Applied Thermal Engineering 216 (2022) 119002

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

Economical design optimization of a phase change material active system

equipped with a solar collector
Misagh Ebrahimpour, Gohar Gholamibozanjani, Wei Yu, Brent Young, Mohammed Farid *
Department of Chemical and Materials Engineering, University of Auckland, Private Bag 92019, Auckland, New Zealand

A R T I C L E I N F O A B S T R A C T

Keywords: Coupling solar air collectors and active systems of phase change materials is an efficient approach for improving
PCM storage building heating supply. Through design optimization, such a system can become economical and energy-
Solar air collector efficient in different climates and for different types of buildings. However, previous works have not paid
Dynamic modeling
much attention to the economic feasibility of such systems during designing. In this paper, economical design
Economical design optimization
optimization of a system including a solar collector and a PCM heat exchanger was explored. For this purpose, a
model that integrated the solar collector, PCM storage, and hut dynamics was developed for the first time. The
average mean square error between the measured and predicted hut temperatures over 11 days was 4 ◦ C. The
validated model was then used for the design optimization of the system for three different scenarios: office,
domestic, and service, with different schedule times and comfort temperatures in the cold season of Auckland.
The results showed that the optimum surface area of the solar collector was the same at 1 square metre for all the
scenarios; however, the optimum amounts of PCM mass for service, domestic, and office scenarios were 35 kg,
20 kg, and zero, respectively. The optimum values indicate that the system can be cost-effective when used as an
auxiliary heating system to meet a portion of the total heating demand.

materials (PCMs) can be applied easily and provides reasonable energy

storage density at an almost constant temperature [5].
1. Introduction Design optimization of the systems, including solar collectors and
PCM-based energy storage, can make them more economical and
Buildings were responsible for almost 36 % of energy consumption energy-efficient. Design optimization of such systems has been studied
and nearly 40 % of annual global greenhouse gas emissions in 2017 [1]. in different researches. In [6], design optimization of heat pipe evacu­
This is mainly due to the improvement of living standards and occu­ ated solar tube collector with PCMs was carried out in two different
pants’ comfort demands, which has led to more cooling and heating in modes of operation, normal and on-demand. Numerical models were
buildings [2]. Total energy consumption in the sector is expected to rise used to simulate the system and determine the best design for improving
3 % annually in the foreseen future [3]. the system’s thermal performance.Heat pipe position in the glass tube
Solar energy has been recognized as one of the reliable energy and PCM were the design varivales for enhancing the thermal energy
sources for supplying the global energy demand. However, solar energy storage of the system. An efficient and optimized PCM storage unit for a
is intermittent, which would cause a mismatch between the availability collector storage water heater system was explored in [7]. The storage
of the energy and the actual demand. Integrating thermal energy storage unit was a multichannel flat tube and rectangular fins as heat exchanger
(TES) with the solar sources of heat could solve this issue. For decades, elements. The theoretical analysis method was used for optimizing the
researchers have been attracted by thermal energy storage (TES) to storage unit by studying the influence of structural parameters of rect­
reduce building energy consumption and improve thermal comfort [4]. angular fins, as the design variables, on the charge/discharge process.
TES can be used for storing energy in the storage medium, which can be The above researches obtained the best design by analyzing the nu­
released on demand, reducing the mismatch between supply and de­ merical models, which would not lead to the optimum solution.
mand. TES can be applied using four different methods: sensible heating, Design optimization was also investigated by applying optimization
latent heat storage, thermo-chemical (reversible reactions), and physical algorithms for obtaining the optimum design variables. In [8] a multi-
absorption/adsorption. Latent heat storage using phase change

* Corresponding author.
E-mail address: m.farid@auckland.ac.nz (M. Farid).

https://doi.org/10.1016/j.applthermaleng.2022.119002
Received 4 December 2021; Received in revised form 14 June 2022; Accepted 10 July 2022
Available online 21 July 2022
1359-4311/© 2022 Elsevier Ltd. All rights reserved.
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Nomenclature V valve (dimensionless)

x horizontal direction (dimensionless)
A area (m2) y vertical direction (dimensionless)
Cp heat capacity (J/ kg oK)
dH hydraulic diameter (m) Greek symbols
E energy (kW) Δ delta (dimensionless)
f darcy friction factor (dimensionless) η efficiency (dimensionless)
G global solar radiation (W/m2) ρ density (kg/m3)
h entalphy (KJ/kg) ν air velocity (m/s)
i nodal count in horizontal direction (dimensionless) Subscripts
I global solar radiation on a horizontal plate (W/m2) a ambient
j nodal count in vertical direction (dimensionless) d diffuse solar radiation
K clearness index (dimensionless) ele electricity
L length (m) in inlet
Lf lifetime (years) LC lower comfort temperature
lc loss coefficient (W/m2 oK) out outlet
M mass (kg) PCM phase change material
N number of plates (dimensionless) pl plate
OF objective function (USD/year) SAH outlet of solar collector
p pressure (pa) sc solar collector
P price (USD/kg) sol solar
Q heat gain through solar radiation (J) UC uppepr comfort temperature
Q volumetric flowrate (m3/s) UH upper heating range temperature
t time (s)
T Temperature (◦ K)

objective optimization was carried out for TES systems, including PCMs absorber plate, the phase change temperature, and a parameter in the
for solar air systems in a lab-scale test rig. Thermal storage design var­ effective heat capacity curve were used as the optimization design
iables such as the number of the air channels and the number of PCM variables. The objective function thermal performance parameters, but
bricks were optimized. The optimum values increased the average heat the authors concluded that the techno-economic constraints need to be
transfer effectiveness and effective PCM charging time of the system. In added in future studies. The authors of [14] studied the optimal design
[9], a two-level model-based strategy was used to optimize a system of a renewable cooling and heating system that included a desiccant
including a Photovoltaic/thermal (PVT) collector with a centralized wheel, PVT, and a thermal storage unit was carried out. The objective
PCM thermal energy storage. The electrical and thermal performance of function was a twenty-year life cycle cost which included PVT price,
the system were increased by obtaining the optimum air flow rate and PCM price, electricity purchase price, and electricity sale price. The
optimum slope and orientation of the PVT collector. The design of a PVT results showed that the optimized design decreased the system’s life
with a passive PCM was optimized in [10]. PCMs were embedded into cycle cost by 32.4 % and 31.2 % compared to two other design cases. It
the building envelopes while the heated air from the PVT collector was was found the electricity sale price had a significant influence on the
used for heating the building. The Taguchi-Fibonacci search method was optimization results. However, the results cannot be helpful in designing
used for the optimization. The optimized variables were PCM air flow solar air collectors without PV panels, which only generate heat.
rate, PCM type, PCM layer thickness, and additional wall insulation. The Economical design optimization of systems including the coupling of
objective function maximized the signal-to-noise ratio of the coefficient solar air collectors and PCM energy storage, which minimize the system
of the thermal performance enhancement (CTPE) of the building simu­ cost could facilitate the more deployment of such systems.
lated in TRNSYS. For design optimization purposes, a validated model that integrates
In the above studies, the design optimization was focused only on the dynamics of the whole system, including solar collector, PCM stor­
improving the system’s technical performance. However, ignoring the age, and the building, is needed. In [15], the authors numerically
initial cost of solar collectors and PCM energy storage could lead to investigated a PCM-based cascaded energy storage unit with a solar air
designing systems with long payback times, which could lead to un­ collector. Three paraffin-based materials (RT50, RT65, and RT80) were
feasible economical systems. For instance, in [11], parametric analysis used as PCM for the energy storage unit. The thermal energy storage unit
was used for designing a solar heating air system, including an air and the solar collector were modeled. The model was validated by the
vacuum tube solar collector and a concentric-tube latent heat thermal experimental data with high accuracy (R-squared was equal to 0.94). In
energy storage. The design of the energy storage system was optimized another research, simulation of net-zero energy (NZE) was first devel­
regarding the air outlet temperature of the storage and peak shift of the oped using TRNSYS [16]. The authors investigated the energy flexibility
heat supply. The results showed that the optimal mass of PCM was and performance of NZE houses using a solar-assisted heating, ventila­
150–200 kg/m2. This amount of PCM introduces a high initial cost that tion, and air conditioning (HVAC) system with thermal energy storage
leads to a very long payback time. The cost of implementing PCM is (TES), PVT collector, and demand-side management (DSM) strategies. A
needed to be compared against the benefits of using such systems before simulation system of the NZE house was developed using TRNSYS to
making decisions [12]. The importance of economic factors was also evaluate its performance under various conditions. The major compo­
mentioned in [13]. The optimal design of a solar collector integrating nents of the system, such as PVT collector and air-based PCM storage,
PCM thermal storage was carried out in this study. A front and back solar were modeled and validated individually [17,18]. In particular, the air-
air collector (SAC) with a PCM-based absorber plate was considered. The based PCM storage model was developed [18] to represent the thermal
objective function minimized the root mean square error between the dynamics of the storage for cooling purposes in summer using the free
solar air collector and the set temperature. The thickness of PCM in the ambient cooling at night to solidify PCM. A novel solar thermal heater

2
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

coupled with an active PCM heat storage wall was proposed in [19]. Hot was used to insulate the floor, while glass wool was used to insulate the
water heated by the parabolic trough solar collectors was flown through walls and ceiling.
the copper tube to discharge the cool load stored in the PCM wallboard. The hut includes fans, and control valves for directing the heated air,
Also the excess heat could be stored in the PCM wallboard to meet the as shown in Fig. 2.
indoor thermal demand when solar energy is insufficient. The transient The equipment’s specifications can be found in Table 1.
model of the active PCM storage was combined with the TRNSYS model, A PCM heat exchanger is also integrated with the solar collector in
which included the hut model. The indoor hut temperature was used to the hut. The PCM heat exchanger was made up of 19 sets of aluminum
validate the accuracy of the model. The RSME during seven days of macro-encapsulated PCM panels (0.45 m × 0.30 m × 0.01 m) filled with
operation was 0.7 ◦ C. Despite that there were studies on modeling solar 9.5 kg RT25HC (manufactured by Rubitherm GmbH) with melting
collectors with PCM storage, it appears that a model that integrates the temperature range of 22–26 ◦ C and heat of fusion of 230 kJ/kg. A side
solar air collector, and PCM storage, and building models has not been view of the heat exchanger unit is shown in Fig. 3. There was a 5 mm gap
previously reported. between the trays. Air enters the duct from one side and flows parallel to
An economical design optimization was developed in this paper, for the trays, exchanging heat with the PCM in this configuration. To ensure
the first time, for a heating system with a solar air collector and a PCM a uniform flow, a distributor was made and installed in the pipe leading
heat exchanger, supplying heat to an office-size hut. The optimization to the heat exchanger. A 20 mm layer of PVC/NBR black rubber foam
was carried out using an integrated model representing the whole sys­ was used to insulate the entire assembly from the environment (thermal
tem’s dynamics and interactions. The model was validated by the conductivity of about 0.037 W/m⋅K). More details can be found in [20].
experimental data of a winter season in Auckland. The validated model Temperatures from the hut, solar collector, and PCM heat exchanger
was controlled by applying an algorithm tested experimentally. Using were fed into a temperature controller. The temperatures were
the controlled virtual system, the optimum solar collector area and measured using T-type thermocouples calibrated against a reference
amount of PCM used were obtained by using the hut under different thermometer. The air velocity in pipes was measured using a digital
application scenarios such as office, domestic, and service. The objective anemometer. In pipes with a diameter of 10 cm, the air velocity was 3
function was defined by considering the energy cost saving of the system m/s. A pyranometer was used to measure solar radiation. Table 2 shows
during the cold season of New Zealand, which was penalized by both the the different types of measurement instruments and their accuracy.
system capital and operating costs. The data were logged using a Compact Reconfigurable (Compac­
tRIO) Data Acquisition System (NI Crio-9012, National Instruments,
2. Methodology USA) and LabVIEW software. The system analog inputs received the
temperatures and returned the decisions on the solar collector and PCM
A hut heated by a solar collector and a PCM heat exchanger was heat exchanger operation by analog outputs. These decisions were
considered in this work. PCM heat exchanger could store diurnal solar applied by manipulating valves and fans based on an ON/OFF control
energy for the later demand. The experimental setups and dynamic algorithm as described below.
model are described in the next sections. Model validation was carried There are three possible pathways for supplying heat, which are
out by using two sets of experimental data from winter 2020. controlled through three valves (V1, V2, V3) shown in Fig. 4.
Pathway 1 (V1 and V3 open, V2 closed), heating the hut using solar
2.1. Experimental setup energy: The solar collector sucks the hut air in and then sends the heated
air back to the hut, using an electricity-driven fan (Fan 1 in Fig. 4).
The setup includes an experimental hut equipped with a solar col­ Pathway 2 (V1 and V2 open, V3 closed), charging the PCM: The air
lector and a PCM heat exchanger, as shown in Fig. 1. It is located near heated by the solar collector is sent to the PCM heat exchanger before
Ardmore airport in Auckland (37.0314◦ S, 174.9724◦ E), New Zealand. sending it back to the hut. This circulation was done by an electricity-
The hut’s external dimensions are 2.7 m × 2.7 m × 2.7 m, and both have driven fan, Fan 2, installed at the outlet of the PCM heat exchanger.
a single-glazed window (0.8 m × 0.8 m) facing north. Polystyrene foam Pathway 3 (V2 and V3 open, V1 closed), heating the hut using PCM

Fig. 1. Experimental set up including a PCM heat exchanger (A) inside a hut equipped with a solar collector (B).

3
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Fig. 2. Hut set up details. The arrows show different pathways. Air flow is adjusted by adjusting valves and fans.

stored heat was released later (discharging) during the cooler hours (hut
Table 1
temperature lower than 19 ◦ C) through sucking the hut air via Pathway
List of equipment used in the experimental hut.
3.
Equipment Description Operating Electric power
conditions (W)

Solar air 1 m2 flat plate – Maximum: 550 2.2. Dynamic model

heater
Fan 100 mm Plastic Duct Flowrate: 130 12 Individual numerical sub-models describing system components
Booster Inline Fan m3/h
make up the dynamic model coded in either MATLAB or EnergyPlus
Valve PVC Fully open/fully –
closed software. In EnergyPlus, PCM can be defined as a layer on the building
envelopes that is suitable for passive applications. Heat exchangers
containing PCM materials, on the other hand, cannot be defined in
discharging: The hut air is circulated into the PCM heat exchanger and EnergyPlus. To compensate for EnergyPlus’s lack of flexibility in this
sent back to the hut, using the electricity-driven fan, Fan 3. case, the PCM heat exchanger was modelled in MATLAB and linked to
The desired lower and upper temperature bounds for the huts were EnergyPlus via an EnergyPlus Co-simulation Toolbox interface that
set to be 19 ◦ C and 25 ◦ C, respectively. These desired bounds were
maintained by adjusting air flow through the valve opening. If the Table 2
temperature of hut was lower than the air coming from the solar col­ Measurement instrumentation.
lector, then the valve was opened, and the fan switched on to circulate
Instrument Accuracy range Opearting range Model
the solar collector air to the hut and provide heating. The fan was
switched off after reaching the desired upper bound (25 ◦ C), and none of Thermocouples ±0.45 oC 0–50 ◦ C T-type
Reference thermometer 0.02 ◦ C 0–50 ◦ C Ebro TFX430
the pathways are used in this case. If after providing heating through Digital anemometer (± 2 % + 0.2 m/s) 0.4–30 m/s AM-4201
pathway 1, the hut temperature exceeded the desired upper bound, the Pyranometer <10 W/m2 0–2000 W/m2 VAEQ08E
air was circulated through Pathway 2 for melting PCM (charging). The

Fig. 3. Schematic side view of the heat exchanger unit with one magnified metal container [20].

4
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Fig. 4. Schematic view of the heating control. The temperatures of solar air collector, hut and heat exchanger are the inputs, and the three valves, and three fan
settings are the outputs of programmable logic controller (PLC).

controlled data flow between the two software systems. The sub-models solar collector outlet temperature (blue dashed line) was sent back to the
for the solar collector, heat exchanger, and hut interact together as hut models in EnergyPlus through the interface code. This is the heat
shown in Fig. 5. supply Pathway 1 described in Section 2.1.
The models for hut, solar collector and PCM heat exchanger were The second regime (charging) was when the PCM was charged, and
named Sub-model 1, 2 and 3 respectively. The solar collector and PCM the hut air was calculated in sub-model 1 and passed from EnergyPlus to
heat exchanger were modelled in MATLAB, while the hut was simulated MATLAB via the interface code. The hut temperature was introduced to
using EnergyPlus (E+ in Fig. 5). Sub-model 1, 2 and 3 are fully inter­ sub-model 2 along with the ambient temperature and solar radiation
acted through an interface for exchanging the data between MATLAB data. The calculated solar collector outlet temperature was the inlet to
and EnergyPlus software. In this case, the dynamic model operated the heat exchanger. In sub-model 3, the outlet temperature was the
under four different regimes as described below. output which was sent back to EnergyPlus through the interface code.
The first regime was when the PCM was charged, and the hut air was This cycle continued until the end of the charging phase. This is the heat
calculated in Sub-model 1 and passed from EnergyPlus to MATLAB via supply Pathway 2 described in Section 2.1.
the interface. The hut temperature was introduced to sub-model 2 along In the third regime (discharging), the hut air was circulated through
with the ambient temperature and solar radiation data. The calculated the PCM heat exchanger to discharge the PCM without going through

Fig. 5. Dynamic model structure. Dashed lines show the alternative heat supply pathways.

5
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

the solar collector. Therefore, the hut temperature determined from Table 4
EnergyPlus was directly introduced to sub-model 3 which is shown by Fraction of diffuse radiation according to clearness index.
the black dashed line in Fig. 5. The rest of the data flow was the same as Id Clearness index (kT)
for the charging regime. This is the heat supply Pathway 3 described in I
Section 2.1. 1 – 0.09* kT kT ≤ 0.22
The fourth regime is when neither charging nor discharging happens. 0.9511 – 0.16*kT + 4.388*kT 2 –16.638* k3T + 12.336* k4T 0.22 < kT ≤ 0.8
In this case, the heat supply was stopped, and the hut simulation was 0.165 kT > 0.8

carried out using EnergyPlus.

The start and end times of the charging and discharging regimes or
periods were determined from the experimental data and applied to the η = η0 − lc
Tin − Ta
(5)
simulation. Due to the stability problems, the simulation time step was G
considered one minute. Qsol = η G A (6).Where η is the efficiency of the collector and
subscript “a” refers to the ambient temperature, G is the solar radiation
2.2.1. Weather conditions and A is the collector surface area. η0 is the intercept (maximum) effi­
The ambient weather conditions were measured using the BRANZ ciency and lc (W/m2 oK) is the first order loss coefficient.
[21] weather station. The measurements included dry bulb temperature,
relative humidity, pressure, wind speed, solar radiation, and dew point 2.2.3. PCM heat exchanger model
temperature. The information was used in the weather file using the The model inputs came from the hut (Pathway 3) or the solar col­
Elements software [22] which is an open soure software tool for creating lector (Pathway 2). The output passed to the hut model.
and editing custom weather files for building energy modeling. A two-dimensional explicit finite difference model was used to model
Cloud cover data were recorded from the satellite information which the heat exchanger in MATLAB [20]. Non-linear algebraic equations
was available on the Meteorological Service of New Zealand Ltd.’s describing the heat transfer medias, which include PCM, metal
weather forecast website [23]. As only global radiation was measured, container, and air, were solved using the finite differences method.
the normal and diffuse solar radiation were calculated using cloud cover Nodal distribution of the model is shown in Fig. 6. The nodal counts in
data for estimating the clearness index [24]. The sky condition was the x and y directions are denoted by i and j, respectively. Δx refers to
defined based on the cloud coverage data, which aimed to determine the nodal discretization in x direction and Δy in y direction. dp and dc, show
clearness index as shown in Table 3. For every sky condition, a linear half of PCM thickness and metal container thickness, respectively. L is
function correlated the opaque cloud coverage and clearness index the length of the metal container in the flow direction. Tp,centre represents
ranges. PCM temperature in the center of the PCM layer in x and y directions.
Table 4 shows how the calculated clearness index was used to PCM behaviour was studied using the enthalpy method, which simplifies
determine the fraction of the hourly radiation (I) on a horizontal plate the heat transfer process during the phase change [26]. The enthalpy
which is diffuse (Id ) [25]. After introducing diffuse and global solar ra­ information was provided by Rubitherm GmbH [27].
diation, normal radiation was calculated in the Elements software. During the model’s development, a number of assumptions were
taken into account. The air between the metal containers was assumed
2.2.2. Solar collector model to be distributed evenly. The heat transfer between the air and the
The input to this model was the hut temperature, and the output containers was modeled in two dimensions in PCM and containers, but
either went to the hut (Pathway 1) or PCM heat exchanger (Pathway 2). the airflow direction was taken into account for the heat transfer be­
The thermal dynamics of the collector were modeled in MATLAB using tween the air and the containers. Additionally, the air density was
the following equation: assumed to be unchanged during the air flow in the ducts.
dhsc The advantage of this comprehensive model over previous models is
Msc
dt
= Fin hin − Fout hout + Qsol the inclusion of natural convection in the melted PCM, the thermal mass
of the PCM container’s wall, and the effect of PCM volume expansion.
Where F is the air mass flow rate in the collector and M is the solar The model was validated at the lab scale with a good agreement of less
collector mass. h is the enthalpy and subscripts “sc”, “in”, and “out” than 8 % average deviation between the model and experimental PCM
refer to collector, collector inlet and outlet air streams respectively. and air temperature measurements [20].
These enthalpies were calculated based on the following equations: The inlet air flow rate and temperature were introduced to the heat
exchanger model. The PCM initial temperature was measured prior to
hin = Tin Cpin (2)
charging. In the discharging phase, the PCM initial temperature was set
to be the same as the temperature reached at the end of the charging
hout = Tout Cpout (3)
phase.
hsc = Tsc Cpsc (4)
2.2.4. Hut model
Where T is the temperature and Cp is the specific heat capacity. As an The model input either came from the solar collector model
approximation, the collector temperature was assumed to be equal to (Pathway 2) or the PCM heat exchanger model (Pathway 3). EnergyPlus
the average inlet and outlet temperature. was used to model the huts. More details about the model equations can
Qsol is the heat gain from the solar radiation on the collector which be found in the software documentation [28]. The details of hut enve­
was calculated by Eqs 5 and 6. lope materials and their properties were similar to the hut used in [29].
To consider the heat supply from the PCM heat exchanger, a
“ZoneHVAC:IdealLoadsAirSystem” object was defined in EnergyPlus.
Table 3
This object’s air flow rate and temperature were introduced from the
Clearness index according to cloud cover. PCM heat exchanger model in MATLAB to EnergyPlus via the MATLAB-
E+ interface code.
Sky condition Opaque cloud coverage Clearness index (kT)
Inside the hut was a heat exchanger and its connections to the solar
Cloudy 88–100 % kT < 0.35 collector, which could store heat during the day and release it during the
Partially cloudy 6–87 % kT ≥ 0.35 and kT ≤ 0.65
cooler hours. This would have an impact on the hut’s temperature
Clear sky 0–5 % kT > 0.65

6
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Fig. 6. Nodal distribution in the 2-D system of current study [20].

dynamics. As a result, the mass of the equipment inside the hut was performance. The first set of data was collected from the 5th to the 10th
defined using EnergyPlus’ “InternalMass” object. of July 2020. During this period, the integrated model components, and
A curtain was draped across the north side of the hut’s window. This individual models of the solar collector, PCM heat exchanger, and hut
had a significant impact on heat gain from solar radiation. The “Win­ were compared with experimental data. To further analyse the system
dowMaterial:Shade” object was used to define the window shade ma­ integrated model, validation was carried out from 13th to 17th of July
terials’ properties, which represented the radiation transmission and 2020 by comparing the measured temperatures with those predicted by
reflection through the curtain. the hut models. The outside temperature and solar radiation for the first
The hut is near other structures that provide shade to the outside and second data sets are shown in Fig. 7, respectively. Solar radiation
walls. This is particularly important on the hut’s east external wall, varied from early morning to late afternoon and reached a maximum of
which impacts the heat gained through radiation. The sun exposure of 700 (W/m2). The lowest temperature was close to 2 ◦ C, in data set 2, and
the east external wall was defined using the “BuildingSurface:Detailed” the maximum temperature was around 18 ◦ C.
object.
The air infiltration through the door was also added as a “ZoneIn­
flitration:EffectiveLeakageArea” object. This model is based on Sherman 3.1. Solar air collector model
and Grimsrud’s work [30] which needs the effective air leakage area.
The other parameters were set based on the default values in For simulating the solar collector outlet temperature, the air inlet
EnergyPlus. flow rate and temperature along with solar radiation were introduced to
the solar collector model. The air inlet temperature was generated from
2.2.5. Integrated model interface the hut simulation in EnergyPlus. The values of parameters η0 and lc in
The interface exchanged data between MATLAB and EnergyPlus. On Eq. (5) were estimated as 0.42 and 2.3, respectively. These values were
one side of the interface, the hut temperature came from the hut model obtained by fitting a line to equation (5), where the solar collector
in EnergyPlus, and on the other side of the interface, either the solar supplied heat was calculated by the heat transferred to the air. The solar
collector or PCM heat exchanger outlet temperature was fed to the hut collector validation was carried out by comparing simulation results and
model in EnergyPlus. The interface employed between MATLAB and experimental data, which are presented for the 7th and 9th of July when
EnergyPlus was the EnergyPlus Co-simulation Toolbox [31]. The air the solar collector provided heat from early morning to late afternoon.
temperature and air mass flow rate of the “ZoneHVAC: Ideal­ Fig. 8 shows the comparison for the 9th of July. In fact, the collector
LoadsAirSystem” object was defined as three variables exchanged be­ efficiency was expected to be higher since the heat loss was not
tween MATLAB and EnergyPlus. The exchange of these variables was considered in the model. The heat loss was mainly in the pipeline be­
built by defining “EnergyManagementSystem: Actuator” and “Exter­ tween the hut and the collector inlet that was exposed to the ambient
nalInterface: Actuator” objects in EnergyPlus. These two objects were temperature, which could decrease the inlet temperature to the collec­
linked together by programming using the “EnergyManagementSystem: tor. The root mean square errors between the model output and the solar
program” object. collector outlet temperature for the 7th and 9th of July were 2.89 ◦ C and
2.75 ◦ C, respectively.
3. Modelling results and discussion The results show that the model can simulate the gross dynamic
behaviour of the solar collector. The difference between experimental
Two sets of data were considered for validating the model measurements and simulation results may be firstly due to the mismatch
between the specified inlet air temperature obtained from EnergyPlus

7
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Fig. 7. Solar radiation and ambient temperature from the 5th to the 10th (first row) and from the 13th to the 17th (second raw) of July 2020.

Fig. 8. Solar collector outlet temperature for 9th of July 2020. The root mean square errors is 2.75 ◦ C.

and the experimentally measured hut air temperature. Secondly, the air collector, while discharging, the hut air was circulated through the
solar collector was composed of different materials with different heat exchanger. Heat exchanger simulation was carried out for 4 days from
capacities. Considering the imprecise thermal mass of these individual the 5th to the 8th of July. PCM was not charged for the 9th and the 10th
components in the model could affect the overall heat capacity of the of July since the solar heat was not enough to melt the PCM. The solar
collector and consequently its thermal dynamics, leading to the collector outlet was only introduced to the hut during these two days.
observed mismatches. This also caused a slight time lag between the Fig. 9 shows the simulation results for day 1.
measurements and simulation in Fig. 8. This time lag is of the order of Fig. 9 illustrates the non-linear evolution of the PCM temperature,
5–10 min. which is mainly due to the absorbing and releasing of the latent heat
during the charging and discharging phases. Fig. 9 shows a good
agreement between the simulated and measured PCM temperature
3.2. PCM heat exchanger model during both charging and discharging phases. The heat exchanger outlet
temperature was very similar to PCM temperature since the PCM tem­
The heat exchanger model was simulated during both charging and perature was measured at the end of the plate which was close to the
discharging phases for hut 2. During charging, the inlet air temperature heat exchanger outlet.
to the heat exchanger was set equal to the outlet temperature of the solar

8
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Fig. 9. PCM temperature and heat exchanger outlet temperature during charging (left column) and discharging (right column) phases for day 1.

During the charging phase, the highest error is less than 1.5 ◦ C, between the exchanger surface and the air. Moreover, the PCM tem­
which happened in the charging phase of day 1. The root mean square perature was only measured from a plate at the end of the exchanger
error between the measured PCM temperature and the model output was which was considered the initial temperature for all plates that could
1.17 ◦ C, the highest error (Table 3). Although there was a mismatch have led to observed simulation errors.
between simulation and experimental PCM temperature during
charging, the PCM temperature at the end of the charging phase on day 1 3.3. Hut model
was less than 1 ◦ C. The PCM temperature at the end of the charging
phase is important because it determines the amount of stored heat As described in the methodology section, the above models were
available during the discharging phase. coupled with the hut model of EnergyPlus through the E + Co-simula­
During the discharging phase, the hut air was the inlet to the heat tion toolbox interface.
exchanger. The highest mean square error was 1.2 ◦ C which occurred
during day 2 (Table 5). The difference between simulated hut temper­ 3.3.1. First data set
ature and the experimental data was the main reason for the mismatch Fig. 10 shows the capability of the dynamic model in simulating the
between the simulated and experimental heat exchanger temperatures. huts’ temperature dynamics. The root mean square error between the
The simulated hut temperature during days 1 and 3 was higher than huts’ temperatures and the models’ outputs was 1.53 ◦ C. The maximum
the experimental temperature right from the beginning of the dis­ mismatch is around 2 ◦ C, which happened at noon and early morning, as
charging phase (as shown in Fig. 9), which led to a higher inlet air shown by the rectangular and circle shapes in Fig. 10.
temperature to the heat exchanger and consequently higher outlet
temperature. During day 2, the heat exchanger outlet temperature was 3.3.2. Second data set
lower than the experimentally measured temperature since the inlet air For further validation, a comparison was made between the dynamic
temperature from hut 2 was lower than that measured experimentally. model hut temperatures and the measured data using the second data
During day 4, however, hut air temperature was very close to the set. Fig. 11 shows these results. As Fig. 9 shows, the hut temperature
measured value which resulted in a good match between the simulated reached below the melting point of PCM (22 ◦ C), which indicates that
and experimental heat exchanger temperatures. PCM was fully melted to provide the required heat to the hut.
Furthermore, it should be noted that the non-uniform distribution of The root mean square error between the hut temperatures and the
the air inside the exchanger could lead to non-uniform heat transfer model outputs was 2 ◦ C. The rectangular shapes in Fig. 11 show the
maximum temperature difference between the simulation and experi­
mental data, which happened at noon.
Table 5
Simulation root mean square error (◦ C) for four days. 4. Discussion
Day PCM during Heat exchanger PCM during Heat exchanger
charging during charging discharging during discharging In both data sets, the maximum mismatch happened during the
1 1.17 1.11 0.2 0.13 midday and early morning. The mismatch may be due to the solar ra­
2 0.18 0.13 1.21 1.17 diation and sky temperature data which are related to the cloud cover.
3 0.13 0.1 0.23 0.31 However, the cloud cover data are not measured online but are pre­
4 0.008 0.85 0.013 0.12
dicted based on the satellite information, which is not updated online.

9
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Fig. 10. Hut temperature from 5th to 10th of July 2020.

Fig. 11. Hut temperature from 13th to 17th of July 2020.

Additionally, the cloud cover was used for clearness index calculation radiation, which is expected to contribute to the hut’s error in the solar
which determined the diffuse fraction of hourly radiation. The correla­ heat gain. The higher difference between the real and interpolated solar
tion between the fraction of hourly radiation, which was diffuse, and the radiation happens on partially cloudy days when solar radiation changes
clearness index is well established but approximate, which can be considerably. For instance, Fig. 12 shows the solar radiation variation on
another source of mismatch. days 13th to 15th of July. The solar radiation variation for July 15th is
Moreover, another reason for mismatches was the way solar radia­ smoother than July 13th, which was a partially cloudy day. On July
tion data are defined in EnergyPlus. This data is provided via a weather 14th, a significant change in solar radiation happened between 12 pm
file every hour, and solar radiation is interpolated at each simulation and 1 pm, as indicated by the orange circle in Fig. 12, which interpolated
time step. This causes differences between real and interpolated solar data cannot be represented.

Fig. 12. Solar radiation variation for three days.

10
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

5. Design optimization Table 6

Building schedule times and comfort temperatures [32].
The validated model was applied for the design optimization of the Building Comfort Temperature Range (◦ C) Schedule time
system. Three types of scenarios were considered to simulate office,
Service 20–25 24 h
domestic, and service buildings. The control strategy was defined to Office 20–24 8 am–4 pm
maintain the temperature at comfort ranges. The design optimization Domestic 20–24 6 pm–12 am
was formulated to minimize the cost of the system while a control
strategy was applied. The optimal design was obtained for the weather
conditions of the cold season in Auckland, which was averaged over four Zealand cold season from May until the end of August.
years. The data flow between the optimization and simulation envi­ The first two terms in equation 7 are PCM and solar collector capital
ronments can be found in Fig. 13. costs. The third term is the fan usage cost which is the cost of circulating
the air in the PCM heat exchanger.
5.1. Control system ∑
n
Mpcm Ppcm ASC PSC
OF = + + Efan Pele − Esave Pele
Lfpcm LfSC
The schedules for the three types of buildings are different. The
k=o

schedule represented the time of day that HVAC system needed to The last term is the saved energy cost through using the PCM heat
operate to maintain the space temperature at a comfort range. Table 6 exchanger and solar collection. The saved energy was calculated by
depicts the details of the schedules applied in the simulation. summating the heat energy supplied by the solar collector and the PCM
The upper and lower temperatures were set for the control system. heat exchanger.
According to the PCM, hut, and solar collector temperatures, the control The weight parameters were considered for obtaining the objective
logic forced air flow in different pathways to keep the hut temperature in function value annually. The last two terms were calculated for one cold
a specific range. The control logic is different from the control strategy in season which leads to an annual amount. To consider the annual amount
Section 2.1 since the backup heater is also added to the system. Fig. 14 for the first two terms, the capital cost of PCM and solar collectors was
shows the control algorithm. For each scenario, the control was applied divided by their lifetimes.
considering the period shown in Table 6. The design variables are the amount of PCM,Mpcm , and the area of the
A solar air collector, an electric heater, and a PCM heat exchanger collector,ASC , which are presented in detail in Table 7. The upper and
unit supplied heat to the hut. The air was circulated from the solar lower bounds for the constraints were specified based on our knowledge
collector to the hut so long as the hut temperature was within the from the hut’s experimental data. The lower bound for solar collector
comfort level and stopped once it reached the upper bound of thermal area is one, by assuming that solar collector is certainly a component of
comfort. On the other hand, the backup heater would be started when the system. Details of the parameters in the objective function are given
the hut temperature dropped below the lower bound of the thermal in Table 8. Esave is the total energy supplied by the solar collector and
comfort range. Hut received energy from the solar collector and stored it PCM heat exchanger.
in PCM so long as the hut temperature was within comfort level. The values were obtained from quotes of prices requested from
The energy stored in PCM was supplied the heat to the hut in the commercial vendors. Based on some of these large values, one may be
following cooler hours. The backup heater would be used when the hut skeptical, however, they were quoted from the manufacturers and it is
temperature dropped below the comfort level. Once the hut temperature out of the scope of this study to challenge them. It should be noted that
reached the upper bound of the thermal comfort, all energy sources were PPCM is the cost of the PCM material plus the heat exchanger body, which
stopped. If the hut temperature exceeded the upper bound of comfort was estimated per kilogram of PCM material on a commercial produc­
temperature and the PCM temperature was lower than the hut temper­ tion scale basis. The wholesale price of a PCM with a melting point of 23
ature, then the PCM heat exchanger units collected the extra heat from
◦
C was asked from a PCM manufacturing company in the USA. The
the hut through Pathway 3. Fig. 14 shows the flowchart of the control freight cost to New Zealand was also included in the price. The heat
strategy used in the hut. exchanger wholesale price was estimated by asking the heat exchanger
containers and body price from manufacturers. It should be mentioned
5.2. Objective function that the plastic containers were considered in the design that is more
economical than the aluminum containers used in the experiments.
The optimization problem minimizes the objective function in Efan is the electricity consumed for circulating the air through the
equation 7, which is the summation of the minimized cost over the New solar collector and PCM heat exchanger. It is calculated by equation:

Fig. 13. Data flow for design optimization.

11
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

Fig. 14. Control algorithm for heating the hut [33].

Table 7 fLρv2t
Δp = (9)
Optimization design variables details. dH
Symbol Design variable Type Domain Unit
Where f is the Darcy friction factor, L is the length of the plate containg
MPCM Amount of PCM Continuous [0, 100] Kg the PCM, ρ is the air density, vt is the air velocity and dH is the hydraulic
ASC Area of collector Continuous [1,5] m2
diameter. Δp changes by the amount of PCM (Mpcm ) as the number of
plates changes accordingly to fit the PCM amount, as per equation (10).

Table 8 N=
Mpcm
(10)
Objective function parameter details. Mpl
Symbol Parameter Value Unit In which N is the number of plates and Mpl is the mass of the PCM in
PPCM PCM storage price 15 USD/kg each plate.
PSC Solar collector price 800 USD/m2
Lf PCM PCM lifetime 100 Years 5.3. Optimization method
Lf SC Solar collector lifetime 30 Years
Pele Electricity price 0.22 USD/kwh
The objective function was minimized using fminsearch, in MATLAB.
This free derivative method was chosen as the hut model equations in
Δp EnenrgyPlus were not explicitly available. The objective function was
Q̇fan = Q̇air (8)
ηfan calculated by applying the control system to the simulation environment
given in Fig. 14. The simulation was carried out for the cold season
Q̇air is the air volume flow rate and is system overall pressure drop which begins in May and finishes at the end of August in Auckland. The
which is calculated using Darcy’s Law for parallel plates in the PCM heat ambient condition for this period was obtained by averaging four years
exchanger and the pressure drop equation as follows: of weather data.

12
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

5.4. Results and discussion Table 9

Optimum value of the design variables.
The computational time needed for finding the optimum points was Building type MPCM (kg) ASC (m2)
3 h using a Corei5-7600 CPU computer with 16 GB RAM. The optimum
Service 35 1
values of design variables for different scenarios are given in Table 9. Office 0 1
The optimum mass of PCM for service type building is higher than Domestic 20 1
the other types of building use since the bigger saved energy during the
day can be used at any time during the day. However, a lower amount of
PCM is needed in the domestic building type as the stored energy is only optimization will be investigated for a system controlled by a price-
required from 6 pm to 12 am. In the office scenario, the PCM heat based control to add the potential of PCM for peak load shifting.
exchanger is not needed as the scheduled time is from 8 am to 4 pm
when the solar collector can provide the heat directly to the hut. The Declaration of Competing Interest
solar collector area was one square meter (variable lower limit) for all
types of building use which is due to the high capital cost of the solar The authors declare that they have no known competing financial
collector. This shows that the solar collector capital cost is one of the interests or personal relationships that could have appeared to influence
barries toward inceasing the deployment of such systems. The optimum the work reported in this paper.
values also show that the system can be cost-effective if used as an
auxiliary heating supply for providing a portion of the total heating Data availability
demand.
Aside from schedule differences, the number of people in the hut for The authors are unable or have chosen not to specify which data has
various scenarios would vary. This may affect the internal load and the been used.
ventilation rate, which would consequently change the optimum design
variables. The hut occupancy in the EnergyPlus model was not taken References
into account because the validation data was gathered in an unoccupied
[1] International Energy Agency, Statistics, 2017.
situation. Additionally, including occupancy by adding the different [2] N. Soares, J.J. Costa, A.R. Gaspar, P. Santos, Review of passive PCM latent heat
numbers of people for the small hut of this study would not be realistic. thermal energy storage systems towards buildings’ energy efficiency, Energy Build.
EnergyPlus, on the other hand, allows users to specify the number of 59 (2013) 82–103.
[3] Energy Technology Perspectives – Executive Summary, 2017.
people and their activities for the hut model. [4] Dincer, Ibrahim, Marc A. Rosen, Thermal energy storage systems and applications,
Furthermore, the optimum values of the variables were obtained John Wiley & Sons, 2021.
considering a fixed price for electricity all day long. These values can [5] D. Mewes, F. Mayinger, Heat and cold storage with PCM: An up to date
introduction into basics and applications, 2008.
change for regions where the on-peak and off-peak electricity prices can [6] V.R. Pawar, S. Sobhansarbandi, Design optimization and heat transfer
be different. In that case, heaters can be used to charge the PCM heat enhancement of energy storage based solar thermal collector, Sustain. Energy
exchanger with a lower price and release that in cold on-peak hours with Technol. Assess. 46 (2021) 101260, https://doi.org/10.1016/j.seta.2021.101260.
[7] C.Q. Chen, Y.H. Diao, Y.H. Zhao, Z.Y. Wang, L. Liang, T.Y. Wang, Y. An,
a higher price. Optimization of phase change thermal storage units/devices with multichannel flat
Although the price of solar air collectors is unlikely to change, the tubes: A theoretical study, Renew. Energy 167 (2021) 700–717.
price of PCM is likely to fall as more commercial production is antici­ [8] W. Lin, Z. Ma, H. Ren, S. Gschwander, S. Wang, Multi-objective optimisation of
thermal energy storage using phase change materials for solar air systems, Renew.
pated in the future. This would have an impact on the objective func­ Energy 130 (2019) 1116–1129.
tion’s weight parameters as well as the optimum design values. [9] W. Lin, Z. Ma, S. Wang, M.I. Sohel, E. Lo Cascio, Experimental investigation and
two-level model-based optimisation of a solar photovoltaic thermal collector
coupled with phase change material thermal energy storage, Appl. Therm. Eng.
6. Conclusions 182 (2021) 116098, https://doi.org/10.1016/j.applthermaleng.2020.116098.
[10] W. Lin, Z. Ma, Using Taguchi-Fibonacci search method to optimize phase change
This paper presents the dynamic modeling and economical design of materials enhanced buildings with integrated solar photovoltaic thermal collectors,
Energy 106 (2016) 23–37.
a system, including an active PCM storage system and a solar collector in
[11] C. Arkar, S. Medved, Optimization of latent heat storage in solar air heating system
an office-sized hut. An integrated model was obtained by coupling with vacuum tube air solar collector, Sol. Energy 111 (2015) 10–20.
EnergyPlus with MATLAB via an EnergyPlus Co-simulation Toolbox [12] E. Markarian, F. Fazelpour, Multi-objective optimization of energy performance of
interface. Experimental validation of the integrated model proved the a building considering different configurations and types of PCM, Sol. Energy 191
(2019) 481–496.
reliability of the model in representing the dynamics of the systems. The [13] M. Zálešák, L. Klimeš, P. Charvát, Design optimization of a solar air collector
importance of accurate measurements of influencing variables such as integrating a phase change material, Chem. Eng. Trans. 81 (2020) 211–216.
solar radiation on the mismatch between the model and experimental [14] H. Ren, Z. Ma, W. Li, V.V. Tyagi, A.K. Pandey, Optimisation of a renewable cooling
and heating system using an integer-based genetic algorithm, response surface
data was discussed. The validated model was used for economical design method and life cycle analysis, Energy Convers. Manage. 230 (2021) 113797,
optimization of the system equipment. The highest amount of PCM mass https://doi.org/10.1016/j.enconman.2020.113797.
(35 kg) was obtained for a service type building use and the lowest [15] S. Nekoonam, R. Roshandel, Modeling and optimization of a multiple (cascading)
phase change material solar storage system, Therm. Sci. Eng. Prog. 23 (2021)
(zero) for office building use. This was due to the higher heating demand 100873, https://doi.org/10.1016/j.tsep.2021.100873.
for service building use (24 h) compared to office building use where [16] H. Ren, Y. Sun, A.K. Albdoor, V.V. Tyagi, A.K. Pandey, Z. Ma, Improving energy
heat is needed only during the daytime (8 am − 4 pm). However, the flexibility of a net-zero energy house using a solar-assisted air conditioning system
with thermal energy storage and demand-side management, Appl. Energy 285
optimum value of the solar collector area was the same at one square (2021) 116433, https://doi.org/10.1016/j.apenergy.2021.116433.
meter for all the building types due to the high price of solar collectors. [17] M. Fiorentini, P. Cooper, Z. Ma, Development and optimization of an innovative
This demonstrates that the capital cost of solar collectors is one of the HVAC system with integrated PVT and PCM thermal storage for a net-zero energy
retrofitted house, Energy Build. 94 (2015) 21–32.
barriers to the deployment of more of these systems. The optimum
[18] J.P. Arzamendia Lopez, F. Kuznik, D. Baillis, J. Virgone, Numerical modeling and
values also indicate that the system cannot supply all the heating de­ experimental validation of a PCM to air heat exchanger, Energy Build. 64 (2013)
mand of the hut. Therefore it cen be economic if used as an auxiliary 415–422.
heating demand providing a portion of the total heating demand in [19] H. Li, J. Li, X. Kong, H. Long, H. Yang, C. Yao, A novel solar thermal system
combining with active phase-change material heat storage wall (STS-APHSW):
combination with a heating supply system such as HVAC. The results Dynamic model, validation and thermal performance, Energy 201 (2020) 117610,
were obtained by considering a fixed electricity price, however PCM https://doi.org/10.1016/j.energy.2020.117610.
potential for peak load shifting could increase the system cost saving for
varying electricity prices during the day. In future research, design

13
M. Ebrahimpour et al. Applied Thermal Engineering 216 (2022) 119002

[20] G. Gholamibozanjani, M. Farid, Experimental and mathematical modeling of an [28] U.S. department of Energy, EnergyPlus documnetation, https://energyplus.net/
air-PCM heat exchanger operating under static and dynamic loads, Energy Build. documentation (accessed March 28, 2022).
202 (2019) 109354, https://doi.org/10.1016/j.enbuild.2019.109354. [29] P. Devaux, M.M. Farid, Benefits of PCM underfloor heating with PCM wallboards
[21] BRANZ, https://www.branz.co.nz/, (accessed March 28, 2022). for space heating in winter, Appl. Energy 191 (2017) 593–602.
[22] Elements, https://bigladdersoftware.com/projects/elements/, (accessed March 28, [30] Ashrae, Ashrae Handbook Fundamentals, and Ga Atlanta. “American society of
2022). Heating.” Refrigerating and Air-Conditioning Engineers 1 (2009).
[23] MetServices, https://www.metservice.com/, (accessed March 28, 2022). [31] J. Dostal, T. Baumelt, S.I. Tanabe, H. Zhang, J. Kurnitski, M.C. Gameiro da Silva,
[24] M.C. Kotti, A.A. Argiriou, A. Kazantzidis, Estimation of direct normal irradiance I. Nastase, P. Wargocki, G. Cao, L. Mazzarela, C. Inard, “Model predictive control
from measured global and corrected diffuse horizontal irradiance, Energy 70 for buildings with active one-pipe hydronic heating.” E3S web of conferences, EDP
(2014) 382–392. Sciences 111 (2019) 04050, https://doi.org/10.1051/e3sconf/201911104050.
[25] J.A. Duffie, W.A. Beckman, Solar Radiation, Solar Eng. Therm. Process. 4 (2013) [32] G. Gholamibozanjani, J. Tarragona, A.d. Gracia, C. Fernández, L.F. Cabeza, M.
3–42. M. Farid, Model predictive control strategy applied to different types of building
[26] B. Zalba, J.M. Marı́n, L.F. Cabeza, H. Mehling, Review on thermal energy storage for space heating, Appl. Energy 231 (2018) 959–971.
with phase change: materials, heat transfer analysis and applications, Appl. Therm. [33] G. Gholamibozanjani, M. Farid, A comparison between passive and active PCM
Eng. 23 (3) (2003) 251–283. systems applied to buildings, Renew. Energy 162 (2020) 112–123.
[27] Rubitherm Technologies GmbH, Macroencapsulation - CSM, (n.d.). https://www.
rubitherm.eu/en/index.php/productcategory/organische-pcm-rt (accessed march
28, 2022).

14