energies

Article
An Object-Oriented R744 Two-Phase Ejector
Reduced-Order Model for Dynamic Simulations
Michal Haida * , Rafal Fingas, Wojciech Szwajnoch, Jacek Smolka, Michal Palacz ,
Jakub Bodys and Andrzej J. Nowak
Institute of Thermal Technology, Silesian University of Technology, Konarskiego 22, 44-100 Gliwice, Poland;
rafal.fingas@gmail.com (R.F.); wojciech.szwajnoch@gmail.com (W.S.); jacek.smolka@polsl.pl (J.S.);
michal.palacz@polsl.pl (M.P.); jakub.bodys@polsl.pl (J.B.); andrzej.j.nowak@polsl.pl (A.J.N.)
* Correspondence: michal.haida@polsl.pl; Tel.: +48-237-2810




Received: 1 March 2019; Accepted: 27 March 2019; Published: 3 April 2019


Abstract: The object-oriented two-phase ejector hybrid reduced-order model (ROM) was developed
for dynamic simulation of the R744 refrigeration system. OpenModelica software was used to
evaluate the system’s performance. Moreover, the hybrid ROM results were compared to the results
given by the non-dimensional and one-dimensional mathematical approaches of the R744 two-phase
ejector. Accuracy of all three ejector models was defined through a validation procedure for the
experimental results. Finally, the dynamic simulation of the hybrid ROM ejector model integrated
with the R744 refrigeration system was presented based on the summer campaign at three different
climate zones: Mediterranean, South American and South Asian. The hybrid ROM obtained the
best prediction of ejector mass flow rates as compared with other ejector models under subcritical
and transcritical operating conditions. The dynamic simulations of the R744 ejector-based system
indicated the ejector efficiency variations and the best efficiency at the investigated climate zones.
The coefficient of performance (COP) varied from 2.5 to 4.0 according to different ambient conditions.
The pressure ratio of 1.15 allowed a more stabilised system during the test campaign with an ejector
efficiency from 20% to over 30%.

Keywords: R744; two-phase ejector; refrigeration; dynamic simulation; low-order model; object-oriented
modelling

1. Introduction
In all highly developed societies, energy conversion processes need to be carried out with the
highest possible energy efficiency together with a reduction in energy consumption but also with great
concern for the environment. This approach is nowadays known as so-called “good industrial practice”,
and it is utilised in all branches of industry. It is supported and, in many situations, even enforced
by the European Commission Regulations. As a typical example, the issue of high energy efficiency
of all refrigeration cycles is combined with the reduction of greenhouse gas emissions. Succeeding
regulations regarding environmental protection, for instance, as reported in Kigali Amended [1],
are very strict regarding the use of synthetic coolants, which are harmful for the environment. One of
the solutions is to use natural refrigerants due to their negligible environmental impact and high market
availability [2]. This has led many researchers to examine the refrigeration cycles with natural cooling
media, e.g., carbon dioxide (denoted as R744), which is featured with a Global Warming Potential
(GWP) equal to 1 and an Ozone Depletion Potential (ODP) equal to 0; it is also non-flammable and
non-toxic refrigerant.
The first research on R744 as a working fluid in refrigeration cycle, which was conducted by
Lorentzen [3], showed great potential for application, although the author pointed out a reduction in

Energies 2019, 12, 1282; doi:10.3390/en12071282 www.mdpi.com/journal/energies

Energies 2019, 12, 1282 2 of 24

cooling capacity at elevated temperatures as a result of the low critical temperature of 31.1 ◦ C and high
critical pressure of 73.8 bar. According to Goodarzi et al. [4], the simple vapour compression R744 cycle
results in a much smaller coefficient of performance (COP) in comparison to the synthetic equivalents
working at similar temperature variations. This effect is connected to high thermodynamic losses
during the expansion process due to the transcritical nature of the R744 vapour compression cycle.
Therefore, a modification of the R744 system configuration to improve the system’s energy performance,
especially during transcritical operation, is needed [5]. Moreover, well-designed components installed
in the R744 vapour compression cycle influence the COP improvement [6].
COP improvement was attained through implementation of the liquid separator and flash gas
bypass valve to the R744 refrigeration system. The R744 booster system is commonly installed in cold
and moderate climate zones, especially in the Scandinavian region [7]. Further energy performance
improvement can be achieved through application of the parallel compressor, which sucks fluid
directly from the liquid separator [8]. This system reduces the electric power consumption by size
reduction of the medium temperature (MT) compressor during utilisation in the summer season in
hot climates [6]. Sharma et al. [9] stated that the R744 parallel compression system obtains higher
COP when compared to the conventional R410A direct expansion system in the northern and central
parts of the United States of America. The COP improvement of the R744 parallel compression system
was also observed for a heat pump application [10]. Gullo et al. [11] stated that the R744 “all-in-one”
transcritical system with mechanical subcooling can reduce the system inefficiencies by up to 59%.
However, the high thermodynamic losses during expansion process in the high-pressure expansion
valve limit the possible COP improvement. Hence, one of the solutions is to use the ejector technology
as the main expansion device and to recover potential expansion work [12].
Gullo and Cortella [13] concluded that the use of a system with a two-phase ejector as
the main expansion device is the most efficient choice for upgrading the system’s performance.
Catalán-Gil et al. [14] confirmed COP improvement of the R744 transcritical system equipped with
the gas ejector up to 29.5% when compared to the standard booster system. The main objective of the
ejector application is to recover the pressure-related work of the supersonic motive flow and convert it
to kinetic energy in order to exchange the momentum with the entrained low pressure suction stream.
The consequence is that the pressure at the outlet is higher than the pressure of the suction stream,
which results in higher pressure at the compressor inlet and thus less compression work is required as
compared with the standard booster type cycle. Complex phenomena appear in the R744 two-phase
ejector such as supersonic flow in the converging-diverging nozzle, momentum transfer in the mixing
section, and pressure increment of the mixed flow in the diffuser, necessitating the evaluation of the
ejector performance using a more advanced mathematical approach [15].
Kornhauser [16] was the first to present a 1D model for a synthetic medium system with the
ejector to recover the expansion work. He showed a COP improvement of 21% over the conventional
cycle with the expansion valve for different refrigerants. Li and Groll [17] performed a numerical
investigation of the R744 ejector-based system and obtained a significant COP improvement compared
to the standard direct expansion system or R744 booster system with the flash gas bypass (FGB)
concept. The authors stated that the COP improvement was more than 16% when compared to
the basic transcritical R744 cycle for air-conditioning operating conditions. A 0D model was also
implemented for the dynamic simulation of the R744 ejector-based refrigeration system by Richter [18].
However, that model assumes the efficiency of the ejector given by Elbel and Hrnjak [12], resulting in
low accuracy at the wide ranges found in supermarket applications.
Sumeru et al. [19] showed in their paper that the COP improvement of R744 transcritical
ejector based system was up to 55% for thermodynamic analyses and up to 20% for experimental
investigations. In an experimental study, Elbel and Hrnjak [12] obtained a cooling capacity and COP
improvement of up to 8% and 7%, respectively. This discrepancy between theoretical and experimental
COP improvements is caused by the idealisation of the refrigeration components, especially the
two-phase ejector. Moreover, the ejector models used for such computations do not take the complex
Energies 2019, 12, 1282 3 of 24

ejector geometry and local flow phenomena inside the device into consideration. This suggests a
need for conducting more advanced numerical analyses, such as computational fluid dynamics (CFD)
models, to allow for the evaluation of the flow behaviour inside the ejector.
The commonly used approach in the numerical investigation is the homogeneous equilibrium
model (HEM), assuming mechanical and thermodynamic equilibrium between both phases inside
the ejector [20]. This methodology is mostly derived from the concepts of Kornhauser [16].
Smolka et al. [21] used the homogeneous real fluid approach and applied an energy equation expressed
in terms of the specific enthalpy instead of the standard temperature basis. The equation was employed
in the Ansys Fluent solver utilising the user defined functions (UDF). In this case, the error margin
of CFD results validated against real test data was kept below the level of 10% for most of the
operating points above the critical point. Similar work was done by the previously mentioned study
by Lucas et al. [20], but the authors used an OpenFOAM solver instead of commercial Ansys Fluent.
Unfortunately, the HEM featured increasing inaccuracy with a decreasing motive nozzle temperature
and a decreasing distance to the saturation line.
To extend a regime in which high quality results are obtained, another approach called the homogeneous
relaxation model (HRM) was introduced. The idea was proposed by Bilicki et al. [22], followed later by
comparison of the HEM, HRM results and experimental data provided by Downar-Zapolski et al. [23].
The aforementioned model evaluates the metastable effect during the expansion process by an
additional vapour mass balance governing equation and the semi-empiric relaxation time (RT)
definition [22]. Angielczyk et al. [24] investigated the HRM for the CO2 supersonic two-phase
flow through the ejector motive nozzle and presented a novel correlation for RT by entering
additional information, such as temperature and quality profiles as well as the critical mass flow
rates (MFR). Colarossi et al. [25] applied the HRM for R744 condensing two-phase ejector simulations
for improvement in relation to the HEM in terms of accuracy and a wider operation regime of high
quality results. Palacz et al. [26] validated the HRM CFD model of the R744 two-phase ejector with RT
defined in [24]. The authors indicated that both nozzles had a mass flow rates accuracy improvement
of 5% compared with the HEM. Haida et al. [27] modified the HRM by searching for RT coefficients
based on the genetic algorithm optimisation procedure. The authors stated that the modified HRM
extended the application range of the CFD model in the subcritical region to motive nozzle pressures
above 59 bar.
However, the very time-consuming nature of CFD simulations for single operating points makes
these models unfavourable for implementation into system analysis. Under these circumstances,
many researchers use the 0D or 1D mathematical model of the two-phase ejector. The implementation
of 0D/1D ejector models to system simulations allows system performance evaluation due to several
assumptions of the simplified ejector model. Therefore, the dynamic simulations of the R744
ejector-based refrigeration system were performed either based on the assumptions of the two-phase
ejector, such as fixed ejector efficiency [28] or accepting constant efficiencies of ejector parts for selected
operating regimes [29]. The aforementioned assumptions may cause high discrepancy in theoretical
predictions of the motive and suction mass flow rates when compared with the results of experimental
tests and more advanced CFD approaches.
As already mentioned, the CFD numerical approach and experimental data enable more accurate
evaluation of ejector performance than 0D or 1D models. However, those solutions are not suitable,
particularly for dynamic calculations, because of their high computational cost. Hence, there is still a
need to develop more versatile computational tools for steady-state and dynamic system analysis of
ejector units. One solution is the implementation of a low-order fast and accurate model based on CFD
results combined with the experimental data of the system analysis. One of these approaches is the
reduced-order model (ROM), proposed by Haida et al. [30], which was developed based on the proper
orthogonal decomposition (POD) approximation together with the radial basis interpolation functions
(RBF) [31]. Haida et al. [30] showed that the POD-RBF results in mass flow rates with ±10% accuracy
compared to the experimental data. It was developed on the basis of a 2D axisymmetric CFD model
Energies 2019, 12, 1282 4 of 24

proposed by Smolka et al. [21], which implements an ejector in the R744 transcritical cycle to give a
wide range of operating conditions and a cooling capacity. The single case computational time of ROM
was shown to be below 0.05 s, which allows its utilisation in dynamic simulations. The low-order model
of the R744 two-phase ejector was also enhanced by hybrid combination of the CFD results together
with the experimental data by Haida et al. [32]. The hybrid ROM maintained high accuracy and fast
simulations at a very wide operational envelope for the R744 Heat, Ventilation, Air-Conditioning
and Refrigeration (HVAC&R)supermarket system. The discussed model was used to generate maps
of ejectors performance installed in the multi-ejector module [33] for the R744 supermarket system.
Therefore, the implementation of ROM in the system analyses of the R744 ejector-based refrigeration
cycles, being a novel approach for non-commercial applications, can be very valuable in terms of
the accuracy of computational results in comparison to commonly used methodology, without any
negative effect on the computational time.
The main aim of this study was to integrate the two-phase ejector hybrid ROM with the R744
refrigeration system based on object-oriented modelling to produce a dynamic system simulation.
First, the three mathematical models of the R744 two-phase ejector, i.e., 0D model, 1D model and the
hybrid ROM model, are briefly discussed and confront. Then, the accuracy of all these three ejector
models was evaluated by comparing selected numerical results with the experimental measurements.
In the next section, these three models are applied into system analysis of the refrigeration cycle to
determine its COP. Finally, dynamic simulation of the hybrid ROM ejector model integrated with the
R744 refrigeration system was presented based on the summer campaign in three different climate
zones. In all system analyses carried out in this work, the OpenModelica (OM) software was utilised.

2. The R744 Two-Phase Mathematical Approaches

In this section, all the mathematical approaches for ejector performance evaluations used in this
paper are given. First, the 0D model proposed by Richter [18] is presented with the ejector efficiency
definition given by Elbel and Hrnjak [12]. Then, the 1D model based on the Kornhauser approach [16]
is described. It comprises a set of equations used to calculate the thermodynamic conditions of stream
at the ejector outlet. The third method is the hybrid ROM which uses a POD-RBF approach built on
the CFD results and experimental data to find the motive and suction nozzles mass flow rates for a
given set of input parameters.
The R744 two-phase ejector geometric assembly is shown in Figure 1. It can be seen that the
ejector consists of a converging-diverging motive nozzle, a converging suction nozzle, a pre-mixer with
converging cross-section, a mixer with fixed cross-section and a diffuser. The designed fixed ejector
consists of part of the the multi-ejector module that was experimentally tested by Banasiak et al. [33].
The multi-ejector module contains four R744 vapour fixed ejectors of different ejector capacity (changed
in a binary order 1:2:4:8) to adopt expansion performance for different cooling demands and ambient
conditions. Such solution ensures high energy efficiency of the module [34]. The dimensions of the
investigated ejector are presented in Table 1.

Table 1. The main geometry parameters of the R744 two-phase ejector installed in the multi-ejector
module [33].

Parameter Name Unit Dimension

Motive nozzle inlet diameter mm 3.80
Motive nozzle throat diameter mm 1.41
Motive nozzle outlet diameter mm 1.58
Motive nozzle converging angle ◦ 30.00
Motive nozzle diverging angle ◦ 2.00
Diffuser outlet diameter mm 8.40
Diffuser angle ◦ 5.00
Energies 2019, 12, 1282 5 of 24

Figure 1. Geometry assembly of the R744 two-phase ejector.

Each of the two-phase ejector mathematical models was compared with the experimental data to
define the accuracy range in a wide operating regime. The relative difference was used to define the
discrepancy of the MFR and the mass entrainment ratio of the ejector models:

ṁmodel − ṁexp
δi = · 100%, (1)
ṁexp

where the subscript i denotes the motive nozzle, the suction nozzle or the mass entrainment ratio.

2.1. 0D Model Using the Bernoulli Equation

The performance of the first theoretical ejector model is a 0D iterative approach that assumes
the ejector is adiabatic of constant efficiency and that the mixing of the motive and suction flows is
at constant pressure, lower than that of outlet stream. The ejector efficiency was defined by Elbel
and Hrnjak [12] as the ratio of the recovered ejector expansion work rate to the maximum possible
expansion work rate recovery potential:

Ẇrec h( pOUT , sSN ) − h( pSN , sSN )

ηej = = χ· , (2)
Ẇrec,max h( pOUT , s MN ) − h( p MN , s MN )

where ηej is the ejector efficiency, Ẇ is the expansion work rate in W, h is the specific enthalpy in
J·kg−1 , p is the pressure in Pa and s is the specific entropy in J·kg−1 ·K−1 . The term χ defines the mass
entrainment ratio, which is the ratio of the suction nozzle MFR to the motive nozzle MFR:

ṁSN
χ= , (3)
ṁ MN

where ṁ is the mass flow rate of the motive nozzle (MN) and the suction nozzle (SN) in kg·s−1 .
The main stream flowing through the converging-diverging nozzle of the ejector accelerates due
to the Venturi effect and throttles into the two phase region. According to Richter [18], the mass flow
rate of the motive nozzle can be calculated from Bernoulli’s equation for single phase flow:
q
ṁ MN = Ae f f · 2 · ρ MN · ( p MN − pSN ), (4)
Energies 2019, 12, 1282 6 of 24

where ρ MN is the fluid density at the motive nozzle in kg·m−3 and Ae f f is the effective cross-section
area at the motive nozzle throat in m2 . The effective area was defined as the throat area in the
investigated ejector, as shown in Table 1. The suction nozzle MFR is calculated based on Equation (3).
According to the mass balance governing equation, the outlet mass flow rate is defined as

ṁOUT = ṁ MN + ṁSN . (5)

Finally, the specific enthalpy of stream at the outlet of the ejector is calculated from the following
energy balance:
ṁ MN · h MN + ṁSN · hSN = (ṁ MN + ṁSN ) · hOUT . (6)

The assumed constant ejector efficiency allows the R744 ejector-based system to be simulated
with a defined accuracy level at a specified application and climate zone. Hafner et al. [28] assumed
the efficiency of all investigated ejectors to be up to 20% based on the preliminary experimental
measurements carried out on the prototype ejector. Hence, a similar efficiency of 20% was used for the
investigation presented in this paper.

2.2. 1D Homogeneous Equilibrium Model

The simplified 1D model of the two-phase based on the Kornhauser approach is formulated based
on the following assumptions [16]:

• A negligible pressure drop in the gas cooler, evaporator and all connections;
• There is no heat loss to the environment from the system, except via heat rejection in the gas cooler;
• The liquid and vapour streams outflowing from the separator are saturated;
• The flows across expansion devices are isenthalpic;
• The compressor has a given isentropic efficiency;
• The evaporator has a given superheat degree, and the gas cooler has a given outlet temperature;
• The flow in the ejector is considered to be a 1D homogeneous equilibrium flow;
• The motive and suction streams enter the constant area mixing zone with the same pressure,
and there is no mixing between them before the inlet of the constant area mixing;
• The expansion efficiencies of the motive and suction streams as well as the efficiency of the ejector
diffuser are given constants.

Having specified the aforementioned assumptions, the set of equations describing the ejector
performance proposed by Li and Groll [17] is set up. Denoting the pressure at the mixing zone as p MN ,
the equations describing the ejector section before the inlet to the constant area mixing section (CAMS)
are given. The motive stream pressure drops from the motive nozzle inlet to the mixing section inlet as
it accelerates in the converging nozzle before entering the mixing section. Since this process is assumed
to be isentropic expansion, it is described by

s MNm,is = s MN , (7)

where s MNm,is is the specific entropy in J·kg−1 ·K−1 at the inlet to the CAMS. The specific enthalpy of the
motive stream before entering the mixing zone is determined from the relationship of the properties:

h MNm,is = f (s MNm,is , pbm ), (8)

where the subscript bm determines the beginning of the mixing section. The actual specific enthalpy of
the motive stream at the inlet of CAMS is calculated using the definition of expansion efficiency:

h MN − h MNm
ηm = . (9)
h MN − h MNm,is
Energies 2019, 12, 1282 7 of 24

The velocity of the motive stream at the inlet of the CAMS is determined by assuming the
conservation of energy across the expansion process:
q
u MNm = 2 · (h MN − h MNm ). (10)

The specific volume of the motive stream at the inlet of CAMS can be determined as a function of
the specific enthalpy and pressure:

νMNm = f (h MNm , pbm ), (11)

where ν is the specific volume in m3 ·kg−1 . The area occupied by the motive stream at the inlet of the
CAMS per unit total ejector flow rate is determined using the conservation of mass:
v MNm
a MNm =
. (12)
u MNm · 1 + χ

The set of equations used to determine the suction stream is similar to the one for the motive
stream shown in Equations (7)–(11). The formula for calculating the area occupied by the suction
stream at the inlet of CAMS per unit of the total ejector flow rate is given as
νSNm χ
aSNm = · . (13)
uSNm 1 + χ

In order to calculate the mixing section outlet conditions, the iteration loop is applied. By assuming
that the momentum conservation is satisfied for the mixing process and outlet pressure pmix as a
guessed value, the velocity of the stream at the mixing section outlet is calculated as follows:

1 χ
pbm · ( a MNm + aSNm ) + · u MNm + ·u = pmix · ( a MNm + aSNm ) + umix . (14)
1+χ 1 + χ SNm

The specific enthalpy of stream at the mixing section outlet is calculated by the following equation:
 
1
h MN + χ · hSN = (1 + χ) · hmix + · u2mix . (15)
2

The specific volume of mixing stream can be determined as a function of the specific enthalpy
and pressure:
vmix = f (hmix , pbm ). (16)

The mixing pressure is then iterated until the following equation has been satisfied:

( a MNm + aSNm ) · umix

= 1. (17)
vmix

The next section describes the calculations of the parameters describing the diffuser section of the
ejector. First, the entropy at diffuser outlet is calculated as a function of specific enthalpy and pressure
and set equal to the entropy of mixing stream leaving the mixing section:

smix = f (hmix , pbm ), (18)

sOUT,is = smix . (19)

The specific enthalpy at the diffuser outlet can be calculated by applying the rule of conservation
of energy across the ejector:


1 + χ · hOUT = h MN + χ · hSN . (20)
Energies 2019, 12, 1282 8 of 24

The isentropic specific enthalpy at the diffuser outlet is calculated from the diffuser efficiency:

hOUT,is − hmix
ηOUT = . (21)
hOUT − hmix

The diffuser outlet pressure and stream quality are then obtained from the following functions:

pOUT = f (hOUT,is , sOUT,is ), (22)

xOUT = f (hOUT,is , pOUT ), (23)

where x is the vapour quality. Moreover, the stream quality at the diffuser outlet and the mass
entrainment ratio must satisfy the following condition:

(1 + χ) · xOUT > 1. (24)

The 1D model proposed by Kornhauser [16] requires an iterative solution to reach the
condition (24). The motive nozzle MFR is calculated using Equation (4). In addition, information about
the thermodynamic efficiency of each ejector part has to be given for the ejector performance
calculations. In this paper, an efficiency of 0.8 was assumed for all ejector parts based on the
investigation presented Elbel and Hrnjak [12].

2.3. A Hybrid Reduced-Order Model

The numerical model of the ejector used in this paper is a hybrid ROM of the R744 two-phase
ejector taken from the scientific work done by Haida et al. [32]. The aforementioned model was
developed based on a POD approximation of the CFD results and experimental data. The hybrid
ROM evaluates the real ejector efficiency under very wide operating conditions for R744 HVAC&R
supermarket application within a short computation time while preserving a high level of accuracy.
The operational envelope is presented in Figure 2. The motive nozzle conditions presented in Figure 2a
were defined for a pressure range from 50 to 140 bar and for the temperature close to the saturation
line, above the critical temperature and for different subcooling degrees. The suction nozzle was
defined for different pressures in the range from 26 to 46 bar. Moreover, the different superheat degrees
were defined up to 15 K and a two-phase region was also considered at the quality of 0.8. Thus,
this model can be used for dynamic simulation of ejector-based refrigeration cycles in refrigeration,
air-conditioning and heat pump applications.

Figure 2. The operational envelope of the hybrid reduced-order model (ROM): (a) the motive nozzle
conditions; (b) the suction nozzle conditions.
Energies 2019, 12, 1282 9 of 24

The POD approach constructs the optimal approximation base spanning the set of N sampled
values of the two-phase flow parameters inside the ejector stored in a single vector called the
snapshot [31]. The snapshot vectors are thus related to the input parameters. The POD searches for the
orthogonal matrix Φ reconstructing the rectangular snapshot matrix U utilising a linear combination
of the snapshots. Moreover, the POD model requires an additional interpolation procedure to evaluate
the ejector behaviour continuously for different operating conditions. The radial basis interpolation
functions were applied for the investigated ROM model. In this study, the thin plate spline radial
function with a smoothness factor was employed. The implementation of RBF into the POD model
reduces the dimensionality of the ROM to the number of unknown parameters defined as the boundary
conditions of the CO2 two-phase ejector, listed below:

• Motive nozzle pressure,

• Motive nozzle specific enthalpy,
• Suction nozzle pressure,
• Suction nozzle specific enthalpy,
• Outlet pressure.

The snapshot generation for both experimental and CFD values was prepared in a similar way as
a set of motive and suction nozzle mass flow rates. More details about the POD-RBF approach can be
found in [30].
The CFD results used to build the hybrid ROM were performed using the modified HRM [27]
within the operating regime presented in Figure 2 with a defined pressure difference step of 1 bar and a
temperature difference step of 5 K. Therefore, the total number of the CFD points was 5380, which were
combined with approximately 200 experimental data points to generate the hybrid ROM. Then,
the hybrid ROM was validated for all two-phase ejectors installed in the multi-ejector module [32].
The use of experimental data with the CFD results to generate the hybrid ROM allowed the prediction
of the MFR of both nozzles with an accuracy level within ±1% at each validated point. The very high
accuracy of the hybrid ROM allowed the implementation of the aforementioned model into the R744
supermarket system simulations to evaluate the energy performance of the ejector-based system under
different operating conditions and cooling demands.

3. Object-Oriented Modelling of the R744 Transcritical System

The object-oriented model of the R744 transcritical system equipped with the two-phase ejector
allows dynamic simulation for performance evaluation and control strategy definition. Moreover,
the integration of different ejector models with the R744 cycle allows the sensitivity of the ejector model
to be observed with respect to modifications of dynamic system parameters. Hence, the validation
procedure of the object-oriented ejector models gives information about the accuracy of each
investigated model to select the proper model for system implementation.
The software used for this work was OM, which is the open source environment based on the Modelica
3.3 language for simulating, optimising and analysing complex systems for industrial and academic
usage [35]. Moreover, it allows the performance of steady-state as well as dynamic simulations of
constructed systems and the analysis of the system as a whole using the transfer behaviour between
components as opposed to other well-known equation solvers used for thermodynamic applications.
This approach is called object-oriented modelling which is a construction of complex systems based on
single objects with well-defined properties or transfer behaviour as a set of interacting and inter-related
objects. This kind of modelling allows the visualisation and analysis of complex systems; they can
also be modified and their construction can be changed in an easy way. Object-oriented designs are
also more maintainable. The CoolProp package [36] containing thermophysical properties of R744 was
used to evaluate the flow behaviour inside each system component.
The 0D model and 1D model of the R744 two-phase ejector were directly implemented to the
Modelica software due to the simplicity of both approaches. However, the more complex hybrid
Energies 2019, 12, 1282 10 of 24

ROM was compiled using C-code external functions. Each external function was defined to calculate
either the motive nozzle or suction nozzle mass flow rate given from the hybrid ROM under specified
operating conditions.
The mathematical approaches of the two-phase ejector were integrated to the R744 transcritical
refrigeration system to investigate the system model derivation and ejector work during annual
operation. The layout of the R744 ejector-based transcritical system is shown in Figure 3. The R744
loop was designed as a booster type system. The superheated vapour R744 stream from the evaporator
outlet is compressed in the semi-hermetic compressor to the discharge pressure level. The high-side
pressure is set according to the ambient conditions, either under subcritical or transcritical conditions.
The discharged flow outside the compressor rejects the heat in the gas cooler heat exchanger due
to the heat transfer between the refrigerant and auxiliary flow, i.e., the glycol–water mixture. Then,
the internal heat exchanger (IHX) allows heat transfer between the high-side pressure and intermediate
pressure to increase the sub-cooling degree under subcritical conditions and decrease the outlet
temperature under transcritical conditions. The sub-cooled flow is expanded either by the high
pressure electronic expansion valve (HPV) or by the two-phase ejector. Moreover, the ejector entrains
the vapour flow from the evaporator outlet and the mixed flow enters the liquid receiver together
with the throttled flow from the HPV. In the liquid receiver, the liquid saturated phase discharges
the evaporator, and the vapour phase of R744 flows through IHX to absorb heat from the high-side
pressure stream. The vapour outflowing from the IHX is expanded by the flash gas bypass and enters
the compression suction side. The R744 liquid outflowing from the liquid receiver is expanded by the
metering valve to the MT pressure and absorbs the heat in the evaporator. Finally, the R744 stream
outflowing from the evaporator experiences superheat conditions, which prevents any liquid droplets
forming in the compressor.

Figure 3. The P & ID diagram of the R744 transcritical cycle equipped with the two-phase ejector.
Energies 2019, 12, 1282 11 of 24

The system performance analysis was done based on the first-law analysis. Therefore, COP was
defined as the ratio of the heat absorbed in the evaporator to the internal work rate of the compressor:

Q̇ev
COP = , (25)
Ẇcomp

where COP is the coefficient of performance, Q̇ev is the heat rate absorbed in the evaporator in W and
Ẇcomp is the internal work rate of the compressor in W. The heat rate and the internal work rate are
defined as follows:
Q̇ev = ṁ11 · [h12 ( p12 , t11,sat + SH ) − h11 ( p11 , t11,sat )] , (26)

Ẇcomp = ṁ1 · [h2 ( p2 , t2 ) − h1 ( p1 , t1 )] , (27)

where SH is the superheat degree in K and the indexes are defined according to Figure 3. The specific
enthalpy at the compressor outlet was defined based on the isentropic efficiency in the following form:

h2s ( p2 , s1 ) − h1 ( p1 , t1 )
ηis,comp = . (28)
h2 ( p2 , t2 ) − h1 ( p1 , t1 )

The rate of heat in the gas cooler was calculated as follows:

Q̇ gc = ṁ2 · [h2 ( p2 , t2 ) − h3 ( p3 , t3 )] . (29)

The outlet conditions are defined either by the set of sub-cooling degrees or by a set of the
temperatures and pressures under transcritical conditions. The energy balance equation was defined
for IHX in the following form:
ṁ3 · (h3 − h4 ) = ṁ7 · (h8 − h7 ). (30)

All valves presented in Figure 3 were defined to control the expansion process and discharge the
evaporator as follows: s
pin − pout
ṁi = OD · k v · ρin · , (31)
ρin

where OD is the opening degree of the valve and k v is the valve flow coefficient in m3 ·s−1 . The liquid
receiver was evaluated based on the mass balance and energy balance equation:

dρ
ṁ5 + ṁ6 = ṁ10 + ṁ7 + V · , (32)
dτ
dρ
ṁ5 · h5 + ṁ6 · h6 = ṁ10 · h10 ( p10 , x = 0) + ṁ7 · h7 ( p7 , x = 1) + V · h · , (33)
dτ
dρ
where V is the volume of the liquid receiver in m3 , dτ determines the mass accumulation in the
dρ
liquid receiver in kg·s−1
and V · h · is the energy accumulation in W. Finally, the ejector work was
dτ
calculated using the ejector efficiency defined in Equation (2). Moreover, the ejector performance was
also evaluated based on the pressure ratio Π between the outlet and the suction nozzle:
pOUT
Π= . (34)
pSN

Figure 4 presents the flowchart of the computation procedure of the R744 transcritical system
at a single time step. The global input data is a weather data to evaluate system work at different
test campaigns, i.e., daily or annual. The object-oriented system solved the governing equations for
all considered components. The object-oriented ejector model is combined with the C-code hybrid
ROM using external functions. Hence, the motive nozzle and the suction nozzle mass flow rates were
reached by calling the hybrid ROM at defined input parameters. Based on the input parameters and
Energies 2019, 12, 1282 12 of 24

results given by the hybrid ROM, the ejector model computed continuously the ejector efficiency and
the mass entrainment ratio. Finally, the COP value of the investigated system was calculated.

Figure 4. The computational procedure flowchart of the R744 transcritical system at a single time step.

4. Comparison of the Investigated R744 Two-Phase Ejector Numerical Models

In order to compare all three investigated mathematical models of the R744 two-phase ejector,
the accuracy of the motive nozzle and the suction nozzle MFRs under different operating conditions
(that are typical for refrigeration and air-conditioning applications) are tested. The set of the
operating conditions is shown in Table 2. The investigated points were selected from an experimental
test campaign carried out at the R744 multi-ejector test rig at the SINTEF/NTNU laboratory in
Trondheim [33]. Moreover, the selected points were previously used to validate the HRM CFD model
in [26]. The motive nozzle conditions were defined from approximately 54 bar up to approximately
95 bar. Therefore, the accuracy comparison of the ejector models MFRs was defined for ejector
utilisation under subcritical conditions, close to the critical point, and transcritical conditions. Moreover,
the motive nozzle temperature varied from approximately 279 K to over 308 K. The suction nozzle
conditions were defined for refrigeration and air-conditioning applications. Hence, the suction nozzle
pressure varied from approximately 27 bar up to approximately 32 bar and the temperature was in the
range from approximately 273 K to below 280 K. The outlet pressure was from 32 bar to approximately
39 bar for the evaluation of ejector performance at different ratios of pressure between the outlet and
the suction nozzle.
Energies 2019, 12, 1282 13 of 24

Table 2. The set of operating conditions adapted from [26].

Motive Nozzle Inlet Suction Nozzle Inlet Outlet

ID pM N TM N pSN TSN pOUT
bar K bar K bar
#1 94.46 308.43 27.21 275.75 32.85
#2 86.04 304.48 27.32 273.61 32.90
#3 91.91 304.13 31.41 278.43 38.24
#4 87.86 301.55 31.55 278.66 38.29
#5 80.62 299.40 31.58 278.49 38.48
#6 78.45 301.71 31.72 278.86 38.28
#7 76.56 301.49 27.33 274.01 32.87
#8 75.79 301.22 28.17 275.73 36.80
#9 66.51 295.56 28.21 275.36 34.85
#10 66.62 295.53 27.87 274.93 32.88
#11 61.79 293.42 29.93 276.73 33.87
#12 59.27 291.58 29.14 277.44 34.83
#13 58.41 283.15 27.82 277.71 34.83
#14 53.93 279.48 27.30 278.85 34.23

All selected operating conditions are also presented in Figure 5 in a pressure-specific enthalpy
diagram of R744. In the same figure, the pressure ratio parameter is shown relative to the suction
nozzle pressure. The different sub-cooling degrees under subcritical conditions and the different
temperatures under transcritical conditions of the motive nozzle were set for validation. The pressure
ratio varied from 1.12 to 1.32 due to the pressure difference between the outlet and the suction nozzle.
In addition, different pressure ratios were reached at similar suction nozzle conditions, which were
strongly related to the value of the mass entrainment ratio.

Figure 5. Selected operating conditions: (a) the motive nozzle conditions on the R744 pressure-specific
diagram; (b) the pressure ratio in terms of the suction nozzle pressure.

Table 3 presents the comparison results of each two-phase ejector model together with the
experimental data. The relative differences between the motive nozzle and the suction nozzle MFRs
together with the mass entrainment ratio were set to evaluate the accuracy of the investigated ejector
models. According to the experimental data, the motive nozzle MFR varied from approximately
0.07 kg·s−1 to over 0.1 kg·s−1 and the suction nozzle MFR was in the range from approximately
0.01 kg·s−1 to 0.035 kg·s−1 . As the result of both MFRs, the mass entrainment ratio was below 0.42.
The 0D model using a constant value of ejector efficiency obtained a motive nozzle MFR
discrepancy from approximately −31% under subcritical conditions to 27% under transcritical
conditions. The suction nozzle MFR accuracy of the 0D model was in the range from approximately
Energies 2019, 12, 1282 14 of 24

−43% to 36.5%. As a result, for the MFR predictions of both nozzles, the relative difference of the mass
entertainment ratio varied from −42% to over 47%.
A similar accuracy was reached by the 1D model due to similar calculations of the motive nozzle
MFR using Equation (4). The 1D ejector model predicted the suction nozzle MFR with an accuracy in
the range from approximately −32% to below 60%. Moreover, the mass entrainment ratio discrepancy
of this model was from approximately −21% to over 72%.
The hybrid ROM obtained very high accuracy of the motive nozzle MFR below 1%, which
confirmed very good agreement of the results given by the hybrid ROM compared to the experimental
data. Very low discrepancy in the hybrid ROM was reached for the suction nozzle MFR and the mass
entrainment ratio for all investigated points.

Table 3. The results given by each mathematical model.

Experimental Data 0D Model 1D Model Hybrid ROM

ID ṁ M N ṁSN χ δṁ M N δṁSN δχ δṁ M N δṁSN δχ δṁ M N δṁSN δχ
kg·s−1 kg·s−1 - % % % % % % % % %
#1 0.084 0.035 0.417 27.0 5.8 −16.7 27.0 13.1 −11.0 0.9 1.0 1.0
#2 0.079 0.032 0.409 20.3 −12.5 −27.3 20.3 0.7 −16.3 0.8 0.5 0.7
#3 0.095 0.033 0.344 22.6 −15.3 −30.9 22.6 −0.7 −19.0 0.5 0.6 0.6
#4 0.097 0.032 0.326 18.1 −23.0 −34.8 18.1 −6.3 −20.7 0.4 0.2 0.3
#5 0.090 0.025 0.278 11.7 −29.0 −36.4 11.7 −0.6 −11.0 0.1 0.8 0.5
#6 0.073 0.026 0.349 9.9 −35.9 −41.6 9.9 −13.7 −21.5 0.5 0.1 0.3
#7 0.067 0.028 0.411 8.6 −33.1 −38.4 8.6 −13.5 −20.4 0.1 0.5 0.3
#8 0.067 0.011 0.166 6.6 −8.3 −14.0 6.6 44.2 35.3 1.0 0.9 1.0
#9 0.072 0.014 0.192 −11.9 −37.1 −28.6 −11.9 19.7 35.8 0.5 0.8 0.7
#10 0.072 0.022 0.304 −10.9 −42.7 −35.7 −10.9 −21.1 −11.4 0.6 0.5 0.6
#11 0.072 0.019 0.259 −26.8 −40.0 −18.0 −26.8 −32.4 −7.6 0.2 0.4 0.3
#12 0.076 0.009 0.116 −31.3 −26.5 7.1 −31.3 2.6 49.4 0.8 0.1 0.5
#13 0.103 0.007 0.064 −8.9 −11.1 −2.5 −8.9 36.4 49.7 0.1 0.5 0.3
#14 0.100 0.003 0.031 −7.4 36.5 47.4 −7.4 59.4 72.1 0.5 0.1 0.3

Figure 6 presents the relative difference between numerical result produced by each R744
two-phase ejector model and appropriate measurement under all selected operating conditions. The 0D
and 1D models reached a motive nozzle MFR relative difference within ±10% for operating conditions
close to the critical point (#5÷#8) and for a motive nozzle pressure below 60 bar at #13 and #14,
which is shown in Figure 6a. Moreover, both ejector models overestimated the motive nozzle MFR
under transcritical conditions and underestimated it under subcritical conditions. Hence, the higher
inaccuracy represented by the relative difference for the motive nozzle MFR by the 0D and 1D ejector
models was strongly related to the suction nozzle MFR and the mass entrainment ratio predictions.
The motive nozzle MFR relative difference shown in Figure 6a confirmed the better agreement of the
hybrid ROM. The relative difference for the mass entrainment ratio is shown in Figure 6b together with
the ejector efficiency given by the experimental data. The 0D model reached the best agreement close
to the ejector efficiency of 0.2 at #12 and #13 due to the assumed constant efficiency based on previous
literature [28]. However, the efficiency varied from approximately 0.14 to almost 0.35, which caused
underestimation of the mass entrainment ratio for ejector efficiencies above 20% and overestimation for
efficiencies below 20% for the 0D model. The 1D model obtained an underestimated mass entrainment
ratio under transcritical conditions (#1÷#7), and it also obtained a very low pressure ratio at #10 and
#11. The very high pressure ratio above 1.24 caused significant overestimation of the mass entrainment
ratio, and the inaccuracy was above 30% for #8 and #9 and for motive nozzle pressure below 60 bar
(#12÷#14). The hybrid ROM reached satisfactory agreement at each investigated point due to the
hybrid combination of the CFD results and the experimental data.
Energies 2019, 12, 1282 15 of 24

Figure 6. Relative difference between numerical result for each the R744 two-phase ejector mathematical
model and measurements: (a) the motive nozzle mass flow rate (MFR); (b) the mass entrainment ratio
together with the ejector efficiency given from the experimental data.

The comparison discussed above proved high accuracy of the hybrid ROM for a very wide
operating region. Therefore, the performance evaluation of the R744 two-phase ejector implemented in
the refrigeration system using the hybrid ROM allowed ejector efficiency values close to those recorded
during experimental tests under different ambient conditions.

5. The System Energy Performance Comparison Using Different Two-Phase Ejector Models
Figure 7 presents the results of the energy performance comparison of the R744 vapour
compression system. The comparison was done based on the experimental data given by
Banasiak et al. [33]. Two test series were carried out on the R744 system equipped with the
two-phase ejector presented in Table 1. The operating conditions are shown in Table 4. The system
performance calculations based on the hybrid ROM obtained similar COP values when compared to
the experimental data of approximately 2.54 for Test A and 3.3 for Test B. Moreover, the COP accuracy
of the hybrid ROM was within ±1% for both cases. The 1D ejector model obtained lower COP values
of approximately 2.25 and 2.75 when compared to the experimental data, respectively. Hence, the COP
accuracy of 1D model was over −10% for Test A and over −15% for Test B. Similar to the 1D ejector
model, the 0D model underestimated COP by up to 2.2. for Test A and approximately 2.73 for Test B.
As the results for COP were lower when compared to the experimental data, the COP accuracy of the
0D model was close to −15% and approximately −18% for Tests A and B, respectively. The very high
energy performance accuracy of the R744 system based on the hybrid ROM of the two-phase ejector
allowed a system performance evaluation under arbitrary ambient conditions and cooling demands
due to the wide operating regime of the hybrid ROM.

Table 4. Operating conditions of the R744 vapour compression system adapted from [33].

Parameter Test A Test B

Evaporation temperature, ◦C −5 8
Evaporator outlet superheat, K 10 10
Gas cooler outlet temperature, ◦ C 25 30
Liquid separator pressure, bar 34 35
Energies 2019, 12, 1282 16 of 24

Figure 7. The energy performance of the R744 system based on different two-phase ejector models
under the operating conditions given by Banasiak et al. [33]: (a) coefficient of performance (COP);
(b) COP accuracy.

6. Dynamic Simulations of the R744 Two-Phase Ejector Integrated With the Refrigeration System
The implementation of the object-oriented two-phase ejector hybrid ROM to the R744 refrigeration
system also allowed dynamic simulations of the ejector-based system for different ambient conditions
and cooling demands. Moreover, detailed information about the ejector performance during
continuous utilisation was observed. Dynamic investigation of the R744 refrigeration system equipped
with the two-phase ejector was completed in three different regions characterised by different ambient
temperature ranges: Mediterranean zone (measured in Naples, Italy), South America zone (measured
in Rio de Janeiro, Brazil) and South Asia zone (measured in New Delhi, India). The weather data were
obtained from the free open-source EnergyPlus platform [37]. The analysis was performed during
the summer campaign over a period of two months, which is equivalent to approximately 1488 h.
Moreover, the dynamic simulations were done with a time interval of 60 s to evaluate the reaction of
the system’s operating conditions to the ambient variation.
Figure 8 presents the ambient temperature and the motive nozzle pressure variations during the
investigated campaign at each selected climate region. The motive nozzle pressure was strongly related
to the gas cooler outlet pressure; therefore, the relationship between the ambient conditions and the
motive nozzle pressure controls the ejector and gas cooler utilisation. The ambient temperature in the
Mediterranean region shown in Figure 8a was in the range from approximately 15 ◦ C to below 34 ◦ C.
Hence, the ejector was utilised under subcritical conditions during the summer campaign. However,
the R744 refrigeration system reached a maximum pressure of 85 bar at two temperature peaks halfway
through the summer campaign. In Figure 8b, the ambient temperature is shown to vary from 20 ◦ C
to approximately 37 ◦ C for the South America region; this variation is smaller than the temperature
variation shown in Figure 8a. Hence, transcritical conditions were reached for approximately 50%
of time of the investigated campaign. Moreover, the lowest motive nozzle pressure of 60 bar for the
South America region was approximately 10 bar higher than the lowest pressure obtained in the
Mediterranean region. A highest ambient temperature was reached for a long period of time in the
South Asia region, as presented in Figure 8c. The temperature varied from 24 ◦ C to approximately
42 ◦ C. As a result of the high ambient conditions, the two-phase ejector was utilised for transcritical
conditions for most of the investigated time. Moreover, the motive nozzle pressure was in the range
from 65 bar to over 105 bar, which was strongly related to the system performance as well as the
ejector efficiency.
The variations in the motive nozzle conditions in terms of the varied ambient conditions were
strongly related to the ejector performance. Moreover, the poor control of the two-phase ejector caused
the ejector to have low efficiency, and the system control was far from the optimum COP. One of the
most important parameters during ejector operation is the pressure ratio parameter due to the liquid
Energies 2019, 12, 1282 17 of 24

receiver pressure control. Therefore, investigation of the ejector performance during the summer
campaign in different regions was done for three different pressure ratios. In addition, the R744
ejector-based system was utilised in a refrigeration application, and the MT evaporator pressure was
set to approximately 28 bar with the defined superheat degree of 15 K.

Figure 8. The relationship between the ambient conditions (left) and the gas cooler outlet pressure/motive
nozzle pressure (right) during the summer campaign in different regions: (a) Mediterranean zone;
(b) South America zone; (c) South Asia zone.

Figure 9 presents the ejector efficiency for different pressure ratios during the investigation in the
Mediterranean region. The lowest pressure ratio of 1.07 was linked to the lowest efficiency of below 0.2
due to the low potential to recover the expansion work. The small value of the pressure ratio caused
the ejector to be more stabilised and the efficiency variation was approximately 0.03. An increase
in the pressure ratio increased the oscillations of the ejector performance. However, the two-phase
ejector with the pressure ratio of 1.15 resulted in an ejector efficiency in the range from approximately
0.22 to below 0.31. The efficiency drop was reached after approximately 300 h as the effect of the
significant ambient temperature decreased up to 15 ◦ C. The highest efficiency of the two-phase ejector
Energies 2019, 12, 1282 18 of 24

of approximately 0.33 was obtained for pressure ratio of 1.22. However, the ejector efficiency dropped
down to 0.19 following a decrease in the motive nozzle pressure. Hence, the pressure ratio should
be related to the ambient conditions and the cooling demand to maintain the high efficiency during
temperature decrease.

Figure 9. The ejector efficiency at different pressure ratios during the summer campaign in the
Mediterranean region.

The performance of the R744 two-phase ejector during the summer campaign in the South
America region confirmed the possibility of controlling the ejector in an efficient way, which is shown
in Figure 10. Similar to the results presented in Figure 9, the ejector was utilised at the lowest pressure
ratio of 1.07 which resulted in the lowest efficiency found during the investigated campaign of below
0.2. Moreover, higher ambient temperatures above 35 ◦ C caused a higher efficiency drop to almost
0.1. The significant efficiency improvement was observed for higher pressure ratios. The two-phase
ejector with a pressure ratio of 1.15 reached an efficiency in the range from approximately 0.25 to
0.31. In addition, the ejector at a pressure ratio of 1.15 was more stable compared with the ejector at a
pressure ratio of 1.22. The highest ejector efficiency value of approximately 0.33 was obtained at the
highest pressure ratio of 1.22, although the ejector performance dropped down to approximately 0.24.
Therefore, the possibility to define the ejector operating conditions during continuous work allows
stable and high performance of the two-phase ejector to be achieved.
Figure 11 presents the ejector efficiency for different pressure ratios during the investigated
campaign in the South Asia region. The worst ejector performance was noticed for the pressure
ratio of 1.07, which agrees with the analyses performed for the Mediterranean region and South
America region. Moreover, the very high ambient temperature of above 40 ◦ C strongly influenced
the efficiency degradation for the pressure ratio of 1.07. As for the results for highly varied ambient
conditions, the two-phase ejector with the pressure ratio of 1.07 obtained an efficiency in the range
from approximately 0.07 to below 0.2. An increase in the pressure ratio of up to 1.15 significantly
improved the ejector efficiency. However, similar performance degradation was observed for very high
Energies 2019, 12, 1282 19 of 24

ambient temperatures for which the efficiency varied from approximately 0.17 to 0.31. The highest
value of the pressure ratio of 1.22 caused the best performance for most of the investigated time
during the summer campaign. The two-phase ejector reached efficiency in the range from 0.2 up
to 0.33. Further improvement of the ejector performance at very high ambient temperatures and a
motive nozzle pressure above 100 bar can be achieved by an increase in the pressure ratio. Hence,
the R744 two-phase ejector should be controlled to optimise the ejector efficiency by proper setting of
the pressure ratio for a specified application.

Figure 10. The ejector efficiency at different pressure ratios during the summer campaign in the South
America region.

The characteristics of the ejector performance during the summer campaign in different hot
climate zones are strongly related to the COP value of the system, especially during work in the
daytime. Hence, Figure 12 presents the COP comparison for the R744 ejector-based system at different
climate regions for 24 h during the summer campaign. The pressure ratio was set to 1.22 to achieve the
highest ejector efficiency. The R744 refrigeration system equipped with the two-phase ejector obtained
the highest COP values of above 3.5 in the Mediterranean region due to the subcritical operation for
most of the investigated time according to Figure 8a. The lowest COP during the day was observed in
the afternoon at approximately 4:00 p.m., whereas the highest peak of COP was reached at 4:00 a.m.
The R744 ejector-based system experienced a lower COP in the South America region due to the higher
daily ambient temperature, which is shown in Figure 8b. The significant COP drop below 3.0 was
observed at approximately 2:00 p.m. The lowest COP of the R744 refrigeration system was obtained
for the South Asia climate zone due to the very high ambient temperature above 40 ◦ C. Therefore,
the COP dropped from approximately 3.8 to less than 2.5, although the energy performance of the
R744 ejector-based system was higher between the second hour and sixth hour as a result of the ejector
performance. The significant difference in COP between the selected climate regions forced the control
process of the R744 refrigeration system to be improved to operate close to the optimum COP.
Energies 2019, 12, 1282 20 of 24

Figure 11. The ejector efficiency at different pressure ratios during the summer campaign in the South
Asia region.

Figure 12. The coefficient of performance for 24 h of the summer campaign in different hot climate zones.
Energies 2019, 12, 1282 21 of 24

7. Conclusions
The object-oriented hybrid ROM of the R744 two-phase ejector was developed for dynamic
simulations. The proposed model was compared with two simpler mathematical approaches of
the ejector that have commonly been used in the literature for refrigeration system investigations.
Moreover, a dynamic simulation of the R744 refrigeration system equipped with the two-phase ejector
for different climate zones was performed to evaluate the ejector and the system’s performance.
The object-oriented hybrid ROM was implemented in OpenModelica free software using the
external function built in C-code programming. The proposed model allowed the evaluation of
ejector performance under arbitrary operating conditions within the operational envelope defined
for the HVAC&R supermarket system. Apart from the hybrid ROM, the 0D ejector model assumed
a constant efficiency, and the 1D ejector model with the assumption of homogeneous equilibrium
two-phase fluid flow was implemented for the comparison of different mathematical approaches
to describe ejector performance. In comparison, procedure accuracy, represented by the relative
difference between numerical result produced by each two-phase ejector model and appropriate
experimental measurement, is determined. The 0D model obtained satisfactory prediction of both
nozzles’ MFRs only at values close to the assumed ejector efficiency. The more accurate 1D model
reached an accuracy level within ±10% for motive nozzle pressures close to the critical point. However,
very high discrepancy was observed for high pressure ratios. The hybrid ROM obtained the best
accuracy for all investigated points due to the hybrid combination of the experimental data with the
CFD results that was used to build the ROM basis. Therefore, the integration of the hybrid ROM
allowed the evaluation of the R744 refrigeration system equipped with the designed two-phase ejector.
The investigation of the ejector performance during the summer campaign demonstrates the
variation in the work of the ejector. Three different climate zones were selected for investigation:
a Mediterranean region (Italy), a South American region (Brazil), and a South Asian region (India).
The R744 two-phase ejector was mostly utilised under subcritical conditions in the Mediterranean
climate zone. Transcritical conditions were achieved for most of the summer campaign for zones
with higher ambient temperatures in South America above 30 ◦ C and in South Asia above 40 ◦ C.
The ejector efficiency is strongly related to the pressure ratio, especially during conditions of significant
temperature difference. An increase in the pressure ratio allowed a high efficiency of above 0.2 up
to 0.35 to be maintained. However, the pressure ratio should be well controlled to achieve an ejector
performance close to the optimum during system utilisation. According to the results, the pressure
ratio of 1.15 allowed the greatest stabilisation of the system.
The evaluation of energy performance of the R744 refrigeration system equipped with the
two-phase ejector confirmed the influence of the high ambient temperature on COP. The highest
COP value was obtained in the Mediterranean region as a result of the system operating in subcritical
conditions for most of the investigated time. An increase in the ambient temperature generally caused
the COP degradation. Hence, the lowest COP value was observed in the South Asia climate zone.
Energy performance improvement for the R744 ejector-based system can be achieved by optimisation of
the liquid receiver pressure and outlet conditions of the gas cooler as well as the internal heat exchanger.

Author Contributions: Conceptualisation, M.H. and J.S.; methodology, M.H.; resources, M.H., W.S. and R.F;
investigation, M.H., W.S. and R.F.; validation, M.H., R.F., W.S., J.S., M.P., J.B. and A.J.N.; writing—original draft
preparation, M.H.; writing—review and editing, M.H., W.S., R.F., J.S., M.P., J.B. and A.J.N.; supervision, J.S.
and A.J.N.
Funding: The authors gratefully acknowledge the financial support of the National Science Centre through
project No. 2017/27/B/ST8/00945. Moreover, the work of M.H. was funded by the Rector’s research grant
No. 08/060/RGJ18/0157 provided by Silesian University of Technology.
Conflicts of Interest: The authors declare no conflict of interest.
Energies 2019, 12, 1282 22 of 24

Nomenclature ef f effective cross-section area

ev evaporator
a flow rate ratio, m2 ·kg−1 ·s−1
exp experimental data
h specific enthalpy, kJ· kg−1
fv flash valve
lr liquid receiver
kv valve flow coefficient, m3 · s−1
max maximal value
ṁ mass flow rate, kg· s−1
mix mixing section
p pressure, Pa
MN motive nozzle
s specific entropy, kJ· kg−1 · K−1
T temperature, K MNm motive stream at the inlet of CAMS
u stream velocity, m·s−1 OUT outlet
U snapshot basis matrix, - rec recovered ejector expansion work rate
ν specific volume, m3 ·kg−1 SN suction nozzle
Ẇ work rate, W SNm suction stream at the inlet of CAMS
x vapour quality, − v valve
Greek Symbols Abbreviations
χ mass entrainment ratio, − CAMS Constant Area Mixing Section
δ relative difference, % COP Coefficient of Performance
η ejector efficiency, % GWP Global Warming Potential
Π pressure ratio, − HEM Homogeneous Equilibrium Model
ρ density, kg· m−3 HPV High Pressure Valve
τ time, s HRM Homogeneous Relaxation Model
Subscripts I HX Internal Heat Exchanger
amb ambient MFR Mass Flow Rate
bm beginning of the mixing section ODP Ozone Depletion Potential
comp compressor ROM Reduced-Order Model

References
1. Briefing Note on. UN Environment Ozone Secretariat in Nairobi, Kenya: 2017. Ratification of the Kigali
Amendment. Available online: http://conf.montreal-protocol.org/meeting/oewg/oewg-39/presession/
briefingnotes/ratification_kigali.pdf (accessed on 27 September 2018).
2. ASHRAE. ANSI/ASHRAE Standard 34, Designation and Safety Classification of Refrigerants; ASHRAE: Atlanta,
GA, USA, 2016.
3. Lorentzen, G. Throttling, the Internal Haemorrhage of the Refrigeration Process; Institute of Refrigeration:
London, UK, 1983; Volume 80.
4. Goodarzi, M.; Gheibi, A.; Motamedian, M. Comparative analysis of an improved two-stage multi-inter-
cooling ejector-expansion trans-critical CO2 refrigeration cycle. Appl. Therm. Eng. 2015, 81, 58–65. [CrossRef]
5. Kim, M.H.; Pettersen, J.; Bullard, C.W. Fundamental process and system design issues in CO2 vapor
compression systems. Prog. Energy Combust. Sci. 2004, 30, 119–174. [CrossRef]
6. Bansal, P. A review—Status of CO2 as a low temperature refrigerant: Fundamentals and R&D opportunities.
Appl. Therm. Eng. 2012, 41, 18–29. [CrossRef]
7. Sawalha, S.; Karampour, M.; Rogstam, J. Field measurements of supermarket refrigeration systems. Part I:
Analysis of CO2 trans-critical refrigeration systems. Appl. Therm. Eng. 2015, 87, 633–647. [CrossRef]
8. Chesi, A.; Esposito, F.; Ferrara, G.; Ferrari, L. Experimental analysis of R744 parallel compression cycle.
Appl. Energy 2014, 135, 274–285. [CrossRef]
9. Sharma, V.; Fricke, B.; Bansal, P. Comparative analysis of various CO2 configurations in supermarket
refrigeration systems. Int. J. Refrig. Revue Int. Du Froid 2014, 46, 86–99. [CrossRef]
10. Rony, R.U.; Yang, H.; Krishnan, S.; Song, J. Recent Advances in Transcritical CO2 (R744) Heat Pump System:
A Review. Energies 2019, 12, 457. [CrossRef]
11. Gullo, P. Advanced Thermodynamic Analysis of a Transcritical R744 Booster Refrigerating Unit with
Dedicated Mechanical Subcooling. Energies 2018, 11, 3058. [CrossRef]
12. Elbel, S.; Hrnjak, P. Experimental validation of a prototype ejector designed to reduce throttling losses
encountered in transcritical R744 system operation. Int. J. Refrig.-Revue Int. Du Froid 2008, 31, 411–422.
[CrossRef]
Energies 2019, 12, 1282 23 of 24

13. Gullo, P.; Cortella, G. Comparative Exergoeconomic Analysis of Various Transcritical R744 Commercial
Refrigeration Systems. In Proceedings of the 29th International Conference on Efficiency, Cost, Optimisation,
Simulation and Environmental Impact of Energy Systems, Portoroz, Slovenia, 19–23 June 2016; pp. 19–23.
14. Catalán-Gil, J.; Sánchez, D.; Llopis, R.; Nebot-Andrés, L.; Cabello, R. Energy Evaluation of Multiple Stage
Commercial Refrigeration Architectures Adapted to F-Gas Regulation. Energies 2018, 11, 1915. [CrossRef]
15. Elbel, S.; Lawrence, N. Mathematical modeling and thermodynamic investigation of the use of two-phase
ejectors for work recovery and liquid recirculation in refrigeration cycles. Int. J. Refrig. 2015, 58, 41–52.
[CrossRef]
16. Kornhauser, A. The use of an ejector as a refrigerant expander. International Refrigeration and Air Conditioning
Conference; Purdue University, Purdue ePubs: West Lafayette, IN, USA, 1990; pp. 1–11.
17. Li, D.; Groll, E.A. Transcritical CO2 refrigeration cycle with ejector-expansion device. Int. J. Refrig.
2005, 28, 766–773. [CrossRef]
18. Richter, C. Proposal of New Object-Oriented Equation-Based Model Libraries For Thermodynamic Systems.
Ph.D. Thesis, Braunschweig University of Technology, Braunschweig, Germany, 2008.
19. Sumeru, K.; Nasution, H.; Ani, F.N. A review on two-phase ejector as an expansion device in vapor compression
refrigeration cycle. Renew. Sustain. Energy Rev. 2012, 16, 4927–4937. [CrossRef]
20. Lucas, C.; Rusche, H.; Schroeder, A.; Koehler, J. Numerical investigation of a two-phase CO2 ejector.
Int. J. Refrig. 2014, 43, 154–166. [CrossRef]
21. Smolka, J.; Bulinski, Z.; Fic, A.; Nowak, A.J.; Banasiak, K.; Hafner, A. A computational model of a transcritical
R744 ejector based on a homogeneous real fluid approach. Appl. Math. Model. 2013, 37, 1208–1224. [CrossRef]
22. Bilicki, Z.; Kestin, J. Physical Aspects of the Relaxation Model in Two-Phase Flow. Proc. R. Soc. Lond. A
Math. Phys. Eng. Sci. 1990, 428, 379–397. [CrossRef]
23. Downar-Zapolski, P.; Bilicki, Z.; Bolle, L.; Franco, J. The non-equilibrium relaxation model for one-dimensional
flashing liquid flow. Int. J. Multiph. Flow 1996, 22, 473–483. [CrossRef]
24. Angielczyk, W.; Bartosiewicz, Y.; Butrymowicz, D.; Seynhaeve, J.M. 1-D modeling of supersonic carbon
dioxide two-phase flow through ejector motive nozzle. In International Refrigeration and Air Conditioning
Conference; Purdue University, Purdue ePubs: Braunschweig, Germany, 2010; Volume 2362, pp. 1–8.
25. Colarossi, M.; Trask, N.; Schmidt, D.P.; Bergander, M.J. Multidimensional modeling of condensing two-phase
ejector flow. Int. J. Refrig. 2012, 35, 290–299. [CrossRef]
26. Palacz, M.; Haida, M.; Smolka, J.; Nowak, A.J.; Banasiak, K.; Hafner, A. HEM and HRM accuracy
comparison for the simulation of CO2 expansion in two-phase ejectors for supermarket refrigeration systems.
Appl. Therm. Eng. 2017. [CrossRef]
27. Haida, M.; Smolka, J.; Hafner, A.; Palacz, M.; Banasiak, K.; Nowak, A.J. Modified homogeneous relaxation
model for the R744 trans-critical flow in a two-phase ejector. Int. J. Refrig. 2018, 85, 314 – 333. [CrossRef]
28. Hafner, A.; Forsterling, S.; Banasiak, K. Multi-ejector concept for R-744 supermarket refrigeration. Int. J.
Refrig. Revue Int. Du Froid 2014, 43, 1–13. [CrossRef]
29. Liu, F.; Groll, E.A.; Li, D. Investigation on performance of variable geometry ejectors for CO2 refrigeration
cycles. Energy 2012, 45, 829–839. [CrossRef]
30. Haida, M.; Smolka, J.; Hafner, A.; Ostrowski, Z.; Palacz, M.; Nowak, A.J.; Banasiak, K. System model
derivation of the CO2 two-phase ejector based on the CFD-based reduced-order model. Energy 2018,
144, 941–956. [CrossRef]
31. Ostrowski, Z.; Białecki, R.A.; Kassab, A.J. Solving inverse heat conduction problems using trained POD-RBF
network inverse method. Inverse Probl. Sci. Eng. 2008, 16, 39–54. [CrossRef]
32. Haida, M.; Smolka, J.; Hafner, A.; Ostrowski, Z.; Palacz, M.; Madsen, K.B.; Försterling, S.; Nowak, A.J.;
Banasiak, K. Performance mapping of the R744 ejectors for refrigeration and air conditioning supermarket
application: A hybrid reduced-order model. Energy 2018, 153, 933–948. [CrossRef]
33. Banasiak, K.; Hafner, A.; Kriezi, E.E.; Madsen, K.B.; Birkelund, M.; Fredslund, K.; Olsson, R. Development
and performance mapping of a multi-ejector expansion work recovery pack for R744 vapour compression
units. Int. J. Refrig. 2015, 57, 265–276. [CrossRef]
34. Gullo, P.; Hafner, A.; Banasiak, K.; Minetto, S.; Kriezi, E.E. Multi-Ejector Concept: A Comprehensive Review
on its Latest Technological Developments. Energies 2019, 12, 406. [CrossRef]
35. Fritzson, P. Principles of Object-Oriented Modeling and Simulation with Modelica 3.3: A Cyber-Physical Approach,
2nd ed.; Wiley-IEEE Press: Piscataway Township, NJ, USA, 2015.
Energies 2019, 12, 1282 24 of 24

36. Bell, I.H.; Wronski, J.; Quoilin, S.; Lemort, V. Pure and Pseudo-pure Fluid Thermophysical Property
Evaluation and the Open-Source Thermophysical Property Library CoolProp. Ind. Eng. Chem. Res. 2014,
53, 2498–2508. [CrossRef] [PubMed]
37. EnergyPlus Weather Data. EnergyPlus, 2018. Available online: https://energyplus.net/ (accessed on
1 February 2019).

c 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).