Energy Conversion and Management 126 (2016) 228–246

Contents lists available at ScienceDirect

Energy Conversion and Management

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

Effect of non-condensable gas on heat transfer in steam turbine

condenser and modelling of ejector pump system by controlling the gas
extraction rate through extraction tubes
Dušan Strušnik a,⇑, Marjan Golob b, Jurij Avsec a
a
University of Maribor, Faculty of Energy Technology, Hočevarjev trg 1, SI-8270 Krško, Slovenia
b
University of Maribor, Faculty of Electrical Engineering and Computer Science, Smetanova ulica 17, SI-2000 Maribor, Slovenia

a r t i c l e i n f o a b s t r a c t

Article history: The paper describes the impact of non-condensable gas (NCG) on heat transfer in a steam turbine con-
Received 1 June 2016 denser (STC) and modelling of the steam ejector pump system (SEPS) by controlling the gas extraction
Received in revised form 20 July 2016 rate through extraction tubes. The ideal connection points for the NCG extraction from the STC are iden-
Accepted 31 July 2016
tified by analysing the impact of the NCG on the heat transfer and measuring the existing system at a
thermal power plant in Slovenia. A simulation model is designed using the Matlab software and
Simulink, Neural Net Work, Fuzzy Logic and Curve Fitting Toolboxes, to control gas extraction rate
Keywords:
through extraction tubes of the gas pumped from the STC, thus optimising the operation of the steam
Condenser
Ejector
ejector pump system (SEPS). The gas extraction rate from the STC is controlled in the extraction tubes
Fuzzy logic by pumping only the NCG to the maximum extent. The SEPS is optimised by selecting a Laval nozzle
Non-condensable gas of appropriate size to reduce the steam for the operation of the SEPS, whereby the amount of the
Operating principle extracted NCG is maintained. As the SEPS motive steam is produced in a boiler, the consumption of coal
Pumped gas for the production of the SEPS motive steam is reduced as well as the greenhouse gas environmental
pollution.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
A STC is an important subset of a condensing steam turbine. Its
main purpose is to maintain the prescribed vacuum condition of
around 0.01 MPa by evacuating exhaust gases from the steam tur-
bine. Exhaust gases are multiphase gases, comprising a condens-
able gas (CG) and a NCG. The CG includes dry and wet vapour.
Water vapour is removed from the STC through its condensation
and by pumping the condensate into the boiler feeding system.
NCG is evacuated by means of a SEPS. If no NCG evacuation takes
place from the STC, the condensation area in the STC would be
filled with the NCG and the condensation process would stop.
Fig. 1 shows the principle of exhaust gas evacuation from the
STC by means of the SEPS [1,2].
It is evident from Fig. 1 that cooling water from a nearby river is
used for the CG condensation. The condensate is collected at the
bottom of the STC and pumped into the boiler feeding system using
the condensate pump, while the NCG phase is pumped into the
⇑ Corresponding author.
E-mail address: dusan.strusnik@gmail.com (D. Strušnik).
atmosphere through the connection tube and using the SEPS. The
connection element is installed in the upper part of the STC.
SEPS are devices designed to use the pressure energy of a work-
ing fluid for the transport of another working fluid, whereby no
mechanical work is supplied or recovered. SEPS can be operated
with incompressible fluids (liquids), and in this application they
are normally referred to as jet pumps or educators. They are used
as vacuum compressors or vacuum pumps in order to produce vac-
uum in steam turbine systems, in refrigeration systems, for bulk
material transport etc. The actual efficiency is low, ranging from
0.1 to 0.35 [1,2]. The process is non-reversible due to mixing of
two flows. Some other authors also analysed the SEPS in the follow-
ing papers [3–7]. Water vapour from the turbine steam extraction 1,
previously expanded in the steam turbine, thus emitting part of its
energy, is used for the operation of the SEPS. The quality of water
vapour, travelling to steam extraction 1 of the steam turbine is reg-
ulated with pressure amounting to approx. 0.9 MPa and tempera-
ture to approx. 570 K. In our case, a SEPS is a two-stage flow-type
compressor of a primary and secondary stage [1,2]. In the primary
stage, i.e. the condensation stage, the pumped gas from the STC is
compressed at a pressure of approximately 0.01 MPa. A mixture
of the pumped gas from the STC and motive steam from the primary

http://dx.doi.org/10.1016/j.enconman.2016.07.082
0196-8904/Ó 2016 Elsevier Ltd. All rights reserved.
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 229

Nomenclature

Abbreviations mNCG share of non-condensable gas in pumped-out mixture,

ANN artificial neural network %
ANFIS adaptive neural fuzzy inference system


!
NCG;z condensable gas molar flux density at a distance of z,
B, C article kmol/(m2 s)
Cl close nNCG/A amount of non-condensable gas per surface area unit,
CG condensable gas kmol/m2
FLC fuzzy logic controller p pressure, MPa
F-Cl fast close pNCG-z partial pressure of non-condensable gas at a distance
F-Op fast open along the z axis, MPa
HP high pressure pNCG-1 partial pressure of non-condensable gas on the con-
LP low pressure densate layer surface, MPa
MAE mean absolute error pac actual pressure, MPa
MP middle pressure pex experimental pressure, MPa
MSE mean square error pAir partial air pressure, MPa
NCG non-condensable gas pH2O partial steam pressure, MPa
Op open ps saturation water vapour pressure, MPa
R2 correlation coefficient px, p1 pressure in the mixing section, MPa
RMSE root mean square error p0 inlet motive steam pressure, MPa
SCADA supervisory control and data acquisition p2 diffuser inlet mixed gas pressure, MPa
SEPS steam ejector pump system p3 exhaust ejector mixed gas pressure, MPa
STC steam turbine condenser p4 pumped gas pressure, MPa
pCG(298 K) saturation water vapour pressure at a temperature of
Parameters 298 K, MPa
A surface, m2 pCG(313 K) saturation water vapour pressure at a temperature of
Ai inner tube cross-section, m2 313 K, MPa
AL narrowest Laval nozzle cross section area, m2 qQt heat transfer difference of STC tube
At tube surface area, m2 qPgen generated power in the case of SEPS motive steam
A2 inlet diffuser cross sectional area, m2 expansion in the turbine, kW
c mixture molar density, kmol/m3 qmi mass flow of working fluid cross individual turbine
cpc cooling water specific conductivity, J/(kg K) component, kg/s
c1 steam speed at the exit from the Laval nozzle, m/s qmt amount of cooling water through a tube of the STC, kg/
c2 inlet diffuser mixed gas speed, m/s s
D molecular diffusivity, (MPa m2)/s qmnon-out mass flow of the pumped non-condensable gas from
DCG-NCG diffusion of condense gas in binary mixture, condense the STC, kg/s
gas in non-condensable gas, (MPa m2)/s qm0 motive steam mass flow through the Laval nozzle, kg/s
DCG-NCG,Tex diffusion of condensable gas in non-condensable gas qm4 pumped gas mass flow, kg/s
at Tex, (MPa m2)/s R gas constant, kJ/(kg K)
DCG-NCG,T2 diffusion of condensable gas in non-condensable gas R0 motive steam gas constant, kJ/(kg K)
at T2, (MPa m2)/s Rm universal gas constant, kJ/(kmol K)
DNCG-CG diffusion of non-condensable gas in a binary Rmix2 gas constant of the gas mixture in the mixing section,
mixture of condensable gas in non-condensable gas kJ/(kg K)
(MPa m2)/s rNCG-o outer radius of non-condensable gas layer around
D(pex,Tex) experimental value of diffusivity of component at pex tube, m
and Tex, (MPa m2)/s ri tube inner radius, m
D(p2,T2) diffusivity of a component at p2 and T2, (MPa m2)/s ro tube outer radius, m
Di tube inner diameter, m Tac actual temperature, K
Do tube outer diameter, m Tex experimental temperature, K
dTNCG temperature difference of condensate boundary layer Tm temperature of the mixture, K
due to non-condensable gas layer, K Tmean average temperature of extracted gases, K
dqQNCG-loss heat loss difference due to the non-condensable gas Tmix2 diffuser inlet mixed gas temperature, K
layer around the tube, W Tov temperature of tube outer wall, where the steam con-
dnNCG differential of non-condensable gas quantity, kmol denses, K
g gravitational acceleration, m/s2 Ts water vapour saturation temperature, K
h gas enthalpy, kJ/kg Tp-1 temperature of extracted gas of tube-i, K
h0 specific enthalpy of motive steam, kJ/kg T0 inlet motive steam temperature, K
h1 specific enthalpy of steam Laval nozzle expansion, kJ/ T2 corrected temperature of the binary mixture, K
kg Vi valve position of tube -i, %
h1s specific enthalpy of steam isentropic Laval nozzle v2 inlet diffuser specific volume of the mixed gas, m3/kg
expansion, kJ/kg ai heat transfer coefficient on tube inner side, W/(m2 K)
h3 diffuser outlet mixed gas specific enthalpy, kJ/kg ao heat transfer coefficient on tube outer side, W/(m2 K)
h30 isentropic specific enthalpy of diffuser outlet mixed Dhco water vapour steam condensation enthalpy, kJ/kg
gas, kJ/kg DTAir driving air temperature difference, K
h4 specific enthalpy of the pumped gas, kJ/kg DTc inlet-outlet cooling water temperature difference of
k heat transfer through a tube, W/(m2 K) the STC, K
230 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246
DTH2O driving water vapour temperature difference, K
DTln logarithmic mean temperature difference between
steam and cooling water in the STC (K),
DT1 temperature difference between steam and cooling
water on the outlet side of the STC, K
DT2 temperature difference between steam and cooling
water on the outlet side of the STC, K
Dhi difference of specific enthalpy of working fluid of indi-
vidual component, kJ/kg
Dz difference of distance in the direction of z axis from
boundary condensation surface, mm
wCG molar fraction difference of condensable gas, kmol
kNCG heat conductivity of non-condensable gas, W/(m K)
kc heat conductivity of cooling water, W/(m K)
kco heat conductivity of condensate layer, W/(m K)
kh heat conductivity of tube, W/(m K)
lc dynamic viscosity of cooling water, MPa s
lco dynamic viscosity of condensate layer, MPa s
Uc volume flow of cooling water through tube, m3/s
qc cooling water density, kg/m3
qco condensate density, kg/m3
ejector stage is led to the primary cooler. Most of the CG is con-
densed here and returned to the condenser through a special
barometer loop. The mixture remaining in the primary cooler after
the condensation is pumped at the steam pressure of the secondary
(atmospheric) stage and compressed to the pressure slightly higher
than the atmospheric pressure, then led, together with the motive
steam from the second stage, to the secondary cooler. The CG is also
condensed in this cooler. The condensate passes through the con-
densate pot to the STC and the residual NCG to the atmosphere.
Analysis of the NCG impact on condensation process and STC heat
transfer has been studied by many authors [8–11]. The results show
that the presence of even a small amount of NCG in steam substan-
tially reduces the STC heat transfer.
Artificial neural networks (ANN) can be used as an alternative to
analytical modelling approach. ANN is a model of a complex sys-
tem, where a large number of equal elements react to signals in
relation to one another. These are mathematical models, imitating
the structure and functionalities of biological neural networks [12].
The essential property of an ANN is that it is capable of finding
itself correlative dependence or regression between the input
and output values. When the ANN is created, it works in situations
not encountered during the creation process, which means that it
can resolve the tasks, where the solution does not exist, in the form
of a sequence of steps, such as in computer algorithms; neverthe-
less, there is a risk of unpredictable performance [13–17].
Fuzzy logic controller (FLC) is based on expert knowledge
expressed in terms of rules and can thus be employed to predict
the behaviour of many uncertain systems. The FLC advantage is
that it does not require knowledge of the underlying physical mod-
els as a precondition for its application. The FLC design is relatively
simple and is based on criteria correlation of input-output vari-
ables [13]. The correlation between the inputs-outputs is described
with If-then rules, allowing the implementation of complex non-
linear system in the manner in which people think [13]. The FLC
functioning and design are described in more detail by the authors
of the following papers [18–23].
Neural networks can learn from data. Opposite, FLC cannot
learn, but they are easy to understand. To improve the modelling
approach, hybrid intelligent systems that increase the capability
of learning and adapting automatically have been used by
researchers for many different purposes in a variety of engineering
systems. Among the many neural network system, one of the most
used and powerful is the Adaptive neural fuzzy inference system
Subscripts and Superscripts
d turbine component number
id turbine component iteration
i, j iterations
M Mach number
M2 Mach number in the mixing section
Nu Nusselt number
n number of learning data sets
o output value
Pr Prandl number
Re Reynold number
S pattern
t target value
z axis
gd diffuser isentropic efficiency
gLaval-is Laval nozzle isentropic efficiency
,0 motive steam specific heat ratio
,mix2 specific heat ratio of the gas mixture in the mixing sec-
tion
(ANFIS). ANFIS shows very good learning and prediction capabili-
ties, which makes it an efficient tool to deal with encountered
uncertainties in any system. It was used by researchers in various
engineering systems [24–27].
The objective of this investigation is to establish FLC and ANN
based model for simulation the operation of the SEPS, serving to
extract the NCG from the STC. The simulation model of the control
of the amount of the pumped gases through the extraction tubes
was designed by means ANN, FLC, Curve Fitting Toolbox, and x-
steam toolbox. The simulation model works so that the dimension
of the Laval cross-section of the Laval nozzle of the primary and
secondary stage of the SEPS changes until the required and actual
amount of the pumped gases become equal. The FLC calculation
unit is intended to control the open positions of control valves,
by means of input signals, i.e. temperatures of the pumped gas
through extraction tubes. Nonlinear ANN computes temperature
and pressure of water vapour used for the SEPS’s drive by means
of 3 input data items, i.e. temperature, pressure and water vapour
mass flow, entering the steam turbine, The ANN was designed by
using real process data from the thermal power plant.
2. Analysis of the NCG impact on condensation process and STC
heat transfer
The analysis of the NCG impact on condensation and the STC
heat transfer comprises two parts. In the first part, the NCG impact
on the condensation process of the isolated tube in the STC was
analysed, where diffusivity of CG through the layer of NCG and
CG binary mixture was examined. In the second part, the NCG
impact on heat transfer through the isolated pipe was analysed.
Using the results of the analyses and the measurements performed
of the existing system, a temperature profile of the STC was devel-
oped, from which ideal connections points were identified for
extraction of NCG from the STC. The measurements of the existing
system were obtained from the supervisory control and data acqui-
sition system (SCADA) [28] at a thermal power plant in Slovenia in
the 2013/2014 heating season.
2.1. Analysis of NCG impact on isolated tube condensation process in
the STC
In the condensation process, the temperature driving force
transports exhaust gases from a steam turbine to the cooled
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 231

Fig. 1. Principle of the steam turbine exhaust gas evacuation from the STC by means of the SEPS.

Fig. 2. Condensation of a steam turbine exhaust gases in a water-cooled STC.

surfaces of the STC, where NCG cannot condensate. As a conse-
quence, a layer of NCG is formed around the cooled surfaces, thus
reducing the heat transfer [9–11]. It is essential to continuously
remove NCG from the turbine condenser; otherwise, the STC
becomes saturated with NCG and the water vapour condensation
process stops. Fig. 2 shows the condensation of exhaust gases from
a steam turbine in a water-cooled STC, namely water vapour con-
densation in the presence of NCG and condensation of pure water
vapour [29].
The temperature profile (top r.h. diagram in Fig. 2) shows that
during the condensation of water vapour or CG in the presence
of NCG, the condensate temperature drops from 298 K to 283 K.
As a result, the temperature driving force DTAir between the con-
densate and cooling water is halved and the quantity of condensate
is reduced by the same amount. This means that for the condensa-
tion of the same quantity of water vapour, in the presence of NCG,
the amount of cooling water increases. The condensate becomes
undercooled and as such it is pumped into the boiler feedwater
system. This results in losses due to increased cooling water quan-
tity, losses due to reheating of the undercooled condensate and
losses due to a lower output of the steam turbine. The steam tur-
bine output is lower because the NCG increases pressure in the
232 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246

STC and water vapour in the steam turbine in unable to expand to p = pCG(313 K) = 7.375
10 3 MPa at a temperature of 298 K, which
the prescribed pressure level in the STC. This results in a lower out- is p = pCG(298 K) = 3.166
10 3 MPa. In the absence of NCG, the molar
put of the process [29]. flux density of CG is 3
10 4 kmol/(m2 s) [30]. In Table 1 of binary
The effect of NCG on water vapour condensation is described by system molar diffusivity a reading is taken for water vapour-air,
Fick’s law of constitutive diffusion of a substance [29]. Molecular namely DCG-NCG = 2.92
10 6 (MPa m2)/s at a temperature of
diffusion in a binary mixture is a consequence of uncontrolled 313 K, which means the water vapour diffusivity or CG into air or
movement of molecules [30–32]. Transmission of substances NCG. Using Eq. (1), the above value is corrected by the temperature
results from the changing of positions of molecules in a space difference T2 = ((313 K + 298 K)/2) = 305.5 K (corrected tempera-
rather than from collisions between the molecules. Collisions ture of the binary mixture) to obtain the calculation of actual value
between the molecules provide a basis for the transfer of heat, of diffusivity of CG into NCG:
energy and momentum, but they decelerate the transfer of sub-  74  7
stance. Molecular diffusion is more intense in gases than in liquids T2 6 305; 5 4
DCG NCG;T2 ¼ DCG NCG;Tex ¼ 2:92
10

and solids. The intensity of the phenomena of transfer in the flow T ex 313
of fluids and gases is also affected by turbulence [30]. The bases of ¼ 2:8
10 6 ðMPa
m2 Þ=s ð3Þ
a turbulent transfer are temporal fluctuation of speed in individual
places, resulting in eddy currents within fluids. When there is a where DCG-NCG,T2 is diffusivity of CG into NCG at a temperature of T2,
temperature gradient in the current of fluids, thermal energy is DCG-NCG,Tex is diffusivity of CG into NCG at a temperature of Tex from
also transferred in the same way with the swirling currents. Fick’s Table 1, Tex is temperature of a binary gas mixture for air-H2O pair
law states that a substance is diffused relative to a current of mix- from Table 1 and T2 is the corrected temperature of the binary mix-
ture in the direction of a decrease in concentration. The index of ture. Given that in our case the condensation is one-sided one-
Fick’s molecular diffusion of compound DCG-NCG, defines diffusion dimensional diffusion in the direction of the z axis, the equation
of CG into a binary mixture of CG and NCG [30,33,34]. Diffusivity of the Stefan’s law (Eq. (2)) is used to compute the density of the
DNCG-CG is defined by diffusion of NCG into a binary mixture of NCG flux and transformed, taking into account the boundary condi-
CG and NCG. Similarly, diffusion in a binary mixture was computed tions and integration, into the equation [30]:
by the authors in the following papers [35–37]. Table 1 shows the

! DCG NCG p
experimentally obtained values of molecular diffusivity in various
ln NCG z



NCG;z ¼ ð4Þ
binary gas mixtures and at various temperatures [30]. Rm
T m pNCG 1

Experimentally obtained values of molecular diffusivity apply where Rm is the universal gas constant, Tm is temperature of the
only to selected temperature and pressure conditions. The remain- mixture, pNCG-1 is partial NCG pressure on the surface of the con-
ing values are calculated using equation [30]: densate layer and pNCG-z is partial pressure of the NCG at a distance
 7 along the z axis (Fig. 2). Partial pressure of NCG as a function of the z
pex T ac 4
Dðp2;T2Þ ¼ Dðpex;TexÞ

ð1Þ coordinate z (Fig. 2) is described by equation [30]:
pac T ex

!


!
where D(p2,T2) is diffusivity of a component at actual pressure pac NCG;z
Rm
T c
pNCG z ¼ pNCG 1
exp ð5Þ
and actual temperature Tac in the STC, D(pex,Tex) is experimental DCG NCG
value of diffusivity of a component at pressure pex and temperature
In assessing the total amount of NCG per unit of surface nNCG/A,
Tex, pex/pac is the pressure ratio between the actual and experimen-
we derive from the gas equation in its differential form and the
tal value in the STC and Tac/Tex is the temperature ratio between the
integration is used for solving it [30]:
experimental and actual value in the STC.

! Z 1
Rm
T c
Z nNCG
To calculate the molar flux density of the component N CG;z ,



pNCG z ðzÞdz ¼ dnNCG ; ð6Þ
Stefan’s law of density of molar flux, one-sided diffusion and 0 A 0
one-dimensional geometric conditions in the direction of z axis
where A is surface and dnNCG is differential of the NCG amount. By
(Fig. 2), computed by means of [30]:
taking into account the pressure distribution the amount of NCG per

!


c
DCG NCG DwCG unit of surface is calculated [30]:
NCG;z ¼
ð2Þ
1 wCG Dz
!



!
N CG;z
Rm
T c

!
1
Z 1 Z 1
where NCG;z is molar flux density of CG at a distance along z axis


DCG NCG
nNCG pNCG 1
¼ pNCG z ðzÞdz ¼ exp ðzÞdz
(Fig. 2), c is molar density of a mixture, DCG-NCG is diffusivity of A Rm
T c 0 Rm
T c 0
CG into a binary mixture of CG and NCG, DwCG is molar fraction dif- ð7Þ
ference of CG and Dz difference of distance in the direction of z axis
from the boundary condensation surface (Fig. 2). Eqs. (6) and (7) are used to compute the trend of the partial
In our case, the tables or manuals for water vapour are used pressure of NCG and the change of the total amount of NCG per
to read for given conditions the values of pressure of water unit of surface nNCG/A, at different temperatures of the condensate
vapour saturation at a temperature of 313 K, which is boundary layer as a function of the distance along z axis (Fig. 2).
Table 2 shows the data used in the calculations.
Fig. 3 shows the trend of NCG partial pressure as a function of
the distance from the surface of the condensate layer at different
Table 1 temperatures of the boundary layer. It is evident from the Fig. 2
Fick’s molecular diffusivities in binary gas mixtures [30].
that the partial pressure of the NCG increases while the tempera-
System Temperature, Tex (K) Molecular diffusivity, D (MPa m2)/s ture of the condensate boundary layer decreases. In our case, the
Air-CO2 276.2 1.42
10 6 highest partial pressure of the NCG amounts to 6.148
10 3 MPa
Air-H2O 289.1 2.82
10 6
at a temperature of the condensate boundary layer of 283 K and
6
Air-H2O 313 2.92
10 decreases exponentially. At a temperature of 298 K, the partial
6
Air-Benzene 298.2 0.96
10
6 pressure of the NCG on the condensate boundary layer amounts
CO-N2 273.0 1.77
10
to 4.209
10 3 MPa and decreases to the temperature of the
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 233

Table 2
Data on water vapour condensation on the condensate boundary layer at different temperatures [30].

Condensate Steam saturation pressure Temperature of mixture, Diffusivity, Partial air pressure, Flux density,
temperature (K) p = ps (MPa) Tm (K) DCG-NCG (MPa m2)/s pNCG-1 (MPa) N CG;z (kmol/(m2 s))



!

3
313 7.375
10 313 2.92
10 6 0 3
10 4
3
311 6.674
10 312 2.9
10 6 0.701
10 3
2,8
10 4

3
308 5.622
10 310.5 2.88
10 6 1.753
10 3
2.4
10 4

3
303 4.241
10 308 2.84
10 6 3.134
10 3
1,9
10 4

3
298 3.166
10 305.5 2.8
10 6 4.209
10 3
1,5
10 4

3
293 2.337
10 303 2.76
10 6 5.038
10 3
1,1
10 4

3
288 1.704
10 300.5 2.72
10 6 5.671
10 3
0,7
10 4

3
1.227
10 298 2.68
10 6 6.148
10 3
0,4
10 4

x 10
-3 As it is impossible to carry out the pumping of the NCG precisely at
6 Condensate Layer Temperature 283 K the above indicated points, the suction pipe of the SEPS is con-
NCG Partial Pressure (MPa)

Condensate Layer Temperature 288 K

nected to the coolest points of the STC.
Condensate Layer Temperature 293 K
5
Condensate Layer Temperature 298 K
Condensate Layer Temperature 303 K
4 Condensate Layer Temperature 308 K
2.2. Analysis of the effect of the NCG on the heat transfer through the
Condensate Layer Temperature 311 K isolated tube of the STC
3

2 The condensation heat in the STC is released into the environ-

ment by means of cooling water flowing on the inner side of the
1
tube. The tubes in the STC are integrated into tube bundles. The
0 number of tubes in a tube bundle depends on the amount of the
0 5 10 15 20 25 30 35 40
exhaust steam from the steam turbine, the design and the tube
Distance From The Condensate Layer Surface (millimetres)
material. In our case, losses will be analysed in more detail, occur-
Fig. 3. Ratio between the partial pressure of the NCG and the distance from the ring when heat passes through an isolated tube of the STC. The
surface of the condensate layer at different temperatures of the condensate losses occurring when heat passes through the isolated tube result
boundary layer. from the heat resistance of the tube and the NCG layer around it.
The tube heat resistance depends on the tube material, which is
closely linked to the tube price. A layer of NCG around the tube
considerably reduces the heat transfer as it acts as an insulator.
-3
x 10
2.5
CG 13%, NCG 87% At 283 K
The impact of NCG on heat transfer through a tube was described
Quantity Of The NCG Per Unit

CG 23%, NCG 77% At 288 K

2 CG 37%, NCG 63% At 293 K in more detail by the authors in the following papers: [10,11,38–
CG 50%, NCG 50% At 298 K
CG 63%, NCG 37% At 303 K 41]. The heat transfer through the isolated STC tube is calculated
Area (kmol/m2)

1.5 CG 80%, NCG 20% At 308 K

CG 93%, NCG 7% At 311 K
using the equation:
CG 100% At 313 K
1
1 k¼ ð8Þ
ro
ai
ri þ kro
ln rro þ rNCG
kNCG
0

ln rNCG
ro
o
þ a1o
h i
0.5
where k is heat transfer through the tube, ro is outer tube radius, ai
0 is heat transfer coefficient on the inner side of the tube, ri is inner
0 5 10 15 20 25 30 35 40 45 50 tube radius, kh is conductivity of the tube, rNCG-o is outer radius of
Distance From The Condensate Layer Surface (millimetres)
the NCG layer around the tube, kNCG is heat conductivity of the
Fig. 4. Amount of the NCG per unit of surface at different concentrations of the NCG and ao is heat transfer coefficient on the external side of the
mixture and temperatures of the condensate boundary layer. tube. The following values were used in the calculations: heat con-
ductivity of the tube, brass kh = 120 W/(m K) and heat conductivity
of NCG, kNCG = 0.0265 W/(m K). The heat transfer coefficient on the
condensate boundary layer of 313 K, where no NCG is present on inner side of the tube ao is computed using the Dittus-Boelter equa-
the boundary layer. tion [10,11,38]:
Fig. 4 shows the concentration of the NCG per unit of surface
nNCG/A computed by means of Eq. (7). kc
Nu kc
0; 023
Re0;8
Pr0;4
ao ¼ ¼
Fig. 4 shows that the concentration of the NCG decreases as a Di Di
function of the distance of the condensate boundary layer, /c

D
q
0;8  0;4
i c
kc
0; 023
i l
cpkc
clc
A
whereby the decrease is the slowest when the share of the NCG c
in a binary mixture is the largest. In this case, the transfer of heat ¼ ð9Þ
Di
of the binary mixture is the lowest, as the concentration of the NCG
is the highest close to the condensate boundary layer that prevents where kc is heat conductivity of cooling water, Di is inner tube diam-
diffusion of the CG into the NCG layer. As the distance from the eter, Uc is volume flow of cooling water through the tube, Nu is the
boundary layer of the condensate layer increases, the amount of Nusselt number, Re is the Reynolds number, Pr is the Prandtl num-
the NCG per unit of surface decreases exponentially and after sev- ber, Ai is inner tube cross-section, qc is cooling water density, lc is
eral millimetres it becomes equal to the pressure in the STC. At the dynamic viscosity of cooling water, cpc is specific conductivity of
points where a phase transformation takes place of the physical cooling water. The following values were used in the calculations:
state of the CG (condensation), the concentration and the partial heat conductivity of cooling water kc = 0.598 W/(m K), volume flow
pressure of the NCG are the highest. The above points are the most of cooling water through the tube Uc = 5.444
10 3 m3/s, inner
appropriate connecting points for pumping the NCG from the STC. tube cross-section Ai = 3.4618
10 4 m2, cooling water density
234 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246

qc = 998 kg/m3, dynamic viscosity of cooling water lc = 1.003 -

10 9 MPa
s and specific conductivity of cooling water cpc = 4182 -
J/(kg K) [38]. The heat transfer coefficient on the external side of
the isolated tube ao in condensation of water vapour saturation,
where a condensation layer film is obtained, is calculated by means
of:
!0;25
g
Dhco
q2co
k3co
ao ¼ 0; 728

Do
lco
ðT s T ov
where Do is outer tube diameter, lco is dynamic viscosity of conden-
sate layer, Ts is water vapour saturation temperature, Tov is tube
outer wall temperature, where water vapour condensates, g is grav-
itational acceleration, Dhco is water vapour condensation enthalpy,
qco is condensate density and kco is condensate layer heat conduc-
tivity. The following values were used in the calculations: gravita-
tional acceleration g = 9.81 m/s2, water vapour condensation
enthalpy Dhco = 2.38
106 J/kg, condensate density qco = 992 kg/
m3, condensate layer heat conductivity kco = 0.627 W/(m K), con-
densate layer dynamic viscosity lco = 0.6531
10 9 MPa
s, water
vapour saturation temperature Ts = 312.5 K and tube outer wall
temperature where water vapour condensates Tov = 311 K. The heat
transfer of the isolated tube of the STC is calculated using:
" #
DT 1
qQ t ¼ k
At
DT ln ¼ k
At

ln DDTT 12
where qQt is heat transfer difference of the STC tube, At is tube sur-
face, DTln is logarithmic mean temperature difference between
water vapour and cooling water in the STC, DT1 is temperature dif-
ference between water vapour and cooling water on the outlet side
of the STC, DT2 is temperature difference between water vapour and
cooling water on the inlet side of the STC, qmt is amount of cooling
water through the STC tube, cpc is specific heat of cooling water, DTc
difference due to the NCG layer around the tube and kNCG is heat
conductivity of the NCG.
The STC heat power of the existing system in a thermal power
plant in Slovenia is taken into consideration in the analysis,
amounting to 2.38
107 W, where 10 kg/s of CG condenses on aver-
age from the steam turbine. The cooling water flow rate from the
cooling tube bundle of the STC is 258.7 kg/s. The cooling water
temperature at the inlet of the STC is 285 K and 307 K at the outlet.
ð10Þ The STC tube bundle comprises 4750 tubes. The tube inner diame-
ter is 0.021 m and the tube outer diameter 0.023 m, whereby the
tube length is 6.4 m. The analysis results show that even a small
layer of NCG around the STC tube considerably reduces the heat
transfer. Fig. 5 shows thermal transmittance and heat transfer
when the NCG layer around the STC tube changes.
Fig. 5 shows that the isolated tube of the STC has the highest
thermal transmission when the thickness of the NCG around the
tube is the smallest. By increasing the thickness of the NCG layer
around the tube, heat transmission and heat passing through the
tube are reduced as the NCG layer around the tube acts as an insu-
lator. If the NCG layer around the tube is 0.6
10 6 m, the heat
transfer through the tube is reduced by 60% and by 82% if the
NCG layer is 2
10 6 m. The NCG layer around the tube reduces
water vapour diffusion to the cooled surface and lowers the tem-
perature of the condensate boundary layer. Fig. 6 shows the tem-
perature of the air layer and condensate boundary layer when
DT 2
¼ qmt
cpc
DT c ð11Þ the thickness of the NCG layer around the STC tube increases.
Fig. 6 shows that the condensate boundary layer temperature is
the highest when there is no NCG layer around the STC tube and it
amounts to 313 K. As the thickness of the NCG layer around the
313.02
313
NCG Temperature (K)
Temperature (K)

312.98
Condensate Bundary Layer Temperature (K)
is inlet-outlet temperature difference of the cooling water of the 312.96
STC. The following values were used in the calculations: logarithmic 312.94
temperature difference of the STC DTln = 12.65 K and amount of
312.92
cooling water through the STC tube qmt = 0.054463 kg/s. The tem-
perature difference of the condensate boundary layer due to the 312.9

NCG layer around the isolated tube is calculated by means of [38]: 312.88

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

dqQ NCG loss x 10
-6
dT NCG ¼ ð12Þ
2
p
kNCG
r NCG o Thickness Of The NCG Layer Araund The Tube (metres)

where dTNCG is temperature difference of the condensate boundary Fig. 6. Temperature of the air layer and of the condensate boundary layer when the
layer due to the NCG layer around the tube, dqQNCG-loss is heat loss thickness of the NCG layer around the STC tube increases.

5000 900 Tube Heat Transfer (W/m2K)

Heat Through The Tube (W)

800
Heat Through The Tube (W)

4000
Heat Transfer (W/m2K)

700

600
3000
500

400
2000
300

1000 200

100

0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2

Thickness Of The NCG Layer x 10-6 Thickness Of The NCG Layer x 10-6
Araund The Tube (meters) Araund The Tube (meters)

Fig. 5. Thermal transmittance and heat transfer in case of increase of the thickness of the NCG layer around the STC tube.
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246
tube increases, the temperature of the condensate boundary layer
decreases. When the thickness of the NCG layer is 2
10 6 m, the
temperature of the boundary layer is reduced by 0.125 K to
312.875 K. The temperature variation of the air layer around the
tube is negligible.
The overall losses resulting from the heat transfer through the
STC tube are the losses arising from the thermal resistance of the
tube material and the losses arising from the NCG layer around
the tube. They are shown in Fig. 7. The largest share of the losses
of the STC is caused by the NCG layer around the tube, namely
90% of the total losses arising from the transfer of heat of the
STC. The remaining losses are the losses caused by the thermal
resistance of the material and cannot be avoided. The losses may
be reduced by choosing a more appropriate tube material, which
is linked to the costs. Any excessive losses arising from the NCG
layer around the tube are avoided by preventing the atmospheric
air from entering the STC and by pumping the NCG from the STC
in a more reliable manner.
The calculation shows that when the total amount of NCG
around the tube is 2.195
10 4 m3, the thermal power of the STC
is reduced by 20%. If the overall amount of the NCG around the
tube is 8.78
10 4 m3, the thermal power of the STC is reduced
by 50%. If, however, the total amount of the NCG around the tube
is 2.195
10 3 m3, the thermal power of the STC is reduced by as
much as 74%. The NCG pumping from the STC has to be ensured
even before the above gas reaches the tubular cooling surfaces.
The NCG concentration is the highest on the coldest places of the
STC. The coldest places of the STC were identified by means of an
16
14
12
Loss (W/K)
10
Totall Loss (W/K)
8 Losses Due To NCG Layer Araund Tube (W/K)
Losses Due To Thermal Resistance (W/K)
6 of the Laval cross-section of the Laval nozzle of the primary and
4 secondary stage of the SEPS changes until the required and actual
2 amount of the pumped gases become equal. The required amount
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
-6
x 10
Thickness Of The NCG Layer Araund The Tube (metres)
Fig. 7. STC losses when the thickness of the NCG layer around the tube increases.
Fig. 8. Temperature profile of the STC and ideal connecting locations for NCG pumping.
235
analysis of the temperature profile of the existing STC in a thermal
power plant in Slovenia, shown in Fig. 8. The circles numbered
from 1 to 3 in Fig. 8 show the ideal connecting points for the
NCG pumping. Some other authors analysed various systems using
a temperature profile [42–45].
The connecting points 1, 2 and 3 are located so as to create a
complete circle of connections used for gas pumping. Each con-
necting point has 3 connections. Two connections are on the exter-
nal side of the cooling tube bundle and one is in the central part of
the cooling tube bundle. Fig. 9 shows a schematic diagram of the
control of the quantity of pumped gases through the extraction
tubes. The connecting locations interconnect the extraction tubes
for STC gas pumping. The extraction tubes are fitted with 3 control
valves to control the amount of the pumped gas depending on the
temperature of the pumped gas. The amount of the pumped gas
increases through the extraction tubes, where the pumped gases
are cooler and decreases, at the same time, through the extraction
tubes, where the pumped gases are warmer. As a result, pumping
of a larger amount of NCG is ensured and of a smaller amount of
CG, given that the NCG concentration is the highest on the colder
places. This way, the total amount of the pumped gases from the
STC can be reduced, the SEPS operates more efficiently and con-
sumes less energy for its operation.
3. Simulation model of the control of the amount of pumped
gases through extraction tubes
The simulation model of the control of the amount of the
pumped gases through the extraction tubes was designed by
means of the Matlab software and a set of tools provided by the
above software: Simulink, ANN, FLC, Curve Fitting Toolbox, and
x-steam toolbox. The simulation model comprises the following
auxiliary simulations models: nonlinear ANN unit, x-steam calcu-
lation unit, SEPS calculation unit, FLC calculation unit and pumped
gas calculator. The simulation model works so that the dimension
of pumped gases is calculated using the functional equation as a
2 function of the average temperature of the pumped gases, devel-
oped by means of the results of the existing system analyses. The
actual amount of the pumped gases is comprised of the gases
pumped by the SEPS from the STC depending on the dimension
of the Laval cross-section. When both amounts of the pumped
236 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246

Fig. 9. Control of the amount of the pumped gases through extraction tubes.

Fig. 10. Algorithm of the simulation model of the control of the amount of the pumped gases through extraction tubes.

gases are equal, the simulation model displays the results of the
Laval nozzle diameter of the primary and secondary stage of the
SEPS, the amount of the pumped gases, the quality and quantity
of the water vapour passing through the Laval nozzles. Fig. 10
shows the simulation model algorithm.
3.1. Auxiliary simulation model: nonlinear ANN unit
Nonlinear ANN unit (Fig. 10) is a nonlinear ANN system, which,
by means of 3 input data items, i.e. temperature, pressure and
water vapour mass flow, entering the steam turbine, computes 2
output data items, i.e. temperature and pressure of water vapour
used for the SEPS’s drive. The ANN was designed by using real pro-
cess data of the 2013/2014 heating season from the SCADA system
[28] in a thermal power plant in Slovenia. SCADA contains more
than 850 groups of data concerning the district heating system,
constantly recorded on an hourly basis 365 days a year. Some other
authors used in their research the SCADA system [46–50]. Three
groups of input data were used in the creation of the nonlinear
ANN unit, expressed in the [3  4793] matrix form and 2 groups
of output data, expressed in the [2  4793] matrix form. As a result,
each group contains 4793 data items. The nonlinear ANN unit algo-
rithm structure comprises an input layer, two hidden layers and an
output layer. The input layer contains 3 neurons, the first hidden
layer contains 35 neurons, the second hidden layer 5 neurons
and the output layer 2 neurons. The nonlinear ANN unit algorithm
structure (Fig. 10) was selected through validation and it has min-
imum errors between the results provided by the nonlinear ANN
algorithm structure and the real process results.
The X-steam calculation unit is not to be specifically introduced.
Its role is to calculate the enthalpy and entropy values of water
vapour used as the SEPS motion steam.
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 237

3.2. Auxiliary simulation model: SEPS calculation unit Table 3

The geometrical data of the existing SEPS [2].

Once the thermodynamic properties of water vapour used for Parameter Primary ejector Secondary ejector
the SEPS operation are known, the SEPS calculation units compute Narrowest Laval nozzle diameter (m) 0.0098 0.0121
the mass flow of water vapour used for the SEPS operation, water Laval nozzle outlet diameter (m) 0.0396 0.0264
vapour speed in the narrowest Laval diameter, speed of water Diffuser inlet diameter (m) 0.0496 0.0323
vapour from the Laval nozzle and quantity of removed NCG from Final diameter of a diffuser (m) 0.1025 0.0872
Diffuser expansion angle (°) 10 10
the STC [2]. The calculations recorded in the primary and sec- Nozzle efficiency (%) 0.97 0.97
ondary calculation units are based on the thermodynamic transfor- Diffuser efficiently (%) 0.75 0.75
mation of the SEPS shown in Fig. 11. Several other authors
described the thermodynamic transformation of the SEPS in a sim-
ilar way [3,51,52].
The motive steam enters the ejector at point 0 and flows isentropic Laval nozzle expansion. Steam speed at the exit from the
through the Laval nozzle, where the steam expands at supersonic Laval nozzle is calculated as [2,53–55]:
speed. In the mixing section x, the expanded motive steam pumps
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
c1 ¼ 2
gLaval is
ðh0 h1s Þ ð15Þ
gases from the STC 4, where they mix with each other from point 1
to point 2. The mixture of gases enters a diffuser section, where
where c1 is steam speed at the exit from the Laval nozzle.
kinetic energy of gas transforms into pressure energy and at point
The motive steam expanded in the Laval nozzle and gases
3, the gases exit at a higher pressure and a lower speed. The anal-
pumped from the STC are mixed in the mixing section. The ther-
ysis is based on the facts of conservation of momentum, mass and
modynamic analysis is described by means of the mass, momen-
energy balances in each part of the ejector (Laval nozzle, mixing
tum and energy balance equations [1,2]. Mass balance equation
and diffuser sections) [1,2]. The following facts are taken into con-
[2,53–55]:
sideration in the calculation: there is no heat transfer in the trans-
formations in the Laval nozzle, the motive steam expands in the A2
c 2
qm0 þ qm4 ¼ ð16Þ
Laval nozzle from the initial state of p0 up to pressure in mixing v2
section, px (Fig. 11), we assume that pressure px in the mixing sec-
where qm4 is pumped gas mass flow, A2 is inlet diffuser cross sec-
tion equals the gas pumping pressure p4 (Fig. 11), the gases in the
tional area, c2 is inlet diffuser mixed gas speed and v2 is inlet dif-
mixing section mix with each other, potential energy is negligibly
fuser specific volume of the mixed gas. Geometrical data of the
low and is not taken into consideration and outlet speed from the
existing SEPS is indicated in Table 3.
diffuser is extremely low and is therefore neglected [2].
Momentum balance equation [2,53–55]:
The motive steam mass flow through the Laval nozzle is calcu-
lated using the equation [2,52]: qm0
c1 þ p4
A2 ¼ ðqm0 þ qm4 Þ
c2 þ p2
A2 ð17Þ
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u  ,,0 þ11 where p4 is pumped gas pressure and p2 is diffuser inlet mixed gas
AL
p0 t,0 2
u 0
qm0 ¼ pffiffiffiffiffi


gLaval is ð13Þ pressure. Energy balance equation [2]:
T0 R0 ,0 þ 1  
c2
where qm0 is motive steam mass flow through the Laval nozzle, AL qm0
h0 þ qm4
h4 ¼ ðqm0 þ qm4 Þ
h2 þ 2 ð18Þ
2
is narrowest Laval nozzle cross section area, p0 is inlet motive steam
pressure, T0 is inlet motive steam temperature, ,0 is motive steam where h4 is specific enthalpy of the pumped gas. Pressure in the
specific heat ratio, R0 is motive steam gas constant and gLaval-is. is mixing section px = p1 = p4 is taken into consideration in the calcu-
Laval nozzle isentropic efficiency. The Laval nozzle isentropic effi- lation. The diffuser conditions are calculated using the energy bal-
ciency is defined as [2,53–55]: ance equation [2,53–55]:

h0 h1 c32
gLaval is ¼  0; 97 ð14Þ h2 þ ¼ h3 ð19Þ
h0 h1s 2
where h0 is specific enthalpy of motive steam, h1 is specific enthalpy where h3 is diffuser outlet mixed gas specific enthalpy. By introduc-
of steam Laval nozzle expansion and h1s specific enthalpy of steam ing isentropic efficiency of a diffuser we obtain [2,53–55]:

Fig. 11. Thermodynamic transformation of the SEPS [2].

238 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246


,mix2
R
Tmix2 c22

h30 h2 ¼ gd
ðh3 h2 Þ ð20Þ p2
qm0
h0 þ qm4
h4 ¼ A2
c2
þ
where h30 is isentropic specific enthalpy of diffuser outlet mixed gas
Rmix2
T mix2 ,mix2 1 2

and gd is diffuser isentropic efficiency. Gas pumped mass flow qm4 ,mix2 ,mix2 2
¼ A2
c2
p2 þ
M2 ð26Þ
is calculated using a derived equation where thermodynamic prop- ,mix2 1 2
erties of the mixed gas are taken into consideration, Mach number From Eq. (26) we express c2 and insert it into Eq. (17) to obtain a
in the mixing section and geometrical properties of the diffuser. The quadratic equation for the calculation of the mass flow rate of
Mach number in the mixing section is calculated [2,53,54]: pumped gas from STC, whose solution is [2]:
c2
M 2 ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð21Þ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
n h ioffi
,mix2
Rmix2
T mix2 B B2 4
h4 qm20
h0 C
,,mix2 1
A2
p2
1 þ ,mix2 1

M 2
mix2 2 2
where M2 is Mach number in the mixing section, ,mix2 is specific qm4 ¼
2
h4
heat ratio of the gas mixture in the mixing section, Rmix2 is gas con-
ð27Þ
stant of the gas mixture in the mixing section and Tmix2 is diffuser
inlet mixed gas temperature. Using Eqs. (16), (17) and (21) the nec- Part B in Eq. (27) is calculated [2]:
essary diameter A2 is expressed and calculated [2,53]:
qm0
c1 B ¼ qm0
ðh0 þ h4 Þ ð28Þ
A2 ¼ h i ð22Þ
p2
,mix2
M 22 þ1 p4 Part C in Eq. (27) is calculated [2]:

Eqs. (19)–(21) are used to obtain the equation [2]: C ¼ qm0
c1 þ p4
A2 p2
A2 ð29Þ
 ,mix2 1
p3 ,mix2 gd
M22
½,mix2 1Š The pumped gas mass (qm4) is a two-phase mixture comprising
¼1 ð23Þ
p2 2 condensable steam and non-condensable gas. The analyses of the
existing SEPS [1] showed that the ratio of the pumped two-phase
where p3 is pressure of exhaust ejector mixed gas. If the inlet dif-
mixture is 1:4 on average. The mass flow of the non-condensable
fuser cross-section area A2 is known, it is possible to express M2
gas is calculated using the equation [2]:
as a function of p2 from Eq. (22) to obtain [2]:
qm0
c1 p p2 qmnon out ¼ qm4
0:75 ð30Þ
M 22 ¼ þ 4 ð24Þ
p2
A2
,mix2 p2
,mix2 where qmnon-out is mass flow of the pumped non-condensable gas
By inserting Eq. (23) into Eq. (21) we obtain [2]: from the STC.

 ,mix2 1  
p3 ,mix2 gd
ð,mix2 1Þ qm0
c1 þ p4
A2 3.3. Auxiliary simulation model: FLC calculation unit

1 1¼0
p2 2
,mix2 p2
A2
The FLC calculation unit is intended to control, by means of
ð25Þ
input signals, i.e. temperatures of the pumped gas through extrac-
If p3 is known, Eq. (24) is solved, and p2 is obtained and then M2 tion tubes or connection points 1, 2 and 3, the open positions of
is calculated from Eq. (23). It is necessary to determine c2, T2 and control valves. Any extraction tube or connection point has its
qm4. Taking into account Eq. (16), the energy Eq. (18) can be own control valve, Fig. 9. FLC calculation unit has 3 input signals,
written as follows [2]: 3 output signals and comprises 8 correction rules, specifying the

Low Ok High F-Cl Cl Ok Op F-Op

Connecting 1 1 1
Sucked Gas Valve 1
Temperature Position
0.5 0.5

1.If is T1 (Low) and T2 is (Ok) and T3 is (Ok)

0 then is V1 (F-Op),V2 is (Cl) and V3 is (Cl) 0
10 20 30 40 50 2.If is T1 (High) and T2 is (Ok) iand T3 is (Ok) 0 20 40 60 80 100
T1 then is V1 (F-Cl), V2 is (Op) and V3 is (Op) V1
3.If is T1 (Ok) and T2 is (Low) and T3 is (Ok)
Connecting 2 1 Low Ok High
then is V1 (Cl), V2 is (F-Op) and V3 is (Cl) 1 F-Cl Cl Ok Op F-Op
Sucked Gas 4.If is T1 (Ok) and T2 is (High) and T3 is (Ok) Valve 2
Temperature then is V1 (Op), V2 is (F-Cl) and V3 is (Op) Position
0.5 5.If is T1 (Ok) and T2 is (Ok) and T3 is (Low) 0.5
then is V1 (Cl), V2 is (Cl) and V3 is (F-Op)
6.If is T1 (Ok) and T2 is (Ok) and T3 is (High)
0 then is V1 (Op), V2 is (Op) and V3 is (F-Cl) 0
7.If is T1 (Low) and T2 is (Low) and T3 is (High) 0 20 40 60 80 100
10 20 30 40 50
then is V1 (F-Op), V2 is (F-Op) and V3 is (F-Cl) V2
T2 8.If is T1 (High) and T2 is (High) and T3 is (Low)
1
F-Cl Cl Ok Op F-Op
Connecting 3 1 Low Ok High then is V1 (F-Cl) and V2 is (F-Cl) and V3 is (F-Op)
Sucked Gas Valve 3
Temperature Rules Position
0.5
0.5

0 0
10 20 30 40 50 0 20 40 60 80 100
T3 V3

Fig. 12. Architecture of FLC calculation unit valve positions of extraction tubes.
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 239

correlation of input-output signals. The FLC calculation unit archi-
tecture is shown in Fig. 12. The input signals are marked Low, Ok
and High and designate the temperature of the pumped gases.
The output signal designations are: F-Cl fast closure, Cl closure,
Ok, Op opening and F-Op fast opening.
3.4. Auxiliary simulation model: pumped out gas calculator
The auxiliary model of the pumped out gas calculator computes
mean temperatures of pumped out gases and the required amount
of pumped out gases. The FLC calculation unit computes the mean
value of the pumped out gases using the equation:
Pn
i¼1 T p i
V i
T mean ¼ Pn
i¼1 V i
and pressure of steam admitted to the turbine) and two set of out-
put data (motive steam temperature and pressure).
Each piece of input data of the ANN has a precisely defined out-
put data item. The neuron weights are defined in the ANN design
so that the final sum of the results of all neurons provides a result
with the minimum possible errors between the input output data.
Several learning algorithms are available to obtain the relation-
ships between the inputs and the outputs. The most widely used
algorithm is the feed-forward back-propagation learning
Levenberg-Marquardt algorithm [60,61]. Furthermore, these equa-
tions were used by some earlier researchers [62–66]. The mean
square error (MSE) is defined as
n
1X
ð31Þ MSE ¼ ðtj o j Þ2 ð33Þ
n j¼1

where Tmean is mean temperature of the pumped out gases, Tp 1 is and the root mean square error (RMSE) is defined as follows
temperature of the pumped out gas of the tube and Vi is the position
" #1=2
of the valve of the tube. Using the results of the mean temperatures n
X 2
of the pumped out gases and the equation, generated with the Mat- RMSE ¼ ð1=nÞ ½t j oj Š ð34Þ
lab Curve Fitting tool, the auxiliary model computes the share of the j¼1

NCG in the pumped out mixture. The equation for the calculation of In addition, the correlation coefficient (R2), mean absolute error
the share of NCG is made on the basis of the analyses of the pumped (MAE) and mean absolute percentage error (MAPE) are respec-
out gas of the existing system and by means of the Sum of Sine tively defined as
model of stage 2: "Pn #
j¼1 ðt j oj Þ2
mNCG ¼ 99:44
sinð0:01493
T mean þ 22:67Þ R2 ¼ 1 Pn 2
ð35Þ
þ 0:823
sinð0:3484
T mean 55:99Þ ð32Þ j¼1 ðoj Þ

where mNCG is the share of NCG in the pumped out mixture. The 1X n

results of Eq. (32) are shown in Fig. 13.
The analysis of the measurements of the existing system shows
that in order to maintain the pressure of 0.01 MPa and temperature
of 312 K in the STC, it is necessary to pump from the STC 0.043 kg/s
of NCG. The auxiliary simulation model computes the required
amount of the pumped out gases as a function of the temperature
of the pumped out gases according to the diagram shown in Fig. 13.
Some other authors also analysed the conditions in the ejector by
means of modelling [56–59].
4. Validation of simulation model
The simulation model validation is carried out by testing differ-
ent ANN architectures and verifying the pressure dynamics in the
STC, provided by the simulation model and real process results.
The ANNs architectures are designed by testing the model with
the best correspondence between the ANNs results and the real
process data [28]. The real process data of the 2013/2014 heating
season in a matrix form [5  4793] was used in the ANN design
[28]. Three sets of input data were used (mass flow, temperature
MAE ¼ jtj oj j ð36Þ
n j¼1
where t is the target value, o is the output value, and n is the
number of learning data sets. Input and output layers are normal-
ized in the ( 1, 1) or (0, 1) range. In addition, the correlation
coefficient ranges between 0 and 1. A very good fit yields an R2
value of 1, whereas a poor fit results in a value near 0 [11].
The ANNs testing results of different architectures feed-forward
back-propagation learning Levenberg-Marquardt algorithms are
indicated in Table 4.
Table 4 shows that the winning nonlinear ANN structure used
in the nonlinear ANN unit has the 35-5 architecture (in bold), as
this architecture has the lowest errors. The architecture of the win-
ning ANN structure was presented in detail in chapter 3 and shown
in Fig. 10. The process of design and regression of the winning non-
linear ANN structure, used in the nonlinear ANN unit, is shown in
Fig. 14.
It is evident from Fig. 14 that the creation of the nonlinear ANN
structure, used in the nonlinear ANN unit, was carried out in 700
epochs. The best validation performance in terms of MSE is

95
The Sucked Out Gas (%)
Proportion Of NCG In

90

85

80

75

70
285 290 295 300 305 310 315 320 325
Mean Temperature Of The Sucked Out Gas K

Fig. 13. Results of Eq. (32).

240 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246

Table 4
The ANNs testing results of learning with the Levenberg-Marquardt algorithm.

Hidden neurons Hidden layers Epochs Data set size MSE RMSE R2 MAE
10-8-5 3 700 4793 43.7962 6.6178 0.9999 3.0765
60-10 2 700 4793 50.1995 7.0852 0.9999 3.2408
35-5 2 700 4793 41.0635 6.4081 0.9999 2.8508
10-5 2 700 4793 45.7717 6.7655 0.9999 3.1879
9-3 2 700 4793 55.2172 7.4308 0.9999 3.6369
120 1 700 4793 52.2107 7.2257 0.9998 2.8141
80 1 700 4793 52.5508 7.2492 0.9998 2.7487
40 1 700 4793 45.6821 6.7588 0.9999 3.0329

2
Training: R =0.99998 Validation: R 2=0.99992
Best Validation Performance is 41.063 at epoch 54

Output ~= 1*Target + 0.18

Output ~= 1*Target + 0.13
1.9 600 600 Y= T
10 Y= T
Train 500 Fit 500 Fit
Validation Data 400 Data
400
Test
300 300
1.8 Best
10 200
200
100
Mean Squared Error (MSE)

100

1.7 200 400 600 200 400 600

10 Target Target
2 2
Test: R =0.99995 All: R =0.99995

Output ~= 1*Target + 0.14

Output ~= 1*Target + 0.16

600 600 Y= T
Y= T
1.6
10 500 Fit 500 Fit
Data 400 Data
400
300 300
1.5
200 200
10
100 100

0 50 100 150 200 400 600 200 400 600

700 Epochs Target Target

Fig. 14. End regression design procedure of the winning ANN nonlinear structure, using the auxiliary simulation model of the nonlinear ANN unit.

640
Temperature (K)

620
600
580
560 Real Proces Motive Steam Temperature
ANN Results Motive Steam Temperature
540
0.93
Pressure (MPa)

0.92
0.91
0.9
Real Proces Motive Steam Pressure
0.89 ANN Results Motive Steam Pressure
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Time (minutes)

Fig. 15. Comparison of results of the nonlinear ANN structure, employed in the auxiliary simulation model of the nonlinear ANN unit and the real process.

41.063 at epoch 54. Regression R2 is 0.99995. Fig. 15 shows a com- The mean values of the analysed period show that between the
parison of the results provided by the winning nonlinear ANN real process data and the results provided by both simulations
structure, employed in the nonlinear ANN unit and the real process models the differences of the pressure in the STC are minimal.
results. The mean value of the pressure of the analysed period in the real
Moreover, the pressure dynamics in the real process STC was process STC amounts to 9.321
10 3 MPa and the mean value of
verified, top diagram in Fig. 16, and the pressure dynamics in the pressure in the STC, provided by the simulation model to
STC, provided by the simulation model, bottom diagram Fig. 16. 9.291
10 3 MPa. The pressure difference is 3
10 5 MPa, which
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 241

0.014 Real Proces Pressure In The STC, Mean Value 0.009321 MPa

Pressure (MPa)
0.012

0.01

0.008

0.014
Simulation Model Results Pressure In The STC, Mean Value 0.009291 MPa
Pressure (MPa)

0.012

0.01

0.008

500 1000 1500 2000 2500 3000

Time (minutes)

Fig. 16. Verification of pressure dynamics in the STC.

308
Temperature (K)

Pipe 1 Pipe 2 Pipe 3

305

302

299

296
90 Valve 1 Valve 2 Valve 3
Valve Position (%)

80
70
60
50
40
30
20
0 500 1000 1500 2000 2500 3000
Time (seconds)

Fig. 17. Temperatures and positions of control valves on extraction tubes.

is negligible. The above facts confirm the quality of the simulation
model.
5. Simulation model results
The results of the simulation model show the control of the
pumped amount of gases from the STC with the intention of pump-
ing a larger amount of NCG and consequently, streamlining the
SEPS operation. The pumping of a larger amount of NCG is ensured
through the control of the pumped out amount of gas through the
tubes. The amount of the pumped out gases through the tubes is
controlled so that the amount of the pumped out gases is increased
in the tube with a lower temperature of the pumped out gases and
decreased in the tube with a higher temperature of the pumped
out gases. The concentration of NCG is the highest on colder sur-
faces, as indicated in chapter 2.3. By installing additional connec-
tion points and controlling the amount of the pumped out gases
through the tubes, additional cold points are obtained where the
concentration of NCG is the highest due to low temperatures.
These sites are ideal for NCG pumping. The simulation model input
data includes simulated temperatures of pumped out gases
through the extraction tubes in the 296.5 K to 308 K range.
Fig. 17 shows the model input data and the control valve positions
on the extraction tubes.
The positions of valves in Fig. 17 are the results of the auxiliary
simulation model of the FLC calculation unit, computed as a func-
tion of the temperature of the evacuated gases. By means of the
temperatures and the positions of the control valves on extraction
tube, the above auxiliary simulation model also computes the
mean temperature of the removed gases from the STC and the
required extraction amount. The simulation model results are
shown in Fig. 18. Point 1 in Figs. 18, 19, 21 and 22 represents the
pumping conditions of the existing SEPS in a thermal power plant
in Slovenia and point 2 the pumping conditions, where the
extracted quantity is controlled in the extraction tubes.
It is evident from Fig. 18 that the minimal required quantity of
extracted gases from the STC is 0.0445 kg/s at the mean tempera-
ture of extracted gases amounting to 299 K. The above temperature
of extracted gases may be achieved by controlling the gas extrac-
tion rate through the extraction tubes. By means of the results of
the required quantity of extracted gases and the results of the
actual quantity of extracted gases the simulation model corrects
the dimension of the Laval nozzle in the Laval cross-section until
the required quantity and the actual quantity of the extracted
gases are equal. The simulation model computes the actual quan-
tity of the extracted gases as a function the quality of water vapour
used for the operation of the SEPS, and the dimension of the Laval
nozzle in the Laval cross-section. The results of the dimensions of
the nozzle in the Laval cross-section of the primary and secondary
ejector stage in various water vapour qualities used for the opera-
tion of the SEPS, are shown in Fig. 19.
The required dimensions of the Laval nozzle in the Laval cross-
sections are reduced depending on the required and actual extrac-
tion quantity and vice versa. The quality of water vapour used for
242 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246

308
Pumped Gas Out Total Temperature

Temperature (K)
305

302

299

296
0.06
Mass Flow (kg/s)
Pumped Gas Out Request Quantity
0.055
1
0.05
2
0.045

0.04
0 500 1000 1500 2000 2500 3000
Time (seconds)

Fig. 18. Overall temperature and required amount of gas extraction.

-3
x 10 Motive Steam 0.9 MPa 553 K
11
Laval Diameter (metre)

Motive Steam 1.1 MPa 593 K

Motive Steam 1.3 MPa 613 K
10 1
Primary Ejector

Motive Steam 1.5 MPa 643 K

9 2

7
Motive Steam 0.9 MPa 553 K
Laval Diameter (metre)

Motive Steam 1.1 MPa 593 K

0.013
Secondary Ejector

Motive Steam 1.3 MPa 613 K

Motive Steam 1.5 MPa 643 K
0.012 1

0.011 2

0.01

0.009

0 500 1000 1500 2000 2500 3000

Time (seconds)

Fig. 19. Results of the model of the Laval cross-sections of the Laval nozzle of the primary and secondary ejector stages in various water vapour qualities.

the operation of the SEPS also has a large impact on the dimension bine, qmi is the mass flow of working fluid cross individual turbine
of the Laval nozzle in the Laval cross-section. In the case of water component and D hi is the difference of specific enthalpy of the
vapour of better quality, a smaller Laval cross-section of the Laval working fluid of individual component.
nozzle is needed and vice versa. Fig. 20 shows the amounts of It is evident from the results in Fig. 21 that the SEPS motive
extracted gases in cases of motive steam of different qualities for steam consumption is the highest at the poorest steam quality
the primary and secondary ejector stage. due to the fact that in this case the nozzle dimension in the Laval
In Fig. 20, the blue curve shows the total amount of extracted cross-section is the largest. The SEPS motive steam consumption
gases pumped by the secondary ejector stage and the red1 curve, is the lowest when the steam quality is the best, due to the fact
however, the total amount of extracted gases pumped by the pri- that in this case the dimension of the nozzle in the Laval cross-
mary ejector stage, whereas the green curve shows the amount of section is the smallest.
NCG fed into the secondary ejector stage 2 and the violet curve the Moreover, the results were analysed of the simulation model of
amount of CG fed into the secondary ejector stage. Fig. 21 shows the generated power in the case of expansion of the amount of
the results of the SEPS motive steam consumption model and the steam used for the operation of the SEPS from the quality of steam
power that would be obtained in the case of expansion of the above entering the steam turbine (9.2 MPa and 793 K) to the quality of
indicated amount of water vapour in the steam turbine. The gener- steam used for the operation of the SEPS. Fig. 22 shows the results
ated power in the case of SEPS motive steam expansion in the tur- of the simulation model.
bine is calculated [67]: Fig. 22 shows that additional steam turbine power produced in
d
X the case of expansion of the amount of steam used for the opera-
qP gen ¼ qmi
Dhi ; ð37Þ tion of the SEPS is the largest in the expansion of water vapour
id ¼1 in the steam turbine up to the quality of 0.9 MPa and 553 K and
where qPgen is generated power in the case of SEPS motive steam the lowest in expansion of water vapour in the steam turbine up
expansion in the turbine, id is the iteration of individual component to the quality of 1.5 MPa and 643 K. The results of the model of
of the turbine, d is the number of individual components of the tur- average values of the generated steam turbine power from Figs. 21
and 22 are shown in Table 5, where Power-1 is the average value of
1
For interpretation of color in ‘Fig. 20’, the reader is referred to the web version of power generated by the steam turbine during the expansion of the
this article. water vapour quantity used for the operation of the SEPS, from the
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 243

0.06 0.06
Motive Steam Quality Motive Steam Quality
0.9 MPa 553 K 1.1 MPa 593 K
0.05 0.05

Mass Flow (kg/s)

Mass Flow (kg/s)
0.04 0.04

0.03 Secondary Ejector Total Gas Sucked Out 0.03 Secondary Ejector Total Gas Sucked Out
Primary Ejector Total Gas Sucked Out Primary Ejector Total Gas Sucked Out
0.02 NCG Sucked Into The Secondary Ejector 0.02 NCG Sucked Into The Secondary Ejector
CG Sucked Into The Secondary Ejector CG Sucked Into The Secondary Ejector
0.01 0.01

0 0
0.06 0.06
Motive Steam Quantity Motive Steam Quality
1.3 MPa 613 K 1.5 MPa 643 K
0.05 0.05
Mass Flow (kg/s)

Mass Flow (kg/s)

0.04 0.04

0.03 Secondary Ejector Total Gas Sucked Out 0.03 Secondary Ejector Total Gas Sucked Out
Primary Ejector Total Gas Sucked Out Primary Ejector Total Gas Sucked Out
0.02 NCG Sucked Into The Secondary Ejector 0.02 NCG Sucked Into The Secondary Ejector
CG Sucked Into The Secondary Ejector
CG Sucked Into The Secondary Ejector
0.01 0.01

0 0
0 500 1000 1500 2000 2500 3000 0 500 1000 1500 2000 2500 3000
Time (seconds) Time (seconds)

Fig. 20. Amounts of extracted gases in cases of motive steam of different qualities for the ejector primary and secondary stage.

0.26 Motive Steam Consumption 0.9 MPa 553 K

Steam Turbine Power (kW) Mass Flow (kg/s)

Motive Steam Consumption 1.1 MPa 593 K

1
0.24 Motive Steam Consumption 1.3 MPa 613 K
Motive Steam Consumption 1.5 MPa 643 K
0.22

0.2 2

0.18
280
Motive Steam Quality 0.9 MPa 553 K
260 Motive Steam Quality 1.1 MPa 593 K
1
Motive Steam Quality 1.3 MPa 613 K
240 Motive Steam Quality 1.5 MPa 643 K

220
2
200

180
500 1000 1500 2000 2500 3000
Time (seconds)

Fig. 21. Results of the SEPS motive steam consumption model and generated power in the case of expansion of the indicated amount of steam in the steam turbine.

Table 5
Results of the model of average values of the generated power of the steam turbine in
Figs. 21 and 22.

Pressure Temperature Consume Power-1 Power-2 Difference

(MPa) (K) (kg/s) (kW) (kW) (kW)
0.9 553 0.2240 234.80 87.344 147.46
1.1 593 0.2045 214.31 65.28 149.03
1.3 613 0.1918 200.97 54.93 146.04
1.5 643 0.1845 193.31 42.06 151.25

Fig. 22. Results of the model of the generated power of steam turbine in the case of of the SEPS, from the quality of steam entering the steam turbine
expansion of the amount of steam to the quality of steam used for the operation of
the SEPS.
to the quality of steam used for the operation of the SEPS.
The results of the simulation model of average values show that
in the case of expansion of the average amount of steam of
0.224 kg/s, used for the operation of the SEPS of the quality of
water vapour quality entering the steam turbine to the quality of 0.9 MPa and 553 K, the steam turbine would additionally generate
the outlet steam from the steam turbine and Power-2 is the aver- 234.8 kW of power. In the case of expansion of the above amount
age value of power generated by the steam turbine during the of water vapour in the steam turbine only up to the quality of
expansion of the amount of water vapour used for the operation steam used for the operation of the SEPS, the steam turbine would
244
Pressure (MPa) Temperature (K)
650
Motive Steam Temperature
600
550
Motive Steam Pressure
1.4
1.2
1
Steam Density In The Laval Diametre
3.6
3.4
Density (kg/m3)
3.2
3
2.8
2.6
2.4
2.2
2 500
0 50
additionally generate 87.34 kW. The difference in power amounts
to 147.46 kW. In the case of expansion of the average amount of
steam of 0.1845 kg/s, used for the operation of the SEPS of the qual-
ity of 1.5 MPa and 643 K, the steam turbine would additionally
generate 193.31 kW of power. In the case of expansion of the above
amount of steam in the steam turbine only up to the quality of
steam used for the operation of the SEPS, the steam turbine would
additionally generate 42.06 kW of power. The difference in power
amounts to 151.25 kW.
By changing the quality of steam used for the operation of the
SEPS the condition of the steam expanded in the Laval nozzle also
changes. Fig. 23 shows the condition of steam expanded in the
Laval nozzle at the Laval cross-section of 0.0074 m of the primary
stage, at the Laval cross-section of 0,009 m of the ejector secondary
stage and sinusoidal oscillation of quality of steam used for the
operation of the SEPS.
The results of the model show that as regards the selection of an
appropriate nozzle and quality of steam for the operation of the
SEPS there are no substantial differences in terms of the improved
performance. An essential difference as regards the performance is
expressed so that for the operation of the SEPS, the steam is used
that has previously lost part of its internal energy or has expanded
in the steam turbine up to the quality of steam envisaged to be
used for the operation of the SEPS. Therefore, the Laval nozzles
are selected of the Laval cross-section, corresponding to the avail-
able steam quality and the required extracted amount. The effi-
ciency of the operation of the SEPS is also ensured through the
extraction of as cold as possible gases, as NCG is larger in such
gases and thereby the total required extraction is lowered.
The measurements and analyses of the SEPS show that when
the quality of the motive steam for the operation of the SEPS is
0.9 MPa, 553 K and the pressure in the STC is 10,000 Pa, the aver-
age amount of gases extracted from the STC amounts to
0.057 kg/s, containing 75% of NCG and 25% of CG. As a result,
0.043 kg/s of NCG is extracted on average from the STC and 0.014
of CG. The results under points 1 and 2 in Fig. 18 also provide sim-
ilar quantities of extracted gases. Point 1 represents the extracted
amount of steam of the existing system in a district heating instal-
lation in Slovenia and point 2, however, the extracted amount of
D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246
0.9
Steam Pressure In The Laval Diameter
0.8
Pressure (MPa)
0.7
0.6
0.5
0.4
1300
1200
1100
Speed (m/s)
1000 Steam Velocity From The
900 Primary Ejector Nozzle
Steam Velocity From The
800 Secondary Ejector Nozzle
Steam Velocity In The Laval Diameter
700
600
100 150 200 250 300 0 50 100 150 200 250 300
Time (seconds) Time (seconds)
Fig. 23. Conditions of steam that expands in the Laval nozzle.
steam by controlling the amount of extracted gases through the
tubes. By using additional connection points and controlling the
extracted amount of gases through extraction tubes, almost only
NCG is extracted from the STC. The results of the simulation model
show that in this case, the SEPS can operate with two Laval nozzles
of a smaller geometrical cross-section. In the primary ejector stage,
the Laval nozzle may be reduced from the initial Laval cross-
section of 0.01 m to 0.0094 m and in the secondary ejector stage,
from 0.013 m of the Laval cross-section to 0.0117 m, points 1 and
2 in Fig. 19. As a result of smaller dimensions of the Laval nozzles,
the consumption of the SEPS motive steam is reduced. The average
consumption of steam for the operation of the existing SEPS is
0.264 kg/s and the average consumption of steam for the operation
of the SEPS with the control of the extracted amount of steam
through the extraction tubes is 0.212 kg/s, points 1 and 2 in
Fig. 21. Therefore, the average savings regarding the steam con-
sumption amount to 0.052 kg/s. The above indicated amount of
steam can be used for other purposes. If the amount of steam saved
expands in the steam turbine, the turbine would additionally gen-
erate 53 kW of power. If the additionally generated power is mul-
tiplied by the steam turbine operating hours, i.e. 7200 h/year, the
additionally generated electricity amounts to 381600 kWh.
6. Conclusion
This paper presents a simulation model that streamlines the
operation of the SEPS, serving to extract the NCG from the STC.
The operation of the SEPS is streamlined by adequately dimension-
ing the Laval cross-section of two Laval nozzles in the primary and
secondary ejector stage. The SEPS consumes less motive steam if
the dimensions of the Laval nozzles are properly selected. The
streamlining of the SEPS is based on a well thought-out selection
of the quality of the steam used by the SEPS for its operation, a well
thought-out selection of points of connection through which the
NCG are pumped out of the STC and by controlling the amount
of pumped out gases through the points of connection or extrac-
tion tubes. A well thought-out selection of the motive steam for
the operation of the SEPS is based on the fact that the steam has
 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246 245

previously expanded in the steam turbine to the quality of steam [25] Petković D, Shamshirband S. Soft methodology selection of wind turbine
parameters to large affect wind energy conversion. Electr Power Energy Syst
needed for the operation of the SEPS. The well though-out selection
2015;69:98–103.
of the connecting locations is based on the analysis of the impact of [26] Shamshirband S, Petković D, Saboohi H, Anuar NB, Inayat I, Akib S, et al. Wind
NCG on the transfer of STC heat and the temperature analysis of turbine power coefficient estimation by soft computing methodologies:
the STC in the existing system of the district heating plant in Slove- comparative study. Energy Convers Manage 2014;81:520–6.
[27] Petković D, Ćojbašič Ž, Nikolić V. Adaptive neuro-fuzzy approach for wind
nia. The control of the amount of gases extracted through the turbine power coefficient estimation. Renew Sustain Energy Rev
extraction tubes facilitates the pumping of almost only NCG from 2013;28:191–5.
the STC, thus reducing the total amount that has to be pumped [28] Supervisory Control and Data Acquisition (SCADA). <http://www.te-tol.si/>.
[29] Islam MR, Jahangeer KA, Chua KJ. Experimental and numerical study of an
out from the SEPS in order to maintain the required pressure in evaporatively-cooled condenser of air-conditioning systems. Energy
the STC. By carrying out the above measures and choosing the 2015;87:390–9.
appropriate dimension of the Laval cross-section of the Laval noz- [30] Škerget L. Prenos gibalne količine, toplote in snovi. Univerza v Mariboru,
Fakulteta za strojništvo; 2010.
zle the consumption of the motive steam for the operation of the [31] Abbasi A, Eslamloueyan R. Determination of binary diffusion coefficients of
SEPS is reduced and the efficiency of the operation of the SEPS hydrocarbon mixtures using MLP and ANFIS networks based on QSPR method.
improved. Chemomet Intell Lab Syst 2014;132:39–51.
[32] Wyrwał J, Świrska J, Marynowicz A. Some estimates of the diffusion stress
tensor and the diffusion energy for a binary mixture of water vapour and dry
air. Int J Heat Mass Transfer 2010;53:4233–7.
References [33] Kremer GM. Diffusion of relativistic gas mixtures in gravitational fields.
Physica A 2014;393:76–85.
[34] Zhu Q, Moggridge GD, D’Agostino C. A local composition model for the
[1] Strušnik D, Avsec J. Using steam as an alternative motive fluid in the existing
prediction of mutual diffusion coefficients in binary liquid mixtures from
turbine ejector system of the Ljubljana District Heating Plant. J Energy Technol
tracer diffusion coefficients. Chem Eng Sci 2015;132:250–8.
2013;2:31–48.
[35] Gupta VK, Torrilhon M. Comparison of relaxation phenomena in binary gas-
[2] Strušnik D, Marčič M, Golob M, Hribernik A, Živić M, Avsec J. Energy efficiency
mixtures of Maxwell molecules and hard spheres. Comput Math Appl
analysis of steam ejector and electric vacuum pump for a turbine condenser air
2015;70:73–88.
extraction system based on supervised machine learning modelling. Appl
[36] Ryzhkov II, Stepanova IV. On thermal diffusion separation in binary mixtures
Energy 2016;173:386–405.
with variable transport coefficients. Int J Heat Mass Transfer 2015;86:268–76.
[3] Jafarian A, Azizi M, Forghani P. Experimental and numerical investigation of
[37] Li H, Wilhelmsen Ø, Lv Y, Wang W, Yan J. Viscosities, thermal conductivities
transient phenomena in vacuum ejectors. Energy 2016;102:528–36.
and diffusion coefficients of CO2 mixtures: review of experimental data and
[4] Yan G, Chen J, Yu J. Energy and exergy analysis of a new ejector enhanced auto-
theoretical models. Int J Greenhouse Gas Control 2011;5:1119–39.
cascade refrigeration cycle. Energy Convers Manage 2015;105:509–17.
[38] Saari J, Kaikko J, Vakkilainen E, Savolainen S. Comparison of power plant steam
[5] Reddick C, Sorin M, Sapoundjiev H, Aidoun Z. Carbon capture simulation using
condenser heat transfer models for on-line condition monitoring. Appl Therm
ejectors for waste heat upgrading. Energy 2016;100:251–61.
Eng 2014;62:37–47.
[6] Ariafar K, Buttsworth D, Al-Doori G, Sharifi N. Mixing layer effects on the
[39] Pieve M, Salvadori G. Performance of an air-cooled steam condenser for a
entrainment ratio in steam ejectors through ideal gas computational
waste-to-energy plant over its whole operating range. Energy Convers Manage
simulations. Energy 2016;95:380–92.
2011;52:1908–13.
[7] Li X, Huang H, Zhao W. A supercritical or transcritical Rankine cycle with
[40] Hu H, Du X, Yang L, Yang Y. POD based modeling on flow and heat transfer of
ejector using low-grade heat. Energy Convers Manage 2014;78:551–8.
air-cooled condenser influenced by natural wind. Int J Heat Mass Transfer
[8] Gasser L, Wellih B, Hilfiker K. Exergy analysis for increasing the efficiency of
2014;74:431–40.
air/water heat pumps. Lacern University of Applied Sciences and Art; 2008.
[41] Mirzabeygi P, Zhang C. Turbulence modeling for the two-phase flow and heat
[9] Bloch HPA. Practical guide to steam turbine technology. McGraw-Hill; 1996.
transfer in condensers. Int J Heat Mass Transfer 2015;89:229–41.
[10] Haseli Y, Dincer I, Naterer GF. Entropy generation of vapor condensation in the
[42] Kylili A, Fokaides PA, Christou P, Kalogirou SA. Infrared thermography (IRT)
presence of a non-condensable gas in a shell and tube condenser. Int J Heat
applications for building diagnostics: a review. Appl Energy 2014;134:531–49.
Mass Transfer 2008;51:1596–602.
[43] Dassisti M, Intini F, Chimienti M, Starace G. Thermography-enhanced LCA (Life
[11] Li JD. CFD simulation of water vapour condensation in the presence of non-
Cycle Assessment) for manufacturing sustainability assessment. The case
condensable gas in vertical cylindrical condensers. Int J Heat Mass Transfer
study of an HDPE (High Density Polyethylene) net company in Italy. Energy
2013;57:708–21.
2016:1–12.
[12] Desel J, Erwin T. Modeling, simulation and analysis of business processes. Bus
[44] Sousa J, Villafañe L, Paniagua G. Thermal analysis and modeling of surface heat
Process Manage, Lect Notes Comput Sci 2000;1806:129–41.
exchangers operating in the transonic regime. Energy 2014;64:961–9.
[13] Strušnik D, Avsec J. Artificial neural networking and fuzzy logic exergy
[45] Linfenga Z, Quana Z, Minb L, Yaxinga D. A new analytical model for the
controlling model of combined heat and power system in thermal power
underground temperature profile under the intermittent operation for
plant. Energy 2015;80:318–30.
Ground-Coupled Heat Pump systems. Energy Procedia 2015;75:840–6.
[14] Strušnik D, Avsec J. Artificial neural networking model of energy and exergy
[46] Lacko R, Drobnič B, Mori M, Sekavčnik M, Vidmar M. Stand-alone renewable
district heating mony flows. Energy Build 2015;86:366–75.
combined heat and power system with hydrogen technologies for household
[15] Strušnik D, Golob M, Avsec J. Artificial neural networking model for the
application. Energy 2014;77:164–70.
prediction of high efficiency boiler steam generation and distribution. Simul
[47] Radhakrishnan MB, Srinivasan D. A multi-agent based distributed energy
Model Pract Theory 2015;57:58–70.
management scheme for smart grid applications. Energy 2016;103:
[16] Keçebas A, Yabanova I, Yumurtaci M. Artificial neural network modeling of
192–204.
geothermal district heating system thought exergy analysis. Energy Convers
[48] Karavas CS, Kyriakarakos G, Arvanitis KG, Papadakis G. A multi-agent
Manage 2012;64:206–12.
decentralized energy management system based on distributed intelligence
[17] Shamshirband S, Iqbal J, Petković D, Mirhashemi M. Survey of four models of
for the design and control of autonomous polygeneration microgrids. Energy
probability density functions of wind speed and directions by adaptive neuro-
Convers Manage 2015;103:166–79.
fuzzy methodology. Adv Eng Softw 2014;76:148–53.
[49] Smrekar J, Assadi MA, Fast M, Kuštrin I, Dec S. Development of artificial neural
[18] Wang LW. A course in fuzzy systems and control. International Edition,
network model for a coal-fired boiler using real plant data. Energy
Prentice-Hall International; 1997.
2009;34:144–52.
[19] Tomsovic K, Chow MY. Tutorial on fuzzy logic applications in power systems.
[50] Dai J, Liu D, Wen L, Long X. Research on power coefficient of wind turbines
Prepared for the IEEE-PES winter meeting in Singapore; 2000.
based on SCADA data. Renew Energy 2016;86:206–15.
[20] Lygouras JN, Botsaris PN, Vourvoulakis J, Kodogiannis V. Fuzzy logic controller
[51] Reddick C, Sorin M, Sapoundjiev H, Aidoun Z. Carbon capture simulation using
implementation for a solar air-conditioning system. Appl Energy
ejectors for waste heat upgrading. Energy 2016;100:251–61.
2007;84:1305–18.
[52] Xing M, Yu J, Liu X. Thermodynamic analysis on a two-stage transcritical CO2
[21] Jin L, Huang GM, Fan YR, Wang L, Wud T. A pseudo-optimal inexact stochastic
heat pump cycle with double ejectors. Energy Convers Manage
interval T2 fuzzy sets approach for energy and environmental systems
2014;88:677–83.
planning under uncertainty: a case study for Xiamen City of China. Appl
[53] Marietta M. Investigation of an ejector heat pump by analytical methods. Oak
Energy 2015;138:71–90.
Ridge National Laboratory, United States Department of Energy, University of
[22] Shamshirband S, Petković D, Tong CW, Al-Shammari ET. Trend detection of
Tennessee; 1984. p. 7–12.
wind speed probability distribution by adaptive neuro-fuzzy methodology.
[54] Hanafi AS, Mostafa GM, Waheed A, Fathy A. 1-D mathematical modeling and
Flow Meas Instrum 2015;45:43–8.
CFD investigation on supersonic steam ejector in MED-TVC. Energy Procedia
[23] Petković D, Shamshirband S, Tong CW, Al-Shammari ET. Generalized adaptive
2015;75:3239–52.
neuro-fuzzy based method for wind speed distribution prediction. Flow Meas
[55] Bai T, Yan G, Yu J. Thermodynamic analyses on an ejector enhanced CO2
Instrum 2015;43:47–52.
transcritical heat pump cycle with vapor-injection. Int J Refrig 2015;58:22–34.
[24] Nikolić V, Petković D, Shamshirband S, Ćojbašič Ž. Adaptive neuro-fuzzy
[56] Banasiak K, Palacz M, Hafner A, Buli Z, Smołka J, Nowak AJ, et al. A, CFD-based
estimation of diffuser effects on wind turbine performance. Energy
investigation of the energy performance of two-phase R744 ejectors to recover
2015;89:324–33.
246 D. Strušnik et al. / Energy Conversion and Management 126 (2016) 228–246
the expansion work in refrigeration systems. An irreversibility analysis. Int J
Refrig 2014;40:328–37.
[57] Sözen A, Arcaklioğlu E. Exergy analysis of an ejector-absorption heat
transformer using artificial neural network approach. Appl Therm Eng
2007;27:481–91.
[58] Maghsoodi A, Afshari E, Ahmadikia H. Optimization of geometric parameters
for design a high-performance ejector in the proton exchange membrane fuel
cell system using artificial neural network and genetic algorithm. Appl Therm
Eng 2014;71:410–8.
[59] Sözen A, Arcaklioglu E, Ozalp M. Performance analysis of ejector absorption
heat pump using ozone safe fluid couple through artificial neural networks.
Energy Convers Manage 2004;45:2233–53.
[60] Suzuki Y, Ovaska S, Furuhashi T, Roy R, Dote Y. Soft computing in industrial
applications. London-Springer-Verlag; 2000.
[61] Arslan O, Yetik O. ANN based optimization of supercritical ORC-Binary
geothermal power plant: Simav case study. Appl Therm Eng 2011;31:3922–8.
[62] Shen CC, Cao GY, Zhu XJ. Nonlinear modeling of MCFC stack based on RBF
neural networks identification. Simul Model Pract Theory 2010;10:109–19.
[63] Rezaei J, Shahbakhti M, Bahri B, Aziz AA. Performance prediction of HCCI
engines with oxygenated fuels using artificial neural networks. Appl Energy
2015;138:460–73.
[64] Keles D, Scelle J, Paraschiv F, Fichtner W. Extended forecast methods for day-
ahead electricity spot prices applying artificial neural networks. Appl Energy
2016;162:218–30.
[65] Rostek K, Morytko L, Jankowska A. Early detection and prediction of leaks in
fluidized-bed boilers using artificial neural networks. Energy 2015;89:914–23.
[66] Alvarez ME, Hernandez JA, Bourouis M. Modelling the performance
parameters of a horizontal falling film absorber with aqueous (lithium,
potassium, sodium) nitrate solution using artificial neural networks. Energy
2016;102:313–23.
[67] Bejan A. Entropy generation minimization. N.W, Boca Raton,
Florida: Corporate Blvd.; 2000.