Sunteți pe pagina 1din 21

HVAC&R Research

ISSN: 1078-9669 (Print) 1938-5587 (Online) Journal homepage: http://www.tandfonline.com/loi/uhvc20

A Simplified Dynamic Model for Chilled-Water


Cooling and Dehumidifying Coils—Part 1:
Development (RP-1194)

Xiaotang Zhou & James E. Braun

To cite this article: Xiaotang Zhou & James E. Braun (2007) A Simplified Dynamic Model for
Chilled-Water Cooling and Dehumidifying Coils—Part 1: Development (RP-1194), HVAC&R
Research, 13:5, 785-804

To link to this article: https://doi.org/10.1080/10789669.2007.10390986

Published online: 25 Feb 2011.

Submit your article to this journal

Article views: 211

Citing articles: 21 View citing articles

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=uhvc21
©2007, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Published in HVAC&R Research, Vol. 13,
No. 5, September 2007. For personal use only. Additional reproduction, distribution, or transmission in either print or digital form is not permitted
without ASHRAE’s prior written permission.
VOLUME 13, NUMBER 5 HVAC&R RESEARCH SEPTEMBER 2007

A Simplified Dynamic Model for


Chilled-Water Cooling and Dehumidifying
Coils—Part 1: Development (RP-1194)
Xiaotang Zhou, PhD James E. Braun, PhD
Associate Member ASHRAE Fellow ASHRAE

Received September 13, 2006; accepted March 13, 2007

This paper is based on findings resulting from ASHRAE Research Project RP-1194.

Dynamic coil models are useful in the development and evaluation of feedback controllers and
fault detection and diagnostics algorithms. Existing models in the literature have not been vali-
dated fully using experimental data and have not been evaluated in terms of trade-offs between
model accuracy and computational requirements. This paper develops a computationally fast,
distributed model for the transient behavior of cooling and dehumidifying coils. The model uti-
lizes several steady-state performance indices that simplify dynamic modeling, including fin effi-
ciency, air-side effectiveness, and water-side effectiveness. Predictions from the simplified
model were compared with a detailed reference model and with other simplified models from
the literature. Comparisons with detailed experimental measurements are given in a companion
paper (Zhou and Braun 2007). The model provides very accurate predictions of coil transient
performance and executes more than 1000 times faster than real-time on a 1.2 GHz personal
computer. As a result, it is very practical for implementation within system simulation programs
that execute over long time periods, such as a year.

INTRODUCTION
Dynamic models are often used in the development of advanced feedback controllers, control
strategies, and diagnostic methods. In the initial design and validation of these algorithms, the
use of computer models is much more cost-effective than running real-time experiments. Bour-
douxhe et al. (1996) presented a very detailed review of transient models in the HVAC field.
They found a lot of literature on transient modeling of heat exchangers but relatively little work
on transient modeling of cooling and dehumidifying coils.
In some situations, quasi-static models may be appropriate for transient simulations if the
dynamic variations of the inputs are slower than the response of the cooling coil. The model
developed by Braun et al. (1989) is an example of a simplified quasi-static model, which utilizes
effectiveness relationships for heat transfer only and heat and mass transfer. This lumped model
was validated through comparisons with detailed numerical solutions and manufacturer’s catalog
data and works very well in predicting air and water outlet states under steady-state conditions.
For heat exchangers without dehumidification, Gartner and Harrison (1963) developed an
analytical model for the transient behavior of cross-flow heat exchangers. They derived three
partial differential equations, which describe energy storage in the primary fluid (air), heat
exchanger material, and secondary fluid (water) with the assumption of steady flow for both

Xiaotang Zhou is an Engineering Leadership Program associate for the Carrier Corporation, Syracuse, NY. James E.
Braun is a professor of mechanical engineering for Ray W. Herrick Laboratories, Purdue University, West Lafayette, IN.

785
786 HVAC&R RESEARCH

flow streams. The solutions were presented in terms of transfer functions that relate outlet air
and water temperatures to variations of inlet air and water temperatures under the condition of
constant fluid flow rates. The validity of the model was demonstrated through comparisons with
transient experimental data. This model has been the basis for many other transient heat
exchanger models, including models described by Gartner and Harrison (1965), Gartner and
Daane (1969), Tamm (1969), Tamm and Green (1973), Bhargava et al. (1975), and Jawadi
(1988). Gartner and Daane (1969) developed a mathematical model for serpentine cross-flow
heat exchangers that considers changes in fluid velocities, and Jawadi (1988) simplified this
model by neglecting transients associated with air.
Dehumidification adds a significant complication, and only a small number of papers appear
in the literature for dynamic modeling of cooling and dehumidifying coils. Based on the level of
detail, the models can be categorized into reference, lumped, and filter models as proposed by
Bourdouxhe et al. (1996). A reference model is based on a fairly detailed understanding of phys-
ical phenomena and usually involves partial differential equations. A lumped model is also
based on a physical representation but with certain simplifications so that the mathematical
expressions are reduced to a smaller number of ordinary differential and algebraic equations. A
filter model is based on the use of a time constant added to the outputs of a steady-state model.
McCullagh et al. (1969) presented a reference model that considers the dynamic behavior of
chilled-water cooling and dehumidifying coils. At any position within each row of tubes, tem-
peratures of both the tube and fin materials are assumed to be uniform in the direction normal to
the water flow. Between rows of tubes, the air is considered to be completely mixed. Energy bal-
ances applied to the air, tube, fin, and water lead to partial differential equations that are solved
using a finite-difference numerical solution. Steady-state predictions of the model were com-
pared to experimental results under wet conditions and showed reasonably good agreement for
both sensible and latent heat transfer except for two-row coils where the assumption of complete
mixing of air between rows was poor. The transient response of the model was not validated
using experiments.
Clark (1985) developed a filter model for cooling and dehumidifying coils configured in a
counter cross-flow arrangement. The model is based on the steady-state model presented by
Elmahdy and Mitalas (1977). In dry conditions, the heat flow rate between moist air and water is
calculated on the basis of logarithmic mean temperature difference; in wet cases, logarithmic
mean enthalpy difference is used to calculate the heat flow rate. The coil dynamics are modeled
very simply by adding a single time constant to the steady-state air outlet temperature and
humidity and to water outlet temperature. This time constant is a function of heat capacitance
associated with coil material and overall heat transfer coefficient. Use of the model to represent
the dynamics of coils with fewer than four rows was not recommended. The dynamics of the
coil model were investigated experimentally only for dry conditions with changes in water flow
rate and water inlet temperature. The agreement between model and experiment was reasonably
good for the conditions considered.
Ding et al. (1990) developed a lumped model for cooling and dehumidifying coils that have
counter cross-flow configurations. A single capacitance is used to represent the total energy
storage of the coil metal material and water, and the energy transfer rates are determined using
effectiveness models similar to those presented by Braun et al. (1989). Air dry-bulb tempera-
tures are used as driving potentials for dry cases, and air wet-bulb temperatures are used for wet
conditions. The model was only compared with experimental data for dry coils. The agreement
between model prediction and experimental measurement was good for air outlet temperature
but poor for water outlet temperature.
Chow (1997) developed a lumped model that arises from energy balances applied to finite
control volumes within a chilled-water cooling and dehumidifying coil. The accuracy of the
VOLUME 13, NUMBER 5, SEPTEMBER 2007 787

model was demonstrated through comparisons with three other coil models developed by
Clark (1985), Holmes (1988), and Stoecker (1975) for both dry and wet coils. The results for
air outlet temperatures from the different models were close, while the results for water outlet
temperatures had large variations. No experimental validation was performed for this model.
Furthermore, the models used as a basis for comparison have not been validated fully using
experiments under dry and wet conditions with transient operation.
None of the previous studies explored trade-offs between model accuracy and computational
requirements to identify an appropriate model to be used within transient simulations for cooling
and dehumidifying coils. Furthermore, experimental validation for transient coil performance
has not been performed under both dry and wet conditions. There is a need for an accurate
dynamic coil model that executes quickly so that it can be used within system simulations for
long time horizons, such as a cooling season or year.

Table 1. Characteristics of Test Coils

Coil Physical Parameter Eight-Row Coil Four-Row Coil

Fin geometry Wavy Louvered


Coil depth, m 0.264 0.132
Number of fins per inch (per 0.0254 m) 8 12
Coil face width, m 0.6096 0.6096
Coil face height, m 0.6096 0.6096
Tube material Copper Copper
Tube outer diameter, m 0.0127 0.0127
Tube thickness, m 0.0004 0.0004
Tube longitudinal pitch, m 0.033 0.033
Tube transverse pitch, m 0.0381 0.0381
Fin material Aluminum Aluminum
Fin thickness, m 0.0002 0.0002

Figure 1. Schematic illustrations of coil flow circuiting for four-row and eight-row test coils.
788 HVAC&R RESEARCH

The current paper develops a simplified dynamic coil model that takes advantage of
steady-state performance indices to reduce the number of control volumes needed to accurately
characterize a cooling coil. The model is compared with a more detailed reference model and
with simplified models from the literature. Comparisons between reference and simplified
model results are presented for two coils representative of medium to larger commercial applica-
tions. Table 1 provides parameters that characterize the two coils and Figure 1 depicts their flow
circuiting. Detailed experimental validation of the simplified model for both wet and dry condi-
tions was performed for these coils in a companion paper (Zhou and Braun 2007).

REFERENCE COOLING COIL MODEL


Figure 2 depicts a counter cross-flow cooling coil that is considered in this study. Chilled
water enters at the back row of the coil and flows in a serpentine arrangement. Air flows over
finned tubes and is cooled and possibly dehumidified due to contact with cold surfaces. A
detailed reference model is developed in this section by applying energy balances to the air,
water, tube, and fin material within the cooling coil. The following assumptions are employed in
the model development:

• Water is incompressible
• Ideal gas mixture for air and water vapor
• Constant densities and specific heats for air, water, fin, and tube material
• Negligible conduction in air and water flow directions for both fluids
• Negligible conduction for fin and tube material in the water flow direction
• Quasi-steady water and airflows
• Uniform air velocity across the coil cross section
• Negligible energy storage within air
• Negligible effect of water condensate retained on fin and tube outer surfaces when dehumidi-
fication occurs
• Lewis number of unity for heat and mass transfer
• The temperature profile within fins in the fin-height direction follows the steady-state profile
allowing the use of heat transfer and combined heat and mass transfer fin efficiencies

Figure 2. Schematic of a counter cross-flow cooling coil.


VOLUME 13, NUMBER 5, SEPTEMBER 2007 789

The first eight assumptions are utilized by most of the models presented in the literature and
have been validated indirectly through comparisons of model predictions with measurements for
dry coils. The quasi-steady assumption for water and airflow implies that there are no
time-derivative terms involving these flows. However, the flows are boundary conditions for the
models that have transient changes that may be imposed during the numerical solution of the
model. Zhou and Braun (2005) evaluated the last three assumptions and found they have negli-
gible effect on accuracy for typical cooling and dehumidifying coils. Zhou and Braun (2005)
and Zhou et al. (2007) also presented an improved method for characterizing fin efficiency for
heat transfer when condensation occurs that is incorporated into the modeling approaches
described in the current paper.

Dry Coil Section


At any point within the cooling coil, it is necessary to consider whether moisture condenses.
For a dry section, energy balances applied to water and coil material (tube and fin) result in the
following differential equations when utilizing the set of assumptions listed above:

∂T w ∂T w 1
C ′w ⋅ ---------- + C· w ⋅ ---------- + ------- ⋅ ( T w – T c ) = 0 (1)
∂t ∂x R ′
w

dT c 1 1
C ′c ⋅ --------- + ------ ⋅ ( T c – T a, in ) + ------- ⋅ ( T c – T w ) = 0 (2)
dt R ′ R′
a w

where T w and T c are the local “bulk” temperatures of water and coil material, T a, in is the local
air inlet temperature, C ′w and C ′c are the heat capacitances of water and coil material per unit
length of tube (in the water flow [x] direction), C· w is the heat capacitance rate associated with
the water flow stream, and 1 ⁄ R ′w and 1 ⁄ R ′a are thermal conductances for heat transfer between
water and coil material and coil material and air, respectively, per unit length of tube. Capaci-
tance is the product of the specific heat and mass. The coil capacitance is the sum of the capaci-
tances for the tube and the fins. The local distribution of temperatures within the fin material is
handled through use of an overall fin efficiency so that local coil temperature, T c , in Equations
1 and 2 is the tube outer surface temperature at the base of the fins. Zhou (2005) evaluated this
assumption through comparison with numerical solutions to two-dimensional fin models and
found excellent agreement.
The water-side thermal resistance is

1
R ′w = ---------------- , (3)
hw ⋅ Pt

where h w is the convection heat transfer coefficient for water flowing through the tube and P t is
the tube internal perimeter.
Since energy storage is neglected within the air, the air behaves quasi-statically and responds
instantaneously to changes in coil surface and inlet air temperature. A quasi-static model that is
consistent with the use of air inlet temperature in Equation 2 is the effectiveness-NTU method.
With this approach, the air-side thermal resistance is

1
R ′a = ---------------- , (4)
ε a ⋅ C· ′a
790 HVAC&R RESEARCH

where εa is the air-side heat transfer effectiveness, and C· ′a is the heat capacitance rate of the air-
flow stream per unit length of tube. The effectiveness accounts for spatial variation in air tem-
perature as the air flows across the local finned-tube section of the coil. With this in mind, the
local air-side heat transfer per unit length of tube is assumed to occur between a spatially vary-
ing air temperature and a uniform tube surface temperature such that the air-side heat transfer
effectiveness is determined with the following:

– NTU a
εa = 1 – e (5)

where NTU a is the air-side NTU , calculated as

η a ⋅ h a ⋅ A ′a
NTU a = -------------------------- (6)
C· ′a

and where η a is the overall fin efficiency for heat transfer, h a is the convection coefficient for
air-side heat transfer, and A ′a is the air-side surface area per unit length of tube.
The local outlet air temperature is determined with the following:

T a, out = T a, in + ε a ⋅ ( T c – T a, in ) (7)

which is the local inlet air temperature for the position in the next row just downstream in the
direction of airflow. The outlet air enthalpy is determined using a psychrometric function with
the outlet air temperature and the inlet humidity ratio. The outlet air temperature and enthalpy
for the coil are averages across the entire exit face of the last row.

Wet Coil Section


If moisture is condensing locally on the surface of the coil and the Lewis number is unity,
then the driving potential for combined heat and mass transfer is the difference between air
enthalpy and saturation air enthalpy at the coil surface temperature. In this case, Zhou and Braun
(2005) have shown that Equation 2 can be replaced by the following:

· dT c 1 1
C 'c ⋅ --------- + ------- ⋅ ( h s, c – h a, in ) + ------′- ⋅ ( T c – T w ) = 0 (8)
dt R *′ Rw
a

where h a, in is the local air inlet enthalpy, h s, c is the saturation air enthalpy at T c , and 1 ⁄ R*'
a is
the conductance for heat and mass transfer between coil material and air per unit length of tube.
The air behaves quasi-statically and responds instantaneously to changes in coil surface and inlet
air conditions. For the wetted surface case, a heat and mass transfer effectiveness model is
employed that is consistent with the use of air inlet enthalpy in Equation 8 such that

1
R*'a = --------------------
· ,
- (9)
ε a ⋅ M a'
*

·
where ε *a is the effectiveness for combined heat and mass transfer and M a' is the air mass flow
rate per unit length of tube. The heat and mass transfer effectiveness accounts for spatial varia-
tion in air enthalpy as the air flows across the local finned-tube section of the coil. The local dis-
tribution of the saturated air enthalpies at the interface between the air and water on the fins is
handled through the use of an overall heat and mass transfer fin efficiency. With this in mind,
the air-side heat and mass transfer effectiveness is determined with the following:
VOLUME 13, NUMBER 5, SEPTEMBER 2007 791

– NTU a*
ε *a = 1 – e (10)

where

η a* ⋅ h a ⋅ A'
NTU a = ---------------------------a- (11)
C· ′a

and where η*a is the overall fin efficiency for combined heat and mass transfer (different from
η a , which is for heat transfer only). In this case, h a is the convection coefficient for air-side
heat transfer under dehumidification.
The local outlet air enthalpy is

h a, out = h a, in + ε *a ⋅ ( h s, c – h a, in ) , (12)

which is the local inlet air enthalpy for the position in the next row that is just downstream in the
direction of airflow. The outlet air enthalpy for the coil is the average enthalpy across the entire
exit face of the last row.
If moisture is condensing, Equation 7 is still valid for determining the outlet air temperature.
However, the heat transfer fin efficiency is different than the dry fin efficiency because the fin
temperature distribution is influenced by the heat and mass transfer process as described by
Zhou and Braun (2005) and Zhou et al. (2007).

Heat Transfer Coefficients and Fin Efficiencies


Four-row and eight-row cooling coils that were considered are described in Table 1, Figure 1,
and by Zhou and Braun (2005, 2007). The interior of the tubes was smooth, and water-side con-
vection coefficients were determined using available correlations based on the water-side Rey-
nolds number, Re w. For laminar flow ( Re w ≤ 2300), the Sieder-Tate correlation was employed;
for fully turbulent flow ( Re w ≥ 10000), the Dittus-Boelter equation was used; and for transition
flow ( 2300 ≤ Re w ≤ 10000 ), the Gnielinski correlation was applied. In order to have smooth tran-
sitions between the correlations, linear interpolation was employed in transition regions near the
boundaries of the correlation ranges.
Air-side convection coefficients depend on several factors, including fin and tube geometries,
tube circuiting arrangement, and air-side dry or wet surface conditions. Although there are many
correlations available in the literature, none of them were suitable for the coils considered in this
study. For this study, specific correlations were determined for each coil using experimental data
obtained under the steady-state operating conditions as described by Zhou and Braun (2005).
Figure 3 shows experimental data for air-side convection coefficients under dry and wet con-
ditions for the eight-row and four-row coils as a function of air velocity. These results indicate
that there is an approximate 10% to 15% improvement in heat transfer coefficient when mois-
ture is condensing as compared with a dry coil. Also, it is apparent that the four-row coil has a
significantly higher heat transfer coefficient than the eight-row coil. The four-row coil has a
higher fin density and also utilizes louvered fins as compared with wavy fins for the eight-row
coil. The uncertainties in the heat transfer coefficients were less than 15% and are illustrated for
the eight-row coil in Figure 3. This figure also shows that the air-side heat transfer coefficients
are nearly linear with air velocity for the range considered. Linear correlations were employed
for the results of this study.
792 HVAC&R RESEARCH

Figure 3. Dry and wet air-side convection coefficients for eight-row and four-row coils.

The overall air-side heat transfer fin efficiency for dry surfaces used in Equation 6 is deter-
mined using the following:

η a = 1 – AR f ( 1 – η f ) (13)

where AR f is the ratio of fin area to total surface area and ηf is the individual fin efficiency. For
the coils considered in this study, individual fin efficiencies were determined using the method of
Schmidt (1949) for hexagonal fin arrays with staggered tubes as described by ASHRAE (2005).
For combined heat and mass transfer fin efficiencies used in Equation 11, fin efficiency relation-
ships for heat transfer only are utilized but with modified properties as described by Braun et al.
(1989). When the fins are wet, it is still necessary to utilize a heat transfer fin efficiency to deter-
mine the exit area temperature using Equation 7. In this case, the individual heat transfer fin effi-
ciency used in Equation 13 is modified as described by Zhou and Braun (2005) and Zhou et al.
(2007) to account for the combined effects of heat and mass transfer on the fin temperature distribu-
tion. Existing models utilize relationships developed for dry fins (i.e., no condensation) in deter-
mining convective heat transfer under wet conditions. However, the fin temperature distribution is
different for wet and dry fins and therefore the use of a dry fin efficiency relationship for convective
heat transfer is not strictly correct under wet conditions. Zhou and Braun (2005) and Zhou et al.
(2007) developed a correction factor for existing fin efficiency relationships that allows a better
estimate of convective heat transfer fin efficiencies under wet conditions.

Numerical Solution
For this study, finite differencing in space and time was applied to the differential equations,
resulting in the following discretized equations for each control volume:

0
( Tw – Tw ) · ( T w – T w, in ) 1
C ′w ⋅ ------------------------ + C w ⋅ ------------------------------ + ------- ⋅ ( T w – T c ) = 0 (14)
Δt Δx R′
w

0
( Tc – Tc ) 1 1
C ′c ⋅ ---------------------- + ------- ⋅ ( T c – T a, in ) + ------- ⋅ ( T c – T w ) = 0 (15)
Δt R'a R′ w
VOLUME 13, NUMBER 5, SEPTEMBER 2007 793

0
( Tc – Tc ) 1 1
C ′c ⋅ ---------------------- + ---------- ⋅ ( h s, c – h a, in ) + ------- ⋅ ( T c – T w ) = 0 (16)
Δt R*' a R′
w

where Δx is the spacial step along the water flow direction, Δt is the time step, and the super-
script 0 represents the value at a previous time.
For any given control volume and time step within a transient simulation, a dry analysis is
performed first because the computational requirements are less than those for a wet analysis.
A dry analysis involves the solution of Equations 14, 15, and 7 for T w , T c , and T a, out , respec-
tively, given initial conditions, inlet conditions, and other required parameters. The equations
are linear in these three state variables, facilitating an analytic solution. The outlet enthalpy is
determined using a psychrometric function with the outlet air temperature and the inlet
humidity ratio.
In order to evaluate whether a wet analysis should be performed, the coil surface temperature
from the dry analysis is compared with the dewpoint of the inlet air, which is determined from the
control volume inlet air enthalpy and temperature using a psychrometric function. If the coil sur-
face temperature is greater than the dewpoint temperature, then the dry analysis is appropriate.
Otherwise, it is necessary to solve Equations 14, 16, and 12 for T w , T c , and h a, out , respectively,
with the given initial conditions, inlet conditions, and parameters. This requires an iterative solu-
tion because of the nonlinear dependence of h s, c on T c . At each iteration, h s, c is evaluated using
a psychrometric function at T c and a relative humidity of 100%. The outlet air temperature is
then determined using Equation 7.
If the coil incorporates parallel water flow through all tubes in each row and the airflow inlet
conditions are uniform, then it is only necessary to consider a single tube for each row of the coil
since all tubes see the same boundary conditions. At each time interval, the overall coil analysis is
iterative and begins at the air inlet to the coil. Solutions for the state variables are determined for
all control volumes within the first row using values of water inlet temperature from the previous
time step. The outlet air conditions for each control volume from the first row analysis are used as
inlet conditions for control volumes immediately downstream in the next row. After all rows have
been evaluated, the process is repeated using the most recent water inlet condition for each con-
trol volume until changes in all state variables are less than a convergence tolerance.
It is necessary to specify initial conditions for all state variables at the beginning of a transient
simulation. In this study, initial conditions were assumed to be the states that result from
steady-state operation at the specified initial inlet air and water conditions. A steady-state ver-
sion of the reference cooling coil model described by Zhou and Braun (2005) was utilized for
this purpose.
A numerical study was performed for the two test coils to determine the number of control
volumes and time step to utilize for accurate results with the reference model. Models with ten
control volumes per each row of tubes and a one-second time step were found to provide accu-
rate results and were used throughout this study for the reference model.

Simplified Cooling Coil Model


The reference cooling coil model is probably too computationally slow to be useful within
long-term system simulations. The computational speed could be reduced considerably if the num-
ber of required control volumes could be reduced. Furthermore, the number of control volumes
can be reduced significantly with little effect on accuracy by utilizing an appropriate temperature
profile within each control volume instead of a uniform temperature distribution. The profile
should converge to the steady-state profile that would result for constant boundary conditions.
794 HVAC&R RESEARCH

The simplified model involves applying an overall energy balance to each row within the
cooling coil and utilizing an effectiveness model to estimate the heat transfer rate between the
coil surface and water. In addition, it is assumed that the time derivative for water temperature is
the same at each point within the water flowing through the row. This assumption also implies
that the time derivative for the spatially averaged water temperature is equal to the time deriva-
tive for the outlet temperature of the row or

dT w dT w, out
---------- = ------------------
- , (17)
dt dt

where T w and T w, out are the water mean and outlet temperature for the row.
The use of water-side and air-side effectiveness ensures that the model approaches steady-state
or quasi-steady behavior when subjected to static or slowly changing boundary conditions and
allows the use of relatively few state variables as compared with the reference model.

Dry Row
For each discrete time interval, each row is treated as either all wet or all dry. For a dry row,
the following differential equations result from an overall energy balance using the condition of
Equation 17:

dT w, out
C w ⋅ ------------------- + C· w ⋅ ( T w, out – T w, in ) + ------- ⋅ ( T w, in – T c ) = 0
1
(18)
dt Rw

dT c 1 1
C c ⋅ --------- + ------ ⋅ ( T c – T a, in ) + ------- ⋅ ( T c – T w, in ) = 0 (19)
dt R a Rw

where for each row, T w, in is the water inlet temperature, T c is the mean temperature for coil
material, T a, in is air inlet temperature, C w and C c are total heat capacitances of water and coil
material, C· w is the total heat capacitance rate associated with the water flow stream, and R w and
R a are total thermal resistances for heat transfer between water and coil material and coil mate-
rial and air, respectively.
The water-side heat resistance for each row is determined with the following:

1
R w = ----------------- (20)
ε w ⋅ C· w

where ε w is the water-side heat transfer effectiveness for the row expressed as

– NTU w
εw = 1 – e (21)

and

hw ⋅ Aw
NTU w = ----------------- (22)
C· w

where A w is the total internal surface area of the tubes within the cooling coil row. In this study,
water-side heat transfer coefficients were determined as described for the reference model.
VOLUME 13, NUMBER 5, SEPTEMBER 2007 795

The air-side thermal resistance for each row is

1
R a = ---------------
- , (23)
ε a ⋅ C· a

where C· a is the total air heat capacitance rate and where the air-side heat transfer effectiveness,
ε a , is determined with Equation 5. In this study, air-side heat transfer coefficients and fin effi-
ciencies were determined as outlined for the reference model.
The row outlet air temperature is determined with Equation 7, which is the inlet air tempera-
ture for the next downstream row. The outlet enthalpy is determined using a psychrometric func-
tion with the outlet air temperature and the inlet humidity ratio.

Wet Row
For a wet row, Equation 19 is replaced with the following:

dT c 1 1
C c ⋅ --------- + --------- ⋅ ( h s, c – h a, in ) + ------- ⋅ ( T c – T w, in ) = 0 (24)
dt R*a Rw

where for each row, h a, in is the air inlet enthalpy, h s, c is the saturation air enthalpy at the mean
coil temperature T c , and R*a is the total resistance for heat and mass transfer between coil mate-
rial and air determined with

1
R*a = ----------------- , (25)
ε *a ⋅ M· a

where M· a is the total air mass flow rate and where the air-side heat and mass transfer effective-
ness, ε *a , is determined with Equation 10. In this study, fin efficiencies for combined heat and
mass transfer were determined as described for the reference model.
The row outlet air enthalpy is determined using Equation 12, which is the inlet air enthalpy
for the downstream row. Equation 7 is used for determining the outlet row air temperature.

Numerical Solution and Computational Speed


A variety of numerical schemes could be applied to integrate Equations 18, 19, and 24. In this
study, both implicit and explicit time differencing schemes were investigated. A description of
the implicit scheme was presented for the reference model and is not repeated here. The finite
difference equations for the explicit method are given as follows:

0
( T w, out – T w, out )
Cw ⋅ ------------------------------------------ + C· w ⋅ ( T w, out – T w, in ) + ------- ⋅ ( T w, in – T c ) = 0
0 0 1 0 0
(26)
Δt Rw

0
( Tc – Tc ) 1 0 0 1 0 0
C c ⋅ ---------------------- + ------ ⋅ ( T c – T a, in ) + ------- ⋅ ( T c – T w, in ) = 0 (27)
Δt Ra Rw

0
( Tc – Tc ) 1 0 0 1 0 0
C c ⋅ ---------------------- + --------- ⋅ ( h s, c – h a, in ) + ------- ⋅ ( T c – T w, in ) = 0 (28)
Δt R*a Rw

where the superscript 0 represents the value at the previous time.


796 HVAC&R RESEARCH

The numerical solution of the explicit finite difference equations is somewhat similar to that
for the reference model but does not involve any iteration due to use of an explicit finite differ-
encing approach. First of all, a steady-state version of the simplified model as presented by Zhou
and Braun (2005) is solved to provide quasi-static initial coil states for the specified initial
boundary conditions. At any subsequent time, the analysis begins at the first row downstream of
the coil inlet airstream. For each row, a dry analysis is performed first, which involves the deter-
mination of T w, out , T c , and T a, out from Equations 26, 27, and 7 for given initial conditions,
inlet conditions, and other required parameters, respectively. The equations are explicit in these
three state variables. The outlet enthalpy is determined using a psychrometric function with the
outlet air temperature and the inlet humidity ratio.
To evaluate whether a wet analysis should be performed for a particular row, the coil surface
temperature is compared with the dewpoint of the inlet air, which is determined from the control
volume inlet air enthalpy and temperature using a psychrometric function. If the coil surface
temperature is greater than the dewpoint temperature, then the dry analysis is appropriate. Other-
wise, Equations 26, 28, and 12 are used to determine explicitly T w, out , T c , and h a, out with the
given initial conditions, inlet conditions, and parameters. The outlet air temperature is deter-
mined using Equation 7.
The choice of numerical integration scheme has an impact on computational speed. Computa-
tional speed is a critical issue for dynamic coil models since they are often used together with
other transient models of HVAC&R equipment to simulate system performance with small time
steps (e.g., one second) for long time periods (e.g., one year). The execution speed associated
with the simplified model was investigated for both the implicit and explicit schemes for the
eight-row coil. Representative results are presented for step changes in airflow rate for two dif-
ferent inlet air conditions that result in dry and wet coils as specified in Table 2. Table 3 presents
comparisons for computational speed for the different cases in terms of the ratio of real-time to
computational time. The results were determined for the model implemented on a laptop having
a 1.2 GHz Pentium III processor. The model was implemented as a dynamic-link library and
called as a procedure from MatLab. An integration time step of one second was employed with
the implicit method to realize accuracy comparable to the reference model. The explicit method
utilized a time step of 0.1 seconds to achieve similar accuracy.
Table 3 shows that the model runs three to four orders of magnitude faster than real-time and is
very suitable for dynamic system simulations. It also runs significantly faster for dry coil condi-
tions than for wet coil. The performance for partially dry and partially wet coils is somewhere in
between these two cases. It is also apparent that the implicit method is superior for dry coils and

Table 2. Inlet Conditions for Example Transient Comparisons with Airflow Step Change
M· , T a, in , RH a, in , ω a, in , M· , T w, in ,
Case a w
kg/s °C % g/kg kg/s °C
1 0.6–1.2 26.67 20 4.38 0.5 4.44
2 0.6–1.2 26.67 60 13.31 0.5 4.44

Table 3. Computational Speeds for Implicit and Explicit Integration Schemes


Computational Speed
Condition (Real-Time/Computational Time)
Implicit Explicit
Dry 10,000 5000
Wet 750 2500
VOLUME 13, NUMBER 5, SEPTEMBER 2007 797

inferior for wet coils in terms of computational speed when compared to the explicit method. For
dry coils, the implicit finite-difference equations are linear and the method does not require itera-
tion for state variables within each row. Furthermore, the implicit method can utilize a larger time
step than the explicit method. However, for wet coils the implicit difference equations are nonlin-
ear and an iterative solution is required to determine state variables. Since the wet coil computa-
tional requirements are significantly greater than the dry coil requirements, the explicit method
with a time step of 0.1 seconds is recommended and was utilized for all subsequent results for the
simplified model presented in this paper.

Reference and Simplified Model Comparisons


The accuracy of the simplified model was evaluated through comparisons with the reference
model for the two test coils subjected to step changes in boundary conditions, including step
changes in air and water flow rate, air inlet temperature and humidity, and water inlet tempera-
ture. This section presents representative results that were determined for step changes in air-
flow rate for two different inlet air conditions that result in dry and wet coils. The inlet
conditions used for the presented results are given in Table 2.
Figures 4 through 7 show example comparisons between transient predictions from the sim-
plified and reference models for Cases 1 and 2 from Table 2 applied to the two test coils. In all

Figure 4. Comparisons of reference and simplified model predictions for an eight-row


dry coil with Case 1 boundary conditions from Table 2.

Figure 5. Comparisons of reference and simplified model predictions for a four-row


dry coil with Case 1 boundary conditions from Table 2.
798 HVAC&R RESEARCH

Figure 6. Comparisons of reference and simplified model predictions for an eight-row


wet coil with Case 2 boundary conditions from Table 2.

Figure 7. Comparisons of reference and simplified model predictions for a four-row


wet coil with Case 2 boundary conditions from Table 2.
VOLUME 13, NUMBER 5, SEPTEMBER 2007 799

cases, the simplified model did an excellent job of predicting the exit air and water states. Very
similar results were obtained for other step changes and dynamic variations in inlet conditions.

Comparisons with Existing Models


The simplified distributed model developed in this paper was compared with existing simpli-
fied models developed by Braun et al. (1989), Clark (1985), and Ding et al. (1990) These three
existing models are lumped models and do not account for the distributed nature of the transient
terms. Furthermore, the model of Braun et al. (1989) is quasi-static. The purpose of the compar-
isons was to identify if any of these lumped models could simulate coil transients reasonably
well with less computational requirements. The models presented by McCullagh et al. (1969)
and Chow (1997) were not compared since these two models have the same order of complexity
as the current model and, thus, no advantage in execution speed.
For the model comparisons, the transient predictions were characterized in terms of variations
in air outlet temperature and humidity and water outlet temperature. These output variables were
normalized using the outlet conditions associated with the maximum and minimum values for
the transient using the following:

T a, out – T a, out, min


Θ a, out = -------------------------------------------------------- (29)
T a, out, max – T a, out, min

ω a, out – ω a, out, min


Ψ a, out = ---------------------------------------------------------- (30)
ω a, out, max – ω a, out, min

T w, out – T w, out, min


Θ w, out = ---------------------------------------------------------- (31)
T w, out, max – T w, out, min

where the subscripts min and max refer to minimum and maximum values.
Two representative transient conditions are presented for the eight-row test coil subjected to
step changes and sinusoidal variations in water flow rates. The inlet conditions are specified in
Table 4 and the corresponding comparison results are given in Figures 8 and 9. The model
developed by Braun et al. (1989) predicts step changes in coil outlet parameters in response to
step changes in water flow rate, which is expected since this model is quasi-static and contains
no transient terms. For the sinusoidal variations, the response of the quasi-static model is much
faster than the transient models, especially for predictions of water outlet temperatures. Also, in
comparing the normalizing factors it is apparent that this model overpredicts the amplitude of
the transient responses for outlet states.
The Clark (1985) model improves the transient predictions significantly simply by adding a
single time constant to the quasi-static outlet states. However, the air-side outlet parameters
respond more slowly and the water-side outlet temperature responds faster in comparison with
the simplified model presented in this paper (referred to as the Zhou model in Figures 8 and 9).
The Clark (1985) model gives reasonable transient results, but a single time constant cannot
accurately characterize transients of cooling coils because the time constants for the air-side and
water-side are different.
The lumped model developed by Ding et al. (1990) demonstrates almost the same air-side
transients as the Clark model, so the curves for the dimensionless air outlet temperature and
humidity are overlapped in both figures. For the water-side outlet temperatures, the Ding
(Ding et al. 1990) model shows initial jumps in transient responses when there is a step
change in water flow rate. Although this unrealistic behavior is not seen for sinusoidal varia-
800 HVAC&R RESEARCH

Table 4. Inlet Conditions for Model Comparisons


M· , T a, in , RH a, in , ω a, in , M· , T w, in ,
Case a w
kg/s °C % g/kg kg/s °C
3 1.2 26.67 60 13.31 Step changes shown below 4.44
4 1.2 26.67 60 13.31 Sinusoidal changes shown below 4.44

tions in water flow, the model does not accurately predict the amplitudes in outlet state varia-
tions for this case as compared with the simplified model presented in this paper (referred to as
the Zhou model in Figures 8 and 9). The Ding (Ding et al. 1990) model considers a coil to
either be fully dry or fully wet depending on whether an equivalent coil surface temperature is
greater than or less than the air inlet dewpoint temperature. In addition, this model uses
wet-bulb temperature differences as driving potentials instead of enthalpy differences in the
wet regime. Both approximations can cause errors in predicting the total energy transfer.
When calculating the air-side outlet temperature, the air-side effectiveness for total energy
transfer is applied rather than the sensible effectiveness, which also leads to errors in the air
outlet conditions.

CONCLUSIONS
A simplified dynamic coil model was developed that provides very accurate predictions at sig-
nificantly reduced computational requirements as compared with more detailed finite-difference
models. The model executed between three and four orders of magnitude faster than real-time on
a typical personal computer with accuracy comparable to a detailed reference model. The compu-
tational advantages of the model were realized through incorporation of steady-state performance
indices that characterize steady-state temperature profiles in the fins (fin efficiency), air (air-side
effectiveness), and water (water-side effectiveness). The use of these performance indices allows
the employment of relatively few state variables to characterize performance as compared with
typical finite-differencing approaches and ensures the model approaches steady-state or
VOLUME 13, NUMBER 5, SEPTEMBER 2007 801

Figure 8. Transient model comparisons for step changes in water flow rate.

quasi-steady behavior when subjected to static or slowly changing boundary conditions. Predic-
tions from the simplified model were also compared with existing simplified models from the lit-
erature and showed significant improvement in terms of accuracy. A companion paper (Zhou and
Braun 2007) demonstrates the accuracy of the simplified model as compared with detailed tran-
sient experiments.
The accuracy of the model in predicting the time lag associated with the transport time of the
fluid through a heat exchanger depends on the size of the control volume selected. This paper
recommends a control volume that encompasses an entire row of the coil because the model
incorporates a water temperature profile that converges to the steady-state solution under steady
conditions. This approach works well for the four- and eight-row coils considered in the current
study for a variety of step changes in inlet conditions including water temperature. However, it
may not work as well for step changes in water inlet temperature for more shallow coils such as
one- or two-row coils or when the water flow is lower. For these cases, a larger number of con-
trol volumes could be employed. However, in practice it is very difficult to realize a situation
where a step change in water inlet temperature would occur. Generally, the water inlet tempera-
ture changes are relatively slow in comparison with the response time of a cooling coil. It is
802 HVAC&R RESEARCH

Figure 9. Transient model comparison results for sinusoidal variations in water flow rate.

much more common to have rapid changes in air or water flows due to changing fan/pump or
damper/valve controls or in air states due damper mode changes. Therefore, it is believed that
the use of a single control volume for each row will work well in most situations.
It should also be noted that the model developed in this study may not be readily adapted to
direct expansion evaporators that are subjected to ON/OFF compressor cycling. For these types of
systems, the evaporator can experience a significant warming during off periods that leads to
re-evaporation of water from the coil surface. The simplified model does not include any water
retention mechanism on the coil surface and would not predict this phenomenon.

ACKNOWLEDGMENTS
We greatly appreciate the financial support of ASHRAE under RP-1194 and the technical
support provided by the Project Monitoring Subcommittee, which includes Michael Brande-
muehl, Richard Hackner, Philip Haves, and John Mitchell.

NOMENCLATURE
A = area, m2 NTU = number of heat transfer units
C = specific heat, kJ/kg·K or heat capaci- P = perimeter, m
tance kJ/K R = heat transfer resistance, K/W
C· = heat capacitance rate, kW/K RH = relative humidity, %
h = convection coefficient, W/m2·K or t = time, s
specific enthalpy kJ/kg T = temperature, °C
M· = mass flow rate, kg/s x = position within water flow stream, m
VOLUME 13, NUMBER 5, SEPTEMBER 2007 803

Greek symbols
Δ = time or spatial increment ω = humidity ratio, kg/kg dry air
ε = effectiveness Θ = dimensionless temperature
η = fin efficiency Ψ = dimensionless humidity

Subscripts
a = air or air side out = outlet
c = coil s = saturation
f = fin
t = tube
in = inlet
min = minimum tot = total
max = maximum w = water or water side

Superscripts
* = combined heat and mass transfer 0 = previous time interval
' = per unit length of tube

REFERENCES
ASHRAE. 2005. 2005 ASHRAE Handbook—Fundamentals. Atlanta: American Society of Heating,
Refrigerating and Air-Conditioning Engineers, Inc.
Bhargava, S.C., F.C. McQuiston, and L.D. Zirkle. 1975. Transfer functions for crossflow multi-row heat
exchangers. ASHRAE Transactions 81(2):294–314.
Bourdouxhe, J.P., M. Grodent, and J. Lebrun. 1996. Reference Guide for Dynamic Models of HVAC Equip-
ment. Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.
Braun, J.E., S.A. Klein, and J.W. Mitchell. 1989. Effectiveness models for cooling towers and cooling
coils. ASHRAE Transactions 95(2):164–74.
Chow, T.T. 1997. Chilled water cooling coil models from empirical to fundamental. Numerical Heat
Transfer Part A—Applications 32(1):63–83.
Clark, D.R. 1985. Type 12: Cooling or dehumidifying coil. HVACSIM+ Reference Manual. National
Bureau of Standards (now NIST), Gaithersburg, MD, pp. 63–68.
Ding, X., J.P. Eppe, J. Lebrun, and M. Wasacz. 1990. Cooling coil model to be used in transient and/or
wet regimes. Proceedings of the Third International Conference on System Simulation in Building,
December 3–5, Liége, Belgium, pp. 405–41.
Elmahdy, A.H., and G.P. Mitalas. 1977. A simple model for cooling and dehumidifying coils for use in cal-
culating energy requirements for buildings. ASHRAE Transactions 83(2):103–17.
Gartner, J.R., and H.L. Harrison. 1963. Frequency response transfer function for a tube in crossflow.
ASHRAE Transactions 69:323–30.
Gartner, J.R., and H.L. Harrison. 1965. Dynamic characteristics of water-to-air crossflow heat exchangers.
ASHRAE Transactions 71:212–24.
Gartner, J.R., and L.E. Daane. 1969. Dynamic response relations for a serpentine crossflow heat exchanger
with water velocity disturbance. ASHRAE Transactions 75(2):53–68.
Holmes, M.J. 1988. HVAC component specification: Heating and cooling coils. Energy Conservation in
Building and Community Systems Programme, Annex X: System Simulation. Paris: International
Energy Agency.
Jawadi, Z. 1988. A simple transient heating coil model. Proceedings of the Winter Annual Meeting of
American Society of Mechanical Engineers, November 22–December 22, Chicago, IL, pp. 63–69.
McCullagh, K.R., G.H. Green, and S.S. Chandra. 1969. An analysis of chilled water cooling dehumidifying
coils using dynamic relationships. ASHRAE Transactions 75(2):200–09.
Schmidt, T.E. 1949. Heat transfer for extended surface. Refrigerating Engineering 4:351–57.
Stoecker, W.F. 1975. Procedures for Simulating the Performance of Components and Systems for Energy
Calculations. Atlanta: American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.
Tamm, H. 1969. Dynamic response relations for multi-row crossflow heat exchangers. ASHRAE Transac-
tions 75(1):69–80.
804 HVAC&R RESEARCH

Tamm, H., and G.H. Green. 1973. Experimental multi-row crossflow heat exchanger dynamics. ASHRAE
Transactions 79(2):9–18.
Zhou, X., and J.E. Braun. 2005. Dynamic modeling of chilled water cooling coils. ASHRAE RP-1194
Final Report, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.,
Atlanta.
Zhou, X., and J.E. Braun. 2007. A simplified dynamic model for chilled water cooling and dehumidifying
coils—Part 2: Experimental validation. HVAC&R Research 13(5):805–18.
Zhou, X., J.E. Braun, and Q. Zeng. 2007. An improved method for determining heat transfer fin efficien-
cies for dehumidifying cooling coils. HVAC&R Research 13(5):769–84.

S-ar putea să vă placă și