Energy Conversion and Management 47 (2006) 2659–2672

www.elsevier.com/locate/enconman

A simple dynamic model of cooling coil unit

a,*
Guang-Yu Jin , Wen-Jian Cai a, Yao-Wen Wang a, Ye Yao b

a
School of Electrical and Electronic Engineering, Nanyang Technological University, Nanyang Avenue, Singapore 639798, Singapore
b
Institute of Refrigeration and Cryogenics, Shanghai Jiao Tong University, Shanghai 200330, PR China

Received 25 April 2005; accepted 14 October 2005

Available online 29 November 2005

Abstract

In this paper, a dynamic cooling coil unit (CCU) model is developed by extending the CCU engineering model and by
combining the model with the mass and energy balance equations. Commissioning information is then used to estimate, at
most, six model parameters by a nonlinear on line identification method. Unlike other existing dynamic CCU models for
control and optimization, this modeling method is relatively simple and exactly captures the nonlinear characteristics over
a wide operating range of the CCU without requiring geometric specifications, which is very convenient in real time engi-
neering control practice. Experiments on a centralized heating, ventilation and air conditioning (HVAC) pilot plant are
conducted to show that the method is robust and gives an accurate match to real performance over the wide range of oper-
ating conditions. This method is expected to work well for real time control of an operating HVAC system. Even more, the
modeling methodology can also be extended to other heat exchangers.
 2005 Elsevier Ltd. All rights reserved.

Keywords: Heating; Ventilation and air conditioning system; Cooling coil unit; Air handling unit; Engineering model; Mass and energy
balance; Dynamic modeling; Experiments

1. Introduction

The cooling coil unit (CCU), as shown in Fig. 1, is the basic equipment in an air handling unit (AHU) of
heating, ventilating and air conditioning (HAVC) systems. The CCU plays an essential role [2,3] to transfer
the cooling load from the air loop to the chilled water loop by forcing the air to flow over the coil and into
the space to be conditioned. The performance of the coils, which is embodied through their heat transfer char-
acteristics, directly influences the performance of the HVAC system. Therefore, accurate predication of the
CCU performance under transient conditions is required for the optimized control strategies of the entire
HVAC system [4,5].
Various approaches have been suggested, and these can generally be placed in two categories: finite differ-
ence models and lumped parameter models. The finite difference approach results in a large number of equa-
tions that are suitable only for numerical simulation. For this approach, Myers et al. [6] and Kabelac [7] used

*
Corresponding author. Tel.: +65 6790 6538; fax: +65 6793 3318.
E-mail address: jing0002@ntu.edu.sg (G.-Y. Jin).

0196-8904/$ - see front matter  2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enconman.2005.10.020
2660 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672

Nomenclature

A area of flow (m2)

C specific heat (J/kg C)
F film coefficient (W/m2 C)
Q cooling load (W)
R heat resistance (C/W)
T temperature (C)
V volume (m3)
a, b parameters
c1–6 parameters
h enthalpy of moist air (J/kg)
m_ mass flow rate (kg/s)
p pressure (kPa)
q heat exchange quantity of element (W)
u velocity of mass flow (m/s)
s a certain time interval (s)
q density of flow (kg/m3)

Subscripts
a air
chw chilled water
i inlet or on-coil
o outlet or off-coil
tube tube
ma moist air
wb wet bulb
total sum of variables

the governing differential equations to study the dynamic response of a CCU by assuming that one fluid had
an infinite capacitance. Gartner and Daane [8] found transfer function relations for the different coil param-
eters. Bocanegra [9] and Khan [10] developed a model to analyze the performance of a counter flow cooling
and dehumidification coil. Quite recently, Yao [11] presented a rigorous analysis of the effect of perturbations
of the relevant parameters on the thermal quality of a CCU under different initial conditions. The obstacles for
these models being applied in industry applications are that the modeling methodologies require comprehen-
sive information on the structure of the CCU and the physical properties of the fluids, such as fin and tube
thickness, diameter and spacing, which often may not be available from the manufacturersÕ catalog. Another
drawback is that they require large computational effort and potentially cause numerical instabilities.
The lumped parameter approach results in fewer equations but frequently ignores some dynamics due to
the complex heat exchanger behavior. Specifically, the entire CCU dynamics is often modeled as a single sys-
tem, which ignores the important dynamics associated with the moving boundary between the wet surface
region and the dry surface region. Lebrun et al. [12] derived a first order differential equation on the basis
of an energy balance to represent the dynamics of a coil with lumped thermal mass. This approach has been
used by several authors for simulation purpose and occasionally for control analysis purposes [13]. Extending
this approach, Bi et al. [14] presented a lumped parameter empirical model using a robust identification
method and applied the modeling method to HVAC control loop auto-tuning [15].
Recently, Wang and Hibara [16] presented a distinct method (called the equivalent dry bulb temperature
method) to simplify the calculation in each region. This approach was improved by Wang et al.Õs [1] work.
They showed that the complex heat and mass transfer characteristics of the CCU could be further simplified,
and their proposed model still predicts the system correctly. In this paper, we will extend the previous work [1]
 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672 2661

Return Air

Return Fan

Louver Control
Damper Grille
Filter Cooling Coil

Mixed Air Supply Air

Supply Fan VAV Box

Control Valve (VSD)
CHWS Pump (VSD) VSD Pump

Fig. 1. Schematic diagram of typical AHU of HVAC pilot plant.

to develop a simple dynamic model to predict the performance of the CCU in time varying operation condi-
tions. Based on rigorously analyzing the heat transfer mechanism and the energy balance with some approx-
imations, we propose a dynamic model with no more than six parameters that represent the lumped geometric
terms. On a pilot plant of a HVAC system, experiments have been setup to validate the model. The proposed
model is expected to be applied for intelligent auto-tuning dynamic control.

2. Model development

2.1. Modeling based on heat transfer mechanism

A water CCU uses chilled water as the coolant inside the tubes. The chilled water cools and dehumidifies
the moist air that flows over the external surface of the tubes and fins, as shown in Fig. 2. To maintain a high
rate of heat transfer, the air and water normally follow a counter flow configuration; i.e., the low and high
temperature air meets the high and low temperature water correspondingly. During heat exchange with the
air outside the CCU tubes, the chilled water flows from the inlet to the outlet of the CCU forced by the chilled
water pump with inlet temperature Tchw,i and mass flow rate m_ chw , and the outlet temperature of the chilled
water rises to Tchw,o. The air flows from the inlet to the outlet of the CCU forced by the supply air fan with
the on coil, dry bulb temperature Ta,i, wet bulb temperature Ta,i,wb and mass flow rate m_ a , and likewise, the off
coil, dry bulb and wet bulb air temperatures descend to Ta,o andTa,o,wb, respectively.
During the process of heat transfer, the quantity of heat that moves from the hot air element through a
metal pipe wall (usually made of copper or aluminum alloy) into the chilled water element can be given
by
T a  T chw
q¼ . ð1Þ
Rtotal
2662 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672

Tchw, i m· chw
supply chilled water
coil fin

Ta, i Ta , o

on-coil air off-coil air

Ta , i , wb Ta , o , wb
m· a m· a

return chilled water

Tchw, o m· chw
l
Ta,i l Ta ,o
on-coil air off-coil air
Ta,i, wb Ta , o , wb
ua ua

return Tchw, o T chw, i supply

cooling water u chw uchw cooling water

dx
x
Fig. 2. Schematic diagram of typical finned tube CCU.

As shown in Fig. 3, the overall heat resistance Rtotal consists of three parts: the heat resistance of the chilled
water convection, the heat resistance of the metal wall conduction and the heat resistance of the air con-
vection. However, as the metal materials of the walls used in the CCU are good conductors of heat, their

1 2 3

Chilled Air Side

Water
.
ma
Side
.
mchw
Ta

Tchw

Water Film Air Film

Metal Wall
Fig. 3. Heat transfer mechanism of coil.
 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672 2663

resistance can be considered as a constant during operation. The conduction effect can be lumped into the
unknown coefficients and neglected in the following development. Therefore, the overall heat resistance is
Rtotal ¼ Rchw þ Ra . ð2Þ
The quantity of heat thus transferred depends on the nature of the surface and the fluid, the geometry of the
cooling coil and the velocity of the fluid over the surface as well as the temperature differences. The overall
heat resistance of the chilled water coil can then be written as [1]
1 1 bchw Achw m_ ‘chw þ ba Aa m_ ‘a
Rtotal ¼ þ ¼ ; ð3Þ
F chw Achw F a Aa bchw Achw m_ ‘chw ba Aa m_ ‘a
where Fchw, Fa, Achw and Aa are the film coefficient of the chilled water, the film coefficient of the air, the heat
transfer area of the chilled water side convection and the heat transfer area of the air side convection, respec-
tively, and bchw, ba are considered as constant parameters to be identified.
By combining Eqs. (1) and (3), we obtain:
bchw Achw m_ ‘chw ba Aa m_ ‘a
q¼ ðT a  T chw Þ. ð4Þ
bchw Achw m_ ‘chw þ ba Aa m_ ‘a

2.2. Modeling based on energy and mass conservation

For modeling convenience, we assume that:

1. Dry air and the water vapor in the air are treated as a non-reacting mixture of ideal gases.
2. The specific heat and the density of wet air are considered as constants in the processes of heat and mass
transfer.
3. Since the heat and mass exchange between the moist air and the external surface of the CCU tubes arrives
at a balance, the Lewis coefficient is considered as a constant.
4. Both the air and the water are well mixed in the cross section normal to its flow. Therefore, the only gra-
dients for each fluid exist in their respective flow directions [17].
5. The humidity of saturated wet air is approximately linear with its temperature when the temperature
changes are within a small range.

Let the total lengths of the tubes and the CCU depth be n and l, and then, the velocity of the air flow and
the equivalent velocity of the cooling water in the counter air flow direction are
l l m_ chw
u0chw ¼ uchw ¼ ; ð5Þ
n n qchw Atube
m_ a
ua ¼ ; ð6Þ
qa Aa
respectively.
Using the energy and mass conservation laws, the dynamic change of the chilled water temperature and the
air temperature in an infinitesimal volume can be expressed as
 
oT chw oT chw
qchw V chw C chw þ C chw u0chw ¼ q; ð7Þ
ot ox
 
oha oha
qa V a þ ua ¼ q; ð8Þ
ot ox
where Vchw and Va are constants representing the volume of the chilled water and the air element (Fig. 2) in the
mass and heat transfer processes in the X direction (Fig. 4).
There is a dry-wet boundary inside the CCU that divides it into sections of dry surface and wet surface,
as shown in Fig. 5 [18], where the solid curves represent the true temperature and enthalpy profiles.
2664 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672

Fig. 4. Temperature changes in the X direction.

T , ˚C
h, J/kg

Dry-wet boundary
ha ,i (t – τa)

Air e
nthal
Tchw,o (t) py
ha,o (t)
Water te
mperatu
re
Tchw,i (t – τchw )
Dry surface Wet surface

l, m
Fig. 5. Temperature and enthalpy profiles of cooling coil.

Approximating Tchw and ha by a linear equation between the two boundaries (dotted line), then any point
along the air movement direction can be expressed by
T chw;o  T chw;i
T chw ¼ x þ T chw;i ; ð9Þ
l
ha;o  ha;i
ha ¼ x þ ha;i . ð10Þ
l
Taking the first derivatives of Eqs. (9) and (10), respectively, we obtain:
oT chw T chw;o  T chw;i
¼ ; ð11Þ
ox l
oha ha;o  ha;i
¼ . ð12Þ
ox l
In practice, the relative humidity of the on coil air usually fluctuates within a small range, and therefore, the
enthalpy of the moist air can be approximated by the equivalent dry bulb (EDT) temperature method [16], i.e.,
 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672 2665

ha  C ma T a ; ð13Þ
ha;o  ha;i  C ma ðT a;o  T a;i Þ. ð14Þ

By combining Eqs. (7)–(14) we have:

 
dT chw 0 T chw;o  T chw;i
qchw V chw C chw þ C chw uchw ¼ q; ð15Þ
dt l
 
dT a T a;o  T a;i
qa V a C ma þ C ma ua ¼ q. ð16Þ
dt l

2.3. Dynamic modeling

Since q represents a heat exchange quantity of an infinitesimal volume of a CCU, it can not be measured or
calculated. However, we can replace q in Eqs. (15) and (16) by Eqs. (4) with the result:

dT chw c2 m_ ‘a
þ c1 m_ chw ðT chw;o  T chw;i Þ ¼  ‘ ðT a  T chw Þ; ð17Þ
dt m_ a
1 þ c3
m_ chw
dT a c5 m_ ‘a
þ c4 m_ a ðT a;o  T a;i Þ ¼   ‘ ðT a  T chw Þ; ð18Þ
dt m_ a
1 þ c3
m_ chw
l 1 ba A a b a Aa 1 ba A a
where c1 ¼ , c2 ¼ , c3 ¼ , c4 ¼ , c5 ¼ , and ‘ are constants that can
n qchw Atube qchw V chw C chw bchw Achw qa Aa qa V a C ma
be determined from the manufacturerÕs data or by real time experimental data.
Notice that it will take a certain amount of time for an individual air and water fluid element at the new
temperature to traverse the distance l and n from the inlet to the outlet of the CCU, therefore, any temper-
ature changes at the air and water inlet will not be registered instantaneously at the outlet. We denote schw
and sa as the time intervals that the air and chilled water fluid elements, respectively, take for traversing
n nq Atube
from the inlet to the outlet of the CCU, which are defined, respectively, by schw ¼ ¼ chw and
uchw m_ chw
l lqa Aa
sa ¼ ¼ .
ua m_ a
Finally, the engineering dynamic model using only boundary conditions can be written as

dT chw;o ðtÞ c2 m_ a ðtÞ‘

þ c1 m_ chw ðtÞ½T chw;o ðtÞ  T chw;i ðt  schw Þ ¼  ‘ ½T a;o ðtÞ  T chw;o ðtÞ; ð19Þ
dt m_ a ðtÞ
1 þ c3
m_ chw ðtÞ
‘
dT a;o ðtÞ c5 m_ a ðtÞ
þ c4 m_ a ðtÞ½T a;o ðtÞ  T a;i ðt  sa Þ ¼   ‘ ½T a;o ðtÞ  T chw;o ðtÞ. ð20Þ
dt m_ a ðtÞ
1 þ c3
m_ chw ðtÞ

3. Parameter identification

In practice, the number of sensors mounted on an actual HVAC system is limited. Several variables have to
be obtained by indirect methods if needed [1].

1. The chilled water flow rate can be obtained by measuring the pressure difference across the CCU from
Dp ¼ b0 þ b1 m_ chw þ b2 m_ 2chw ; ð21Þ
2666 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672

where b0, b1, b2 are unknown parameters to be determined through catalog data or on line testing, Dp is the
chilled water inlet–outlet pressure difference of a coil. Therefore, the following chilled water flow rate can be
derived from Eq. (21):
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b1 þ b21  4b2 ðb0  DpÞ
m_ chw ¼ . ð22Þ
2b2
2. By employing the method proposed in Ref. [1], the cooling load can be obtained through the chilled water
flow rate and the temperature difference between the CCU inlet and outlet, and it is given by
Q ¼ C chw m_ chw ðT chw;o  T chw;i Þ. ð23Þ
3. According to the energy balance, the air flow rate can then be estimated by
Q
m_ a ¼ . ð24Þ
ha;i  ha;o
The on coil and off coil air enthalpy can also be approximated by measuring the dry bulb temperature
with assumptions for the on coil air relative humidity of 60% and the off coil air relative humidity of
95%, respectively. The use of ha  CmaTa eliminates the need for separately calculating heat transfers for
the dry coil and the wet coil since both the dry bulb and wet bulb air temperatures are considered in cal-
culating m_ a in Eq. (25):
Q
m_ a ¼ C ma . ð25Þ
T a;i  T a;o
4. Using estimations or measurements of the variables Tchw,i, Ta,i, Tchw,o, Ta,o, m_ chw and m_ a , we can identify the
unknown parameters c1–c5 and ‘ in Eqs. (19) and (20). Otherwise, if ‘ = 0.8 (an engineering rule of thumb
value) is adopted, the model has only five parameters to be determined.

According to the different characteristics of the parameters in Eqs. (19) and (20), the parameter identifica-
tion procedure can be separated into two parts: (1) steady state identification and (2) dynamic identification.

3.1. Step 1. Steady state identification

dT chw;o ðtÞ dT a;o ðtÞ
At steady state, Ta,i and Tchw,i can be consider as constants. Consequently, ¼ 0 and ¼ 0.
dt dt
Then, we can rewrite Eqs. (19) and (20) as
c2
c1
m_ ‘a
m_ chw ðT chw;o  T chw;i Þ ¼  ‘ ðT a;o  T chw;o Þ; ð26Þ
m_ a
1 þ c3
m_ chw
c5 ‘
m_
c4 a
m_ a ðT a;o  T a;i Þ ¼   ‘ ðT a;o  T chw;o Þ. ð27Þ
m_ a
1 þ c3
m_ chw
By combining Eqs. (23) and (25), we have,
Q ¼ C chw m_ chw ðT chw;o  T chw;i Þ ¼ C ma m_ a ðT a;o  T a;i Þ. ð28Þ
Substituted by Eqs. (26) and (27), we have follow expression:
c2 c5
C chw m_ ‘a C ma m_ ‘a
c1 c4
Q¼  ‘ ðT a;o  T chw;o Þ ¼  ‘ ðT a;o  T chw;o Þ. ð29Þ
m_ a m_ a
1 þ c3 1 þ c3
m_ chw m_ chw
If we denote:
c2 c5
c6 ¼ C chw ¼ C ma . ð30Þ
c1 c4
 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672 2667

We obtain the steady state model expression:

c6 m_ ‘a
Q¼  ‘ ðT a;o  T chw;o Þ. ð31Þ
m_ a
1 þ c3
m_ chw
Employing the identification procedure used by Wang et al. [1], we can obtain the parameters c3, c6 and ‘.

3.2. Step 2. Dynamic identification

For dynamic identification, a simple, yet robust, identification method for a linear process, proposed by Bi
et al. [14], can be used to estimate the model.
c6 c6
From the estimated c6, we can express c2 ¼ c1 and c5 ¼ c4 . Then, we can rewrite Eqs. (19) and (20)
C chw C ma
as
dT chw;o ðtÞ
¼ c1 /chw ðtÞ; ð32Þ
dt
dT a;o ðtÞ
¼ c4 /chw ðtÞ; ð33Þ
dt
where
c6
m_ a ðtÞ‘
cchw
/chw ðtÞ ¼  ‘ ½T a;o ðtÞ  T chw;o ðtÞ  m_ chw ðtÞ½T chw;o ðtÞ  T chw;i ðt  schw Þ
m_ a ðtÞ
1 þ c3
m_ chw ðtÞ
and
c6
m_ a ðtÞ‘
ca
/a ðtÞ ¼  ‘ ½T a;o ðtÞ  T chw;o ðtÞ þ m_ a ðtÞ½T a;o ðtÞ  T a;i ðt  sa Þ.
m_ a ðtÞ
1 þ c3
m_ chw ðtÞ
To identify the parameters c1 and c4, the process has to be reset at a steady state. Starting from this initial
state at t = 0, step inputs of the mass flow rates of the chilled water and air flow are given for testing. Assum-
ing that the intercept of the tangent to the step response that has the largest slope with respect to the horizon-
tal axis gives H and integrating the differential terms in Eqs. (32) and (33) from t = 0 to t = s (s P H) results in
the following equation:
Z s Z s
dTðtÞ
dt ¼ h uðtÞ dt; ð34Þ
0 dt 0
     
c1 0 T chw;o ðtÞ /chw ðtÞ
where h ¼ , TðtÞ ¼ and uðtÞ ¼ .
0 c4 T a;o ðtÞ /a ðtÞ
With the sampling interval of Ts, and s = nTs, we can define H(kTs) = DT(kTs) = T(kTs)  T((k  1)Ts),
k = 1, 2, 3,    , n. When the perturbation of the input variable is very small, we iteratively have the following
equations:
Z Ts
Ts
HðT s Þ ¼ h uðtÞ dt  h½uðT s Þ þ uð0Þ;
0 2
Z 2T s
Ts
Hð2T s Þ ¼ h uðtÞ dt  h½uð2T s Þ þ uðT s Þ;
Ts 2 ð35Þ
..
.
Z s
Ts
HðnT s Þ ¼ h uðtÞ dt  h½uðnT s Þ þ uððn  1ÞT s Þ.
ðn1ÞT s 2
2668 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672

By collecting Eq. (34) for all the sampled T(t) after t = 0, a system of linear equations is obtained as
Wh ¼ C þ D; ð36Þ

where D is noise or measurement error and

2 T
3 2 3 2 3
½uðT s Þ þ uð0Þ HðT s Þ dðT s Þ
6 7 6 7 6 7
6 T 7 6 Hð2T s Þ 7 6 dð2T s Þ 7
6
Ts 6 ½uð2T s Þ þ uðT s Þ 7 6 7 6 7
7
W¼ 6 7; C¼6
6 .
7;
7 D¼6
6 . 7.
7
26 .. 7 6 .. 7 6 .. 7
6 . 7 4 5 4 5
4 5
½uðnT s Þ þ uððn  1ÞT s ÞT HðnT s Þ dðnT s Þ

The best estimate of ^

h can be found by using the standard least squares method by Eq. (37), and the best
estimates of c1 and c4 can then be obtain from ^
h.

h ¼ ðWT WÞ1 WT C.
^ ð37Þ

4. Experiment

The experimental testing is conducted on a pilot centralized HVAC system as shown in Picture 1. The sys-
tem has three chillers, three zones with three AHUs, three cooling towers and flexible partitions, up to twelve
rooms. All motors (fans, pumps and compressors) are equipped with variable speed devices (VSDs). The sys-
tem is made very flexible so that these three units can be configured to form different schemes.
The CCU for the system has face dimensions of 0.25 m · 0.25 m (10 in. · 10 in.), four tube rows and 328
aluminum fins per meter (100 fins/foot), as shown in Picture 2. The measurement signals for the experiment
are the water and air flow rates, on coil and off coil air dry bulb/web bulb temperatures and the CCU inlet and
outlet water temperatures. The experiment is conducted under the following conditions:

1. Fresh air is used to keep a relatively constant inlet air temperature.

2. The chilled water supply temperature is changed slowly; the cooling load variation is achieved through the
air and water flow rates.

Picture 1. Pilot plant of centralized HVAC systems.

 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672 2669

Picture 2. An AHU and a cooling coil of the pilot plant.

In order to study the dynamic relationships between the parameters and the CCUÕs heat exchange, real time
varying conditions for testing are given (Fig. 6).
Through Step 1 (steady state identification) and Step 2 (dynamic identification), we can obtain the param-
eters c1–c5 and ‘ by experimental fitting, which are listed in Table 1.
The results calculated by the six parameter model and the simplified five parameter model are compared
with the experimental measured data in Figs. 7 and 8, respectively.

Fig. 6. Time varying temperature and mass flow rate of inlet air and chilled water temperature.
2670 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672

Table 1
Estimated parameters of the six parameter and five parameter models
Model Load range (kW) ‘ c1 c2 c3 c4 c5
Six-parameter 0.60–1.20 0.6078 6.7142 0.7412 0.7021 8.9936 3.9722
Five-parameter 0.60–1.20 0.8 6.9431 1.1993 0.8021 9.3274 6.4431

Fig. 7. Comparison of six parameter model calculations with the measured data.

Fig. 8. Comparison of five parameter model calculations with the measured data.
 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672 2671

Table 2
Performances of six parameter model and five parameter model
Model RMS (off coil air temperature) RMS (chilled water temperature)
Six parameter 0.2845 C 0.1719 C
Five parameter 0.2362 C 0.1321 C

Finally, to assess the effectiveness of the model, the calculated results of the model are evaluated through
the root mean square error (RMS) defined by
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
PN 2
i¼1 ðT calculation;i  T experiment;i Þ
RMS ¼ ;
N
where N is the number of fitted points, Tcalculation,i is the temperature of point i predicted by the model and
Texperiment,i is the temperature of point i acquired from the experimental data. The performances of the two
models are given in Table 2.
From the experiment results, we may conclude:

1. The proposed dynamic model accurately reflects the details of the transient performance of the CCU. With
the constant parameters, the models work appropriately in a relatively wide operation range: both the dif-
ference of the outlet chilled water temperature and the off coil air temperature are about 3 C, which is the
normal operating range for a CCU.
2. The linear approximation of the dry and wet CCU does not generate significant error (RMS index) but dra-
matically simplifies the calculation procedure compared with the complex numerical modeling methodolo-
gies. Of course, this model may work even better on the totally dry or totally wet regime.
3. The responses of the off coil air and outlet chilled water temperatures naturally have time delays from the
changes of the on coil air and inlet chilled water temperatures. The testing results do not show this obvi-
ously because the total lengths of the tubes and the CCU depth, being n and l, are relatively small versus the
velocities of the cooling water and air flow, being uchw and ua.
4. The six parameter model results in a little better performance over the entire operation range but at the
price of more complicated calculations compared with those of the five parameter model. Therefore, it
may be worth sacrificing a little accuracy for simplifying the parameter identification procedure in real engi-
neering applications.

5. Conclusion

An accurate, yet simple, dynamic engineering model for a CCU was derived based on heat transfer prin-
ciples and mass and energy balances. The modeling method can be used in the real HVAC plant to handle
real time information. Since the parameters of this model can be obtained from the manufacturerÕs catalog
or experimental data without going into the detailed physical properties of the CCU, it can be easily applied
to real time HVAC system controls. Experimental results were provided to show that both the six parameter
model and the simplified five parameter model can achieve good and accurate estimation over a wide operat-
ing range, and the calculation for the on line parameters identification procedure can be much more simple for
the simplified five parameter model. Further work in this area will concentrate on imbedding this model into
an AHU and the supply air system and testing them on the pilot HVAC system.

References

[1] Wang YW, Cai WJ, Soh YC, et al. A simplified modeling of cooling coils for control and optimization of HVAC systems. Energy
Convers Manage 2004;45:2915–30.
[2] Skimin GK. TechnicianÕs guide to HVAC systems. New York: McGraw-Hill, Inc.; 1995.
[3] Kreith F, West RE. CRC handbook on energy efficiency. Boca Raton: CRC Press; 1997.
2672 G.-Y. Jin et al. / Energy Conversion and Management 47 (2006) 2659–2672

[4] Lu L, Cai W, Soh YC, Xie L. Global optimization for overall HVAC systems—Part II. Problem solution and simulations. Energy
Convers Manage 2005;46:1015–28.
[5] Lu L, Cai W, Soh YC, Xie L, Li S. HVAC system optimization-in-building section. Energy Build 2005;37(1):11–22.
[6] Myers GE, Mitchell JW, Lindeman CF. The transient response of heat exchanger having an infinite capacitance rate fluid. ASME J
Heat Transfer 1970;92(2):269–75.
[7] Kabelac S. The transient response relations for a serpentine cross-flow heat exchanger with water velocity disturbance. ASHRAE
Trans 1989;32(6):1183–9.
[8] Gartner JR, Daane LE. Dynamic response relations for a serpentine cross-flow heat exchanger with water velocity disturbance.
ASHRAE Trans, Part 1 1969;75:53–68.
[9] Bocanegra LM, Khan AY. Parametric analysis of heat and mass transfer performance of a counter flow cooling and dehumidification
coil. In: 92-WA/HT-1, ASME winter annual meeting, November 8–13, 1992.
[10] Khan AY. Heat and mass transfer performance analysis of cooling coils at part-load operating conditions. ASHRAE Trans
1994;100(2):54–62.
[11] Yao Y. Thermal analysis of cooling coils based on a dynamic model. Appl Therm Eng 2004;24:1037–50.
[12] Lebrun J, Ding X, Eppe JP, Wasacz T. Cooling coil models to be used in transient and/or wet regimes—Theoretical analysis and
experimental validation. In: Proceedings of system simulation in buildings Õ90. Liege, Belgium, 1990.
[13] Wang S. Dynamic simulation of a building central chilling system and evaluation of EMCS on-line control strategies. Build Environ
1998;33:1–20.
[14] Bi Q, Cai W-J, et al. Robust identification of first-order plus dead-time model from step response. Control Eng Pract 1999;7(1):71–7.
[15] Bi Q, Cai W-J, Wang Q-G, et al. Advanced controller auto-tuning and its application in HVAC systems. Control Eng Pract
2000;8(6):633–44.
[16] Wang J, Hihara E. Prediction of air coil performance under partially wet and totally wet cooling conditions using equivalent dry-bulb
temperature method. Int J Refrig 2003;26:293–301.
[17] Stevens RA, Fernandez J, Woolf J. Mean temperature difference in one-, two- and three-pass cross-flow heat exchangers. Trans
ASME 1957;79:287–97.
[18] Wang SK. Handbook of air conditioning and refrigeration. 2nd ed. New York: McGraw-Hill Inc.; 2000. p. 15.45–65.