Vous êtes sur la page 1sur 12

Development and validation of a desiccant wheel model calibrated

under transient operating conditions


Muzaffar Ali
a, b,
*
, Vladimir Vukovic
a
, Mukhtar Hussain Sahir
b
, Daniele Basciotti
a
a
Energy Department, Austrian Institute of Technology, Gienggasse 2, 1210 Vienna, Austria
b
Mechanical Engineering Department, University of Engineering and Technology Taxila, Pakistan
h i g h l i g h t s
Developed a desiccant wheel model in equation-based object-oriented program, Dymola/Modelica.
The model is calibrated and validated using the transient measurements of a real system.
Model estimates the actual and optimal performance in both dehumidication and enthalpy modes.
Resulted max and min RMSE (PME) validation error is 3.6 kJ/kg (4.6%) and 1.9 kJ/kg (0.2%), respectively.
a r t i c l e i n f o
Article history:
Received 26 November 2012
Accepted 7 August 2013
Available online 28 August 2013
Keywords:
Equation-based object-oriented modeling
Desiccant wheel
Optimal rotation speed
Model calibration and validation
Transient conditions
a b s t r a c t
The current study presents model based predictions of a desiccant wheel performance using the tran-
sient measurements obtained from a real system. The model is based on a set of equations to simulate
the optimal and measured transient performance as a function of measurable input variables related to
the desiccant wheel material and structure. The model is adapted to analyze the inuence of different
working conditions on the desiccant wheel performance: rotation speeds, air velocity, inlet temperature,
and inlet air humidity for both process and regeneration air. The model is capable of estimating the
optimal rotation speed and pressure drop of the desiccant wheel. Moreover, the developed model can be
applied in both, dehumidication and enthalpy modes. The model is validated in comparison to the
published data and measurements from the real building desiccant wheel installation. The specic
enthalpy at the outlet of process air is considered performance parameter. The obtained results are in
agreement with the published data, while the resulting maximum and minimum validation root mean
square error (mean percentage error) between the simulated and measured transient performance is
3.6 kJ/kg (4.6%) and 1.9 kJ/kg (0.2%), respectively.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Performance analysis of energy recovery devices is critical to
predict the energy transfer in heating, ventilation and air-
conditioning (HVAC) design applications. However, such analysis
under the test conditions of ARI Standard 1060-2003 [1] has proven
to be very expensive and prone to large uncertainties [2]. In the ARI
standard, two inlet conditions for summer and winter are used to
certify the performance of air-to-air recovery devices [3]. Since the
variations in the ambient air conditions are signicant over the
year, HVAC designers need to be concerned with the optimal design
and system performance to minimize the system cost and energy
requirements. Furthermore, the performance characterization of
the energy wheels is more difcult than air-to-air heat exchangers
due to the coupled heat and mass transfer aspects associated with
energy wheels and the need to keep steady state balances of air
ow, water vapor ow and energy. In such a situation, it is chal-
lenging and time consuming to maintain small uncertainties for all
the energy wheel parameters [2]. Several studies for performance
analysis of energy exchangers are performed to determine the
effectiveness of heat exchangers under steady state operating
conditions [4e6]. However, the analysis of the desiccant wheel
performance under transient operating conditions is seldom
studied [2,7,8]. For example, a transient test method was proposed
based on experimental setup to determine the transient response
* Corresponding author. Energy Department, Austrian Institute of Technology,
Gienggasse 2, 1210 Vienna, Austria. Tel.: 43(0) 50550 6484; fax: 43(0) 50550
6613; mobile: 43(0) 6505624913.
E-mail addresses: muzaffarali79@yahoo.com, ali.muzaffar@ait.ac.at (M. Ali),
vladimir.vukovic@ait.ac.at (V. Vukovic), mukhtar.sahir@uettaxila.edu.pk
(M.H. Sahir), daniele.basciotti@ait.ac.at (D. Basciotti).
Contents lists available at ScienceDirect
Applied Thermal Engineering
j ournal homepage: www. el sevi er. com/ l ocat e/ apt hermeng
1359-4311/$ e see front matter 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.applthermaleng.2013.08.010
Applied Thermal Engineering 61 (2013) 469e480
of energy wheels [7,8]. In addition, the effect of humidity and
temperature sensors transient characteristics are also analyzed to
measure performance of energy wheels. It was concluded that the
performance analysis of energy wheels based on transient response
is essential for the optimal performance in the real operation.
In solid desiccant cooling system, desiccant wheel is the key
component and the coefcient of performance (COP) of the system
can be improved by enhancing the performance of the desiccant
wheel [9]. The performance of the desiccant wheel is affected by
several variables. The key factors include the rotation speed along
with the inlet air temperature, moisture content, and face velocity
for both process and regeneration air. In addition to the operating
conditions, effects of desiccant wheel channel thickness and rota-
tion speed are analyzed [5,10e13]. Such studies are mainly focused
on the model-based performance analysis of different types of
desiccant wheels. Wide range, from simplied to complex, desic-
cant wheel models are available with respect to single or composite
adsorbent materials. A study presented detailed review of different
models of desiccant wheels [13]. However, despite of numerous
studies, complete understanding of the design and operational
aspects of desiccant wheels remains a challenge, relying primarily
on the manufacturers practical experience, rather than theoreti-
cally and experimentally proven principles. Such theoretical and
experimental works require solutions to differential equations or
interpretation of specic parameters derived from heat and mass
governing equations. The approach could lead to large errors under
particular operating conditions despite of a good performance for a
small range of other conditions. Therefore, a method was presented
to predict the performance of desiccant wheels in actual use [14].
However, the proposed model is only applicable for desiccant
wheel performance when the rotation speed is lower than the
optimal speed.
Recent advancements in the eld of physical modeling offer
opportunities to use equation-based object-oriented modeling and
simulation approach for building energy systems [15,16]. Using
such approach the present work developed a desiccant wheel
model in Dymola/Modelica [17,18]. The existing models lack the
capabilities to handle real-time control strategies with respect to
wheel operational modes and sensor functionalities. Additionally,
the referred model could only analyze the wheel performance
when the rotation speed is lower than the optimal and only predict
system performance in dehumidication mode. Therefore, the
current study enhances a steady state model available in the liter-
ature [14] to incorporate real-time control strategies for both
dehumidication and enthalpy modes under the transient oper-
ating conditions that are encountered in the commercial system
operation. In addition, the developed model is also capable of
nding the optimumwheel rotation speed, N
opt
, and estimating the
wheel performance for any rotation speed, either lower or higher
than the optimum.
The study consists of three sections. The rst section is related to
the model development in terms of mathematical and Dymola/
Nomenclature
A cross sectional area of rotor (m
2
)
a channel height of desiccant wheel (m)
A
sp
specic surface area of desiccant wheel (m
2
/m
3
)
A
ws
surface area of desiccant wheel (m
2
/kg-rotor)
b channel base of desiccant wheel (m)
C adsorption constant (e)
C
pc
specic heat capacity of desiccant wheel (kJ/kg K)
C
pw
specic heat capacity of water (kJ/kg K)
D diameter of desiccant wheel (m)
D
0
coefcient of diffusivity (m
2
/s)
D
sf
surface diffusivity (m
2
/s)
D
eq
equivalent diameter of channels (m)
E
at
activation energy (J/kg)
f Fanning friction factor (e)
G gain in the delay model (e)
H amount of sensible heat removed per hour (kJ/h)
h specic enthalpy of air (kJ/kg)
k loss coefcient due to contraction and expansion of air
K
w
humidity ratio correction coefcient (e)
K
t
temperature correction coefcient (e)
L length of desiccant wheel (m)
m mass ow rate of air (kg/h)
N rotation speed of desiccant wheel (rph)
n adsorption constant (e)
dp pressure drop (Pa)
Q amount of water vapor absorbed onto wheel during an
hour (kg/h)
q
eq
equilibriumamount of water absorbed (kg of moisture/
kg of adsorbent)
q amount of water vapor absorbed onto wheel (kg of
moisture/kg of dry air)
R gas constant (J/kg K)
t
s
adsorption time (s)
T temperature (K)
U supercial velocity of air (m/s)
u air velocity in the channel (m/s)
Greek letters
r density (kg/m
3
)
b fractional area of each zone (e)
u air humidity ratio (kg of moisture/kg of dry air)
4 relative humidity (0e1)
3 effectiveness (e)
s time constant (s)
Subscripts
a air
ad adsorbent
at activation energy
avg average
c corrected
dw desiccant wheel
d distributed
eq equilibrium
in inlet
l local
lt latent
out outlet
opt optimal
pc heat capacity of air at constant pressure
pw heat capacity of water at constant pressure
reg regeneration air
s process air
sn sensible
sf surface
sp specic
tot total
ws wheel surface
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 470
Modelica modeling. The second section compares the results of the
developed desiccant wheel model using silica gel as adsorbent to
the laboratory experiments available in the literature. The third
section is focused on the model calibration and validation in a real
system based on LiCl adsorbent under the transient operating
conditions. In addition, it also contains the model validation with
respect to the optimal rotation speed.
2. Desiccant wheel model development
2.1. Mathematical model
The desiccant wheel is an electrically driven rotating cylinder of
length L and diameter D with small channels in which the adsor-
bent is adhered to the thin walls of matrix material. A typical
desiccant wheel geometry and sinusoidal channel section are
shown in Fig. 1. The mathematical model is based on the set of
algebraic equations [14], in which the enthalpy and humidity
change simultaneously to determine the outlet temperature and
absolute humidity of the process and regeneration air.
Additional correlations are used to determine some specic
properties of moist air based on transient operating conditions, e.g.
relative humidity and specic enthalpy relations to absolute hu-
midity and temperature [19].
In the desiccant wheel, the amount of moisture adsorbed is
characterized by adsorption isotherms, i.e. the amount of adsorbate
on the unit quantity of adsorbent as a function of pressure or
concentration, at constant temperature. The adsorption isotherms
depend on the type of adsorbent material [19,20]. Large variation is
observed for such isotherms because manufacturers use different
strategies to optimize adsorption material for different applica-
tions. However, in the current model, Freundlich equation (1) is
used to determine the equilibrium amount of moisture adsorbed
q
eq
as a function of relative humidity.
q
eq
C4
1=n
(1)
Here, C and n are Freundlich constants and are related to
adsorption capacity and adsorption efciency, respectively.
The isosteric heat of adsorption is calculated through Clausiuse
Clapeyron relation and is determined to be between 2100 and
2300 kJ/kg. The surface diffusivity is correlated with temperature
by the Arrhenius equation (2) in which the activation energy is
measured for diffusion. The activation energy value of 1730 kJ/kg is
about 80% of isosteric heat of adsorption and presents a realistic
estimate [14].
D
sf
D
0
exp
_

E
at
RT
_
(2)
Moreover, the amount of sensible heat exchange between the
hot regeneration air and process air through the rotating wheel per
unit time is calculated by equation (3) [21].
_
H
r
dw
C
pc
_
T
reg
T
sin
_
NL
b
s
r
a
U
s
(3)
In several studies, different desiccant wheel models are devel-
oped that can be categorized based on the consideration of re-
sistances involved during the heat and mass transfer processes. The
gas-side resistance models account for the heat and mass transfer
resistance only in the bulk gas while solid-side resistances are
ignored. The gas and solid-side resistance models consider heat
conduction and mass diffusion within the solid desiccant [22]. The
current study presents a modied model based on mathematical
expressions that assume surface diffusion as a dominant mass
transfer mechanism. Such assumption was also employed in a
number of studies available in the literature [14,20,23,24]. There-
fore, the amount of water vapor adsorbed by the adsorbent material
at the end of adsorption step is calculated by equation (4).
q
_
2
p
_

D
sf
_
t
p
r
ad
A
ws
q
eqsin
(4)
Here the adsorption time is t 3600b
s
/N. Thus, the amount of
water transfer from process air to the outlet of regeneration air
through the desiccant wheel per hour is calculated by equation (5).
_
Q qALr
dr
N
_
2
p
_

D
sf
_

3600b
s
_

N
p
r
ad
A
ws
q
eqsin
ALr
dw
(5)
The outlet conditions of process and regeneration air in terms of
absolute humidity and specic enthalpy are calculated based on
inlet conditions and heat and mass transfer equations (3) and (5)
per unit mass ow rate.
The above equations are used for the desiccant wheel dehu-
midication mode. However, the desiccant wheels from different
manufacturers can also be used as enthalpy wheels under high
rotation speeds. The criterion for operation in the enthalpy mode is
based on the temperature at the inlet of regeneration section.
When such temperature is belowa dened minimum regeneration
temperature, T
reg_min
, the wheel cannot be regenerated. Typically,
T
reg_min
value of 40

C is considered [25]. Therefore, the enthalpy
mode becomes active when the temperature at regeneration inlet is
below T
reg_min
. In such a situation, the outlet temperature and hu-
midity conditions of the process air are calculated from equations
(6) and (7). Otherwise the wheel always operates in the dehu-
midication mode.
Fig. 1. Schematic of desiccant wheel with temperature (T), relative humidity (H) and rotational speed (N) sensors.
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 471
T
sout
T
sin
3
sn
_
T
reg
T
sin
_
(6)
u
sout
u
sin
3
lt
_
u
reg
u
sin
_
(7)
2.1.1. Optimal rotation speed
The rotation speed of the desiccant wheel, N, is the key oper-
ating parameter to signicantly improve the wheel performance.
The optimal rotation speed, N
opt
, achieves the maximum moisture
removal ensuring minimum absolute humidity at the outlet of
process air [14,26,27]. Thus, the rotation speed should be low
enough for complete regeneration or quick cooling of the wheel
but also high enough to keep the adsorbent far from equilibrium
state. However, N
opt
is inuenced by various operating parameters,
such as face velocity, inlet temperature, inlet humidity of the
process and regeneration air. In several research works, the optimal
rotation speed is predicted by a numerical solution of differential
equations governing the underlying processes [26e29]. In the
current study, the line of constant relative humidity is regarded as
an isostere (line of constant concentration of adsorbent). Thus, the
relative humidity at the outlet of process air is considered equal to
that of regeneration air to estimate the N
opt
, as dened in equation
(8) [14].
4
sout
u
sout
; h
sout
4
rin
u
rin
; h
rin
(8)
The equation denes the lowest possible dehumidication point
that can be achieved along the ideal adiabatic dehumidication line
on the psychometric chart.
Equation (9) is obtained by rearranging equation (5) to calculate
the optimal values of the desiccant wheel rotation speed.
N
opt

_
Q
opt
=
_
2
p
_

D
sf
_

3600b
s
_

N
p
r
ad
A
ws
q
eqsin
ALr
dw
_
2
(9)
However, the mathematical model of the proposed method is
able to predict the outlet conditions only when the rotation speed is
equal to or below the optimal. Therefore, equations (10) and (11)
introduce two correction factors in the model, one for tempera-
ture, K
t
, and another for absolute humidity, K
w
, to estimate the
outlet conditions of process air when rotation speed is higher than
optimal. The linear relationships are assumed based on the exper-
imental results presented in the literature [14,27,29].
T
csout
T
sopt
T
sin
K
t
_
N
in
N
opt
_
(10)
W
csout
W
sopt
X
sin
K
w
_
N
in
N
opt
_
(11)
Fig. 2. Dymola graphical representation of the desiccant wheel model.
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 472
2.1.2. Pressure drop
The total pressure drop across the desiccant wheel is given by
equation (12) as the sum of distributed and local pressure drops
[29]. The distributed and local pressure drops are calculated by
equations (13) and (14), respectively. Equation (15) is used to
calculate the equivalent diameter of the sinusoidal desiccant wheel
channels.
dP
total
dP
d
dP
l
(12)
dP
d
4f
_
L
D
eq
__
1
2
_
r
a
u
2
(13)
dP
l
k
_
1
2
_
r
a
u
2
(14)
D
eq
a
_
1:0542 0:4670
_
a
b
_
0:1180
_
a
b
_
2
0:1794
_
a
b
_
3
0:0436
_
a
b
_
4
_
(15)
Finally, equations (16) and (17) are used for process air outlet
temperature and absolute humidity, respectively, to account for a
possible time delayed response of the monitoring sensors.
dT
sout
=dt
_
GT
sin
T
sout
s
t
_
(16)
du
sout
=dt
_
Gu
sin
u
sout
s
w
_
(17)
2.2. Dymola/Modelica model
The mathematical model of desiccant wheel is implemented in
Modelica language and simulations are performed in Dymola. The
graphical representation of the desiccant wheel model is shown in
Fig. 2. Appropriate control strategies are applied to estimate the
wheel performance under various design and operating conditions.
For example, two approaches can be implemented for the rotation
speed control through the conditional input, N
in
. The rst approach
is related to the wheel operationwhen rotation speed is not dened
by the user i.e. N
in
is inactive. In such a case, the model automati-
cally determines the N
opt
and all outlet conditions of the process
and regeneration air would be at the optimal values. In the second
approach, the user can implement the desired rotation speed by
making N
in
active to determine the actual performance of desiccant
under the particular operating conditions. In addition to the
dehumidication mode, the developed model can also be applied
for the enthalpy mode operation. The enthalpy mode is decided by
the minimum value of regeneration temperature that is imple-
mented as a model input parameter, T
reg_min
. The inlet regeneration
temperature is compared with the T
reg_min
through mode decision
maker component, MDMaker. The MDMaker provides a
threshold to decide the operational mode (dehumidication or
enthalpy) based on the T
reg_min
. The applied temperature at the
inlet regeneration section is compared with the T
reg_min
dened as
model input. If the inlet regeneration temperature is lower than the
dened value of T
reg_min
then the model would be in the enthalpy
mode, otherwise it will operate in the dehumidication mode.
Consequently, the MDMaker controls the input signals of Mod-
eswitch1 and Modeswitch2 equation switches for process outlet
temperature and absolute humidity, respectively. If the input signal
of a switch is true then equations related to the dehumidication
mode are considered, otherwise equations for the enthalpy mode
come into play.
However, for the dehumidication mode, an additional control
is provided through DesiccantSwitch which enables handling of
the mode shift situations when the difference between process and
regeneration air inlet relative humidity becomes zero. If the dif-
ference is more than 10% then the outlet temperature and absolute
humidity are calculated from the dehumidication mode equa-
tions, otherwise the outlet conditions are the average of inlet
conditions. Furthermore, a delay model based on equations (16)
and (17) is applied to the simulated outlet values to consider a
probable time delayed response. The pipe model from the LBL
buildings library [30] based on equations (12)e(14) is used to
determine the pressure drop on the process and regeneration sides
across the desiccant wheel. Finally, the resulting wheel model is
developed for calibration under transient operating conditions and
prediction of the real system performance, implicitly including
system time constants through the calibration process.
3. Model parametric performance analysis
In the present work, two wheel types are analyzed based on
silica gel and LiCl as adsorbents. The presented EOO based
modeling provides an easy approach to change the wheel design
parameters for considering wheels of any type. In the current
section, the performance of a desiccant wheel composed of silica
gel adsorbent is analyzed under laboratory conditions. The ther-
modynamic and geometric properties of the desiccant wheel used
in the simulation process are dened in Table 1 [31,32]. The varia-
tion ranges of operating parameters considered in the analysis are
listed in Table 2.
3.1. Effect of process air inlet conditions
The key operating parameters related to the inlet conditions of
the process air include air velocity, temperature, T
sin
, and humidity
ratio, u
sin
. The inuence of these parameters on the outlet humidity
along with moisture removed from the process air is shown in
Fig. 3. In addition, inlet humidity ratio also affects the outlet tem-
perature of the process air.
Fig. 3A conrms that process air velocity signicantly affects the
outlet humidity ratio [27,31]. The outlet humidity ratio increases
with rising air velocity decreasing the moisture removal. Thus, air
with that passes through the wheel with the higher velocity is dried
less deeply as compared to low air velocity. Therefore, achieving
low process air moisture content at low air velocity requires
Table 1
Thermodynamic properties and geometric parameters of the desiccant wheel used
for the performance analysis.
Desiccant material Silica Gel-Type A
Wheel diameter, D (mm) 320
Wheel length, L (mm) 200
Overall cross-sectional area, A (m
2
) 0.08
Fractional area of process zone, b
s
0.465
Channel pitch, axb (mm) 3.2 1.8
Bulk density, r
b
(kg/m
3
) 250
Specic surface area, A
sp
(m
2
/m
3
) 3000
Surface area of desiccant wheel, A
ws
(m
2
/kg-wheel) 12 (3000/250)
Adsorption isotherm constants:
C (e) 0.24
n (e) 1.5
Specic heat capacity, C
pc
(J/kg K) 921
Density of adsorbent, r
ad
(kg/m
3
) 1200
Activation energy, E
at
(kJ/kg) 1730
Diffusion coefcient, D
0
(m
2
/s) 2.27 10
7
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 473
increase in the size of the wheel. The process air inlet humidity,
u
sin
, proportionally inuences the outlet humidity ratio as shown
in Fig. 3B. Fig. 3C shows that desiccant wheel has better dehu-
midication performance at the lower T
sin
resulting in higher
moisture removal. In addition, the outlet temperature of process air
also rises with the increase in inlet humidity, as shown in Fig. 3D.
The Dymola model prediction trends in Fig. 3 are in good agree-
ment with other published studies [27,31,33,34]. However the
actual values are different due to different inlet conditions and
wheel properties. For example, in the study [31] the inlet condi-
tions of process air are 21

C, 8.0 g/kg and zone ratio of process and
regeneration air 3:1 whereas in the current study these parameters
are 31

C, 10 g/kg, and 1:1, respectively. Additionally, the overall
impact of the process air inlet humidity ratio on the outlet condi-
tions presented in Fig. 4, for the inlet temperatures of 25

C and
35

C are similar to the results presented in the literature [31].
3.2. Effect of regeneration air inlet conditions
On the regeneration side, the performance of desiccant wheel is
mainly affected by regeneration temperature, T
reg
, and humidity
ratio, u
reg
, as shown in Fig. 5. Increasing T
reg
increases moisture
removal resulting in lower outlet humidity ratio of process air, as
shown in Fig. 5A. However, each additional degree of T
reg
causes
more moisture removal at low as compared to high T
reg
values.
Contrary to the former aspect, increased air humidity at the
regeneration inlet decreases the dehumidication of the process air
due to reduced absolute moisture difference, as shown in Fig. 5B.
Moreover, Fig. 5C shows that high T
reg
also results in high outlet
temperature of process air. The fact is due to the rise in the heat of
adsorption with more latent-to-sensible heat conversion. The
resulted trends presented in Fig. 5 are similar to the other published
studies [27,31,33,34].
3.3. Effect of rotation speed
The desiccant wheel rotation speed, N, signicantly affects the
wheel dehumidication performance, as shown in Fig. 6. It is
Table 2
Range of operating parameters for the performance analysis.
Parameters Baseline
values
Parametric
variations
Temperature of ambient air, T
amb
(

C) 30 e
Humidity ratio of ambient air, u
amb
(g/kg) 10 e
Inlet temperature of process air, T
sin
(

C) 31 20e40
Inlet humidity ratio of process air, u
sin
(g/kg) 10.3 4e12
Supercial velocity of process air, m (m/s) 2 1e3.5
Regeneration temperature, T
reg
(

C) 80.5 45e120
Inlet humidity ratio of regeneration air,
u
reg
(g/kg)
12.5 25e25
Rotation speed, N (rph) 36 5e65
Fig. 3. Effect of face velocity, inlet humidity ratio, inlet temperature of process air on process outlet humidity ratio (A), (B), (C); effect of process inlet humidity ratio on process
outlet temperature (D).
Fig. 4. Effect of process inlet conditions on process air outlet conditions.
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 474
always desired to operate the desiccant wheel at its optimal rota-
tion speed N
opt
at which the maximum moisture removal is ach-
ieved. If the wheel rotates at high speed then the process side of the
wheel does not have enough time to remove the moisture. Fig. 6A
and B shows that N
opt
increases with respect to u
sin
. High inlet
humidity ratio improves mass transfer capacity that reduces the
time to reach adsorption equilibrium, resulting in high values of
N
opt
.
In addition, low T
sin
is more favorable for high dehumidication
at low rotation speed compared to high T
sin
, as shown in Fig. 6C. At
high T
sin
, the relative humidity decreases at constant absolute hu-
midity that causes less moisture removal capacity and increases
N
opt
. However, an increase in N above N
opt
does not cause any
further dehumidication as the equilibrium state had already been
reached. Moreover, N
opt
also increases with rising T
reg
as shown in
Fig. 6D and E: N
opt
increases from 11 rph to 39 rph when T
reg
changes from 45

C to 110

C and all other parameters remain
constant. At high T
reg
, it is easier to desorb the adsorbent. Thus, the
wheel rotation speed is increased for rapid movement of the wheel
regeneration zone toward the process side. In the current study, the
effects of wheel rotation speed on process air outlet humidity ratio
as a function of different conditions of process and regeneration air
showed the similar impact compared to various other studies
[27,34]. However, the values of N
opt
of each study are different due
to difference in the inlet air conditions and wheel properties.
4. Model calibration and validation
Following the performance evaluation of the desiccant wheel
model in comparison to the published experimental results, typi-
cally based on silica gel, presented in Section 3, the same model is
used for the simulations of a real system composed of LiCl.
Consideration of a different desiccant type is possible due to model
calibration and validation under transient operating conditions of
the real system. In both cases the respective thermo-physical and
geometry properties of the adsorbents were used.
4.1. Model calibration
The desiccant wheel model is calibrated for the identication of
design parameters under the transient operating conditions. Such
key design parameters are often not provided from the manufac-
turer. The considered transient operating conditions are based on
the monitoring data of a commercial desiccant cooling system
installed in the ENERGYbase ofce building in Vienna, Austria [35].
The ENERGYbase was designed as a passive house building with
renewable technologies to cover the energy demands of heating,
cooling, and ventilation. The conditioned area of the building is
7025 m
2
. The desiccant cooling system(DCS) of the building is solar
thermally driven, composed of two identical units of 8240 m
3
/h
design airow each. The DCS consists of desiccant wheel, heat
wheel, and two spray dehumidiers one in the supply and return
sides. The DCS normally operates for 12 h a day from 7:00 to 19:00,
5 days a week. The whole building is equipped with over 500
sensors to continuously measure all relevant parameters of the
building HVAC system along with indoor environmental condi-
tions. The DCS is monitored and controlled with two data moni-
toring tools: DESIGO [36] and JEVis [37]. The desiccant wheel is
equipped with 9 sensors as shown in Fig. 1. Four sensors are tem-
perature sensors and the other four are the relative humidity sen-
sors installed at the inlet and outlet of process and regeneration air.
One sensor monitors the rotation speed of the wheel. The details of
the desiccant wheel sensors in terms of their accuracy and oper-
ating range are available from the manufacturer [38]. In addition,
the other operating conditions, such as ow rates of process and
regeneration air are also acquired from the building monitoring
system. The desiccant wheel can operate in either dehumidication
or enthalpy mode based on the operating conditions. For the
Fig. 5. Effect of inlet humidity ratio and regeneration temperature on process outlet humidity ratio (A) and (B), respectively; effect of regeneration temperature on outlet tem-
perature (C).
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 475
dehumidication mode, the maximumrotation speed is 20 rph and
in the enthalpy mode wheel can be rotated up to 10 rpm. The de-
tails provided by the wheel manufacturer only include basic
properties such as adsorbent material (LiCl), wheel diameter
(1680 mm) and length (450 mm). Consequently, the developed
model is calibrated for the identication of design parameters of
the installed wheel. The wheel design parameters identied
through calibration include the thermodynamic and structure
properties of the wheel matrix, such as constants of adsorption
isotherm, surface diffusivity, specic surface area, bulk wheel
density, and density of adsorbent. Although the wheel material and
geometry are xed, the manufacturers typically do not provide all
the properties of desiccant matrix except wheel diameter and
depth. Therefore the wheel material properties are also used as
optimization variables during calibration. The variation range of all
design parameters used in the calibration phase is shown inTable 3.
The variation range is based on the available data of different
commercial wheels and the thermodynamic properties of LiCl [20].
However, the step size dened in Table 3 is just an initial guess for
the optimization algorithm. The optimization starts its search with
the provided value and considers the dened step size to determine
the initial search direction. Afterward, the algorithm automatically
renes the step size within the upper and lower bounds depending
on the initial results toward minimization of the objective function.
In the current work, the calibration is executed through opti-
mization process by coupling Dymola/Modelica with an optimiza-
tion tool, GenOpt [39]. In such an approach, GenOpt used the
dened set of desiccant wheel design parameters to provide inputs
Fig. 6. Effect of process inlet humidity ratio on optimal rotation speed (A); effect of rotation speed on process outlet humidity ratio with process humidity ratio (B), and process inlet
temperature (C) as parameter; effect of regeneration temperature on optimal rotation speed (D), and effect of rotation speed on process outlet humidity ratio with regeneration
temperature as parameter, (E).
Table 3
Variation range of calibration parameters.
Parameters Initial
values
Step
size
Parametric
variations
Specic area of wheel, A
sp
(m
2
/m
3
) 2500 500 1000e5000
Bulk density of wheel, r
b
(kg/m
3
) 200 50 150e1000
Density of adsorbent, r
dw
(kg/m
3
) 1150 50 700e2000
Diffusion co-efcient, D
0
(m
2
/s) 1.5E-7 1E-7 1E-9 to 6E-7
Activation energy, E
at
(kJ/kg) 1800 10 1500e3500
Adsorption co-efcient, C (e) 0.15 0.1 0.1e10
Adsorption co-efcient, n (e) 1 0.1 1.5e10
Humidity correction factor, K
w
(e) 0.01 0.001 0e0.05
Temperature correction factor, K
t
(e) 0.01 0.001 0.01e0.05
Sensible and latent effectiveness, 3
sn
, 3
lt
(e) 0.6 0.1 0.5e0.9
Time delay of temperature sensors, s
1
(e) 1 50 1e2000
Time delay of humidity sensors, s
2
(e) 1 50 1e2000
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 476
to the Dymola simulation program. Meanwhile, Dymola provided
text les with values of the target functions essential for optimi-
zation. The advantages of such approach are the capabilities of
GenOpt to utilize multi-core processing units for improved opti-
mization performance. Discrete Armijo Gradient optimization al-
gorithm of generalized pattern search methods is applied with the
typical values for the algorithm parameters [37]. The transient
operating conditions from the monitoring data of ENERGYbase
system are appropriately provided to the desiccant wheel model.
The mean squared error (MSE) between the monitored and simu-
lated outlet specic enthalpy of process air is considered as a cali-
bration objective function.
4.1.1. Model calibration data
As the desiccant wheel and the developed model can operate in
both, dehumidication and enthalpy modes depending on the
outdoor conditions, transient data sets for the model calibration are
selected to cover both operating modes of the wheel. Therefore, the
measurement data of four months in 2012 (April to July) are
analyzed and 6 days are found within this period to cover both
operating modes. The minimum response time interval of the
installed sensors is 1 min. The 1 min time interval can be consid-
ered an instrumental limitation of sensors installed in the system.
However, in the present work real transient conditions of a com-
mercial system are used where 1 min time interval in control
systemis generally in practice. Intervals shorter than 1 min become
too sensitive to control the operation of real building systems.
Consequently, the transient data set consists of 12 h with 1 min
time interval measurements over 6 days during the four months of
desiccant operation in the summer of 2012. However, the data may
not be representative of the true wheel performance, obtained
using highly sophisticated monitoring equipment and carefully
controlling all the parameters in laboratory conditions. Instead,
obtained measurements are from a real operating system installed
in a commercial building. The sensors in the ducts are properly
installed by Siemens Company to ensure mixed air at the inlet and
outlet of process and regeneration sides and enable measurements
of mixed air properties. The sensor is positioned in the middle box
with tubes collecting the air from all directions to ensure mixed air
conditions, as presented in Fig. 7.
For the model calibration, extensive simulations are performed
through coupling Dymola with GenOpt for all selected days. Each
calibration error is based on the transient data of each particular
day. The calibration process is completed in three stages due to the
sensitivity of the design parameters. In the rst stage, the wheel
design parameters with respect to properties of matrix material are
considered. The second stage is related with the sensors delay
parameters. In the last stage, the correction coefcients are
considered. The values of the desiccant wheel design parameters
identied by optimization algorithm during the calibration process
are given in Table 4. The resulted matrix of fundamental properties
is almost constant for all calibration days. The calibration RMSE of
each day along with the ambient conditions for 12 h system
operation is given in Table 5. Thus, the calibrated curve t is based
on a single day transient data. Afterward, the tted model is vali-
dated over additional days showing the predictive value of the
model. Furthermore, the calibration results of one exemplary day,
Fig. 7. Sensor position and setup for measurements.
Table 4
The values of calibration parameters determined by the optimization algorithm.
Cal. Stage Parameters Calibration day
30th April 19th June 29th June 2nd July 6th July 9th July
1st Stage A
sp
3021.032 3134.877 3085.816 3075.673 3021.893 3021.145
r
b
245.966 241.496 244.359 245.020 248.122 247.938
r
dw
1201.010 1203.479 1202.257 1202.169 1200.668 1200.681
D
0
2.303E-7 2.623E-7 2.494E-7 2.506E-7 2.323E-7 2.319E-7
E
at
1729.769 1729.439 1729.631 1729.670 1729.883 1729.875
C 0.226 0.312 0.283 0.281 0.247 0.245
n 1.506 1.506 1.503 1.502 1.501 1.501
3
sn
0.892 0.700 0.697 0.691 0.698 0.699
3
lt
0.600 0.698 0.692 0.698 0.691 0.694
2nd Stage s
1
451.223 431.698 384.877 313.343 329.626 417.335
s
2
3.962 122.273 38.040 27.152 50.193 63.030
T
reg_min
312.841 312.720 311.966 312.815 311.836 312.708
3rd Stage K
w
0.0099 0.0099 0.0099 0.00991 0.00991 0.0099
K
t
4.259E-4 9.748E-4 9.544E-4 8.706E-4 9.635E-4 8.898E-4
Table 5
Ambient conditions and calibration errors of different days.
Calibration
day
T
avg
(T
min
, T
max
)
(

C)
4
avg
(4
min,
4
max
)
(%)
u
avg
(u
min
, u
max
)
(g/kg)
Calibration
error (kJ/kg)
30th April 28.0 (15, 33.1) 32.0 (14.4, 94.5) 7.5 (4.1, 9.9) 2.8
19th June 31.8 (23.7, 35.3) 46.7 (33.9, 77.7) 13.6 (10.9,14.1) 1.0
29th June 30.6 (21.2, 36.1) 53.7 (32.4, 100) 14.6 (11.1, 16.7) 1.2
2nd July 32.5 (21.6, 36.6) 47.1 (27.5, 100) 14.3 (10.1, 17.2) 1.3
6th July 31.6 (23.2, 36.7) 47.6 (29.8, 85.1) 13.7 (10.9, 17.5) 1.2
9th July 28.1 (21, 30.1) 62.2 (40.4, 100) 14.6 (10.1, 20.3) 0.8
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 477
i.e. July 2, 2012, in terms of monitored and simulated values of
process outlet specic enthalpy, temperature and humidity ratio
are shown in Fig. 8. The gure also presents the transient conditions
of inlet temperature and absolute humidity of process and regen-
eration air for the same day.
4.2. Model validation
The calibrated desiccant wheel model is validated against the
transient measurements of all the days except the calibration day
considering MSE between the monitored and simulated outlet
specic enthalpy of process air as an objective function. For
example, the desiccant wheel design parameters identied dur-
ing the calibration based on 30th April are used to validate the
model to determine the RMSE against the transient operating
conditions of ve other days. In addition, the mean percentage
error (MPE) is also calculated based on the simulated and
measured specic enthalpy at the outlet of process air. RMSE and
MPE over all monitoring days inclusive of the respective cali-
bration are given in Table 6. The resulted minimum average
validation error is 1.9 kJ/kg on 4 days i.e. 29th June, 2nd, 6th, and
9th July, 2012 and the maximum error is 3.6 kJ/kg on 30th April,
2012. However, the minimum and maximum MPE are 0.2% and
4.6% occurred on 29th June and 30th April, respectively as
highlighted. The predicted results and monitored values with the
minimum and maximum validation errors of 2nd July and 30th
April, respectively in terms of the outlet specic enthalpy,
temperature and humidity ratio of the process air are shown in
Fig. 9.
The percent of measurement points within the 5% MPE over all 6
measurement days were calculated with respect to the selected
calibration days. The parameters resulting from the calibration on
29th June yielded the best simulation results being within 5% of the
measured values 84% of the measurement time over the 6
considered days. Thus, on average 10 out of 12 h per day the model
had less than 5% MPE when calibrated with the measurements
taken on June 29. Similarly, the percentage of time the model re-
sults were within 5% MPE for the other calibration days were: 46%,
77%, 83%, 82%, and 82% for 30th April, 19th June, 2nd, 6th, and 9th
July, respectively.
4.3. Validation of optimal rotation speed
The concept of estimating the optimal rotation speed is also
validated in comparison to the experimental results. The pre-
dictions of N
opt
are carried out under the wide range of operating
conditions at different regeneration temperatures, as shown in
Fig. 10. The considered desiccant wheel design parameters are
given in Table 1. The results showed that the predicted values of
N
opt
are in good agreement with the experimental data [14].
Fig. 8. Comparison between monitored and simulated results of process air outlet: specic enthalpy (A), temperature (B), and absolute humidity (C).
Table 6
Validation errors of different days.
Calibration day Validation error, RMSE [kJ/kg] and MPE (%) Average values
30th April 19th June 29th June 2nd July 6th July 9th July
30th April 2.8 (4.6) 4.3 (3.8) 4.1 (4.8) 3.9 (5.4) 3.4 (4.7) 3.0 (4.5) 3.6 (4.6)
19th June 4.6 (3.8) 1.0 (0.3) 1.7 (0.7) 1.9 (1.2) 2.3 (2.4) 2.8 (3.8) 2.4 (0.7)
29th June 4.4 (4.2) 1.2 (1.0) 1.2 (0.1) 1.3 (0.2) 1.6 (1.4) 1.9 (2.5) 1.9 (0.2)
2nd July 4.3 (4.5) 1.5 (1.4) 1.4 (0.7) 1.3 (0.4) 1.3 (0.6) 1.5 (1.8) 1.9 (0.8)
6th July 4.1 (4.7) 1.9 (1.9) 1.7 (1.4) 1.5 (1.2) 1.2 (0.2) 1.0 (0.6) 1.9 (1.5)
9th July 4.0 (5.0) 2.2 (2.3) 1.8 (1.8) 1.6 (1.7) 1.1 (0.7) 0.8 (0.0) 1.9 (1.9)
Italics and underline values show the calibration day errors, while the remaining values are validation errors.
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 478
Therefore, the developed desiccant wheel model can also predict
N
opt
with good accuracy in most operating conditions.
4.4. Model calibration criteria
During the model calibration and validation process, it was
observed that the errors are inuenced by the average ambient
conditions, as shown in Table 5. Therefore a criterion was dened
for appropriate calibration and validation of the model subject to
different climate conditions. The impact of ambient conditions on
the calibration and validation errors is presented in Fig. 11. The
results showed that both calibration and average validation errors
are smaller at the high ambient absolute humidity ratios. Therefore,
model calibration and validation at high outdoor absolute humidity
levels is recommended.
5. Conclusions and future work
In the current study, a desiccant wheel model is developed in an
equation-based object-oriented environment, Dymola/Modelica,
for performance prediction in the real systemoperation. The model
can estimate actual and optimal operation under wide range of
operating conditions with respect to actual and optimal wheel
rotation speed. Additionally, the developed wheel model is also
capable of predicting accurately the wheel operation in dehumid-
ication and enthalpy modes.
The model is calibrated and validated over 6 days of transient
measurements of a commercial desiccant wheel installed in
ENERGYbase, passive house ofce building in Vienna, Austria. The
mean squared error (MSE) between the monitored and simulated
Fig. 9. Comparison between the simulated and monitored results: process air outlet specic enthalpy (A, B), and temperature and humidity ratio (C, D) (graphs on the left, A and C,
are for the case with minimum validation error, graphs on the right, B and D, are for the case with maximum validation error).
Fig. 10. Prediction of the optimal rotation speed at different regeneration
temperatures.
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 479
outlet specic enthalpy of process air is considered as a calibration
and validation objective function. The results showed that the
model is able to predict the transient performance of desiccant
wheels with good accuracy. However, calibration and validation
results are dependent on the selection of transient measurements
along with the optimization process adopted in the study. The
model was also used to analyze the effects of the inlet conditions of
process and regeneration air on the wheel performance. The ob-
tained trends are in good agreement with the published data.
In the future, the calibrated and validated desiccant wheel
model will be used to analyze the performance of different con-
gurations of desiccant cooling systems.
References
[1] ARI, ARI Standard 1060-2005, Performance Rating of Air-to-air Heat Ex-
changers for Energy Recovery Ventilation Equipment, Air-Conditioning and
Refrigeration Institute, Arlington, VA, 2005.
[2] O. Omobayode Abe, Effectiveness of Energy Wheels from Transient Mea-
surements. MSc Thesis, Department of Mechanical Engineering, University of
Saskatchewan, Saskatoon, Canada, 2005.
[3] ARI, ARI Standard 940-98, Standard for Desiccant Dehumidication Components,
American Air-Conditioning and Refrigeration Institute, Arlington, VA, 1998.
[4] D.L. Ciepliski, C.J. Simonson, R.W. Besant, Some recommendations for im-
provements to ASHRAE Standard 84-1991, ASHRAE Trans. 104 (1B) (1998)
1651e1665.
[5] C.J. Simonson, R.W. Besant, Energy wheel effectiveness: part I e development
of dimensionless groups, Int. J. Heat Mass Transfer 42 (1999) 2161e2170.
[6] C.J. Simonson, R.W. Besant, Energy wheel effectiveness: part II e correlations,
Int. J. Heat Mass Transfer 42 (1999) 2171e2185.
[7] O.O. Abe, C.J. Simonson, R.W. Besant, W. Shang, Effectiveness of energy wheels
from transient measurements: part Idprediction of effectiveness and uncer-
tainty, Int. J. Heat Mass Transfer 49 (2006) 52e62.
[8] O.O. Abe, C.J. Simonson, R.W. Besant, W. Shang, Effectiveness of energy wheels
from transient measurements: part IIdResults and verication, Int. J. Heat
Mass Transfer 49 (2006) 63e77.
[9] T.S. Kang, I.L. Maclein-Cross, High performance solid desiccant cooling cycles,
Trans. ASME 111 (1989) 176e183.
[10] D. Charoensupaya, W.M. Worek, Parametric study of an open-cycle adiabatic,
solid, desiccant cooling system, Energy 13 (1988) 739e747.
[11] W. Zheng, W.M. Worek, Numerical simulation of combined heat and mass
transfer processes in a rotary dehumidier, Numer. Heat Transfer, Part A 23
(1993) 211e232.
[12] F. Nia, D. Van Paassen, M. Saidi, Modeling and simulation of desiccant wheel
for air conditioning, Energy Build. 38 (2006) 1230e1239.
[13] T.S. Ge, Y. Li, R.Z. Wang, Y.J. Dai, A review of the mathematical models for
predicting rotary desiccant wheel, Renewable Sustainable Energy Rev. 12
(2008) 1485e1528.
[14] A. Kodama, T. Hirayama, M. Goto, T. Hirose, R.E. Critoph, The use of psycho-
metric charts for the optimization of a thermal swing desiccant wheel, Appl.
Therm. Eng. 21 (2001) 1657e1674.
[15] G. Fontanella, et al., Calibration and validation of a solar thermal system
model in Modelica, Build. Simul. 5 (2012) 293e300.
[16] M. Ali, V. Vukovic, M.H. Sahir, Methodology for automated optimization of
HVAC system congurations, in: 41st International KGH Congress on HVAC&R,
Belgrade, Serbia, December 1e3, 2010.
[17] Modelica and the Modelica Association. Available from: http://www.
modelica.org (accessed 18.04.11).
[18] Dymola, Dynamic Modeling Laboratory. Available from: http://www.3ds.com/
products/catia/portfolio/dymola (accessed 18.04.11).
[19] ASHRAE, ASHRAE Handbook: Fundamentals, American Society of Heating,
Refrigeration and Air-conditioning Engineers, Inc., Atlanta, 2009.
[20] B.K. Parsons, A.A. Pesaran, D. Bharathan, B. Shelpuk, Evaluation of Thermally
Activated Heat Pump/Desiccant Air Condition Systems and Components, Solar
Energy Research Institute, 1987.
[21] A. Kodama, M. Goto, T. Hirose, T. Kuma, Performance evaluation for a thermal
swing honeycomb rotor using a humidity chart, J. Chem. Eng. Jpn. 28 (1995)
19e24.
[22] R. Narayanan, W.Y. Saman, S.D. White, M. Goldsworthy, Comparative study
of different desiccant wheel designs, Appl. Therm. Eng. 31 (2011) 1613e
1620.
[23] Z. Gao, V.C. Mei, J.J. Tomlinson, Theoretical analysis of dehumidication pro-
cess in a desiccant wheel, Heat Mass Transfer 41 (2005) 1033e1042.
[24] A.A. Pesaran, A.F. Mills, Moisture transport in silica gel packed beds. Available
from: http://www.nrel.gov/docs/legosti/old/2388.pdf.
[25] M. Beccali, F. Butera, R. Guanella, R.S. Adhikari, Simplied models for the
performance evaluation of desiccant wheel dehumidication, Int. J. Energy
Res. 27 (2003) 17e29.
[26] W. Zheng, W.M. Worek, D. Novesel, Performance optimization of rotary
dehumidier, J. Sol. Energy Eng. 117 (1995) 40e44.
[27] T.S. Ge, F. Ziegler, R.Z. Wang, A mathematical model for the performance of a
compound desiccant wheel (a model of compound desiccant wheel), Appl.
Therm. Eng. 30 (2010) 1005e1015.
[28] J.D. Chung, D.Y. Lee, S.M. Yoon, Optimization of desiccant wheel speed and
area ratio of regeneration to dehumidication as a function of regeneration
temperature, Sol. Energy 83 (2009) 625e635.
[29] S.D. Antonellis, C.M. Jappolo, L. Molinaroli, Simulation, performance analysis
and optimization of desiccant wheel, Energy Build. 42 (2010) 1386e1393.
[30] M. Wetter, Modelica library for building heating, ventilation and air-
conditioning systems, in: Proceedings of 7th Modelica Conference, Como,
Italy, September 20e22, 2009.
[31] ASHRAE, ASHRAE Handbook: Systems and Equipment, American Society of
Heating, Refrigeration and Air-conditioning Engineers, Inc, Atlanta, 2008.
[32] A. Kodama, M. Goto, T. Hirose, T. Kuma, Temperature prole and optimal
rotation speed of a honeycomb rotor adsorber operated with thermal swing,
J. Chem. Eng. Jpn. 27 (1994) 644e649.
[33] S.D. White, et al., Characterization of desiccant wheels with alternative ma-
terials at low regeneration temperatures, Int. J. Refrigeration 34 (2011) 1786e
1791.
[34] S. Yamaguchi, K. Saito, Numerical and experimental performance analysis fo
rotary desiccant wheels, Int. J. Heat Mass Transfer 60 (2013) 51e60.
[35] EnergyBase, Energy Concept e Efciency Through Innovation. Available from:
http://www.energybase.at/eng/im_energiekonzept.html (accessed 16.04.11).
[36] Siemens, DESIGO Building Automation System. Available from: http://www.
big-u.org/catalog/siemens/DESIGO%20V4%20system%20description.pdf
(accessed 20.09.10).
[37] P. Palensky, The JEVis Service Platform e Distributed Energy Data Acquisition
and Management, Vienna University of Technology, 2005.
[38] EE Elektronik, Serie EE29/EE31. Available from: http://www.
pruemmfeuchte.de/templates/pruemmtempl/pdf/polymersensoren/e_und_e/
DB_EE31-29_de.pdf (accessed 16.08.11).
[39] M. Wetter, GenOpt Generic Optimization Program, User Manual. Available
from: http://gundog.lbl.gov/GO/download/manual-2-1-0.pdf (accessed
16.08.10).
Fig. 11. Effect of ambient absolute humidity on calibration (Cal_Err) and average
validation (Val_Err) errors.
M. Ali et al. / Applied Thermal Engineering 61 (2013) 469e480 480

Vous aimerez peut-être aussi