the power
supply for the fan, which can be calculated by the application of the
law of conservation of energy on the control volume of the AHU.
This equation can be calculated as follows [27]:
_Qfan ¼ _ms;tcpa
Ts;t À To;t
(9)
According to the energy balance for the indoor conditioned
space of Equation (3), the values of thermal energy flow from (1)
opaque-surfaces, (2) transparent fenestration surfaces, (3) infiltra-
tion, (4) indoor load and (5) ventilation are calculated by using the
steady state conditions of Equation (3), whereby all of the thermal
energy flow values are equal to the cooling load that is extracted by
the HVAC systems or the mechanical ventilation, which equals the
left-hand side of Equation (3); in turn, Equation (3) can be calcu-
lated by summing Equations (6)e(9).
The instantaneous cooling load of the building can be obtained
from the simulation process after modelling the HVAC system.
Additionally, the instantaneous cooling loads of the building
directly impact the outputs of the controller signals. Therefore, the
method of calculation that is employed in this research study is
based on the output signals of the controller. The output signals of
the controller manipulate the valves of the pre-cooling coil, the
main cooling coil, the reheating coil and the dampers of the return
and fresh air to track the objective of the HVAC system. The valves
and dampers are designed according to the heating/cooling load of
the building. The opening position of the valves and dampers is
recorded as a percentage of the fullest extent (as shown in Fig. 3)
that represents the main cooling coil valve's opening position over
24 h. The percentage of the opening position is related to the
maximum flow rate of the valves and dampers. This signal opening
position is implemented in Matlab to obtain the energy con-
sumption of the HVAC system.
The advantage of using Matlab/Simulink is in the ability to use a
graphical programming language that is based on different block
categories with different properties of each block. Matlab and its
Fig. 3. The control signal percentage for the main cooling coil's chilled water valve for each of the three strategies.
Fig. 2. Representation of building energy and mass transfer for prototypical buildings
with HVAC systems.
_mmw;tcpw
À
Two;t À Twin;t
Ázfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
energy absorbed by the coil
¼ MmHecpHe
dTh;t
dt
zfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflffl{
energy accumulation in the metal mass of coil
þ _mm;tcpa
À
Tm;t À Ts;t
Ázfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
sensible energy delivered by air
þ _mm;t
um;t À us;t
hfg
zfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflfflffl{
latent energy delivered by air dehumidification
(7)
R.Z. Homod / Energy 74 (2014) 762e774 767
7. toolboxes are adopted to perform all of the identification processes
and simulations in this work, as well as in our previous works
[28e31]. System identification and control system toolboxes were
used to identify and build the model, while the fuzzy logic toolbox
was used for the TSK model identification. The obtained models are
then introduced in the Matlab/Simulink environment for simula-
tion and analysis. These categories include the input/output,
transfer functions, arithmetic functions, state space models and
data handling. The building model is represented in the form of
ODE (ordinary differential equation) solvers, which are automati-
cally configured during the Simulink model's run-time. The algo-
rithm of the controller is designed by using Matlab m-files,
parameter layer memory and S-functions, which are based on on-
line parameter tuning. The technique for calculating the cooling
loads is easily implementable, whereby the thermal balance
equation is derived from the arithmetic functions, from which the
energy consumption can be obtained.
4. Simulation results and discussion
4.1. Physical and theoretical model description
The simulated building model is a typical one-story house with
a simple structure. The house consists of heavyweight construction
(brick and concrete) that measures 4.5 m in height, with 248.6 m2
of gross ground floor area. The net floor area of the entire building is
195.3 m2
, excluding the garage area; the gross exposed area of the
windows and wall is 126.2 m2
, while the net area of the exterior
wall is 108.5 m2
. The overall volume of the house, excluding the
garage and suspended ceiling space, is 781.2 m3
Table 1 shows the
physical properties of the components of the building. The dry bulb
temperature varies according to the spring season's climate in
Basrah city, which ranges from 18 C to 32 C, and the humidity
ratio varies from 0.01 to 0.01909 kg of moisture per kg of dry air. The
building model's transfer function and the PMV, or thermal comfort
sensor model, are presented in Appendices A and B [32,33].
To reduce the design cost, as well as the cost that is needed to
fabricate the three HVAC systems, simulation methods are imple-
mented in order to test and analyse the results. The identification
approach of the model demonstration is based on the multi-zone
model of the RLF (residential load factor) method. The identified
model is simulated by three different controller strategies in order
to study their levels of indoor thermal comfort and energy con-
sumption. The first system is a conventional control system (the
control variable objectives are temperature and relative humidity).
The second is a conventional system that includes the addition of a
reheating coil and a wet main cooling coil, while the third system is
similar to the first system, but it includes an addition of a pre-
cooling coil and a PMV index in order to measure the objective of
the controller. The three types of systems are run together in order
to study their performance and energy consumption (as shown by
the simulation block diagram in Fig. 4, which presents the simu-
lation in the evaluation of performance and energy consumption of
the three systems).
Fig. 4. Matlab blocks for the simulations of all three systems.
Table 1
Properties of the materials used for construction of the model.
Component Description Factors
Roof/ceiling Flat wood frame ceiling
(insulated with R-5.3 fiberglass)
beneath vented attic with
medium asphalt shingle roof
U ¼ 0.031 18 W=ðm2KÞ
a roof ¼ 0.85
Exterior
walls
Wood frame, exterior wood
sheathing, interior gypsum
board, R-2.3 fiberglass
insulation
U ¼ 51 W=ðm2KÞ
Doors Wood, solid core U ¼ 2.3 W=ðm2KÞ
Floor Slab on grade with heavy carpet
over rubber pad; R-0.9 edge
insulation to 1 m below grade
Rcvr ¼ 0.21 W=ðm2KÞ;
Fp ¼ 85 W=ðm2KÞ
Windows Clear double-pane glass in
wood frames. Half fixed, half
operable with insect screens
(except living room picture
window, which is fixed). 0.6 m
eave overhang on east and west
with eave edge at same height
as top of glazing for all
windows. Allow for typical
interior shading, half closed.
Fixed: U ¼ 2.84 W=ðm2KÞ;
SHGC ¼ 0.67.
Operable: U ¼ 2.87 W=ðm2KÞ;
SHGC ¼ 0.57; Tx ¼ 0.64;
IACcl ¼ 0.6
Construction Good Aul ¼ 1.4 cm2
/m2
R.Z. Homod / Energy 74 (2014) 762e774768
8. The mean radiant temperature is a more complicated quantity
that depends on the temperature of the surrounding surfaces, as
well as on angle factors of the surrounding surfaces. Therefore, the
plant model leads to the output of the plug-in model of the PMV
index, except the mean radiant temperature requires an interme-
diate sub-model where its output is taken into account because it is
one of the main factors that affects thermal comfort. This sub-
model estimates the mean radiant temperature by using two
methods: theoretical and numerical. For the theoretical method,
the mean radiant temperature is estimated from the measured
temperature of the surrounding walls and surfaces and the angle
factors of these surrounding surfaces. All of the indoor surfaces are
assumed to be black because most building materials have a high
emittance ε, and it is assumed that small temperature differences
exist between the surfaces of the enclosure (i.e., linear combination
of system states). Therefore, the following equation is used [34]:
MRT ¼ T1FPÀ1 þ T2FPÀ2 þ / þ TnFPÀn (10)
where MRT is the Mean Radiant Temperature, Tn is the temperature
of surface ‘n’ and Fp-n is the angle factor between a person and
surface ‘n’.
For the numerical estimation, a black-globe thermometer sensor
is used.
4.2. Decoupling results and discussion
The plant model is dynamically subjected by many thermal
disturbance factors, such as the K2 solar radiation, f4 inside sensible,
FDR fenestration, etc. Three simulation sets are conducted over 24 h
and include nominal, noise and sensor deterioration, as well as an
uncertainty operation, for the three systems' behaviours to be
observed and studied for the different conditions. The main
objective of this work is to validate the decoupling of the proposed
strategy.
4.2.1. Nominal operating conditions
Pre-cooling coils are added to the proposed AHU of the HVAC
system in order to economically control the indoor relative hu-
midity in a humid climate. Additionally, the proposed system has
four control variables for an indoor conditioned space (i.e., the
indoor-air temperature, the indoor-air velocity, the indoor-air hu-
midity and the flow rate of fresh air). These control variables are
optimised by the controller to provide economical indoor-air
Fig. 5. PMV comparisons of the results of the three different systems with different objectives and designs.
Fig. 6. Indoor temperature comparisons of the results between the three different systems with different objectives and designs.
R.Z. Homod / Energy 74 (2014) 762e774 769
9. conditioning that yields the desired level of thermal comfort and
indoor air quality, according to ASHRAE and ISO standards; this also
reduces the cooling load during implementation in real-time. The
other systems have two control objectives that are set at certain
desired values for the indoor conditioned space (i.e., the indoor-air
temperature and the indoor-air humidity). Fig. 4 shows the three
designs of the HVAC system. The manipulation of each TSKFF for
the five AHU inputs in the three designs of the HVAC system be-
haves differently. In the proposed system, the input feedback
sensor allows some degree of tolerance instead of requiring a
specific value for the temperature and relative humidity, which is
needed in the conventional HVAC systems. This optimisation
overcomes the coupling effects (temperature and relative humidi-
ty) perfectly by providing the desired level of thermal comfort,
which is shown in Fig. 5. In regard to Fig. 5, it can be observed that
the proposed (the model of the PMV index addition) system can
track the desired objective and can achieve outstanding perfor-
mance, while the system with the added reheating coil acts within
an acceptable thermal comfort range that has an acceptable offset
from the set point. The conventional system was found to violate
the ASHRAE Standard 55-92 [35] and ISO-7730 [36] for the desired
level of indoor thermal comfort. These standards recommend that
the acceptable levels of thermal comfort are limited to a range
between À0.5 PMV 0.5. It is evident that this violation is caused
by the coupling of temperature and relative humidity. The tem-
perature curves of all three systems are similar to the PMV trend
that is shown in the simulation results, which are tabulated in
Fig. 6. The periodic effect of the coupling factors is apparent from
05:30 o'clock to 08:00 o'clock and from 19:00 o'clock to 24:00
o'clock. The effect can be more clearly observed in the behaviour of
the relative humidity (as shown in Fig. 7), in which the conven-
tional system fluctuates within a wide range, whereas the other
systems remain in the range of approximately 50% RH. The rec-
ommended range of RH, according to the ASHRAE Standard 55-92
and ISO-7730 for the indoor comfort condition, is 40%e60%. High
humidity not only causes poor indoor air quality, but it also causes
wood decay, metal corrosion and structural deterioration [37]. The
calculations of energy consumption are based on the controller
Fig. 7. Indoor relative humidity comparisons of the results between the three different systems with different objectives and designs.
Fig. 8. PMV comparisons of the results of the three different systems based on the operating conditions of noise and sensor deterioration.
R.Z. Homod / Energy 74 (2014) 762e774770
10. signals. One of these signals is shown in Fig. 3. Fig. 3 shows the
results of the simulation of the control signal variation for the main
cooling coil of the chilled water valve, with respect to time. In Fig. 3,
the signals for the conventional system with a reheating coil acts
similar to a BangeBang control action. The modulating valve
continuously fluctuates between ON-OFF, which will eventually
wear out the valve and shorten its lifespan. It can be clearly
observed that the proposed system signal works very efficiently,
which provides good control performance. Figs. 5e7 show the
transient response for the initial condition. This took approximately
an hour because the plant model is dynamically affected by the
thermal mass of the building structure and slab floors, which cre-
ates a flywheel effect. The influence of this flywheel effect begins to
fade and becomes less intense after the HVAC system starts, which
can clearly be observed in the signal of valve open position in Fig. 3.
4.2.2. Operating conditions of noise and sensor deterioration
Disturbance mode has tested decoupling through its validation
of the rejection of noise and sensor deterioration. In noise and
sensor deterioration, the controlled process parameters, sensors'
gains, and noise signals are able to change in the same manner for
each system and simulation that is conducted. Here, we suppose
that sensors deteriorate with 20% fault, and the sensors' gain is
changed to 0.8 (sensor gain ¼ 1 when the sensor performance is
100%). Additionally, to test the sensitivity of the proposed method
to noise, each system is subjected to the same noisy environment
by adding a 10% NSR (noise-to-signal ratio), which refers to the
ratio of the continuous noise signal to the controlled signal. The
sensor deterioration set and subjected noise signal are applied at
the start of the simulation. Fig. 8 shows the three different systems
to try to track a PMV set point, which changes under a square wave
from À0.4, 0 and 0.4 during a 24 h time frame. By using this test,
one can clearly observe the three systems' behaviour for the PMV,
where the proposed system provides superior control performance
and does not violate the ASHRAE 55-92 and ISO-7730 Standards. In
contrast, the other systems exhibited deterioration in their per-
formances and, consequently, violated the Standards of the indoor
thermal comfort. Thus, the proposed system achieves significant
results that verify the use of decoupling parameters rather than
adding a reheating coil or using conventional decoupling methods,
Fig. 9. PMV comparisons of the results of the three different systems in regard to the operating conditions when model uncertainties are present.
Fig. 10. Psychometric chart comparisons of the results of the three different systems in regard to the operating conditions when model uncertainties are present.
R.Z. Homod / Energy 74 (2014) 762e774 771
11. which are extremely intricate and too impractical to solve numer-
ically when the plant system model is complicated, which is the
case for HVAC systems.
4.2.3. Operating conditions regarding the presence of model
uncertainties
In regard to robustness validation, the plant model encompasses
a wide range of operating parameters, which vary as the HVAC
systems undergo fluctuating loads due to changes in external dis-
turbances during a typical day's operation. Therefore, in the pres-
ence of uncertainties regarding the modelling of such parameters, it
becomes necessary to use a robust intelligent controller, such as
TSKFF, to obtain efficient operation in the HVAC systems. To vali-
date the robustness of the TSKFF controller, the building heat loss
coefficients, the heat transfer coefficients of the fan-coil units and
pumps and the thermal time constant are changed. Before a
simulation run begins, all of the model coefficients and the time
constant are increased by 20%. Three TSK models of controllers are
tuned based on the nominal plant model and then are integrated
into a control algorithm that manipulates AHU parameters to
control indoor thermal comfort. The conventional strategy leads to
the worst indoor ambient conditions and becomes less intense after
adding a reheating coil, which is shown in Fig. 9. However, both
strategies violate the standard limitation of ASHRAE 55-92 and ISO-
7730. The proposed scheme maintains asymptotic tracking of a
given reference signal, and it occurs in the presence of the same
parameter variations and model uncertainties when it does not use
reheating and wet main cooling coils. For validation, psychometric
charts are the most commonly used tool for indoor studies and
outdoor air conditions. Fig. 10 describes the air states cycle of
physical and thermodynamic properties for indoor conditions over
24 h. The proposed strategy seems to satisfy the TCZ (thermal
comfort zone), whereas the other strategies frequently crossover
TCZ.
4.3. Energy saving results and discussion
The purpose of the model of the PMV index addition to the
proposed system is to change the restricted conventional objective
variables (temperature and relative humidity) of an HVAC system in
Fig. 11. A comparison of the energy consumption results based on the cooling coil load variation between the three different schemes.
Fig. 12. A comparison of the results of the power consumption between the three different system schemes.
R.Z. Homod / Energy 74 (2014) 762e774772
12. addition to increasing its flexibility with respect to the indoor
control parameters (temperature, fresh air flow rate, indoor air
velocity and relative humidity). The model of the PMV index
addition also enables the controller to improve its performance.
The TSKFF controller exploits the flexibility of the control param-
eters by optimising the parameters through the manipulation of
the AHU parameters (inputs) to provide the desired levels of ther-
mal comfort, while simultaneously reducing the energy con-
sumption of the HVAC system.
Furthermore, the velocity of indoor air can reduce the cooling
load. This can be observed from the simulation results of the energy
that is consumed by the cooling coil load, which is shown in Fig. 11.
The simulation techniques that are used to calculate the cooling
loads are straight forward: thermal balance equations are imple-
mented by using arithmetic functions, and then the consumed
energy can be obtained by using Equations (6)e(9). The simulation
results of the consumed energy in the system with a reheating coil
reveal higher energy consumption than the consumption of the
other systems because the cooling process reduces the air tem-
perature to the sub-cooling state before the reheating process
overcomes the coupling effect and meets the demands of indoor
thermal comfort.
Although the conventional system is better in terms of energy
savings than the system that has the addition of a reheating coil, the
conventional system does not meet the desired level of indoor
thermal comfort. However, the proposed system shows more
favourable results than the other two systems with respect to
achieving the desired level of thermal comfort and reducing energy
consumption, simultaneously. Based on Fig. 11, it can be observed
that the differences in energy consumption among the three sys-
tems increase during the times periods that include the presence of
a coupling effect. The average power consumption for the three
different systems (the conventional system, the addition of a
reheating system and the proposed system) are 10.713 kW,
13.27 kW and 9.016 kW, respectively. The average power costs that
accompany the addition of a reheating system are 1.4718 times
higher than that of the proposed system. Based on the data for 24 h
of power consumption, the calculations for energy consumption for
each of the three strategies show that energy consumption in the
proposed strategy is 32.06% lower than the system with an added
reheating coil, which is shown in Fig.12. This result closely matches
the results that were obtained by Yang and Su [38] in which an
intelligent controller was developed to adjust the PMV index,
which led to saving approximately 30% more of the energy con-
sumption than the conventional methods. Furthermore, the simu-
lation of the cooling coil load output is compared to the numerical
results, which are based on the CLF/CLTDC (the cooling load factor
for the glass/corrected cooling load temperature difference)
method [39,40]. The calculation considers the effects of numerous
outdoor environmental parameters on the indoor thermal loads.
The cooling load of the building is calculated every 30 min to obtain
the absolute margin of error between the simulation results (the
proposed system) and the numerical calculation, which was found
to vary between 0.064 and 0.107 kW. To have a clearer assessment
of the error between the simulation and numerical calculations of
the cooling coil load output, in this study, the statistical index of the
coefficient of determination (r2
) was calculated based on Equation
(11) and had a value of r2
¼ 0.974.
r2
¼
½N
P
yiyi À ð
P
yiÞð
P
yiÞŠ2
h
N
P
y2
i
À ð
P
yiÞ2
N
P
y2
i
À ð
P
yiÞ2
i (11)
where yi is the numerical result, yi is the simulation result, and N is
the number of test samples.
5. Conclusion
In this context, the simulation results of the comparison inves-
tigated the use of a PMV model input as an objective optimisation of
controller decoupling and of reductions in the energy consumed by
an HVAC system. This was performed by considering all of the
factors that are affected by indoor thermal comfort, which are re-
flected by the PMV index. The control system that is proposed in
this work includes, as part of its structure, a PMV model for the
optimisation of the deviations in the parameters of indoor thermal
comfort and of the generation of control actions that pertain to the
AHU inputs. The task regarding the PMV index output is, therefore,
to acquire controller output signals more accurately by exploiting
the decision algorithm's flexibility for the PMV index input's ag-
gregation. The weather in Basrah, a southern city of Iraq, was
considered as a case study to test the system. The output controller
signals were adopted to obtain the energy consumption for three
different control objectives and strategies, which were evaluated
with respect to typical and modified HVAC systems. Based on the
results of the performed simulations, we can conclude that when
using the indoor PMV as a variable objective for the HVAC system,
the controller performs better and provides more energy savings,
while still attaining the desired level of indoor thermal comfort.
The multi-input of the AHU is manipulated by the TSKFF controller,
which is characterised by the optimisation of the outputs for energy
savings. Both the conventional and proposed systems show that
energy is saved in comparison to the system that has an additional
reheating coil in the AHU. The conventional HVAC system shows a
savings of up to 19% of energy usage, which is 61.5 kWh/d less than
the energy that is used by an HVAC system that has an additional
reheating coil; in contrast, the proposed strategy can save up to
32.06% of energy usage, which is 102.1 kWh/d less than the energy
that is used by the HVAC system that has an additional reheating
coil. This is because a TSKFF that is equipped with a model of PMV
index reduces the energy consumption of a building by as much as
it can by utilising the outdoor climate in controlling the rate of fresh
air flow. Meanwhile, the conventional control strategy adjusts the
temperature and relative humidity to a predefined strict set point,
which does not allow for optimisation of energy consumption for
indoor thermal comfort. An important finding of this study is that
the proposed strategy economically addressed the coupling prob-
lem in addition to providing the desired level of thermal comfort.
Furthermore, the procedure of using the PMV model and TSKFF is
straightforward and easy to implement.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.energy.2014.07.047.
References
[1] Anastaselos D, Theodoridou I, Papadopoulos AM, Hegger M. Integrated eval-
uation of radiative heating systems for residential buildings. Energy
2011;36(7):4207e15.
[2] Prek M. Thermodynamical analysis of human thermal comfort. Energy
2006;31(5):732e43.
[3] Mossolly M, Ghali K, Ghaddar N. Optimal control strategy for a multi-zone air
conditioning system using a genetic algorithm. Energy 2009;34(1):58e66.
[4] Chiu CC, Tsai NC, Lin CC. Near-optimal order-reduced control for A/C (air-
conditioning) system of EVs (electric vehicles). Energy 2014;66(1):342e53.
[5] Homod RZ, Sahari KSM. Intelligent HVAC control for high Energy efficiency in
buildings. LAP LAMBERT Academic Publishing; 2014.
[6] Tashtoush B, Molhim M, Al-Rousan M. Dynamic model of an HVAC system for
control analysis. Energy 2005;30(10):1729e45.
[7] Zhao L, Michelsen FA, Foss B. Control design and dynamic simulation of an
HMR pre-combustion power cycle based on economic measures. Energy
2013;51(1):171e83.
R.Z. Homod / Energy 74 (2014) 762e774 773
13. [8] Seem JE. A new pattern recognition adaptive controller with application to
HVAC systems. Automatica 1998;34(8):969e82.
[9] Homod RZ. Takagi-sugeno fuzzy modelling and adaptive control of indoor
Thermal Comfort in HVAC systems using predicted mean vote Index. PhD
Thesis. Kajang, Malaysia: University of Tenaga Nasional; 2012.
[10] Gagnon E, Pomerleau A, Desbiens A. Simplified, ideal or inverted decoupling?
ISA Trans 1998;37(4):265e76.
[11] Homod RZ, Sahari KSM, Almurib HAF, Nagi FH. Gradient auto-tuned Takagi-
Sugeno fuzzy forward control of a HVAC system using predicted mean vote
index. Energy Build 2012;49:254e67.
[12] Kusiak A, Li M. Reheat optimization of the variable-air-volume box. Energy
2010;35(5):1997e2005.
[13] Lee J, Kim DH, Edgar TF. Static decouplers for control of multivariable pro-
cesses. AICHE J 2005;51(10):2712e20.
[14] Pomerleau D, Pomerleau A. Guidelines for the tuning and the evaluation of
decentralized and decoupling controllers for processes with recirculation. ISA
Trans 2001;40(4):341e51.
[15] Wallter M, Wallter JB, Wallter KV. Decoupling revisited. Indus Eng Chem Res
2003;42(20):4575e7.
[16] Tavakoli S, Griffin I, Fleming PJ. Tuning of decentralized PI (PID) controllers for
TITO processes. Control Eng Pract 2006;14(9):1069e80.
[17] Xi X-C, Poo A-N, Chou S-K. Support vector regression model predictive control
on a HVAC plant. Control Eng Pract 2007;15(8):897e908.
[18] Garrido J, Vazquez F, Morilla F. An extended approach of inverted decoupling.
J Process Control 2011;21(1):55e68.
[19] Chen P, Zhang W. Improvement on an inverted decoupling technique for a
class of stable linear multivariable processes. ISA Trans 2007;46:199e210.
[20] Seborg DE, Edgar TF, Mellichamp DA. Process dynamics control. New York:
John Wiley Sons; 1989.
[21] Shinskey FG. Process control systems: application, design and adjustment.
New York: McGraw-Hill; 1988.
[22] Wade HL. Inverted decoupling: a neglected technique. ISA Trans 1997;36(1):
3e10.
[23] Shen Y, Cai W-J, Li S. Normalized decoupling control for high-dimensional
MIMO processes for application in room temperature control HVAC sys-
tems. Control Eng Pract 2010;18(6):652e64.
[24] ANSI/ASHRAE Standard 62.1-2004 Ventilation for Acceptable Indoor Air
Quality incorporates a number of significant changes to the previous stan-
dard; 62-2001.
[25] Kristl Z, Kosir M, T.Lah M, Krainer A. Fuzzy control system for thermal and
visual comfort in building. Renew Energy 2008;33:694e702.
[26] Haines RW, Wilson Lewis. HVAC systems design, handbook. 4th ed. New York:
McGraw-Hill; 2003.
[27] Wang SK. Handbook of air conditioning and refrigeration. 2nd ed. United
States of America: McGraw-Hill Companies; 2001.
[28] Homod RZ, Sahari KSM, Almurib HAF, Nagi FH. Double cooling coil model for
non-linear HVAC system using RLF method. Energy Build 2011;43:2043e54.
[29] Homod RZ, Sahari KSM, Mohamed HAF, Nagi F. Modeling of heat and moisture
transfer in building using RLF method. In: Conference on research and
development, IEEE; 2010. p. 287e92.
[30] Homod RZ, Sahari KSM, Almurib HAF, Nagi FH. RLF and TS fuzzy model
identification of indoor thermal comfort based on PMV/PPD. Build Environ
2012;49:141e53.
[31] Homod RZ, Sahari KSM, Mohamed HAF, Nagi F. Hybrid PID-cascade control for
HVAC system. Int J Syst Control 2010;1(4):170e5.
[32] Homod RZ, Sahari KSM. Energy savings by smart utilization of mechanical and
natural ventilation for hybrid residential building model in passive climate.
Energy Build 2013;60:310e29.
[33] ASHRAE. Psychrometrics, handbook-fundamentals, Chp. 6. TC 1.1, thermo-
dynamic and psychrometric. American Society of Heating, Refrigerating, and
Air-Conditioning Engineers; 2005.
[34] ASHRAE. Handbook-fundamentals. Atlanta: American Society of Heating,
Refrigerating, and Air-Conditioning Engineers, Inc.; 2009.
[35] ANSI/ASHRAE Standard 55(1992). Thermal environment conditions for hu-
man occupancy. Atlanta: American Society of Heating, Refrigeration and Air-
Conditioning Engineers; 1993.
[36] ISO 7730. Moderate thermal environments e determination of the PMV and
PPD indices and the specifications of the conditions for thermal comfort.
Geneve, Suisse: International Standard Organization; 2005.
[37] Lü X. Modelling of heat and moisture transfer in buildings: I. Model program.
Energy Build 2002;34(10):1033e43.
[38] Yang KH, Su CH. An approach to building energy savings using the PMV index.
Build Environ 1997;32(1):25e30.
[39] Karan B, Souma C, Ram MG. Development of CLTD values for buildings located
in Kolkata, India. Appl Therm Eng 2008;28(10):1127e37.
[40] Pita Edward G. Air conditioning principles and systems. 4th ed. New York:
McGraw-Hill; 2002.
R.Z. Homod / Energy 74 (2014) 762e774774