Applied Thermal Engineering Vol. 16, No. 7, pp.

61S623, 1996
Copyright 0 1996Elsevier Science Ltd
Pergamon Printed in Great Britain. All rights reserved
1359-4311(95)00040-2 1359-431 l/96 $15.00 + 0.00

MATHEMATICAL MODELLING OF THE THERMAL

PERFORMANCE OF A PHASE-CHANGE MATERIAL (PCM)
STORE: COOLING CYCLE

A. Ki,irklti,* A. Wheldont and P. Hadley?

*Akdeniz University, Faculty of Agriculture, Department of Agricultural Mechanization,
07059 Antalya, Turkey; and University of Reading, Department of Engineering, and School of Plant
Sciences, Reading, UK

(Received 23 August 1995)

Abstract-A mathematical model for the prediction of the thermal performances of a PCM store
containing 1 m long and 38 mm diameter polypropylene tube has been developed in this study. Air was
utilised in the store as the heat transfer fluid. The model was based on an energy balance or the
‘conservation of energy principle’. The results indicate that the agreement between the predicted and
observed temperature or heat transfer data is generally good. The amount of energy used in increasing
the temperature of the PCM at any time during the phase-change process is predicted to be about 3.5%
of the total energy stored. Copyright 0 1996 Elsevier Science Ltd

Keywords--Heat storage, phase-change materials, cooling cycle, modelling.

NOMENCLATURE
A area, m*
CP specific heat, k/kg”C
Dh hydraulic diameter, m
G heat transfer rate, W
h heat transfer coefficient, W/m2”C
Hf latent heat of the PCM, kJ/kg
k thermal conductivity, W/m”C
L length of the node, m
m air mass flow rate, kg/s
Nu Nusselt number
Pr Prandtl number
radius of the volumetric node, m
Re Reynolds number
T temperature, “C
Greek letters
Al time increment, s
P density, kg/m’
Subscripts
a actual
ai air inlet
a0 air outlet
av average
in incoming
1 liquid
out outgoing
P PCM
S solid, or surface
t tube

INTRODUCTION

The use of PCMs for energy storage goes back to the 1950s and their use includes a wide range
of application areas, such as houses [l], aircraft [2] and computer rooms and telephone boxes [3].
Although many different studies exist on the use of PCMs for different purposes and in different
ATE
16,7-E 613
614 804884.OA. Kiirklii et al.

Air

\ Plastic sheet
Insolation (polystyrene)
\ Ply wood
Fig. 1. Schematic and dimensions of the experimental store.

places, only a few studies have appeared in the literature on the modelling of the thermal
performances of the PCM stores. This is because of the fact that the heat transfer process in systems
where melting or freezing takes place is very complex due to factors such as moving phase interfaces
from liquid to solid or vice versa. These types of problems are referred to as moving boundary
problems or Stefan’s problem [4]. There is also the difficulty of the possible existence of convection
heat transfer in the liquid phase in some cases when dealing with these types of problems.
Majumdar and Saidbakshs [5] indicated that most of the analytical solutions of the problem are
for very special boundary conditions, thus it is not possible to extend these to more complex
practical applications. However, numerical models are easy to solve with very small time
increments, since there are powerful computers available today. If, in this case, a numerical model
can be developed, the model can be used for different practical applications with the adjustment
of the parameters for that specific application. Some of the modelling studies carried out were those
of Majumdar and Saidbakhsh [5], Lacroix [6] and Das and Dutta [7]. Each of these studies
considered a different type of heat exchanger and a PCM. However, only the predicted data were
given in the study of Majumdar and Saidbakhsh [5] and there was a poor agreement between
predicted and observed data in the study of Das and Dutta [7]. Although the model of Lacroix
[6], based on an enthalpy method, predicted the observed data reasonably well, the magnitudes of
errors were not given and a much better accuracy is required for the practical use of such models.
Thus the purpose of this study was to develop a numerical heat transfer model for the prediction
of the thermal performance of a PCM store and to have a better understanding of the thermal
behaviour of the PCM store as a whole. Such a model could be used for similar applications
provided that the necessary modifications to the model are carried out for the PCM heat-exchanger
type. It can also be employed to optimise the diameter, length and spacings of the PCM tubes for
a given charge or discharge period, in other words for a given quantity of available storable energy,
which would be costly and time-consuming to determine experimentally.

DESCRIPTION OF THE STORE

A schematic of the square cross-sectioned store is shown in Fig. 1. The store contained a 1 m
long and 38 mm diameter polyproplyene PCM tube in annular air flow. The PCM was a mixture
of sodium sulphate and sodium chloride salts and water, together with additives to prevent
supercooling and phase segregation. The thermophysical properties of the PCM and of the tube
are given in Table 1.
An air-conditioning laboratory unit was used to supply air at a controlled and desired
temperature to the inlet of the store. To the unit was attached a thermostat and a voltage regulator,

Table 1. Thermophysical properties of the PCM and the tube utilised in the experiments’
Material CP* CP S, Sl k, kl HI (kJ/kg)
PCM 3.1 3.6 1480 1360 0.7 0.5 216
Tobe 1.9 - 904 - 0.12 -
*Here S, I and Hf represent solid, liquid and latent heat, respectively.
 Modellingof PCM performance 615

and it was insulated from outside to control the air inlet temperature to the store more accurately.
The structural material of the store was 12 mm thick plywood (k = 0.12 W/mC) and there was
30 mm thick polystyrene inside the store (k = 0.027 W/m”(Z). Another plastic cover was attached
to the store over the insulation layer and thus an air passage was created between these two coats
to facilitate the flow of a small proportion of the air. This air had the same temperature as the
air flowing over the PCM tube, thus transverse heat exchange was minimised.

METHOD OF MEASUREMENT
Air temperature measurement was carried out using four K-type thermocouples connected to
a Yokogawa-Hokushin data logger and the recording was stored every 5 min on an Apple
microcomputer using a Basic program. Two thermocouples were located at the inlet and outlet of
the store in the middle of the space between the PCM tube and the inside surface of the store to
measure air temperature. The other two thermocouples were positioned at the middle and outlet
of the store but in the centre of the PCM tube. When placing thermocouples into the PCM, extra
care was taken to ensure that the location depths were equal and that the insertion spaces in the
tube were sealed perfectly, so that no bursting or water loss of the PCM occurred.
Air velocities for each experiment were measured by a hot wire anemometer. The required air
mass flow rate was provided by the variable-speed centrifugal fan on the air-conditioning unit and
the air velocity was measured at the exit (see ref. [9] for the measurement principles). In the
experiments, air inlet temperatures of 11, 14 and 17°C and air mass flow rates of 0.0033, 0.0058
and 0.0083 kg/s were used. Comparison of the observed and predicted temperature data is
represented here only for the air mass flow rates of 0.0033 and 0.0058 kg/s when the air inlet
temperature was 11“C.
The experimental heat transfer rates between the PCM tube and air were calculated from the
equation below:

Qa = mCp,(T,, - Tai), (1)

MODEL DESCRIPTION

Division of the PCA4 tube into nodal elements

For the determination of the temperatures of the PCM tube and of the air over it, the whole
length of the tube was divided into 25 equal length increments of 0.04 m each. The volume of the
tube (in the radial direction) was initially divided into four equal increments. However, volume
division was then rearranged so that the first node, which was on the surface of the tube, had half
of the volume of an interior node, to determine the thermal conditions near the surface more
accurately. In the final model, the radial volume node arrangement included a half volume node
for the surface, three exact volume nodes for the interior nodes and a half volume node for the
central node. Figure 2(a) and (b) shows the incremental volume and length arrangement. The
resulting temperature prediction points in total were 155, including the air temperature.
Since the thermophysical properties of the tube and that of the PCM were rather different
(Table l), the equation for the first node was used to calculate the decrease of surface temperature
of the tube, (i.e. polypropylene), instead of PCM. The tube thus represented a higher thermal
resistance to heat flow than the PCM. Surface nodes were always subjected to convection from
one side, due to air flow, and conduction from the other side from the PCM. Initially, it appears
that the calculations are made for only one dimension. In fact, when the individual radial nodes
are considered, calculations are seen to be carried out in two dimensions, namely in the radial and
axial directions.

Structural basis of the model

Having divided the tube into five volumes and into 25 equal-length elements, the basic energy
balance equation can be employed to a control volume about an interior node after some
assumptions. When dealing with the freezing process, there are two distinct phases, namely solid
616 A. Kiirklii et al.

I
’'d
\
‘e
(b) , n-l , ” in+1 I

Center of the PCM tube

_VY”

I I
Fig. 2. Incremental division of the PCM tube into (a) volume and length (b) elements.

and liquid. For most inorganic salt hydrate PCMs, there is also a transition, or gel phase, between
solid and liquid phases. For all these three phases, the thermophysical properties of the PCM differ.
Also, the properties of the solid and liquid phases vary with temperature. However, for the case
examined here, since the temperature increase or decrease in the PCM was small, the variation of
the properties with temperature within any phase was ignored. The values for frozen, freezing and
liquid phases were therefore taken as constants. The thermophysical properties for the freezing were
taken here as the average of solid- and liquid-phase values. In summary, the basic assumptions
in the model were as follows:
The phases are homogeneous.
The PCM densities are constant for each of the three phases.
Convection heat transfer does not occur in the liquid phase.
Supercooling does not take place.
Heat loss or gain from the store is neglected.
The volume node which involved freezing here was the one which was adjacent to a solid node
on one side and to a liquid node on the other side. The other nodes were assumed to be either
completely liquid or completely solid. Thus, accounting for changes in thermal energy storage, a
general form of the energy balance equation can be expressed as follows:

Qc - Qout= Qst. (2)

For convenience, the centre point of each node was regarded as the temperature calculation spot.
The air outlet temperature for the first axial node was subsequently taken as the inlet air
temperature to the next node. The calculations were repeated successively to the end of the PCM
tube until the PCM attained a desired temperature, which was about 1°C higher than the air inlet
temperature.
The air outlet temperature from the nth node at the (p + 1l)th step in the calculations was
obtained from:

Ak(Tt,T ’- TiTn’) = mCp,(TP+

w,. ’- Tii;,‘), (3)
 Modelling of PCM performance 617

thus rearranging:
Zl,f = Ei;tnl + [(Ash/mC&J(Z~~’ - EiFn’)]. (4)

Temperature of the Jirst volumetric node at liquid phase

The liquid state is the phase from which the PCM is cooled down until the phase-change
temperature is attained and it is the stage to which the PCM is heated up above melting
temperature. When the energy balance equations are applied to each node until the freezing
temperature is achieved, the resulting simultaneous equations will make it possible to calculate the
temperature of the PCM with nodal volumes, incremental lengths of the PCM tube and with time.
Initially (i.e. at At = 0), the temperature of the PCM was assumed to be equal and known at
all points within the tube. Thus properties for the first node were those of the entire PCM tube.
Since node 1 was subjected to convection from one side and conduction from the other side, the

2I%L(Ti’
- T&n)
1,n - T!,,,)
energy balance equation for the first volumetric node yields:

=
ln(r&-tJ U-C:’
1 pCpJIL(r,Z - ri)
t 1. (5)

After necessary rearrangement, the equation becomes:

I. (6)
ln(r,/rb)2nr,Lhdt(T~i~“’ - T&) - 217ksAt( T$,,”- Ti,,“)
Ts,., + 1 = Z,., + ln(rc/rb)ptCpJ7L(r: - rQ

When choosing At, stability requirements should be satisfied because the explicit method of finite
difference solutions is not unconditionally stable [lo]. Since the calculated values should approach
steady-state values as the time increases, and At should be chosen such that it should not cause
the solution to diverge from actual steady-state conditions, the value of At is dependent on the
thickness of the volume partitions and other parameters, such as conductivity, density and specific
heat capacity. The value of this time-step in the computer program of the model was changed when
necessary for different air flow rates (for example, 15 s for 0.0033 kg/s, but 10 s for 0.0058 kg/s).
Similar equations were written for the interior volumetric nodes to obtain the temperature
variation of that node with time, and to give expressions for the nodal temperatures. The rest of
the equations of the interior nodes at the solid phase are given in the Appendix.

Temperature of the second volumetric node at phase change

When the phase-change temperature is attained, the PCM starts to freeze and the equations for
liquid phase are no longer valid for the internal nodes. For the surface node, the temperature
equation is the same, except for the change in properties of the PCM from liquid to solid. When
the phase-change temperature is attained in a node, the equation for that node changes to a second
equation to calculate the amount of heat transferred from that node. Clearly, the amount of heat
transferred from the radial nodes along the tube will be different. Thus, the rate of heat transfer
from the second volumetric node at the time of phase change was calculated as follows:

(7)

Here, the Q value gives the amount of heat that was released to freeze the fraction of the PCM
in the node under consideration. However, in practice, not all of this heat comes from freezing,
since a fraction F decreases the temperature of the node while phase change is taking place. Thus
the PCM temperature of the node was computed as follows:

PZ.I = T;,n +
Tp+’
FQ:nAt
Cd - rc2)ULpavgCpavg 1’
The assumption, in the model, of only conduction existing is valid since the gelling agent
employed in the PCM prevents mixing of the contents of the PCM by any liquid motion. The
618 A. Kiirklii et al.

temperatures of the other volumetric nodes at phase change can be calculated similarly (see
Appendix for the related equations).

Temperature of the volumetric nodes after phase change

Once the phase-change process (i.e. freezing) is completed for any volumetric node, the equations
are again equivalent to those for the liquid phase. The only difference is that the liquid
thermophysical properties of the PCM are replaced with those for the solid phase. Changes in the
thermophysical properties of the PCM are ignored in the temperature range from complete freezing
to minimum solid PCM temperature (i.e. about air inlet temperature). Heat transfer is still assumed
to be by conduction only.

Selection of the heat transfer coejicient equations

The selection or development of an appropriate heat transfer coefficient equation is an important

part of the modelling study. Due to the air velocities used in the experiments both being in the
laminar (Re I 2300) and turbulent flow regions (Re > 2300) it was necessary to use two types of
convective heat transfer coefficients. In the model, the following equations were used.
For laminar flow [lo]

Nu= 1.86[$q[$
All the values in the equation were evaluated at the average temperature of the air (i.e. at
(T,, + T,,)/2), which was considered an adequate evaluation; pS was evaluated at the local surface
temperature.

0054
For turbulent flow [l l]

Nu = 0.024(Re)0.8(Pr)0.3 F
[ 1.
The properties used were for the average air temperatures and were assumed to be constant.

Development of the computer program

A flow chart of the computer program, written in Quickbasic programming language, is shown
in Fig. 3. The program begins with the insertion of the air inlet temperature and the PCM initial
temperature. All the node temperatures are equated to this initial PCM temperature because the
PCM is initially assumed to be at constant temperature. The program then calculates the necessary
heat transfer coefficients and proceeds to the computation of the temperatures of the nodes
according to the phase situation. The program also defines where to go after each temperature
calculation by comparing this temperature with a predefined temperature condition. When the
temperature of the 25th length node at the central volumetric node is greater than or equal to the
air inlet temperature, the computations are terminated.

RESULTS

Determination of the F value

The numerical mathematical model was first validated with experimental data. The model was
set-up to calculate the temperatures of the air and that of the PCM at the formerly defined volume
nodes. The F value, that is the percentage of energy that was released to decrease the temperature
of the PCM whilst freezing, was found to be 3.8%. The final value of F was defined after several
trial and error computations by the model and then by comparing the predicted temperature values
with the experimental ones. The F value which gave the best agreement with the experimental
complete melting time of the PCM was then chosen.
 Modelling of PCM performance 619

Comparison of measured and predicted PCh4 temperature data

Figure 4 shows the variation of the measured and the predicted temperatures of the PCM with
time at the middle and the end of the tube at the central volumetric node for air mass flow rates
of (a) 0.0033 kg/s and (b) 0.0058 kg/s at the air inlet temperature of 11°C during the cooling cycle

I start 1

/ Input ta1, pcml /

Re, Vel, Dh

1 hc(turbulcnt) 1

TVl,TV?.,-I’M,W4,TW,Tao

Fig. 3. Flowchart of the computer program.

620 A. Kiirklii et al.

0 1 2 3 4 5 6 7 8 9 i0
Time (hr)

(b)
30

25

G
e 20
5
a
5 1s

5
$ 10
Y

I I I 1 I I I

0 1 2 3 4 5 6 I 8

Time (hr)
Fig. 4. Predicted versus measured temperatures in cooling cycle. Tti = 1l”C, (a) m = 0.0033 kg/s and (b)
m = 0.0058 kg/s.

of the PCM. The cooling cycles were commenced when the temperature of the PCM was
about 29°C.
It can be seen from Fig. 4(a) and (b) that the agreement between the predicted and the
experimental data was generally good. The model predicted the complete melting time of the PCM
to a good approximation, as can be seen from the figure. The complete melting time of the PCM
for the air mass flow rate of 0.0058 kg/s was almost half that of 0.0033 kg/s, due to a higher heat
transfer rate. The differences between the predicted and observed temperatures was about 2.5”C
at maximum, which was outside the phase-change region. This temperature difference, however,
does not create a drawback in terms of the application of the model.
As can be seen from Fig. 4(a) and (b), the temperature of the PCM decreased rapidly to about
21°C and then increased to about 23°C (‘V’ shapes in the curves), which was an indication that
the nucleation, or crystallisation (i.e. latent heat release stage) of the PCM started. After this stage,
the temperature of the PCM stabilised at about 22°C with a very small decreasing slope until the
phase change was completed. All of the cooling curves exhibited what was called ‘one- and
two-phase supercooling’ [12]. In this situation, after nucleation, the PCM temperature was unable
 Modelling of PCM performance 621

Table 2. Summary of complete freezing times (hr)

T., (“C) m (kg/s) Time (hr)
11 0.0033 5.5
0.0058 2.3
0.0083 1.3
14 0.0033 8
0.0058 5.3
0.0083 3.2
17 0.0033 11.2
0.0058 6.5
0.0083 3.5

to rise to its melting temperature but levelled off at a value slightly below. However, this amount
of supercooling (about 1’C) will not present any significant problem from the point of view of
application. After the phase-change process, the PCM temperature continued to decrease steadily
but faster than the phase-change stage. Towards the end of the cycle, the slopes of the curves were
very small due to quite a small amount of heat transfer between the PCM and heat transfer fluid
(i.e. air). In the PCM tube, the changes in the PCM temperatures at the outlet of the tube always
lagged those in the middle, because of the nearly constant h, and the decreasing temperature
difference between the PCM and air. As the air mass flow rate increased or the air inlet temperature
decreased, the phase change (or complete freezing) time of the PCM decreased (Table 2).

Comparison of measured and predicted heat transfer data

Good agreement between actual and predicted temperature data is also an indication of the good
agreement of the predicted heat transfer data with the actual ones. Table 3 shows the actual and
predicted amounts of released energy obtained by applying Simpson’s integration rule [13] for the
air inlet temperatures of 11 and 14°C and all air mass flow rates used in the experiments. In general,
the actual amounts of released energy were slightly higher than the predicted ones. This could be
due to the thermal mass of the polystyrene insulation material used inside the PCM store. However,
average relative error was rather small (about 0.04%).
A sensitivity analysis of the model showed that increasing At, for example from 5 to 15 s for
the air flow rate of 0.0033 kg/s and air inlet temperature of 14°C had no significant effect on the
calculation output, whereas it was observed that the model was more sensitive to the changes in
h, for a volume element with constant dimensions.

CONCLUSIONS
In this study, a numerical mathematical model of a PCM store was developed and the model
was validated by using experimental data.
The percentage of the energy released in decreasing the temperature of the PCM during freezing
was found to be 3.8%. The thermal properties of the PCM tube in this study were also found to
have an important effect on the heat transfer, since they represented a higher thermal resistance
than the PCM, due to its lower thermal conductivity. This was either considered equal to the PCM
thermal conductivity or neglected in the other studies mentioned above. Thus consideration of the
tube and container thermal properties needs to be made separately from the PCM, whether the
container be plastic or metal.
The model suggested that it was possible to eliminate some complex assumptions, such as the
change of thermal properties of the PCM with temperature and the existence of convection in the
liquid phase, while maintaining acceptable accuracy. The agreement between the experimental and

Table 3. Actual and predicted amounts of released energy

T, ("C) m (kg/s) Actual (kJ) Predicted (kJ)
11 0.0033 294.3 282.5
0.0058 296.7 283.3
0.0083 298.0 289.0
14 0.0033 285.8 275.9
0.0058 288.0 277.3
0.0083 287.1 282.2
622 A. Kiirklii er al.

predicted data by the model for both the PCM temperature and the energy released was good. The
accuracy of the model could be increased by considering the changes in the thermal properties of
the air with temperature. Furthermore, the model developed here can be a useful tool in choosing
the optimum design for a particular application. This can be achieved by simple modifications in
the model [8].

Acknowledgement-This study was carried out as part of a research project at Reading University, on the use of
phase-change materials for energy management in greenhouses.

REFERENCES

1. G. A. Lane, Solar Heat Storage: Eatem Heat Materials, Volume I:Background and Scientific Principles. CRC Press,
Florida (1983).
2. D. V. Hale, M. J. Hoover and M. J. O’Neil. Phase Change Marerials Handbook. NASA Contractor Report, NASA
CR-61363, (1971).
3. R. Prudhoe, Thermal stability: phase change materials in focus. Australian J. Refrig. Air-cond. Heating 24, April, 33-38
(1991).
4. E. M. Sparrow, S. Ramadhyani and S. V. Patanker, Effect of subcooling on cyclindrical melting. Heat Transfer 100,
395 (1978).
5. P. Majumdar and A. Saidbakhsh. A heat transfer model for phase- change thermal energy storage. Heat Recovery
Systems & CHP 10(5/6), 457 (1990).
6. M. Lacroix, Numerical simulation of a shell-and-tube latent heat thermal energy storage unit. Solar Energy 50(4), 357
(1993).
I. S. Das and T. K. Dutta, Mathematical modelhng and experimental studies on solar energy storage in a phase change
material. Solar Energy 51(5), 305-312 (1993).
8. G. W. C. Kaye and T. H. Laby, Tables of Physical and Chemical Constants and Some Mathematical Functions, p. 411.
Longman (1986).
9. W. C. Osborne and C. G. Turner, Woo& Practical Guide to Fan Engineering, p. 290. Woods of Colchester, UK (1969).
10. F. P. Incropera and D. P. De Witt, Fundamentals of Heat and Mass Transfer. John Wiley and Sons, Chichester (1990).
11. S. Kakac, Heat Transfer with Examples. Middle-East Technical University Press, Ankara (1987) (in Turkish).
12. P. G. Grodzka, Solar Energy Technology Handbook (Part A-Engineering Fundamentals), pp. 795-807. Marcel Dekker,
New York (1980).
13. K. Astroud. Engineering Mathematics, p. 951. MacMillan, India (1991).

APPENDIX
Equations for PCM temperatures at solid phase

For the volumetric node 2,

I.
I(rdlrJZAt~,(T:,,. - T&J - ln(r&J2k,At(T;,. - TL .)
T;:.’ = TrP2.n+
ln(rd/r,)ln(r./rb)p,Cp,(r~ - r?)

For the volumetric node 3,

I.
WJrd)2Atk,(TL - T;,.,) - hr(r,/rb)2k.Ar( T;,,, - T;,,)
T&y”’= Trp1.n+
ln(r,/rb)ln(r,/rd)p.Cp,(r,2 - rj)

For the volumetric node 4,

I.
In(r,)2Afkr(T$,., - T&J - In(rd/r,)2k,Ar( T;,,, - TL)
T;,:.’ = TrII.I + 1
In(rd/n)ln(r,)p,Cp,(ra
- r.)

It must be noted that, in fact, the logarithmic values are equal for the interior nodes, excluding the central volumetric
node for which the logarithmic value is equal to that of the surface node.
For the central volumetric node,

Equations for PCM temperatures at phase-change heat transfer rates at phase change

For the volumetric node 3,

 623
Modelling of PCM performance

For the volumetric node 4,

For the volumetric node 5,

Temperatures at phase change

For the volumetric node 3,

For the volumetric node 4,

1
FQ%.A
T;,:.’ = T%,. +
II (ri - r$lLpp.v,Cp,vg

For the volumetric node 5,