Professional Documents
Culture Documents
vorgelegt von
Ligang Wang
geb. in Qingdao, China
genehmigte Dissertation
Promotionsausschuss
Berlin 2016
Dedicated to my families, my fiance
and
the loving memory of my grandfather (1936-2012).
At twenty years of age, the will reigns;
at thirty, the wit;
and at forty, the judgement.
Benjamin Franklin (1706-1790)
ACKNOWLEDGMENTS
Six years, not a short time! Over the six years, so much sorrow and happiness on the
way! I once walked back and forth, but become stronger after each tough task. Today,
standing at the point with the thesis, I have so much gratitude to all who helped,
supported and accompanied me once or along the way.
I would first thank all my professors. At the age when guidance on life and fu-
ture is needed most, fortunately, I met them. Thank my foresighting, understanding
Prof. Yongping Yang for the support on my decisions and the patience on my growth
since my first day in university. He provided such a high-level international platform
and the chance to compete, cooperate and grow up with world-class researchers.
Thank my kind, learned Prof. George Tsatsaronis. Without his efforts, there is even
no chance to start this thesis. Surprised with his rigor and strictness on research, I was
grateful to his careful modification and suggestions on my papers and the thesis. He
always gave me invaluable emotional support when I felt frustrated in the research
and life in Berlin. His exuberant energy at his sixties impresses me a lot.
Thank my talented, quick-witted Prof. Andr Bardow, my academic role model, for
his criticism and suggestions on my research. His logics, strictness and carefulness on
scientific writing and presentation has accelerated my growth very much.
Thank my ambitious, industry-oriented Prof. Changqing Dong. Still meditate those
in-depth but relaxed talks with him, which inspires me to think more and behave better.
His concerns on technology transfer of research has deeply rooted in my mind.
Thank my warm-hearted, beautiful Prof. Tatiana Morosuk for her help on my publi-
cations and her willing to be the chairlady of my defense.
I would then thank all the colleagues I worked with in Institute for Energy Tech-
nology (Technical University of Berlin), Chair of Technical Thermodynamics (RWTH
Aachen University), Institute for Advanced Energy Systems and National Engineer-
ing Laboratory for Biomass Power Generation Equipment (North China Electric Power
University). In particular, thank Dr. Philip Voll, Dr. Matthias Lampe, Mathias Penkuhn,
Dr. Gang Xu, Dr. Zhiping Yang, Dr. Ningling Wang, Maike Hennen, Shengwei Huang,
Lingnan Wu, Peng Fu and Xiaoen Li for their help.
Finally, I would dedicate my thanks to my solid backing, my family. Their hardwork
gives me more reasons to pursue higher and achieve more. Thank my fiance, Miss
Chunxia Qu, for her four-year persistence, patience and wait with such a long distance
of seven thousand kilometers. I love you all!
Ligang Wang
26 June 2015, Beijing
i
ABSTRACT
A key lever to cope with the well-known issues related to todays extensive use of fos-
sil fuels, e. g., global warming, is to design highly energy-efficient and cost-effective
pulverized coal-fired power plants. Future coal-fired power plants will be with high
temperature and pressure levels, multiple heat sources, multiple products, and many
available technologies integrated to efficiently utilize different-grade heat. This pin-
points the significance of the system improvement and optimization.
This thesis investigates progressively the methodologies for the improvement and
optimization of system configurations of pulverized-coal power plants, i. e., the em-
ployed components and their interconnections. Structural improvement can be effec-
tively proposed by combining exergy-based analysis with engineers judgments. The
structural optimization (optimal synthesis) is not a trivial task and can be best ad-
dressed by mathematical programming (i. e., superstructure-based and -free optimiza-
tion methods), since automatic generation and identification of structural alternatives
is usually involved. This thesis involves further development of the exergy-based anal-
ysis, superstructure-based and -free synthesis approaches and their applications to
pulverized-coal power plants.
Conventional exergy-based analysis can only identify the location and magnitude of
inefficiencies, while an advanced analysis further reveals their source and avoidability
by splitting each inefficiency into endogenous/exogenous and avoidable/unavoidable
parts and their combinations. In the thesis, a new approach of calculating endogenous
exergy destructions is introduced and a modern ultra-supercritical coal-fired power
plant is evaluated in detail and comprehensively. The results show that over half of the
avoidable inefficiencies within most components are endogenous, and the portion dif-
fers significantly for different components. Only nearly 10% of the costs of the whole
system could be avoided for modern industrial designs at present. In addition, mov-
ing convection-leading heating surfaces into the furnace and increasing air preheating
temperature are suggested for performance enhancement.
The superstructure-based synthesis approach requires the user to manually define
a priori the potential solution space through the modeling of a superstructure. In this
thesis, superstructure-based synthesis is applied to investigate classical considerations
for future plant design, e. g., numbers of reheating and feedwater preheating, reheat-
ing pressure and temperature, enthalpy-rise distribution of feedwater in feedwater
preheaters. These promising structural alternatives are less likely excluded from the
well-defined graphical superstructure built by a simulator. Structural alternatives and
other continuous decision variables are continuously manipulated and optimized by
specially-designed mutation and crossover operations, while each alternative is eval-
uated by system simulations. The matches of steam conditions are discussed with
optimal reheating ratios. It is found that only when the throttle pressure reaches a
specific value corresponding to each temperature level could the benefits from increas-
ing steam temperature levels be fully obtained. Moreover, increasing steam conditions
iii
is not cost-effective compared with adjusting other decision variables (e. g., isentropic
efficiencies of turbines) to improve the plant efficiency.
For complex synthesis problems, a superstructure cannot guarantee to contain all
good alternatives (in particular, the optimal solution), while it might consider a large
number of meaningless or even infeasible alternatives. To circumvent the use of super-
structures, a generic superstructure-free synthesis approach tailored for distributed en-
ergy supply systems is extended and re-implemented for thermal power plants in the
thesis. The approach employs hybrid optimization integrating evolutionary and deter-
ministic optimization to enable simultaneous alternatives generation and optimization.
Given structures are mutated based on an energy conversion hierarchy, which classi-
fies energy conversion technologies according their function. This allows the use of a
minimum number of mutation rules. The original set of mutation rules is enriched by
an efficient insertion rule, which enables the addition of any technology that has not
been existing in the parent structures. The applications show that the superstructure-
free approach automatically identifies complex structures with new features, and it is
highly extensible and stable for even very complex problems.
High-quality fronts (not only the Pareto fronts) are obtained by both superstructure-
based and -free approaches for the trade-offs between plant efficiency and the cost of
electricity. The change of the optimal cost of electricity with respect to each possible
plant efficiency is presented. The Pareto solutions near the economically-optimal solu-
tion turn out to have similar thermodynamic and economic performances, suggesting
that there are plenty different good design alternatives. However, for the set of Pareto
solutions far away from the economically-optimal solution, the minimum cost of elec-
tricity increases significantly when increasing plant efficiency, suggesting that these
solutions are not recommended in practice.
Due to the lack of real-world cost functions, the cost functions employed in both
superstructure-based and -free approaches are obtained from limited published cost
data. These cost functions lead to optimal and near-optimal solutions, particularly of
the superstructure-free approach, with more units (e. g., reheaters and de-superheaters)
than currently found in modern power plants. This indicates that there is a need to
revise the cost functions to reflect the industrial practices at present. However, the ap-
plications in the thesis show that the presented superstructure-free synthesis approach
is promising to cope with complex synthesis problems of pulverized-coal power plants
for integrating a number of available energy- and cost-saving technologies.
iv
CONTENTS
Acknowledgments i
Abstract iii
Table of Contents v
List of Figures viii
List of Tables xi
Glossory and Symbols xii
1 introduction 1
1.1 Structure of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 analysis, optimization and synthesis of coal-fired power plants 5
2.1 Trends and challenges of the design of coal-fired power plants . . . . . . 5
2.2 Analysis of energy systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Exergy analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.2 Exergoeconomic analysis . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.3 Advanced exergy-based analysis . . . . . . . . . . . . . . . . . . . . 11
2.2.4 Analysis of coal-fired power plants . . . . . . . . . . . . . . . . . . 14
2.3 Optimization of energy systems . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.1 Mathematical optimization . . . . . . . . . . . . . . . . . . . . . . . 15
2.3.2 Nonlinearity and integrity in energy system optimization . . . . . 19
2.3.3 Parametric selection and optimization of steam cycles . . . . . . . 20
2.4 Synthesis of energy systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4.1 Superstructure-based synthesis . . . . . . . . . . . . . . . . . . . . . 22
2.4.2 Superstructure-free synthesis . . . . . . . . . . . . . . . . . . . . . . 25
2.4.3 Synthesis of steam cycles . . . . . . . . . . . . . . . . . . . . . . . . 26
2.5 Multi-objective optimization . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5.1 Multi-objective optimization techniques . . . . . . . . . . . . . . . . 27
2.5.2 Multi-objective optimization of coal-fired power plants . . . . . . . 31
2.6 Contribution of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3 comprehensive exergy-based assessment of coal-fired power
plants 35
3.1 Framework for comprehensive exergy-based evaluation . . . . . . . . . . 35
3.1.1 Exergy analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1.2 Advanced exergy analysis . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.3 Exergoeconomic analysis . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.4 Advanced exergoeconomic analysis . . . . . . . . . . . . . . . . . . 39
3.2 Calculation of endogenous and unavoidable exergy destructions . . . . . 40
3.2.1 A new approach for calculating endogenous exergy destruction . 40
3.3 Evaluation of a modern ultra-supercritical plant . . . . . . . . . . . . . . . 44
3.3.1 Exergy analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.3.2 Exergoeconomic evaluation . . . . . . . . . . . . . . . . . . . . . . . 50
3.3.3 Advanced exergy analysis . . . . . . . . . . . . . . . . . . . . . . . . 52
3.3.4 Advanced exergoeconomic analysis . . . . . . . . . . . . . . . . . . 53
v
vi contents
bibliography 153
Publications 177
LIST OF FIGURES
viii
List of Figures ix
xi
N O M E N C L AT U R E
Abbreviations
AMPL a mathematical programming language
APH air preheater
AV avoidable
CC capital investment cost
COE cost of electricity
COND condenser
CP condensate pump
CV controlling valve
CWP cooling-water pump
DA de-aerator
DE differential evolution
DSH steam de-superheater
D destruction
EAs evolutionary algorithms
EA evolution algorithm
ECH energy conversion hierarchy
ECO economizer
EG electric generator
EMOA evolutionary multi-objective optimization algorithm
EN endogenous
ET secondary turbine with steam extractions
EX exogenous
FCI fixed capital investment
FC fuel cost
FP feedwater pump
FRHC final reheater (cold)
FRHH final reheater (hot)
FSH final superheater
FUR furnace
FWH feedwater preheater
F fuel
GAMS general algebraic modeling system
GDP generalized disjunctive programming
GRG generalized reduced gradient
HEN heat exchanger network
HPT high-pressure turbine
IAPWS-IF97 international association for the properties of water and steam, indus-
trial formulation 1997
IDF induced draft fan
ID industrial design
xii
Nomenclature xiii
m mechanical efficiency %
s isentropic efficiency %
a constant 1
p pressure ratio 1
best-identified efficiency %
time for finding the optimal solution min
excess air ratio 1
size of the offspring population in evolution algorithms 1
vector of characteristic variables of all components
size of the parent population 1
annual capacity factor 1
structural alternative
c real-valued crossover constant [0,1] 1
f real-valued weighting factor [0,1] 1
r real-valued random number [0,1] 1
time min
exergy efficiency %
a constant 1
0 isobaric expansion coefficient 1/K
Mathematical Symbols
C
C, cost rate (matrix) /s
E exergy flow rate MW
m mass flow rate kg/s
P power generated or consumed MW
Q heat flow rate MW
Z levelized investment cost rate /s
L lower bound of the optimal objective value
n generation number to find best objective min
s optimal solution
U upper bound of the optimal objective value
x optimal solution or Pareto solution
A, I coefficient matrices in exergoeconomic analysis 1
b matrix of levelized investment cost rate /s
F non-dominated front
b binary variables for identifying structural alternatives
c continuous integer variables for identifying structural alternatives
c, c specific exergy cost (matrix) /MJ
cp0 specific isobaric heat capacity of saturated water kJ/kg/K
CC capital investment cost $/yr
COE cost of electricity /kWh
D design variable space
d design variable
F super-objective
f exergoeconomic factor in exergoeconomic analysis %
f objective function in optimization
F efficiency-related factor
Nomenclature xv
k,i,j,r,m,n index
L levelized associated with cost
max maximum
min minimum
ms main steam
n dimension
out outgoing
R real
r non-domination rank in multi-objective optimization
rhi the ith reheating
sat saturation
sh superheated steam
T theoretical
tot total
wf working fluid
1
INTRODUCTION
Nowadays, one of the central challenges we face is that of ensuring the rapidly grow-
ing energy needs in ways that lead to a prosperous, sustainable and secure energy
future (Yergin, 2006; Omer, 2008; Armaroli and Balzani, 2011). The most immediate
course ahead is how efficiently we use existing and new sources of energy (Jacobson,
2009; Chu and Majumdar, 2012). However, due to that the share of carbon-free en-
ergy sources in total consumption of primary energy remained almost constant (13%)
during the past two decades (IEA, 2014) and the prediction that this number would
approach only 19% by 2035 (BP, 2014), the dominating role of fossil fuels is set to con-
tinue in the next few decades (IEA, 2014). In the context of sustainability, the reduction
of CO2 emission from the utilization of fossil fuels has gained increasing acceptance
(Bates et al., 2002; Metz et al., 2005; Solomon, 2007; Graus et al., 2007; Yang et al., 2008;
Samanta et al., 2011). Unfortunately, the capture of CO2 , in general, is highly energy-
and cost-intensive (Rao and Rubin, 2002; Huang et al., 2010; Petrakopoulou, 2011), re-
sulting in it being hardly affordable at present. In this regard, efficiency improvement,
as the only practical tool capable of reducing CO2 emission from the industrial use of
fossil fuels in the short term (Ber, 2007), has been the vital way forward in mastering
challenges of affordability, sustainability, reliability and security.
Coal-fired power plants contribute 40% of global power generation and almost 30%
of total CO2 emissions in 2012 (World Bank, 2014), among which the most mature
and widespread are pulverized-coal power plants (PCPPs). Particularly in China, the
share of PCPPs in total installed generating capacity remains over 70% (819 GWe of
1145 GWe in 2012 (Wei, 2013)). Therefore, the cost-effective improvement of PCPPs
has gained renewed interest (Kraemer et al., 2006; Burnard and Bhattacharya, 2011;
Campbell, 2013; Eaves et al., 2014).
In fact, the efficiency of PCPPs has been improved greatly during the last three
decades (Bugge et al., 2006; Ber, 2007). The state-of-the-art, ultra-supercritical PCPPs,
have reached design efficiencies of 45%1 and beyond (IEA, 2012). This enhancement
is lead to mainly by employing advanced materials and improved design for key sin-
gle components (Blum et al., 2007); however, the system-level design remains almost
unchanged compared with that of the subcritical. Practitioners, such as Blum et al.
(2007), pointed out that there seems still a large potential to improve the steam/water
cycle itself. In addition, many technologies, such as solar thermal utilization (Yan et al.,
2010; Popov, 2011; Jamel et al., 2013) and waste heat recovery (Espatolero et al., 2010,
2014), are conceptually promising to augment conventional thermal power plants but
1 Unless otherwise noted, efficiency notations in this thesis are based on the lower heating value of the fuel.
1
2 introduction
for synthesizing thermal power plants. This framework employs hybrid optimization
integrating evolutionary and deterministic optimization for simultaneous alternative
generation and optimization in an automatic way. The extended framework is evalu-
ated by an illustrative example: the optimal synthesis of a Rankine-cycle-based ther-
mal power plant. Then, the superstructure-free framework is applied for synthesizing
complex pulverized-coal power plants by considering thermodynamic and economic
objectives separately and simultaneously for bi-objective trade-offs.
Finally, in chapter 6, a summary of the thesis and conclusions are drawn. In addition,
topics for future research are identified.
2
A N A LY S I S , O P T I M I Z AT I O N A N D
SYNTHESIS OF COAL-FIRED POWER
PLANTS
280/600/620 AD700
Ni-based
Austenite
270/580/600 State-of-the-art
technology
240/540/565
167/540/540
Mature technology
advanced USC
supercritical or ultra-supercritical
subcritical
Material development milestones
1960 1980 2000 2020
5
6 analysis, optimization and synthesis of coal-fired power plants
Boiler
VHPT HPT-IPT LPT
Steam turbine
Economizer
Figure 2.2: Schematics of the advanced ultra-supercritical coal power plant under development
(Fukuda, 2010; IEA, 2012). (More reheating can be employed to further increase
the plant efficiency. The feedwater preheating system is not included. HPThigh-
pressure turbine, IPTintermediate-pressure turbine, LPTlow-pressure turbine)
HPT LPT
Boiler
...
ET
DA
pump
FWH
Figure 2.3: Key idea of the Master Cycle (adapted from Kjaer and Drinhaus, 2010, FWH
feedwater preheaters, DAde-aerator, ETa secondary turbine supplying bled steam
for feedwater preheaters)
work ability of the extracted steams). To address the potential overheat crisis of feed-
water preheaters and ensure the complete expansion of extracted steams, a modified
reheating scheme (Master Cycle (Silvestri Jr, 1995)) has been proposed. Key idea of the
Master Cycle (Fig. 2.3) is to employ a secondary turbine (ET) that receives non-reheated
steam, drives the boiler feed pump, and supplies bled steam for feedwater preheaters,
so that the superheat degrees of steam extractions can be significantly reduced. How-
ever, the impact of introducing a secondary turbine on the optimal design of the whole
system has been limited studied (Blum et al., 2007; Stepczy
nska
et al., 2012).
New challenges lying ahead are associated with system-level integration. The inte-
gration opportunity flourishes, as multiple fluids are involved with wide temperature
ranges (Fig. 2.4), e. g., flue gas (1301000 C), steam (35700 C), feedwater (25350 C)
and air (25400 C). On the one hand, there is a need to raise the heat utilization to the
level of overall system, which has not been achieved yet due to independent designs
of the boiler and turbine subsystems. On the other hand, the integration of many avail-
Figure 2.4: Fundamental considerations and new challenges for the design of coal power plants
8 analysis, optimization and synthesis of coal-fired power plants
1 The term, exergoeconomic optimization, may be misleading to some readers. In fact, it includes no opti-
mization techniques but only the exergoeconomic formulation of optimization problems, particularly the
objective function of product cost.
10 analysis, optimization and synthesis of coal-fired power plants
complex case considering the separate components of exergy. This approach has be-
come the most widely accepted exergoeconomic analysis method even for complex
energy systems, e. g., (Kanoglu et al., 2011; Kalinci et al., 2011; Al-Sulaiman et al., 2013;
Alkan et al., 2013), and has combined with mathematical algorithms for iterative op-
timization, e. g., (Tsatsaronis and Moran, 1997; Cziesla and Tsatsaronis, 2002; Seyyedi
et al., 2010a).
2 The term, functional analysis, does not imply that particular branch of mathematics, but a formal, docu-
mented determination of the functions of the system (Frangopoulos, 1983).
2.2 analysis of energy systems 11
the Scope-Oriented Thermoeconomics, which identified cost allocation criteria for dissi-
pative components, based on a possible non-arbitrary concept of Scope, and classified
the system components by Product Maker/Product Taker but not by the classical dissi-
pative/productive concepts. The subsequent optimization application, i. e., (Piacentino
and Cardona, 2010a), presented that the method enabled to disassemble the optimiza-
tion process and to recognize the formation structure of optimality, i. e., the specific
influence of any thermodynamic and economic parameter in the path toward the op-
timal design. Banerjee et al. (2012) proposed an extended thermoeconomics to allow
for revenue generating dissipative units and discussed the true cost of electricity for
systems with such potential. Despite these, it seems that the choice of the best residue
distribution among possible alternatives is still an open research line (Piacentino and
Cardona, 2010b).
Efforts were also made to enhance the ability of exergoeconomics. Paulus and Tsat-
saronis (2006) formulated the auxiliary equations for specific exergy revenues based
on SPECO, and presented "the highest price one would be willing to pay per unit
of exergy is the value of the exergy". Cardona and Piacentino (2006) extend exergoe-
conomics to analyze and design energy systems with continuously varying demands
and environmental conditions. Moreover, an advanced exergoeconomic analysis, devel-
oped by the research group of Tsatsaronis (Kelly, 2008; Kelly et al., 2009; Tsatsaronis,
2008; Gaggioli and Reini, 2014), is capable of identifying the sources and availability of
the capital investments and exergy-destruction costs.
Specific unavoidable
Specific unavoidable
invesment cost
entire exergy destruction within the same component (E D,k ) that would still appear
when all other components in the system operate in an ideal (or theoretical) way while
the kth component operates with its real exergetic efficiency ( k ) (Kelly, 2008; Kelly
et al., 2009). The exogenous exergy destruction within the kth component (E D,k
EX ) is the
2.2 analysis of energy systems 13
remaining part of the entire exergy destruction (E D,k ) and is caused by simultaneous
effects of the irreversibilities occurred in the remaining components. The exergy de-
struction E D,k
EX can also be expressed by a sum of the exogenous parts directly caused
by the rth component ( E EX,r ) plus a mexogenous (MX) exergy destruction term (E MX )
D,k D,k
(Morosuk and Tsatsaronis, 2008a; Tsatsaronis and Morosuk, 2010), caused by simulta-
neous interactions of other components. The endogenous and exogenous concepts are
different from malfunction/dysfunction, which are used in thermoeconomic diagnosis
based on the structural theory (for more details, see Kelly, 2008; Kelly et al., 2009).
To calculate the exergy destruction E D EN , an ideal thermodynamic cycle needs to be
defined first and then irreversibility of each component is introduced by turn (Morosuk
and Tsatsaronis, 2009a; Tsatsaronis et al., 2006a,b). This approach, however, is only
appropriate for the system without chemical reactors and heat exchangers, of which
the ideal operations are hard to define.
The endogenous investment cost of the kth component (Z kEN ) is reasonably deter-
mined by exergy product at the theoretical condition and the investment cost per unit
exergy product at the real condition: Z kEN = E P,k E P )k . Subsequently, the endoge-
EN ( Z/
2.2.3.4 Applications
The advanced exergy analysis and exergoeconomic analysis have been initially devel-
oped based on simple systems, including a refrigeration cycle (Morosuk and Tsatsa-
ronis, 2006; Tsatsaronis et al., 2006a,b; Morosuk and Tsatsaronis, 2008b, 2009a), a gas
turbine system (Kelly, 2008; Kelly et al., 2009; Morosuk and Tsatsaronis, 2009b), and
cogeneration systems (Kelly, 2008; Tsatsaronis and Morosuk, 2010). The application to
complex systems started from a series work of Petrakopoulou et al. (2011; 2011a; 2011b;
2012a; 2012b) for comparatively evaluating power plants with CO2 capture, which were
also the first practices of the advanced exergoenvironmental analysis. More recent work
have extended these advanced evaluation methods to various kinds of energy systems,
such as coal-fired power plants by the author (Wang et al., 2012b; Yang et al., 2013), nat-
ural gas electricity-generating facility (Akkalp et al., 2014a,b), a trigeneration system
(Aras et al., 2014), geothermal district heating systems (Hepbasli and Keebas, 2013),
an externally-fired combined-cycle power plant (Soltani et al., 2013) and even a rubber
factory (Vuckovic et al., 2014).
14 analysis, optimization and synthesis of coal-fired power plants
Figure 2.6: Complete splitting of the exergy destruction in an advanced exergetic analysis (Sor-
genfrei, 2016)
2.2.3.5 Limitations
It seems almost all applications to complex systems are combining the thermodynamic
approach with exergy balance approach. However, this has not been adequate for com-
plex systems, particularly when involving heat exchanger network (HEN). An ideal
operation in one heat exchanger could timely lead to defections of its adjacent heat
exchangers. The "reversible" adiabatic heater concept (Kelly, 2008) could only be an
alternative but not a true solution.
A complete splitting of avoidable and unavoidable exergy destructions relies on a
number of simulations by setting different components at corresponding conditions.
Rigorous mathematical formulations on the interactions among components have not
been developed. This would simplify the calculation procedure and for integrating the
concepts into iterative or mathematical optimization.
The exergy-based analysis of thermal power plants began from 1970s, practiced by
main contributors in the field, e. g., Kotas (1985), Tsatsaronis and Winhold (1985a),
Bejan et al. (1996), and Tsatsaronis (1999a). In the last decade, there has also been a
number of applications for coal-fired power plants, e. g., (Horlock et al., 1999; Dincer
and Al-Muslim, 2001; Sengupta et al., 2007; Aljundi, 2009; Erdem et al., 2009; Ray et al.,
2010; Suresh et al., 2010; Singh and Kaushik, 2013) based on exergy analysis, and for
exergoeconomic analysis, e. g., (Rosen and Dincer, 2003; Valero et al., 2002; Zhang et al.,
2006, 2007; Xiong et al., 2012a,b).
2.3 optimization of energy systems 15
Depending on whether discrete (i. e., integer) decision variables are incorporated, the
optimization problems are first classified as continuous and discrete. Then considering
the nature of functions involved, important subclasses are further identified: (continu-
ous) linear programming (LP), (continuous) nonlinear programming (NLP), integer program-
ming (IP), mixed integer linear programming (MILP), mixed integer nonlinear programming
(MINLP), generalized disjunctive programming (GDP), etc.
The algorithms for different optimization problems, either deterministic or metaheuris-
tic (Glover and Kochenberger, 2003), have been well developed and exhaustively re-
viewed in many references, e. g., a comprehensive description of the most effective
16 analysis, optimization and synthesis of coal-fired power plants
3 www.solver.com/linear-quadratic-technology
2.3 optimization of energy systems 17
solutions are still hardly obtained within acceptable computation times, the nonlinear-
ities, if possible, should be avoided at the formulation phase to enable robust (LP) or
(MILP) optimization (Grossmann and Ruiz, 2012).
In addition, state-of-the-art solvers for deterministic optimization have been highly
integrated to several well-developed high-level algebraic modeling environments, e. g.,
GAMS and AMPL, tailored for complex, large-scale applications.
Evolutionary algorithms
Evolutionary algorithms (EAs), inspired by biological evolution, are generic, stochastic,
derivative-free, population-based, direct search techniques. EAs can often outperform
derivative-based deterministic algorithms for complex real-world problems, even with
multi-modal, noncontinuous objective function, incoherent solution space, and discrete
decision variables; moreover, the global optimality, although not guaranteed, can be
closely approached within a limited number of function evaluations.
The basic run (Fig. 2.7) of an evolution algorithm (EA) starts from an initialization, in
which a set of candidate solutions (population and individuals) are proposed and eval-
uated for assigning the fitnesses (the objective function value, if feasible; otherwise, a
penalty value). Afterwards, for evolving the current parent population to an offspring pop-
ulation, the algorithm starts an iteration loop: parent selection, recombination (crossover),
mutation, evaluation and offspring selection. In order to produce each new individual,
based on the fitness values, one or more parents are selected for crossover and mu-
tation: A crossover operation randomly takes and reassembles parts of the selected
parents, whereas a mutation operation performs a small random perturbation of one
individual. The newly born offsprings are then evaluated; finally, a ranking of offspring
2.3 optimization of energy systems 19
Termination Initialization
Selection Recombination
(Crossover)
Evaluation Mutation
(and parent) individuals is performed, so that those individuals with larger possibility
of leading to the optimality survive and are selected as the offspring population. The
iteration continues until certain termination criterion, e. g., a limit of computation time,
fitness-evaluation number, or generation number, is reached.
Selection, crossover and mutation are three genetic operators of evolutionary algo-
rithms for maintaining local intensification and diversification of the search. Different
strategies on these three aspects lead to a variety of evolutionary algorithms. Selection
strategy mainly exerts influence on population diversity. One commonly used strategy
of selection is the ( + )-selection proposed in evolution strategies (Beyer and Schwe-
fel, 2002), where and , satisfying 1 , denote the sizes of parent and offspring
populations, respectively. Selection ranks the fitness of all + individuals and takes
the best individuals.
Depending on the search space and objective function, the crossover and/or the
mutation may or may not occur in specific instantiations of the algorithm (Beyer and
Schwefel, 2002; Glover and Kochenberger, 2003). Crossover intensifies the local search
but is not essential, as self-adaption of individuals is available due to mutation, which,
in contrast, is usually dispensable to avoid premature convergence toward sub-optimal
solutions. There are different mechanisms of crossover and mutation. For example, ge-
netic algorithm (Deb et al., 2002) usually employs bit strings to represent variables.
Typically, the crossover and mutation for integer variables can be readily and straight-
forwardly effected by gray coding, thus genetic algorithm is mainly applied for solving
combinatorial optimization problems. However, differential evolution (DE (Storn and
Price, 1997)), mentioned as the fastest evolution algorithm (Storn and Price, 1997), does
not rely on any coding but directly manipulates real-valued or discrete variables. Ba-
sically, for mutation DE adds the weighted difference between two parents variable
vectors to a third vector, thus the scheme remains completely self-organizing without
using separate probability distribution and has no limitations for implementation com-
pared to other evolutionary algorithms.
The optimization problems of thermal energy systems are usually highly constrained
and nonlinear, thus belong to (NLP) or (MINLP). The nonlinearity and integrity may
be led to by thermodynamic properties of working fluids, design and operational char-
acteristics of components, the investment cost functions of components, energy balance
20 analysis, optimization and synthesis of coal-fired power plants
equations, etc. These need to be well addressed, so that the problems, in the best case,
can be transformed to (LP) or (MILP) for deterministic optimizations.
For the properties of working fluids, particularly water and steam (IAPWS-IF97
(Wagner et al., 2000)), the highly nonlinear exact mathematical formulations can hardly
be employed. One direct means incorporates polynomial approximations of low de-
grees of nonlinearity in expense of accuracy (Savola, 2007; Jdes, 2009; Luo et al.,
2011; Manassaldi et al., 2011). However, inaccurate regressions may result frequently
in non-applicable "optimal" solutions. Moreover, integer variables may be explicitly
introduced if involving zonal or piecewise regressions.
Another approach evaluates the propertys value and associated derivatives of high
accuracy based on reformulated exact formulations or reprocessed steam tables. Exter-
nal evaluations can be fast and stable, such as in freesteam4 (based on reformulated
IAPWS-IF97), or in TILMedia Suite5 (bi-cubic spline interpolation of standard steam
tables Schulze 2014). Moreover, in these libraries, the discontinuities and even jumps
of the thermodynamic properties are smoothed, and the integer variables indicating
the state zones are encapsulated.
The nonlinear (or perhaps discrete) thermodynamic (operational) behavior of com-
ponents can be properly reformulated. For example, for modeling turbine, alternatives
include constant entropy efficiency model, Willans Line (Pachernegg, 1969), Turbine
Hardware Model (Mavromatis and Kokossis, 1998) and Stodola ellipse (Cooke, 1983).
In those models, the set of variables which the isentropic efficiency depends on dif-
fers, thus the predictions of the off-design behavior are also different in accuracy. For
heat exchangers, the logarithmic mean temperature difference can be replaced by a
refinement of the arithmetic mean (Paterson, 1984; Chen, 1987). While for mixers, the
discrete equality nonlinear relationship of the flow pressures between inlets and outlet
can be either relaxed as an inequality nonlinear constraint (Drud, 1994) or linearized
by introducing additional integer variables (FICO, 2009).
The investment cost functions are always needed if an economic objective is involved
in the optimization. A cost function links the purchased equipment cost of one com-
ponents with its key characteristic variables and associated flow parameters; thus, the
function may be of high nonlinearity. To cope with this, cost functions are usually re-
formulated with separable terms of each variable, which are subsequently piecewise
linearized with the aid of integer SOS2 variables (Tomlin, 1988).
Continuous nonconvex bilinear term (v1 v2 ) is another common source of nonlinear-
ity, e.g., the term m h involved in energy balance equations. This nonconvex nonlin-
earity is usually handled by a convex/concave McCormick relaxation (Tsoukalas and
Mitsos, 2014) or a quadratic reformulation. For the latter approach, two new variables
z1 = (v1 + v2 ) /2 and z2 = (v1 v2 ) /2 are introduced to replace the bilinear term
with z21 z22 . The quadratic term can also be further linearized by SOS2 variables.
Given a specific structure, the parametric selection and optimization of a steam cycle
becomes an easy task. Early work on optimizing steam cycles date back to 1950s and fo-
cus on feedwater-preheater train. For example, Haywood (1949) and Weir (1960, 1964)
analytically determined the optimum range of feed heating and the optimum distribu-
tion of heating amongst the cascaded heaters. These were subsequently simplified and
summarized as two commonly used design methods for feed heating: equal enthalpy
rise and equal temperature rise (Potter, 1988). However, the analytical deductions for
optimal solutions basically rely on many assumptions and are limited primarily to
serial connection of feedwater preheaters.
The selection of reheat parameters has also been a classical engineering task and dis-
cussed elsewhere, e. g., (Silvestri et al., 1992; Habib et al., 1995). In particular, Silvestri
et al. (1992) reviewed the importance of various parameters for better plant designs,
and explored the combinations (or matches) of throttle pressure, throttle (main) and
reheat temperature and the number of reheatings. However, these studies are generally
based on sensitivity analysis.
Subsequent studies mostly employ mathematical programming. Uche et al. (2001)
and Xiong et al. (2012a) employed thermoeconomic decomposition for a given steam
cycle, reformulated the optimization problem to an appropriate quadratic programming
(QP) approximation, and performed local optimization for design parameters. Sanaye
et al. (2003) investigated turbine extraction pressures of eight open and closed feed
water heaters by three derivative-free algorithms. Suresh et al. (2011) and Hajabdollahi
et al. (2012) coupled genetic algorithm and artificial neural network to determine the
maximum design or operation efficiency. Espatolero et al. (2014) evaluated several lay-
outs of feedwater heaters and flue gas heat recovery system, and optimized the bleed
pressures by SQP algorithms.
These studies only explore a limited number of design structural alternatives and,
more importantly, the structural options are generated not in a systematic way. Conse-
quently, the best solutions searched may be far away from the optimal solution.
The heuristic and targeting methods are knowledge-based. The heuristic methods
incorporate rules derived from long-term engineering knowledge and experience. The
aims are to propose "reasonable" initial solutions and improve them sequentially. One
influential method in this group is the hierarchical decision procedure for process syn-
thesis (Douglas, 1985), which introduces common concepts for almost any systematic
synthesis method proposed afterwards, such as (Jaksland et al., 1995; Kravanja and
Grossmann, 1997). The method explores the process nature by sequential decomposi-
tion and aggregation for further improvement (Douglas, 1988) and has been extended
for synthesizing complete flowsheet of the separation system (Douglas, 1995). Other
heuristic rules based methods and practices can be found elsewhere, e. g., (Li and
Kraslawski, 2004).
The targeting methods integrate physical principles to obtain, approach and even
reach the targets for the optimal process synthesis. The most widely applied targeting
method is the pinch methodology (Linnhoff, 1993), which is fundamentally developed
for systematic synthesis of HEN. The method has been extended for complete flow-
sheet synthesis of total site utility systems (Mavromatis and Kokossis, 1998; Matsuda
et al., 2009).
To realize automatic and computer-aided synthesis using these guidelines, a number
of knowledge-based expert systems have been developed for various processes and sys-
tems, such as chemical processes (Siirola and Rudd, 1971; Mahalec and Motard, 1977;
Kirkwood et al., 1988), thermal processes (Sciubba, 1998; Manolas et al., 2001; Matelli
et al., 2009) and renewable energy supply systems (Chen et al., 2007). Expert systems
apply various logical inference procedures, e. g., means-end analysis (Siirola, 1996) and
case-base reasoning (Surma and Braunschweig, 1996), to reproduce engineers design
maps, thus suggest the best suited process for a particular application.
The heuristic and targeting methods are generally effective to quickly identify sub-
optimal structural alternatives (Li and Kraslawski, 2004). However, they are unable
to guarantee the optimality, mainly because of the sequential nature and mathemati-
cally non-rigorousness. Thus, much more comprehensive methods, the mathematical
optimization based methods, have been greatly developed.
The optimization-based methods consider simultaneously the structural options, de-
sign and operation conditions, and perform rigorously with any objective function
(Frangopoulos et al., 2002). In these methods, a synthesis task is formulated as a
mathematical optimization problem with an explicit (superstructure-based) or implicit
(superstructure-free) representation of considered structural alternatives, among which
the optimal structure is identified (Grossmann and Guilln-Goslbez, 2010). In the fol-
lowing, the optimization-based synthesis methods are reviewed in more details.
Grossmann and Guilln-Goslbez, 2010), which has been widely applied to a multi-
tude of process synthesis with different levels of detail, such as HEN (Luo et al., 2009;
Escobar et al., 2014), separation and distillation sequences (Skiborowski et al., 2012),
water networks (Ahmetovic and Grossmann, 2011), polygeneration process Liu et al.
(2010a), steam utility systems (Luo et al., 2011, 2012), and thermal power plants (Jiang
et al., 2002; Grekas and Frangopoulos, 2007; Ahadi-Oskui et al., 2006, 2010).
The superstructure-based synthesis aims at locating the optimal solution from all
possible alternatives embedded in the superstructure, which represents all considered
components and the possible links. The fundamental basis of the superstructure-based
synthesis involves three aspects: superstructure representation and generation, super-
structure modeling and mathematical optimization of the problem.
s {0, 1}n ,
where the vector s contains n binary structural variables indicating the (non-)existence
of components for design synthesis and the on/off-state of components (when involv-
ing operation synthesis). Note that a superstructure can be formulated at different
levels of details (Grossmann and Guilln-Goslbez, 2010): (a) aggregate models concern-
ing only major features like energy balance, e. g., (Yee et al., 1990; Wang et al., 2014),
(b) short-cut models considering simple nonlinear models for component performance,
e. g., surrogate models (Henao and Maravelias, 2011), and (c) rigorous models involving
detailed modeling of component performance, e. g., (Manassaldi et al., 2011; Martelli
et al., 2011). The solving algorithms have been introduced in section 2.3.1.
Since the whole model is usually difficult and expensive to solve, many speedup
techniques have been developed for different applications. For instance, several decom-
position methods, e. g., (Kocis and Grossmann, 1989; Bagajewicz and Manousiouthakis,
1992; Frangopoulos, 1992; Papalexandri and Pistikopoulos, 1996; Daichendt and Gross-
mann, 1998), are capable of partitioning the superstructure into several subproblems
of smaller size. Another approach implicitly indicates the existence of considered com-
ponents by using continuous variables, e. g., use zero mass flow rate to bypass the
components for non-existence (Lang and Biegler, 2002; Stein et al., 2004; Krmer et al.,
2009). In this way, the discrete decision variables are eliminated and the synthesis
problems are reformulated to continuous optimization problems; however, the quality
of local solution highly depends on initial specifications.
For addressing the global optimization with many discrete decision variables, hy-
brid algorithms combining metaheuristic algorithms and mathematical programming
(memetic algorithm) become popular. For example, Urselmann et al. (2011a,b) proposed
a two-level memetic algorithm, where in the upper level the integrity constraints and
discontinuous cost functions are handled by genetic algorithm, while in the lower level
continuous sub-problems are efficiently solved by robust solvers of mathematical pro-
gramming for state variables (Urselmann and Engell, 2015).
tation (e. g., S-graph, see section 2.4.1.1) for generating structurally feasible structures,
while the lower level evaluates the generated structures.
For HEN synthesis, Fraga (2009) proposed a set of grammars for a string representa-
tions to add heat exchangers and split streams. With a string rewriting system, genetic
algorithm is capable of generating complex networks. Toffolo (2009) proposed a more
flexible graph representation, with which genetic algorithms were used to perform in-
sertion and deletion of heat exchanger, and swaps of hot and cold sides of two heat
exchangers. However, these approaches are tailored to HEN synthesis.
Wright et al. (2008) performed both mutation and crossover to heating, ventilating
and air conditioning system for an evolutionary synthesis. The mutation swaps two
randomly selected components or their interconnections, while the crossover allows
the offsprings inheriting structural properties and technical specifications from two
parents either separately or in an equal measure. However, this approach is basically
incapable of being extended to other applications.
Toffolo (2014) proposed a hybrid algorithm for complete flowsheet synthesis of ther-
mal power plants. The approach decomposes a thermal system into the heat transfer
section and the remaining parts (basic configuration) by a heat separation decomposition
(Lazzaretto and Toffolo, 2008). The algorithm sequentially synthesizes the basic config-
uration by genetic algorithm and SQP, and the heat transfer section by pinch method.
However, the special but complex codification developed for synthesizing basic config-
uration (see Toffolo, 2014 for more details) are not advantageous for retrofit synthesis.
In addition, the approach is appropriate when considering only four fundamental ther-
modynamic processes (compression, heating, expansion and cooling) but not chemical
reactions.
Emmerich et al. (2001) proposed a S-graph based genetic algorithm making total
flowsheet synthesis of energy and process systems more flexible. A set of symmet-
ric replacement rules for generating the closest neighboring structures are defined as
minimal moves, such as insert a heater parallel to an existing heater or swap a by-product
stream with a recycle stream. The minimal move mutation operator recognizes existing
patterns, such as one component or a set of components, and replaces them with sim-
ilar patterns according to the replacement rules; while the crossover operator recog-
nizes and swaps the subsystems in the parents, which possess the same function and
similar connection patterns. This approach has been applied to chemical process (Em-
merich et al., 2001; Urselmann et al., 2007; Sand et al., 2008) and thermal power plant
(Emmerich, 2002); however, the major problem lies in the problem-specific replace-
ment rules, which largely limit its extendability. To cope with this problem, Voll et al.
(2012) further developed this approach by combing an energy conversion hierarchy,
which allows for generic replacement rules. The energy conversion hierarchy based
superstructure-free approach is promising but currently tailored to the synthesis of
distributed energy supply systems.
For superstructure-based synthesis, Chen et al. (1997) and Hillermeier et al. (2000) ap-
plied a (modified) decision tree of different types of feedwater preheater to define the
superstructure of feedwater-preheater train. Genetic algorithm is then used to explore
2.5 multi-objective optimization 27
all structures embedded in the superstructure. However, in this study, the structural
solution space are largely constrained; moreover, the applications are tailored to syn-
thesizing feedwater-preheater strings while no degrees of freedom are given to turbine
stages and reheatings.
For superstructure-free synthesis, the approaches proposed by Toffolo (2014) and
Emmerich et al. (2001) have been discussed in section 2.4.2. Emmerich (2002) applied S-
graph based evolutionary algorithm to synthesize the feedwater-preheater train; how-
ever, the structure evolution is based on a large set of technology-specific mutation
rules that have to be specified by the user manually. Manual specification of these mu-
tation rules can be considered as difficult and error-prone as the manual definition of
an appropriate superstructure model.
Therefore, the synthesis of steam cycles is still an open area. In this thesis, the fun-
damental design features (section 2.4) are discussed based on a large graphical su-
perstructure; while a more flexible synthesis method for complex cycles is developed
according to the energy conversion hierarchy based superstructure-free approach.
with the vector x (x Rn ) denoting the independent decision variables in the feasible
solution space X. The vector f represents k objective functions f k : Rn R1 . A solution
x becomes a Pareto solution x, once it is Pareto optimal: There does not exist another
feasible solution x, which satisfies the conditions f ( x ) f ( x ), and f i ( x ) < f i ( x ) for
at least one objective function i. In short, no other solutions dominate a Pareto solu-
tion. Thus, the Pareto solutions lie on the boundary of the objective space; however, in
practice, the Pareto front can only be approximated by the boundary of the attainable
objective solution space (Marler and Arora, 2004). Thus, MOO aims at approximating
high-quality Pareto fronts, where Pareto solutions are dense enough and spread evenly.
There are basically three techniques for dealing with MOO: A priori technique incor-
porates the decision makers preferences before the search, such as the weighted sum
28 analysis, optimization and synthesis of coal-fired power plants
method (Marler and Arora, 2010). The interactive technique integrates generating and
choosing of the preferred solutions throughout the search, such as the physical program-
ming method (Messac and Ismail-Yahaya, 2002). Thus, the first two techniques donot
generate the Pareto fronts. However, a posteriori technique first generates the Pareto set,
among which the preferred is chosen later. A posteriori methods are most commonly
applied in practice, such as the weighted sum method, the e-constraint method (Mavro-
tas, 2009), the normalized normal constraint method (Messac et al., 2003) and evolutionary
multi-objective optimization algorithms (EMOAs (Zitzler and Thiele, 1999; Deb, 2001)).
where the weighting factors are positive and normalized (wi (0, 1) and ik=1 wi =
1). For a priori technique, weighting factors are specified before the optimization run,
while the factors are systematically adjusted during the run for a posteriori technique.
However, the method fails to capture solutions on non-convex parts of the Pareto front;
moreover, evenly-distributed Pareto solutions are generally difficult to yield.
min f i ( x ), i = 1, 2, . . . , m, (2.6)
x
s.t. f j ( x ) e j,k , j = 1, 2, . . . , m; j 6= i; k = 1, 2, . . . , n
x X.
For bi-objective problem, the e-constraint is, in fact, a horizontal or vertical line. A
graphical representation for bi-objective problems is shown in Fig. 2.8: After obtaining
the two anchor solutions 1 by minimizing the objective function 1 and 2 by min-
imizing the objective function 2, the objective function space are divided into many
subspaces by subsequently introducing the bounds of objective function 1, e. g., e1 , e2 ,
e3 and en . The solution of each subproblem is a Pareto solution, e. g., s1 for e1 and s2 for
e2 . The method performs effectively; however, the quality of the Pareto fronts obtained
by the method depends on the slope (shape) of the Pareto front and the division of the
objective space: The higher the front slope, the denser the division should be to obtain
evenly-spread Pareto solutions (Fig. 2.8). The value of e usually has to be successively
and algorithmically modified for each division of objective space to find high-quality
Pareto fronts. Additionally, the method fails to solve practical problems with certain
black-box features.
2.5 multi-objective optimization 29
Objective function 2
Objective function 1
NU1
NU2
NU3
Objective function 2
NU4
NU5
Utopia line
Objective function 1
Figure 2.9: Graphical representation of the normal constraint method for bi-objective problems
30 analysis, optimization and synthesis of coal-fired power plants
Reference point
Objective function 2
Hypervolume of
solution
Figure 2.10: Solution selection in SMS-EMOA: three dominant solutions after certain genera-
tions of evolution with the hyper-volumes referring to the reference point, and the
difference between Pareto fronts obtained by both NSGA-II and SMS-EMOA.
2.6 contribution of the thesis 31
It employs also the same non-dominated sorting but uses the S-merit as the secondary
ranking criterion instead of crowding distance. The S-merit of a solution is defined as
the hyper-volume spanned from the solution to the reference point, which is a worst
possible solution. For selection, when the solutions are with identical non-domination
rank, the solution with the smallest S-merit is discarded (e. g., solution x1 in Fig. 2.10),
as the solution is more unlikely to evolve to a good compromise.
In addition, the ideas of non-dominated sorting, and crowding-distance/S-merit as-
signment can be easily integrated to other population-based metaheuristic algorithms,
such as multi-objective differential evolution (MODE (Madavan, 2002; Iorio and Li, 2005)).
There have been many studies on multi-objective optimization of thermal energy sys-
tems. For example, Toffolo and Lazzaretto (2002) and Lazzaretto and Toffolo (2004)
applied an EMOA to the CGAM problem (Valero et al., 1994b) for the trade-off among
three objectives: the thermodynamic, economic and/or environmental performance.
Kavvadias and Maroulis (2010) investigated the multi-objective design of an energy
trigeneration system considering different energy tariffs and operation strategies by
using an EMOA. Fazlollahi et al. (2012) evaluated EMOAs and e-constraint method
combining integer cut constraint for (MILP) of complex energy systems. The combined
integer cut constraints enable the capability of generating more good solutions close to
Pareto front but not only Pareto solutions. Liu et al. (2010b) performed the e-constraint
method to solve a bi-objective synthesis of polygeneration energy systems, modeled as
non-convex (MINLP).
To the authors best knowledge, no work has been published for multi-objective
optimization or synthesis of coal-fired power plants. In addition, except for Pareto
front, practitioners may be also interested in other parts of the objective space for extra
information, which requires modifications of existing algorithms.
power plant is simulated by modeling all its individual components. In particular, the
boiler subsystem is not treated as one single unit but modeled by a coal combustor
and a series of heat exchangers. To my best knowledge, this application would be
the first application, giving such complete information of modern ultra-supercritical
pulverized-coal power plants based on the 2nd law of thermodynamics.
With respect to methodological aspects, a new approach for calculating endogenous
exergy destructions is introduced and further discussed for pulverized-coal power
plants. The key idea of the approach is neglecting how the exergy is transferred or
converted within the reversible components, but only establishing an overall exergy
balance equation for the reversible blackbox. Thereby, the approach avoids modeling
of theoretically-operated chemical reactors and heat exchangers. The approach has
been originally proposed for open systems; however, in the thesis, it is applied for the
first time to pulverized-coal power plants with closed steam cycles, and open steam-
generating and -heating systems.
not only the Pareto front but also any other desired front of the objective space in a
high-quality, smooth and consistent fashion.
3
C O M P R E H E N S I V E E X E R G Y- B A S E D
ASSESSMENT OF COAL-FIRED POWER
PLANTS
This chapter presents a deep and complete discussion of the performance of large-scale
pulverized-coal power plant (PCPP) and potential improvement strategies. The gen-
eral framework for comprehensively evaluating thermal systems based on advanced
exergy (exergoeconomic) analysis is introduced first in section 3.1. In section 3.2, dif-
ferent procedures for calculating endogenous and unavoidable exergy destructions
within different types of components, which are key entries for a complete analysis of
thermal systems, are presented in detail. Subsequently, an existing modern 1000 MW
ultra-supercritical PCPP (section 3.3) is chosen to be assessed. Different from previ-
ous work on the evaluation of PCPPs (see section 2.2.4), the boiler of the selected
ultra-supercritical PCPP is no longer treated as a blackbox but as a coal combustion
chamber plus a series of heat exchangers, thereby the energy balance of the boiler is ful-
filled in detail. Thereafter, the ultra-supercritical PCPP is evaluated by the framework
(section 3.3), with improvement suggestions given in section 3.4. Finally, section 3.5
concludes the evaluation and the capability of the framework.
35
36 comprehensive exergy-based assessment of coal-fired power plants
due to the stream connections, thus the performance of a component is always influ-
enced by the inefficiencies of other components in the same system. Moreover, there
are usually technical and economic limitations in practice on improving a single com-
ponent and the system, which leads to unavoidable thermodynamic inefficiencies and
costs. These two basic splittings can be categorized as avoidable/endogenous, unavoid-
able/endogenous, avoidable/exogenous and unavoidable/exogenous parts.
Details on conventional and advanced exergy (exergoeconomic) analyses are given
in section 3.1.13.1.4. Key quantities for evaluating the performances of a system and
individual components are listed in Table 3.1. The calculation of these quantities is
based on basic energy-balance simulations. All necessary simulations and the entries
to calculate the remaining terms are listed in Table 3.2, where the unavoidable and
theoretical conditions have been discussed in section 2.2.3.1 and 2.2.3.2.
Table 3.2: Simulations needed for calculating all terms (Nc the number of components)
for the overall system (tot) E F,tot E P,tot = E D,tot + E L,tot = E D,k + E L,j , (3.1)
k j
where the subscripts F, P, D and L stand for the fuel, product, destruction and loss,
while the terms E, and y are the exergy flow, exergy efficiency and exergy destruction
ratio, respectively.
The advanced exergy analysis splits the exergy destruction within the kth component,
and evaluates the interactions between two (sets of) components:
UN
E D,k
UN
= E P,k E D / E P k
, (3.8)
E D,k
AV
= E D,k E D,k
UN
, (3.9)
E EX = E D,k E EN ,
D,k D,k (3.10)
UN,EN UN
E D,k = E P,k
EN
E D / E P k
, (3.11)
UN,EX EUN E UN,EN ,
E D,k = D,k D,k (3.12)
AV,EN UN,EN
E D,k = E D,k
EN
E D,k , (3.13)
AV,EX UN,EX
E D,k = E D,k
EX
E D,k , (3.14)
E D,k
EX
= EX,r
E D,k + E D,k
MX
, (3.15)
r
EX,r EN,k +r
E D,k = E D,k E D,k
EN
, (3.16)
UN,EX,r EN,k +r UN UN,EN
E D,k = E P,k E D / E P k
E D,k , (3.17)
AV,EX,r EX,r UN,EX,r
E D,k = E D,k E D,k , (3.18)
where the superscripts UN, AV, EN, EX and MX, are unavoidable, avoidable, endogenous,
EX,r
exogenous and mexogenous. The term E D,k is the part of exergy destruction within the
kth component caused by the irreversibility in the rth component (k 6= r). With the
exergy destruction E D,k already known, the calculation of the terms E D,k EN , E EN,k +r and
D,k
UN
E D / E P
k
(from system simulations IIII in Table 3.2) enable the determination of all
the remaining terms (splittings).
capital investment cost (CC) can be determined (for more details, see (Bejan et al.,
1996)). Thereafter, the COE can be calculated given the levelized fuel cost (FC) and
operation and maintenance cost (OMC):
where the subscripts i and j are flow indices, while the terms Nin and Nout are the
number of inlet and outlet streams of component k. For each component, the incoming
cost flows, investment cost rates and all exergy flow rates are known, whereby Nout 1
auxiliary equations are needed to fully close the equation set. The auxiliary equations
are derived from the F and P principles in SPECO method (Lazzaretto and Tsatsaronis,
2006). The F principle states that the specific cost (cost per exergy unit) associated with
this removal of exergy from a fuel stream must be equal to the average specific cost at
which the removed exergy was supplied to the same stream in upstream components;
while the P principle states that each exergy unit is supplied to any stream associated
with the product at the same average cost.
For a dissipative component k, there is no productive purpose: it is hardly possible
to define a exergy product when the component is considered in isolation. Within
such components, the exergy destructions or losses occur without causing anything
thermodynamically useful directly. However, based on the F principle, the specific
costs of the incoming and outgoing exergy streams are specified as the same. Therefore,
there is a difference between the incoming and outgoing cost rates, termed as C dif,dc :
Nin Nout
E c + Z k + C aux,dc,k = E c + C dif,dc,k ,
in,i out,i
(3.22)
i =1 i =1
where the term C aux,dc indicates the cost rate of additional working fluids. All fictitious
cost rates C dif,dc are charged to the cost of final product in the thesis.
The specific cost of exergy loss is considered as equal to that of fuel exergy of the
component k, from which the exergy loss is ejected to the environment:
C L,j = cF,k E L,j ,
Similarly, all cost rates of the exergy losses are charged to the cost of the final product.
Combining cost balance equations and all auxiliary equations, the linear matrix for-
mulation is explicitly formed:
Ac = b or
I C=b, (3.23)
3.1 framework for comprehensive exergy-based evaluation 39
where the coefficient matrices A and I have dimensions Ns Ns (Ns the number of
exergy flows of the whole system); while the matrices c, C and b (Ns 1) are matrices
of specific exergy costs, flow cost rates and investment cost rates, respectively. The
matrix b (Ns 1) includes Z k for each cost balance equation and 0 for all auxiliary
equations. The linear matrix equation can be straightforwardly solved, thus the specific
cost and cost rate of each stream are obtained.
After knowing the specific costs of all exergy streams and identifying the cost rates
of the fuel and product of each component, the thermoeconomic variables, i. e., aver-
age unit costs of the fuel cF,k and the product cP,k , the cost rate of exergy destruction
C D,k , the summation C D + Z k , the relative cost difference rk and the exergoeconomic
factor f k , are calculated for presenting the formation process of final product cost and
evaluating the components:
Once the cost rates of total exergy destructions C D,tot and total exergy losses C L,tot
are known, the cost of electricity can be calculated in the following way as well:
where the term C P,tot is the sum of cost rates of exergy products of all components.
The costs of electricity calculated from both Eq. 3.19 and Eq. 3.29 should be the same,
which is useful to ensure a correct exergoeconomic analysis.
The entries for splitting all investment costs and exergy destruction costs are terms
E P UN in Eq. 3.30, Z/
E P in Eq. 3.31, Z UN / E P in Eq. 3.34, calculated based on
Z/ k k k
simulations I and II in Table 3.2. With the calculated terms Z kUN , Z kEN and Z kUN,EN , the
remaining splittings of investment costs can be readily obtained as follows:
E P UN ,
Z kUN = E P,k Z/
(3.30)
k
EN EN
Zk = EP,k Z/ EP k , (3.31)
Z AV = Z k Z UN ,
k k (3.32)
ZkEX = Z k
ZkEN , (3.33)
Z kUN,EN = E P,k
EN
Z UN / E P , (3.34)
k
Z kUN,EX = Z kUN Z kUN,EN , (3.35)
Z kAV,EN = Z kEN Z kUN,EN , (3.36)
Z kAV,EX = Z kEX Z kUN,EX , (3.37)
40 comprehensive exergy-based assessment of coal-fired power plants
Given the splitting of exergy destruction, the cost rate of exergy destruction within
each component is split by simply multiplying the average specific exergy cost cF,k and
the amount of exergy destruction split.
The new concept to calculate the endogenous exergy destruction is introduced by Math-
ias Penkuhn in Institute for Energy Technology, Technical University of Berlin. The
concept has been examplarily applied to the CGAM problem (Penkuhn, 2015). The
basis of the new concept is that the nature of an ideal reversible process or system
defines the relation between the exergy input and output. This feature pinpoints that
the details on how the exergy is transferred or converted within a reversible process
is not significant when constructing each system I. Therefore, each system I can be
simplified: The considered component under its real condition is connected with a
3.2 calculation of endogenous and unavoidable exergy destructions 41
pump
preheater 1
preheater
11 exhausted fluegas 7
For the ideal system and all systems I, the following parameters are kept the same as
in the real case: temperature of the feedwater into boiler (stream 1), temperature and
pressure of the main steam (stream 2), pressure of the steam to be reheated (stream
3), temperature of the reheated steam (stream 4), temperatures of exhausted flue gas
(stream 11) and slag (stream 10), excess air ratio, slag ratio, the mass flow ratio m 4 /m 2 ,
and net power output. Note that the influence of the boiler imposed on the turbine
subsystem comes only from the pressure losses of working fluids. When the boiler
works theoretically, the mass flow rates of main and reheated steams are completely
determined by the turbine subsystem (e. g., Eq. 3.43). Therefore, for evaluating the
endogenous exergy destruction within a specific component of the turbine subsystem,
only the exergy balance of the turbine subsystem needs to be considered.
In this approach, for each productive component, three equations are established:
One equation scales the mass flow rate of the fuel of the whole system or subsystems,
while the other two express the exergy balances of the considered component. To solve
these three equations for the endogenous exergy destruction of considered component,
the specific exergies and mass flow rates of involved streams have to be determined.
The involved specific exergies can be calculated either a) after removing all pressure
drops in other theoretically-operated components and coupling with the preceding
theoretically-operated turbine or pump (if exist), or b) by setting the same values as
in the real case. The involved mass flow rates are considered to be proportional to the
main streams (as in the real case) to ensure a fixed power output.
In the following, the equations for determining endogenous exergy destructions
within different components are given in detail. The types of components considered
42 comprehensive exergy-based assessment of coal-fired power plants
a) b)
2 4 2 4
13 . .
Pnet Reversible Pnet
3 12 3
5 energy conversion 5
14 16 15
33
1 Reversible 1 32
energy conversion
17
c) d)
2 4 2 4
. .
Pnet Pnet
3 Reversible 3 Reversible
energy conversion 5 energy conversion 5
24 20
1 23 25 1
22 21 19 18
e) f)
2 4 2 4
27 3 Reversible 3
coal coal energy conversion
Reversible
8 energy 28 26 8
conversion 31
9 29 1 9 1
air air 30
10 11 11 10
slag exhausted fluegas exhausted fluegas slag
Figure 3.2: Systems I for calculating the endogenous exergy destructions within a) steam tur-
bine, b) condenser, c) feedwater preheater, d) pump, e) fluegas-water (steam) heat
exchanger, and f) air preheater
include steam turbine (ST), condenser (COND), pump, FWH, fluegas-water (steam)
heat exchanger, air preheater, and fan. The thermodynamic models for all these com-
ponent types are listed in App. A.
Steam turbine
The steam turbine expands steam from a given pressure to a specific pressure. When
the mass flow rate of the reheat steam is specified to be proportional to that of the main
steam, the systems I for all turbines, whether before or after reheat, can be represented
by Fig. 3.2a. The exergy balance equations are, therefore, given as follows:
(e2 e1 ) m 2 + (e4 e3 ) m 4 E D,ST
EN
= Pnet , (3.44)
(e12 e13 ) m 12 E P,ST
EN
E D,ST
EN
= 0, (3.45)
ST (e12 e13 ) m 12 E P,ST
EN
= 0, (3.46)
where the ratios m 4 /m 2 and m 12 /m 2 are kept the same as in the real case. Given fixed
parameters t1 , p2 , t2 , p3 and t4 , the specific exergies e1 , e2 , e3 and e4 can be determined
after removing all pressure losses in the boiler subsystem: p1 = p2 and p4 = p3 . The
temperatures t3 and t12 depend on the turbine expansion before reheat or the con-
sidered turbine. Therefore, the above equation set is closed for the three unknown
variables m 2 , E P,ST
EN and E EN . Compared with the real case, the mass flow rates of main
D,ST
steam, reheat steam and steam flowing into the turbine decrease in the system I by
multiplying the same fraction.
3.2 calculation of endogenous and unavoidable exergy destructions 43
Condenser
Similarly, the mass flows rates m 4 , m 15 , m 16 are proportional to m 2 ; the specific exergies
e1 e4 are calculated as above mentioned; the variables e15 and h15 depend on an ideal
steam expansion; the specific exergies and enthalpies of streams 16, 17, 32 and 33 are
kept the same as in the real case. Therefore, the above equation set is closed for the
two variables, m 2 and E D,COND
EN .
Feedwater preheater
where the mass flow m 22 is also proportional to m 2 . After removing pressure drops in
all the remaining components, the pressure p22 becomes smaller than that in the real
case, which leads to a smaller pressure p21 as well: p21 = p22 pFWH . Given fixed
temperatures t22 and t21 , the specific exergies e22 and e21 are determined. Finally, the
above equation set is closed for the three variables: m 2 , E D,FWH
EN , and E F,PWH
EN .
For pumps and compressors, there are usually two functions: 1) to overcome flow
resistance of the fluid in successive components, and/or 2) to elevate the fluid pressure
to a required level. Therefore, the exergy destruction caused by the first function is
exogenous, while that lead to by the second is endogenous. For fans, all the exergy
destructions are exogenous. The calculation model for a pump (Fig. 3.2d) is given:
where the mass flow rate m 18 is also proportional to m 2 and the specific exergy e18 is
set the same as in real case. Again, given fixed values of t1 , p2 , t2 , p3 and t4 , the specific
44 comprehensive exergy-based assessment of coal-fired power plants
and E F,PUMP
EN . The same considerations are taken also for compressors.
Similar to feedwater preheaters, the fluegas-water (steam) heat exchangers heat the
working fluids from certain temperature to a specified temperature. Thus, the equa-
tions for calculating its endogenous inefficiency can be given based on Fig.3.2e:
When the turbine subsystem is theoretical, the mass flow m 2 is determined by Eq. 3.56;
however, the specific exergy e1 or e4 are different from those in Fig. 3.2ad, due to
the pressure losses in the considered heat exchangers. The specific exergies e26 and e27
are calculated considering the pressure losses: p27 = p2 and p26 = p27 + pHE (heat
exchangers for main steam), and p26 = p3 and p27 = p26 pHE (heat exchangers for
reheat steam). Finally, the above equation set is closed for the three variables: m 2 , E D,HE
EN ,
and E F,HE
EN .
Air preheater
The handling of air preheater is different from the above heat exchangers, as the mass
flow rates of its hot and cold streams are coupled. Considering the system I in Fig.3.2f,
the calculation model for its endogenous exergy destruction is given as follows:
where the mass flows m 9 , m 10 and m 11 are proportional to m 8 , given the same excess
air ratio and slag ratio. Moreover, specific exergy e11 is the same as in the real case;
while the specific exergy e31 is recalculated based on a new pressure p31 . Finally, the
above equation set becomes closed for the three variables: m 8 , E F,APH
EN and E D,APH
EN .
Table 3.3: Basic specification of the pulverized-coal fired power plant under real conditions
Terms Value
lower heating value (LHV) of 22,000 kJ/kg. The chemical exergy of coal is taken as
1.06 times of its LHV.
The simulations under real and unavoidable conditions are handled by a commer-
cial simulator Ebsilon Professional1 , while all systems I constructed by the approach in
section 3.2.1 are solved by EES2 . Details on the employed component models are given
in App. A. The component specifications for real and unavoidable conditions are listed
in Table B.1, while the stream data of the real case are listed in Table B.2. For the spe-
cific exergy of gas mixture, water and steam, the reference state is set as 1.013 bar and
273.15 K, while the standard molar chemical exergies for different species are referred
to (Szargut et al., 1987). In addition, detailed equations for the exergy and exergoeco-
nomic evaluation of the PCPPs are listed in App. B.1.
The results of conventional exergy and exergoeconomic analyses are listed in Table B.5.
The exergetic performance of different components (particularly heat exchangers), and
spatial distributions of exergy dissipation and costs are deeply investigated.
CV
8
12 FSH 26
52 FRHH 11
70
71
131 134 72
WW 72
120
55 60 65 150
118 75 77 84 88 102
45 92
46 ECO1
47 42 41
119 57 62 79
44 43 67 97 96
93
ECO2 37 35 34 32 30 29 28 27 26 25 24
122 121 132 135
154
DA 153
86 85 89 94 168 167
M
136 FWH8 FWH7 FWH6 FWH5 FWH4 FWH3 FWH2 FWH1
M
152
169
M
Attemperting Air
A
P APH Air preheater FP Feedwater pump HPT High-pressure turbine PAF Primary air fan
H 104 CV Controlling valve FRHC Final reheater (cold) IPT Intermediate-pressure turbine PSH Primary superheater
149 CWP Cooling-water pump FRHH Final reheater (hot) IDF Induced draft fan SAF Secondary air fan
172 113
M SAF M
170 DA De-areator FSH Final superheater LPT Low-presure turbine SSH Screen-type superheater
147 ECO Economizer FWH Feedwater preheater OT Ordinary turbine WW Waterwall
146 145
Chimney
148 PAF 103
IDF
171
M 112
2500 100
2000 80
Exergy efficiency / %
Exergy efficiency
Temperature / C
1000 40
Average temperature
500 20
Water (steam) temperature Air
0 0
WW SSH1 SSH2 FSH FRHC FRHH PRH2 PSH ECO1 PRH1 ECO2 APH
Figure 3.4: Exergy efficiencies and temperature profiles of hot and cold streams of all heat
exchangers in the boiler subsystem
70%. In the successive convection-dominant SSHs and FSH, the gradual decrease in
temperature differences for heat transfer generally promotes their exergy efficiencies.
Interestingly, the reheaters, FRHC, FRHH and PRH2, perform not much better than
the upstream SSHs and FSH, even with lower flue gas temperatures. This is because
of much lower average temperatures of cold-side steams in these reheaters, which may
be caused by the properties of water and steam. In addition, those heat exchangers in
the post flue gas duct, PSH, PRH1 and ECOs perform with higher exergy efficiencies.
Note that the exergy efficiency of APH is lower than all other convection heat sur-
faces in the boiler subsystem, although the heat transfer in APH occurs under almost
the lowest temperature difference among all. This is because of the temperature level of
the cold-side air, which is the lowest cold-side temperature among all heat exchangers.
A low temperature level of cold stream could partially counteract the performance en-
hancement of a heat exchanger obtained by employing a small temperature difference
for heat transfer (see section 3.3.1.3).
500
100 Average temperature
Exergy eff. / %
90 Fluid temperature
400 80
Temperature / C
70
300 60
H1 H2 H3 H4 H5 DA H6 H7 H8
200
100
0 H2 H3 H4 H5 DA FP H6 H7 H8
0 100 200 300
. 400 500 600 700 800
Q / MW
Figure 3.5: Exergy efficiencies and temperature profiles of the heat exchangers in the turbine
subsystem
48 comprehensive exergy-based assessment of coal-fired power plants
100
300
1.0 100 500
80 120 700
C 800
70 B
3.0
60
5.0 140
50 160
0 100 200 300 400 500 600
Thermodynamic average temperature of cold stream / C
Figure 3.6: Performance of heat exchangers on the temperature levels and differences
3.3 evaluation of a modern ultra-supercritical plant 49
1 Tref /Tacold
HE = (3.64)
1 Tref /( Tacold + Ta )
Thus, for a constant temperature difference Ta , the higher the average temperature of
the cold fluid Tacold or the lower the average temperature of the hot fluid Tahot , the larger
the exergy efficiency of the heat exchanger. More importantly, according to the depen-
dencies of exergy efficiency on Tacold and Ta , there are typically three performance
ranges for the heat exchangers in PCPPs (Fig. 3.6).
Range A stands for the working region of feedwater preheaters with Tacold (35250 C)
and Ta (0.55 C). When Tacold is lower than 70 C, the exergy efficiency increases
sharply with the increase in both Tacold and Ta . Then, the Ta becomes the dominating
factor, while the influence of Tacold becomes weak. When the Tacold is over 200 C, the
feedwater preheaters always perform quite well with the exergy efficiencies over 95%.
Based on these insights, it is clear that the low exergy efficiency of FWH1 (only 66%)
is caused by the lowest Tacold among all feedwater preheaters.
Range B represents the working zone of heat surfaces in the boiler subsystem. The
exergy efficiencies of radiation-dominated heating surfaces are usually below 80%. Be-
cause the flue gas temperature has been reduced dramatically in radiation sections, the
successive convection heat sections always have relatively high efficiency (over80%).
Range C includes the working region of air preheater, in which the inlet temperature
of cold fluid is usually as low as the environment temperature. The exergetic efficiency
of air preheaters generally falls within the range from 70% to 80%, below those of
convection heating surfaces.
14.1%
Total Loss 49.6%
72.6% Cooling water
Boiler subsystem 47.2%
12.5%
Exhausted fluegas
Turbine
subsystem
c) d)
Other HEs (ms) 13.4% 3.1% Pumps
5.7% FWHs 3.1% Pipes
29.2% 4.4% HEs (rh)
Water wall 3.1% APH
39.9%
Condenser
57.2% 40.5%
Coal combustion Turbines
Figure 3.7: Spatial distribution of exergy destructions and losses: a) overall distributions of ex-
ergy destructions and losses, b) distribution of exergy losses, c) exergy-destruction
distribution in the boiler subsystem, and d) exergy-destruction distribution in tur-
bine subsystem
The exergy destructions within the turbine subsystem is dominated almost equally
(40%) by all turbines and the condenser (Fig. 3.7d). The remaining exergy destructions
are contributed by feedwater preheaters (over 13%), pumps (3%) and pipes (3%). This
pinpoints the significance of advanced design of turbine blades, particularly the blades
working with wet steam, and also the topology improvement of feedwater preheating
subsystem.
The economic constants used in the economic evaluation include annual effective in-
terest rate (0.1), nominal escalation ratio for OMC (0.03), nominal escalation ratio for
FC (0.035), plant economic life (25 yr), annual operation hours (6900 h/yr), and annual
capacity factor (0.8). The average price of the coal with a LHV of 29,270 MJ/kg was
84.3 $/ton (2.88 $/GJ-LHV) in 2009, leading to a levelized FC as high as 196.5 million$.
The PEC of the entire set of PCPP, i. e., the once-through boiler (including the APH),
FWHs, STs and other separate components, could be obtained from (China Power
Engineering Consulting Group Corporation, 2009). However, the PEC of boiler and STs
needs to be allocated to their sub-components (Fig. 3.3). For the allocation of boilers
PEC, the PEC of rotary APH comprises approximately 2% of PEC of the boiler body
(the boiler subsystem excluding SAF, PAF and IDF). The allocation of the remaining
boilers PEC is based on Table B.3. For allocating the turbine PEC, weighting factors are
1, 1.5 and 2 for HPTs, IPTs and LPTs, respectively. The module factors for estimating
fixed capital investment (FCI) from the PECs of different components are referred to
(Turton et al., 2008) as Table B.4. Then, the total capital cost (TCI) is estimated as around
3.3 evaluation of a modern ultra-supercritical plant 51
828 million$ or a specific cost of 774 $/kW. The levelized CC, OMC and FC contribute
30%, 7% and 63% to total revenue requirement (TRR) (308 million$/yr). The cost of
electricity is 5.20 /kWh.
Exergoeconomic analysis can be conducted for two cases: setting Z k as zero or the
values estimated from an economic analysis. The first case enables us to understand
the formation process of the cost of the final product: how the COE is affected by the
system structure and component types. A classical conclusion from the cost formation
process is that the specific cost of exergy accumulates gradually in course of transfer
and conversion to the final product. This leads to the cost inequivalence of exergy
destructions within different components. Generally, the closer the exergy to the final
product, the higher its specific cost. For the PCPP, the total exergy destruction within
all boilers heat exchangers except those in FUR is almost two times of that within
STs (Table B.5). However, there are no big difference between the hidden costs of the
two sets of components: 1969 $/h for considered heat exchangers and 1794 $/h for all
STs. This suggests that the reduction of downstream inefficiencies closer to the final
product would be more effective than reducing upstream inefficiencies.
The second case, i. e., setting Z k as the estimated values, provides a comprehen-
sive evaluation of the component performances. The component ranking based on
the summation C D + Z could pinpoint the significance for reducing the COE. The
cost-effective design improvement should focus on those components with the high-
est C D + Z . From this perspective, the components in the boiler (particularly the
furnace), HPT1, IPT1 and EG, are those key components (Table B.5).
The components in the boiler subsystem have much larger relative cost differences rk
than the others. The ECOs achieve the greatest values (even as high as 335%). However,
the FUR and other heat exchangers have relatively small rk (from 50% to 150%). In
particular, this exergoeconomic variable is below 63% for components in the turbine
subsystem, mostly within 930%. This suggests that the added costs of exergy streams
through the boiler subsystem are much larger than those by the turbine subsystem
(due to the capital investment). Thus, future boiler design should achieve less capital
investment as well as exergy destruction.
Exergoeconomic factors can identify the major sources of the component cost. The
FUR and APH have far less exergoeconomic factor than any other component in boiler.
The remaining heat exchangers in boiler are investment-cost-dominant components
with high exergoeconomic factors. Potential measure to balance the investment and
exergy destruction costs of these heat exchangers may be: Move part of the heat ab-
sorption in these heat exchangers into the furnace and increase the air preheating tem-
perature by increasing the area of APH. In such a way, more even temperature field of
flue gas can be established in the furnace, which leads to smaller exergy destruction
within furnace, larger area of (and larger exergy destruction within) APH, and smaller
areas of (or larger exergy destructions within) the remaining heat exchangers. This
concept may result in a simultaneous reduction of both capital investment cost of and
exergy destruction within the boiler.
The STs generally have reasonable exergoeconomic factors ranging from 50% to
65% with the exception of the last two parallel STs, LPT3 and LPT6. The exergy de-
structions within these two STs have to be reduced. In addition, the first 5 feedwater
52 comprehensive exergy-based assessment of coal-fired power plants
The results of the advanced exergy analysis of the PCPP are listed in Table B.6, where
the endogenous exergy destructions are calculated considering fixed and varied spe-
cific exergies of streams into and out of the considered components compared with
those of real case. The removal of pressure drops in all theoretically-operated com-
ponents mainly affects the endogenous exergy destructions within heat exchangers:
The larger the pressure drop in the considered component, the greater the influence
on its endogenous exergy destruction. Given fixed temperature increases of cold-side
working fluids, endogenous exergy destructions within those fluegas-water heat ex-
changers (FUR, ECO1 and ECO2) increase. In particular, the FUR causes 20 MW more
endogenous exergy destruction. Endogenous exergy destructions within fluegas-steam
heat exchangers for main steam decrease by almost 1 MW, while those within fluegas-
steam heat exchangers for reheat steam remain nearly unchanged due to small pres-
sure drops of reheat steam. Similarly, for FWHs and APH, there are no big differences
between the values obtained by the two calculation approaches.
The coupling with the preceding theoretically-operated turbine leads to a smaller
temperature of the inlet stream into the considered component, which is common for
the turbine train in PCPP. Smaller temperatures of inlet steams reduces the endoge-
nous exergy destructions within the turbines to varying degrees (Table B.6): Those
within the HPTs before reheat are less affected, while those within the IPTs and LPTs
after reheat are reduced significantly. More importantly, the endogenous exergy de-
struction within the COND is largely reduced due to a small temperature of inlet
steam (less inlet exergy to be removed).
The splitting of exergy destructions (Table B.6) is based on the endogenous exergy
destructions and products calculated considering the removal of related pressure drops
and the possible coupling with turbines. The results show that large parts of the exergy
destructions within most components are endogenous. However, for different types of
components, the share of the endogenous part differs significantly. Almost all exergy
destruction within EG is endogenous. For fluegas-water (steam) heat exchangers and
feedwater preheaters, the shares of endogenous exergy destructions are high (7080%).
The HPTs have large shares of endogenous inefficiency (over 70%) as well, given fixed
conditions of main steam. The thermodynamic inefficiencies within these components
tend to be affected in a limited fashion by the inefficiencies of other components. The
shares of endogenous exergy destruction within each remaining STs vary around 50
60%. The performance of each turbine among these STs are strongly affected by the
inefficiencies of other turbines. Moreover, the share of endogenous exergy destruction
within APH becomes even smaller (only 45%). It suggests the performance of the APH
can also be improved largely by improving the remaining components.
The real potential for improving a component is not fully revealed by its total exergy
destruction but by its avoidable part. A large part (3050%) of the exergy destruction
within heat exchangers in the boiler subsystem except FUR is avoidable. For FUR, the
avoidable exergy destruction of the coupled combustion and radiation heat transfer
3.4 suggestions for improvement 53
comes mainly from a smaller excess air ratio. The avoidable exergy destruction within
each FWH remains below 15% of its total exergy destruction. For pumps and fans,
the ranges of avoidable shares are 5060% and 2035%. For STs, especially the IPTs
and LPTs, the share of avoidable exergy destruction varies significantly from 580%. In
addition, most heat exchangers in the boiler subsystem, IPTs, LPT1, LPT3, LPT6 have
large avoidable exergy destructions (around and over 3 MW).
For most components except FUR, the share of endogenous/avoidable part in avoid-
able exergy destruction agrees with that of endogenous part in the components exergy
destruction. Thus, it suggests that, for most boilers heat exchangers (not including
FUR and APH), FWHs and HPTs, efforts to reduce their avoidable exergy destructions
should be dedicated to improve the components themselves. More importantly, for STs,
the improvement should be focused on the whole turbine train but not only the turbine
considered itself. Over 60% of avoidable exergy destruction within FUR and over 40%
for APH is exogenous. To reduce these avoidable parts, not only the component itself
but also the remaining components should be improved simultaneously.
The splitting of investment costs and exergy destruction costs of all components are
listed in Table B.7 and Table B.8, respectively. The shares of endogenous/exogenous
parts in components total investment cost are identical to the corresponding shares of
endogenous/exogenous exergy destructions. However, the shares of avoidable invest-
ment cost of almost all components range within 8090%. The total avoidable invest-
ment costs of the system (49.9 /s, 10% out of the total investment cost) are contributed
by the heat exchangers in the boiler (28.8 /s), the STs (10.4 /s), the EG (4.6 /s), and
pumps and fans (surprisingly, 4.3 /s in total). Around 90% of EGs contribution is
endogenous, while this figure becomes 70% for boilers heat exchangers, 55% for STs.
As expected, over 70% of the avoidable investment costs from pumps and fans are
exogenous.
The overall avoidable exergy destruction costs of the system (62 /s out of 618 /s)
are contributed most by the STs (27 /s), the boiler (24 /s), pumps and fans (4.8 /s),
and EG (4 /s). Around 60% of the exergy destruction costs of both the boiler and
turbine subsystems are endogenous. Almost all avoidable exergy destruction costs of
EG are endogenous, while those of pumps and fans are mostly exogenous.
further reduce the investment costs. In addition, it will be more beneficial if a large
amount of waste heat of the boiler exhaust and cooling water can be cost-effectively
utilized.
This chapter discusses in depth a series of classical issues for improving the design of
steam cycles, e. g., the matches of the temperature and pressure of main and reheat
steams, number and position of reheating, and the layout of feedwater preheating sys-
tem. In contrast to simulation-based structural and parametric adjustments for highly-
efficient design, this discussion considers various structural alternatives embedded in a
superstructure. After a preliminary investigation on a simple steam cycle (section 4.1),
the superstructure, considering up to two-stage reheating, up to eleven-stage feedwa-
ter preheating (including a de-aerator (DA)) and a secondary steam turbine with steam
extractions (ET), is graphically built in section 4.2 based on a professional simulator.
Afterwards, the synthesis and optimization problems are formulated as MINLP (sec-
tion 4.3). Then, the problems are solved by an enhanced multi-objective differential
evolution (DE/MODE) (section 4.4). Both the single- and double-reheat cases are com-
prehensively discussed considering only the thermodynamic objective (section 4.5) or
along with the economic objective (section 4.6). Finally, several conclusions are drawn
in section 4.7.
57
58 superstructure-based synthesis and optimization of coal-fired power plants
600 C
600 C
26.2 bar / 600 C
Boiler HPT LPT EG
IPT
COND
Feed pump
Figure 4.1: Simple plant for the sensitivity analysis: fixed relative pressure ratio p = p/pin
(-), boiler efficiency B (%), isentropic efficiency s (%), pressure p (bar)
The resulting egg-shell diagram, the cycle efficiency contour (Fig. 4.2a), indicates that
there is a relatively wide range over which the two reheating pressures can be varied
while having a minimal effect on the overall efficiency. Thus, different combinations of
reheat pressures can be properly selected to achieve the same overall efficiency, which
introduces a certain freedom for the system design.
The largest efficiency zone falling in the superheated steam range tends to be far
away from the reasonable quality of exhaust steam considering a fixed condenser pres-
sure of 0.052 bar (Fig. 4.2a). However, clearly from the classical temperature-entropy
diagram of a steam cycle, when fixing the maximum steam temperature level, an in-
crease in the main-steam pressure leads to a decrease in the quality of exhaust steam.
Thus, the largest efficiency zone can be moved to approach a reasonable exhaust quality
range (not too large <0.95 and not too small >0.90) to obtain the best thermodynamic
benefits by adjusting the throttle pressure.
The reheat temperatures also affect both plant efficiency and exhausted-steam qual-
ity (Fig. 4.2b). In the dual-reheat zone above the single-reheat line, both reheating
temperatures affect significantly the overall efficiency; while the second reheat temper-
a) 0.9 b) 750
42
0.95
41 41
0.92
.4 .2
0.8 700 .8
Optimal pressure ratio of reheat 2 / -
0.8
4
39 86
40
0.7 41
6.
.8 .2 .6
0.
40
9
500 0.92 3
0.
0.92
9
0.9
9.
2
0.4 4
5
450
40
3 0.9
.4
9
41
40
0.3 0.88
.
400
.8
2
0.92
0.2 0.95 First-reheat line
0.9
350
9 0.84
0.1 300
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 450 500 550 600 650 700 750
Optimal pressure ratio of reheat 1 / - Temperature of steam out of reheat 1 / C
Figure 4.2: Effects of reheat parameters on plant efficiency (solid line, %) and quality of the ex-
hausted steam (dashed line, 1): a) reheating pressures vs plant efficiency, b) reheat-
ing temperature vs plant efficiency and steam quality at the steam turbine outlet.
(The fixed parameters have been shown in Fig. 4.1)
4 9 Path-rh1 d i n Path-rh2
SW2 5 SW2 10 SW2 e SW2 j SW2 o
11 p
3 8 c h m
a b c d e f g h i j k l m n o p
Block8
Block6 1 2 3 4 5 6 7 8 9 10 11
Block9
M
MT1 MT2 MT3 MT4 MT5 MT6 MT7 MT8 MT9 MT10 MT11 MT12 MT13 MT14 MT15 MT16 MT17 MT18 MT19 MT20 MT21
EG
IDF BOILER
T
To ET
coal
1 2 3 4 5 6 7 8 9 10 11
air 4
5
Block3 (to Block4): possible stream (a-p) of rh2
6 a b c d e f g h i j k l m n o p
CWP
Block2 SW1 7
8
(Path-rh1) 10
9
M
11
Block10
H1 H2 H3 H4 H5 DA H6 H7 H8 H9
f
SW1 g
h
SW1 d
e E1
OT
i
j
a
b CP
c
k
SW1
Block4
l
m DA M
(Path-rh2) p
o
n SW4 SW4 SW4 SW4 SW4 SW4
FP
FWH10 FWH9 FWH8 FWH7 FWH6 FWH5 FWH4 FWH3 FWH2 FWH1
From the cold reheated ET
Path Switches Legends steam of rh1 T
Steam
SW1 SW2
SW4 Water
Electricity
H1 H2 H3 H4 H5 DA H6 H7 H8 H9
To Block10
SW1 SW2 SW3 SW4 Shaft work
Figure 4.3: The superstructure considering up to two-stage reheating, up to 10 feedwater preheaters and a secondary turbine with steam extractions
59
60 superstructure-based synthesis and optimization of coal-fired power plants
ature determines additionally the exhaust steam quality. The deviation of the highest
efficiency zone from reasonable range of steam quality (Fig. 4.2a) would be larger if
the temperatures of throttle and reheat steam increase under the same throttle pres-
sure. Thus, the throttle and reheating pressures must be increased as a compensation
for temperature elevation to obtain the allowed exhaust quality.
Figure 4.4: Possible stream paths controlled by switches SW1SW4 used in Fig. 4.3
4.3 statement of the optimization problem 61
ET, while the rest, heated in the steam generator, flows through block5. In block5, the
SW2 activates stream1 and closes stream211 (in block5). Finally, the reheated steam
reaches MT2 with the aid of the SW3 in block9.
A structural alternative can be activated by specifying associated integer variables:
(1) a global array g including g1 (indicating single or double reheating), g2 (indicating
the maximum number of the employed feedwater preheaters), and a binary integer g3
(indicating the existence of an ET); (2) a continuous integer array c representing the
positions of reheating (rh), rh1 and rh2; and (3) a binary array b controlling whether
the corresponding feedwater preheater is used or not. Therefore, the alternatives ()
embedded in the superstructure can be defined as follows:
= ( g, c, b), (4.1)
g = { gi | g1 {1, 2}, g2 {3, 4, . . . , 10}, g3 {0, 1}},
c = {ci |ci {1, 2, . . . , cmax
rhi }, i = 1, 2},
b = {bi | bi {0, 1}, i = 3, 4, . . . , 9},
where the reheats domain (cmax max
rh1 and crh2 ) can be narrowed by the user to reduce the
searching space. Note that the FWH1, FWH2, and FWH10 (Fig. 4.3) are set in service
for all structural alternatives, which does not change the feedwater preheating layout.
thermodynamic objective is involved, the values of the variables, tms , pms , pex,MT , trh1 ,
trh2 and i , are kept fixed (not decision variables any more). Additional constraints are
imposed to the quality of exhausted steam for the optimization at single design point:
For MT, the steam quality xex,MT can be constrained by type1 (for practical design) or
type2 (Spliethoff, 2010); while for ordinary secondary turbine (OT) and ET only a lower
limit (0.9) is considered. In contrast, if an economic objective (Eq. 3.19, see section 3.1.3)
is involved, the variables, tms , pms , pex,MT , trh1 , trh2 and i become decision variables,
whose bounds are specified and no constraints on steam quality will be imposed. All
component models (including cost functions) are listed in App. A.
The procedure of EAs has been described in section 2.5.1.4. For solving the MINLP
problem (Eq. 4.2), a similar procedure (Fig. 4.5a) is applied: The solution randomizer
in the initialization randomly generates integer and real variables of candidate solu-
tions. Each solution is evaluated by the simulation of the specified superstructure, in
which the stream paths, starting values, and real variables are set. The solution is
added to the initial population, if the simulation is successful and extra constraints
(the four constraints on steam quality in Eq. 4.2) are satisfied; otherwise, the solution
is discarded. Once the initial population is formed, the evolution iteration starts. For
evolving each target solution (si ) in the parent population, a noisy solution (sni ) is first
generated by adding weighted difference between the two randomly selected solutions
(sai and sbi ) to another randomly selected solution (sci ):
where f is the real-valued weighting factor and Ns is the population size. The noisy
and target solutions are recombined to produce a new trial solution (sti ):
s ,
i,j r > c ,
sti,j = i = 1 . . . Ns , j = 1 . . . Nv , (4.4)
sn , r c ,
i,j
where c is the real-valued crossover constant; while r and Nv are a real-valued num-
ber (randomly chosen from 0 to 1) and the number of decision variables, respectively.
After evaluation, the better of the trial and target solutions is kept as a child solu-
tion. The evolution iterations for each solution and population continue until certain
4.4 multi-objective supported differential evolution 63
Figure 4.5: Multi-objective differential evolution and the implementation: a) basic DE/MODE;
b) illustration of desired fronts; c) enhanced selection for the desired fronts
termination criterion is reached. The key controlling parameters of DE and their rec-
ommended values (Storn and Price, 1997; Price et al., 2006) are Ns (ten times of the
number of decision variables), f (0.8) and c (0.9). In addition, different schemes of
mutation and crossover have been developed (Storn and Price, 1997; Price et al., 2006),
e. g., the solution sc in Eq. 4.3 can be the current best-known solution.
The scheme for MODE (Fig. 4.5a) preserves any feasible but not only the better
trial solutions into the off-spring population. After the generation of each whole child
population, the parent and child populations are merged for multi-objective selections.
The non-dominated sorting and crowding distance assignment (see section 2.5.1.4) are
integrated to form a new parent population and to obtain the best known front.
The objective-space fronts desired by the practicians may be more than the Pareto
front itself, for example, the lower front of the objective space F0 + F1 in Fig. 4.5b.
The front F0 + F1 can be obtained by different optimization runs of the MODE in
Fig. 4.5a; however, due to the unclear termination criteria of EAs, the fronts obtained
from different runs may be not continuous with each other (e. g., F1 is obtained from
the first run while F3 is obtained from the second run). More importantly, since EAs
are generally time-consuming, it is better to obtain all desired fronts in a single op-
timization run. Thus, an enhanced multi-objective selection for MODE (Fig. 4.5c) is
proposed: Each merged population is split into two sub-populations according to the
economically-optimal solution (s). If the number of solutions in one sub-population
(assume sub1) is below Ns /2, new child solutions are generated by taking sub1 as the
parent population. Then, the two sub-populations are separately processed by multi-
objective selection techniques. The best Ns /2 solutions from each sub-population form
the new parent population. Therefore, smooth and continuous fronts can be obtained.
64 superstructure-based synthesis and optimization of coal-fired power plants
For generating structural alternatives, the integer variables need to be properly han-
dled. More importantly, for a high generation rate of feasible solutions, an effective
way is favorable to produce reasonable pressure arrays for a turbine train, which donot
lead to simulation failure.
However, the mutation for reheat positions will work ineffectively, if the position carhi
equals cbrhi (occurring frequently as good solutions are preserved in EAs). This muta-
tion failure may lead to an unexpected situation, where the reheat positions (crhi ) of
all solutions in one generation are identical at certain stage of optimization. This is
unfavorable for finding optimal solutions but can be circumvented (Fig. 4.6): For the
selected solutions sa , sb and sc , the new feedwater-preheater train is obtained first for
the noisy solution sn . Then, if the position crhi of these three solutions are different,
the normal mutation procedure for reheat positions (Eq. 4.7) is directly applied; oth-
erwise, the reheat positions can be kept unchanged or adjusted to the outlets of the
neighboring turbines upstream or downstream, which connect an in-service feedwa-
ter preheater in the new train of preheaters. Subsequently, new reheat positions of
the noisy solution can also be adjusted (upstream or downstream) up to the outlet of
the neighboring turbines connecting no feedwater preheaters, e. g., MT1 and MT3 in
Fig. 4.3. Therefore, all the outlets of the turbines are accessible as reheat positions at
any stage of the optimization.
Range of
old new
new old pMTn pMTm
The turbine whose
outgoing steam to
be reheated
Reheat positions of the noisy solution b
Set reheating pressures
1/3 1/3
1/3
Move upstream Move downstream The redundant
turbines
c
new
old
old
new Set outlet pressures of redundant turbines:
New reheat positions of the noisy solution A valid pressure array for the turbine train
Figure 4.6: Mutation of reheat positions Figure 4.7: Generation of pressure arrays for success-
ful simulations during the initialization
happens in both initialization and mutation processes, as too many (totally over 20)
turbines are involved in the superstructure and the bounds of most pressures are var-
ied. However, the repair of the bound violations by resetting the pressure of a turbine
between outlet pressures of the two neighboring turbines upstream and downstream,
is not enough, as the pressures are also constrained by the associated feedwater pre-
heaters: If pressures of the extracted steams of two neighboring feedwater preheaters
are too close, the simulation will fail probably.
An approach for generating a reasonable pressure array (Fig. 4.7) is, thus, proposed:
Given any feasible layout of feedwater-preheater train, the outlet pressures of the tur-
bines connecting in-service feedwater preheaters are generated first. The generation
starts from producing the pressures, whose bounds are fixed. Then, the outlet pres-
sures of intermediate turbines between any two turbines with already-generated outlet
pressures (e. g., MTn and MTm (Fig. 4.7a)) are randomly generated. More importantly,
the intermediate pressures are kept not too close to the known bounds by properly
setting the values of two constants and , say = 0.95 and = 0.05 (the possible
ranges of intermediate pressures are slightly reduced). For any turbine without con-
necting a feedwater preheater but with reheat at the outlet, the outlet pressure (prhi )
is randomly generated between its two neighboring turbines with known outlet pres-
sures, e. g., pMTn and pMTm (Fig. 4.7b). In such a case, the bounds of the reheat pressure
are no longer controlled by the constants and , i. e., = 1 and = 0. Finally,
outlet pressures of the rest (redundant) turbines without connecting any reheater or
in-service feedwater preheaters are set identically to that of its neighboring upstream
turbine, whose outlet pressure has been known.
The above approach (Fig. 4.7) is mainly applied for the initialization; while for muta-
tion the pressures of turbines connecting reheaters and in-service feedwater preheaters
are firstly generated for the noisy solution by
The bound violations are repaired similarly as Fig. 4.7a and b. The pressure array gen-
erated by Eq. 4.8 and 4.9 is specified to the turbines associated with the new feedwater-
preheater train of the noisy solution. Finally, the pressures of all redundant turbines
are handled as Fig. 4.7c.
Three single-reheat cases (Table 4.1) are set to evaluate the single-objective DE algo-
rithm with the settings (Ns = 400, f = 0.8, c = 0.9). The throttle temperature (tms ), the
final feedwater preheating temperature (tfw ), and the number of feedwater preheaters
are specified as the same for all three cases. Case 1 considers varied temperature tms
and trh1 , and an ET; while case 2 does not consider an ET and the temperature trh1 is
Table 4.2: The performance variables of each type of component embedded in the superstruc-
ture for thermodynamical optimization: p relative pressure loss (%), ppressure
loss (bar), s isentropic efficiency (%), m mechanical efficiency (%), efficiency (%),
tup difference between the saturation temperature of hot inlet and the temperature
of cold outlet (C), tlow temperature difference between hot outlet and cold inlet
(C); hFWHhigh-pressure FWH, lFWHlow-pressure FWH
46.4 47
46.2 Generation 50
46.0 46
Case 1: g = (1,7,1), tms = 600 C
Plant efficiency / %
Plant efficiency / %
45.8
45
45.6
Case 2: g = (1,7,0), tms = 600 C, trh1 = 605 C
45.4 44
45.2
43
45.0 Case 3: g = (1,7,0), pms = 262 bar, tms = 600 C
44.8 42 Generation 1
44.6
44.4 41
0 25 50 75 100 125 150 0 100 200 300 400
Generation number / - Solution number / -
Figure 4.8: Plant efficiency versus generation Figure 4.9: Plant efficiency versus solution
number (tfw =300 C for all cases) number in generation 0 (initial
generation) and 50 for case 1
kept constant at 605 C. Case 3 is specified according to an actual power plant in China
with the steam conditions of 262 bar/600 C/605 C. No ET is employed in case 3.
All cases are optimized under constraint type1 (see Eq. 4.2). The associated perfor-
mance variables of each component are specified in Table 4.2. Note that for each case
of the three cases, certain variables are specified as fixed values (Fig. 4.8). In addition,
a fixed temperature of cooling water (20 C) is considered for the superstructure.
The evolution of all candidate solutions of case 1 after 50 generations (Fig. 4.9) in-
dicates that the initially randomly generated candidates disperse almost uniformly
within a wide range, avoiding the search trapped in local optimums. After 50 genera-
tions, most candidates evolve greatly to compete with the best solution of the current
generation; meanwhile, a small amount of solutions unpromising for reaching an opti-
mality are preserved as well to keep the solution diversity.
Since the best solution of the MINLP problem cannot be known in advance, only
near-optimal objectives are obtained when the search terminates at the maximum gen-
eration number (150 for the three cases). The best-known solutions of all generations
(Fig. 4.8) show that the first tens of generations are most important for the evolution, as
the best objective value increase only slightly afterwards. However, to assure reliable
near-optimal solutions, the termination criterion in the following is specified as the
relative variation of the best objective value within 1% during the last 30 generations.
The generation rate of feasible solutions and the computation time are key factors
to evaluate the algorithm. Without the special techniques in section 4.4.2, it is difficult
to produce feasible solutions; however, by employing these techniques, the generation
rate remains at a high level (averagely over 75% (Fig. 4.10)).
The total computation time for each optimization run is large (Fig. 4.10). Fortu-
nately, the time increases linearly in course of evolution. Most importantly, the eval-
uation time of all infeasible solutions is rather small due to the high generation rate
of feasible solutions. Note that the profiles in Fig. 4.10 are similar for all optimization
runs, single-objective or multi-objective, as there is no additional computation effort
for multi-objective problems.
68 superstructure-based synthesis and optimization of coal-fired power plants
100 25
90
Accumulated time / hr
70
solution / %
60 15
50
Total evaluation time
40 10
30
20 5
Evaluation time of feasible solutions
10
0 0
0 50 100 150 200
Generation number / -
Figure 4.10: Performance of one representative optimization run on a platform: Intel Core2 Duo
CPU P8600 2.4 GHz with 4 GB RAM
For current steam conditions of a temperature level below 650 C, additional poten-
tials of energy-savings is expected by optimizing the initial design of reheating and
the layout of feedwater regeneration. Therefore, an exhaustive study on case 3 is con-
ducted. The actual reference industrial design (262 bar/600 C/605 C) reaches a plant
efficiency of 44.41%, with the pressure array of extracted steams being 83.72 bar/63.60
(rh1)/24.05/11.57 (DA)/6.35/2.543/0.669/0.253. The industrial design employs only a
single reheat; however, double reheats are also considered to find the maximum energy-
saving potential under the given steam conditions. The best-known (optimal) solutions
are selected from several optimization runs for the same problem.
45.4
10
45.3 9
Plant efficiency / %
45.2
45.1
45.0
8
44.9
44.7
300 310 320 330 340
Feedwater preheating temperature / C
Figure 4.11: The best-known plant efficiencies under different numbers of feedwater preheaters
and final feedwater temperatures (solid lines: with ET; dashed lines: with OT)
4.5 design optimization for thermodynamic objective 69
Table 4.3: Comparison of the steam extractions after reheating for single-reheat case (a) and
double-reheat case (b) by keeping m ms as 861 kg/s: ppressure (bar), tsat saturation
temperature (C), ttemperature (C), tdecrease in superheating degree (C), m
mass flow rate (kg/s), OOT, EET. (The steam extractions, which has been in two-
phase zone even without an ET, are not listed)
(a) (b)
p tsat t t m p tsat t t m
Therefore, only from the structure adjustment (within the given superstructure), an
additional efficiency improvement by nearly 1 PP can be achieved for single-reheat
units under the current steam conditions.
For higher steam conditions, only double-reheat units are discussed. For comparison
purposes, low steam conditions (steam temperature of 550 C/550 C/550 C and
throttle pressure 170230 bar) are considered as well.
50 1.3 300
type2 750 C (9 FWHs, 1 DA and tfw = 320 C)
49 type1 1.1
250
tms = trh1 = trh2= 750 C
Efficiency penalty / PP
700 C
Plant efficiency / %
C 200
47 0.7
650 C 700 C
46 0.5 150
600 C
650 C
45 0.3 B
100
44 tms = trh1 = trh2 = 550 C 0.1 A
prh1 prh2 50
43 -0.1
(9 FWHs, 1 DA and tfw = 320 C)
42 -0.3 0
170 270 370 470 170 270 370 470
Pressure of the main steam / bar Pressure of the main steam / bar
Figure 4.12: Optimal efficiency for differ- Figure 4.13: The efficiency penalty and the opti-
ent steam conditions under mal reheat pressure under the con-
constraints type1 and type2 straint type2 (Eq. 4.2)
(Eq. 4.2)
with steam conditions of 170 bar/750 C/750 C/750 C, the efficiency loss reaches
almost 1.23 PP. While for the current steam conditions with temperature below 600 C
and pressure lower than 300 bar, there is almost no efficiency deviation (or penalty):
The optimal solutions with such steam conditions are within or near the reasonable
region of the quality of exhausted steams.
To maximize the benefits of improving the temperature levels, the efficiency penalty
illustrated in Fig. 4.12 and Fig. 4.13 needs to be weakened or even eliminated. Note
that increasing the throttle pressure has positive effects on reducing the deviations
(Fig. 4.12) when the temperature level is over 600 C (Fig. 4.12): Given a temperature
level, there would be a transition throttle pressure (pt,ti with ti being the temperature
level (i)), under which the efficiency penalty (the deviation) can be eliminated. For a
temperature level ti , when the throttle pressure increases but keeps below the pressure
pt,ti , the efficiency enhancement results not only from the pressure increase itself but
also from a reduction of the deviation; once the throttle pressure exceeds the pressure
pt,ti , the efficiency improvement will be exclusively due to the pressure increase itself.
The pressure pt,ti could be important for the optimal design of steam cycles with a
temperature level over 600 C. However, a large throttle pressure over the pressure pt,ti
does not contribute significantly to the efficiency improvement. In practice, the throttle
and reheat temperatures vary within a small range (at a fixed temperature level) to
reduce the thermal stress of turbine; thus, it is favorable to select the design throttle
pressure near and below the corresponding pressure pt for maximize the benefit from
temperature-level increases of the main and reheated steams.
The pressure pt,ti can be determined by a logarithmic relationship between the effi-
ciency penalty (ti ,pi ) and the throttle pressure (pi ) based on Fig. 4.13:
where the coefficients ati and bti are associated with each temperature level ti of the
throttle and reheat steams. The coefficients ati and bti can be regressed based on
72 superstructure-based synthesis and optimization of coal-fired power plants
Table 4.4: The transition throttle pressures corresponding to each temperature level
efficiency-penalty lines in Fig. 4.13; thus, by setting the efficiency loss ti ,pi as zero,
the obtained pressure pi becomes the pressure pt,ti .
The coefficients and the pressure pt,ti are determined for three temperature levels
(Table. 4.4). Note that even for steam conditions of 650 C/650 C/650 C, the pres-
sure pt,ti reaches as high as 469 bar. If the temperature level is over 700 C, it seems
impractical to eliminate the efficiency penalty under the present industrially-possible
throttle pressure (350375 bar). For the advanced steam cycles with steam conditions
of 350 bar/700 C/700 C/700 C, almost 0.25 PP efficiency benefit must be sacrificed.
0.4 300
type 1 (9 FWHs, 1 DA and tfw = 320 C) 650 C
Pressure ratio of the 2rd reheat / -
0.15 prh1
100 A B (flat zone) prh2
0.1
50
0.05
0 0
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 170 270 370 470
Pressure ratio of the 1st reheat / - Pressure of the main steam / bar
Figure 4.14: Statistical optimal reheat pres- Figure 4.15: Optimal reheat pressures of con-
sure ratios of zone A and C straint type1 varying with the
throttle pressure
4.6 design optimization for thermodynamic and economic objectives 73
near or over the pressure pt,ti , both pressures prh1 and prh2 will increase greatly. For
zone C, there is no FWH above the first reheat point.
For current industrial designs, the optimal pressure ratio for reheating, i. e., prh1 /pms
and prh2 /prh1 , are usually selected within the range of 0.250.3 for 600 C/ 600 C/
600 C. However, pressure ratios of the optimal solutions for double reheats (Fig. 4.14)
in zone A are confined within the ranges 0.20.35 (for rh1) and 0.150.25 (for rh2).
However, in zone C, these ranges are 0.40.55 (for rh1) and 0.20.3 (for rh2).
When considering constraint type1, three similar zones exist (Fig. 4.15): In zone A,
the pressure prh2 depends only on the temperature level; Zone B moves to small values
of pressure pms with an increase in the temperature level; In zone C, the pressure prh2
rises significantly with an increase in the throttle pressure or the temperature level.
In addition, for all the three zones, the optimal pressure ratios are within the range
0.30.5 (for rh1) and 0.20.3 (for rh2), respectively (Fig. 4.14).
For mathematical optimization, the cost functions of all involved components have
been listed in App. A. With known purchased-equipment costs of all components, the
total capital investment is simply calculated by PECk (the constant is set as
4.75 (Bejan et al., 1996)) but not by a detailed calculation. Similarly, the operation and
BOILER
IPT1-4
IPT1-3
HPT1-2
/
G
EG
FWH7
OT
COND
FWH6
CP
CWP
DA
FWP
FWH5 FWH4 FWH3 FWH2 FWH1
DP
Figure 4.16: The schematic of the power plant with a fixed structure
74 superstructure-based synthesis and optimization of coal-fired power plants
maintenance cost is simply estimated by TCI (the constant is set as 0.06 (Bejan
et al., 1996)). Therefore, the COE can be readily calculated according to the economic
analysis in section 3.1.3, with the same economic constants in section 3.3.2.
A single-reheat, 1000 MW, ultra-supercritical coal-fired power plant (Fig. 4.16) is opti-
mized with respect to the two objectives. Note that the boiler is modeled in a simplified
fashion as its PEC is expressed in a condensed form (App. A). Totally 87 material and
energy streams and 26 components are involved in each simulation.
An industrial design of a Chinese ultra-supercritical plant (Wang et al., 2012a), with
steam conditions of 274 bar/605 C/603 C, is taken as a reference. The plant efficiency
and COE of the plant are 45.8% and 5.20 /kWh (with a coal price of 0.3 /MJ on a
LHV basis), respectively.
For the optimization, 38 decision variables (Table 4.5) are taken into account. The
bounds of the decision variables can be fixed (for most variables) or varied (e. g., the
outlet pressure of some turbine stages). Note that the outlet pressure of HPT1 is con-
sidered with a fixed bound, which confines the final feedwater preheating temperature
within the range from 275 C to 337 C.
Figure 4.17: Cost of electricity versus plant Figure 4.18: Contribution of fuel cost and capital
efficiency at different genera- investment cost to cost of electricity
tion numbers of the front solutions
more uniformly and densely dispersed, leading to a smooth and non-interrupted front
around the 200th generation.
5 1.6
3% ccoal = 0.3 /MJLHV
components / $/s
3 1
9% Boiler
2.5 0.8
2 Electrical generator
56% 0.6
Boiler
21% 1.5
Turbines 0.4
1
Condenser
0.5 0.2
Pumps Condenser
0 0
Electrical generator
41 43.5 46 48.5 51
Feedwater preheaters (including de-aerator) Plant efficiency / %
Figure 4.19: Share of the investment cost Figure 4.20: Distributions of the investment
of each component for the costs of all front solutions
economically optimal solution
with equal temperatures of
main and reheated steams
by the turbines (21%), EG (9%) and COND (9%). The distribution can be validated
by Table 5 in (Xiong et al., 2012b) and also by (China Power Engineering Consulting
Group Corporation, 2009; National Energy Technology Laboratory, 2010).
The levelized cost rates of the four groups of components are compared for all solu-
tions at the front (Fig. 4.20). The PECEG remains unchanged for all frontier solutions
with a fixed power output. However, the summation of all turbine PEC increases al-
most linearly with increasing plant efficiency, which is led to by the increases of steam
temperatures and the isentropic efficiencies of the turbines.
The profile of the cost Z COND results from the pressure change of exhausted steam
from main turbine (4.21). The pressure of the exhausted steam remains at its upper
bound (0.08 bar), when the plant efficiency is below 44%. Thereafter, it decreases until
reaching its lower bound (0.03 bar) for achieving higher plant efficiency. The decrease
in this pressure causes smaller temperature of the exhausted steam. Accordingly, with
1400 0.09
Pressure of exhausted steam / bar
flow rate of main (reheated) steam
0.08
Pressure, temperature or mass
1200
0.07
1000
.
0.06
mms (kg/s)
800 0.05
tsh = trh (C)
600 0.04
0.03
400 psh (bar)
0.02
200
0.01
ccoal = 0.3 /MJLHV
0 0
41 43.5 46 48.5 51
Plant efficiency / %
Figure 4.21: Variables related to purchased equipment costs of the boiler and the condenser
4.6 design optimization for thermodynamic and economic objectives 77
a fixed inlet temperature of the cooling water (as in this case), a larger area (capital
investment) is needed for the reduced temperature difference of heat transfer.
The profile of the cost Z B (Fig. 4.20) is similar to that of the capital cost (Fig. 4.18),
as the boiler has the most contribution (Fig. 4.19). The cost Z B decreases slightly until
an efficiency value of around 44% is reached; afterwards, the cost Z B increases sub-
stantially on the course of increasing plant efficiency. The reasons are investigated
by analyzing the important parameters associated with PECB (Fig. 4.21): Interestingly,
when the plant efficiency is below 44%, both the temperature and pressure of the main
steam stay at their low bounds. This is because the given lowest steam conditions
(230 bar/550 C) is indeed sufficient for reaching a plant efficiency up to 44%; thus,
within this range of plant efficiency (below about 44%), the adjustments of other deci-
sion variables, e. g., the isentropic efficiency of turbine and pump, and the pressures of
steam extractions, become more cost-effective than directly increasing the steam condi-
tions or reducing the pressure of the exhausted steam. However, an efficiency increase
to values above 44% inevitably needs improving the steam conditions. Therefore, the
consideration of the front at efficiency values below 44% (in this case) is not appro-
priate for supercritical power plants. In addition, the profile of the cost Z B with plant
efficiency over 44% is led to by simultaneous changes of the mass flow rate, the levels
of temperature and pressure of main and reheat steams in the course of increasing the
plant efficiency.
tsh sh
out tin m rh rh
out tout tin
rh
Fshrh = 1 + + , (4.11)
tsh
out m sh
out trh
out
sh
tout tref
Ft = 1 + b exp , (4.12)
c
where b and c are two constants. The uncertainty from the two factors are discussed in
the following. Note that the aim of the discussion is not to find more accurate expres-
78 superstructure-based synthesis and optimization of coal-fired power plants
5.6 900
ccoal = 0.3 /MJLHV
5 300
psh (bar)
4.9 200
41 43.5 46 48.5 51
Plant efficiency / %
Figure 4.22: The front and important decision variables of the case with nonequal temperatures
of main and reheated steams
sions of cost functions but to show the mathematical properties of the cost functions.
Practically, more industrial data should be used for reasonable cost functions.
a) 5.6 b) 5.6
a = 247583, c = 50, tref = 850
5.5 5.5
Cost of electricity / /kWh
5.3 5.3
5.2 5.2
base base
b = 5 for all
5.1 5.1
5 a = 145689, b = 20 5
a = 163360, b = 15
4.9 4.9 a = 255067, c = 75, tref = 1100
c = 75, tref = 850 for all a = 250342, c = 75, tref = 1000
4.8 4.8
41 43.5 46 48.5 51 41 43.5 46 48.5 51
Plant efficiency / % Plant efficiency / %
Figure 4.23: The impact of the temperature-related factor FT on the front of the case with iden-
tical temperatures of main and reheated steams: a) variation of the front with
parameter b, b) variation of the front with parameters c and tref
5.20 /kWh, coefficient a is adjusted according to the values of the three coefficients b,
c and tref .
For the sensitivity of the front to coefficient b (Fig. 4.23a), with increasing the coeffi-
cient b, the whole front moves to the lower left with a sharper rise of COE at the Pareto
front (the right part). Therefore, the larger the coefficient b, the lower the COE of the
economically-optimal solution is. In addition, if the coefficient b becomes greater than
ten, the Pareto fronts donot change significantly.
The coefficients c and the reference temperature tref (Fig. 4.23b) have opposing effects
on the front shape. An increase in the coefficient c makes the economically-optimal
solution move to the upper left, when the coefficient c is below 75. With a further
increase in the coefficient c, the front tends to slightly rotate to the left, while the
COE of the economically-optimal solution keeps almost unchanged. However, with an
increasing coefficient tref , the economically-optimal solution moves to the lower right;
meanwhile, the left part of the front becomes straighter: The Pareto front is narrowed.
Table 4.6: Comparison of the decision variables between the industrial design (ID) and the
economically optimal solutions (t /C, p /bar)
designs are much lower than those of the ID, indicating more use of low-pressure
steam for feedwater preheating. In addition, isentropic efficiencies of these turbines
are increased.
For the condenser, the condensing pressure of the optimal designs (around 0.036 bar)
is slightly lower than that of the current design (0.052 bar). For feedwater preheaters, al-
most all temperature differences tup and tlow are reduced, as the increase in the PEC
of feedwater preheaters leads to less use of extracted steam, a higher plant efficiency
(a lower fuel costs), and thus, a lower COE.
More insights on the performances of these designs are given by the exergoeconomic
comparison (Table. 4.7). For most components, the specific costs of both the fuel and the
product in the ID seem to be slightly smaller than those of the design s1 but slightly
larger compared to the design s2 , mainly due to the improvement of the boiler. The
relative cost difference (r, see section 3.1.3) of the boiler in the design s2 is reduced
significantly compared to the ID and the design s1 . However, those of the HPT2 and
LPT4 in both optimal designs increase quite a lot, as the power generation in these two
Table 4.7: Exergoeconomic comparisons between the industrial design (ID) and the economically optimal solutions
Boiler 0.39 0.39 0.39 0.96 0.96 0.94 147 147 142 352 345 318 417 400 402 769 746 720 45.8 46.3 44.2
HPT1 1.03 1.03 1.01 1.17 1.16 1.15 13.9 12.5 13.2 20.3 24.0 21.8 16.6 11.20 10.2 36.8 35.2 32.0 55.0 68.2 68.1
HPT2 1.03 1.03 1.01 1.22 1.27 1.30 18.5 22.9 27.9 8.47 3.40 1.89 2.05 0.83 0.51 10.5 4.23 2.41 80.5 80.4 78.6
IPT1 1.01 1.02 0.99 1.15 1.15 1.12 13.5 12.8 12.7 17.1 23.7 20.5 8.84 9.29 7.76 25.9 32.9 28.2 65.9 71.8 72.5
IPT2 1.01 1.02 0.99 1.17 1.17 1.13 14.9 14.9 14.0 14.8 19.1 18.6 5.77 7.28 7.62 20.5 26.4 26.2 71.9 72.4 71.0
IPT3 1.01 1.02 0.99 1.19 1.23 1.15 17.3 20.9 15.7 12.9 5.71 19.9 3.79 3.10 8.71 16.7 8.81 28.6 77.2 64.8 69.6
LPT1 1.01 1.02 0.99 1.19 1.20 1.19 17.3 17.5 20.0 12.1 15.0 9.78 6.30 8.96 6.88 18.4 23.9 16.7 65.7 62.5 58.7
LPT2 1.01 1.02 0.99 1.20 1.22 1.19 18.3 19.4 20.1 13.6 11.8 13.9 8.07 6.69 6.36 21.6 18.5 20.3 62.7 63.9 68.6
LPT3 1.01 1.02 0.99 1.23 1.23 1.21 20.8 20.9 22.5 9.07 16.0 8.16 5.41 6.28 5.16 14.5 22.3 13.3 62.6 71.8 61.3
LPT4 1.01 1.02 0.99 1.30 1.36 1.33 28.3 33.3 34.2 9.17 3.25 3.39 15.6 3.35 3.66 24.8 6.61 7.04 37.0 49.2 48.1
EG 1.19 1.18 1.15 1.25 1.25 1.22 5.26 5.28 5.39 55.5 55.5 55.5 13.2 13.1 12.8 68.7 68.7 68.3 80.8 80.9 81.2
COND 1.01 1.02 0.99 44.1 52.3 49.9 125 95.7 95.9 169 148 146 26.2 35.3 34.2
CWP 1.25 1.25 1.22 1.86 1.83 1.80 48.7 47.1 48.2 2.10 2.49 2.38 1.98 2.51 2.29 4.07 5.00 4.66 51.5 49.8 50.9
FWH1 1.01 1.02 0.99 1.46 1.41 1.39 44.3 38.5 39.8 1.54 0.37 0.33 3.03 0.22 0.22 4.57 0.60 0.55 33.8 62.3 60.0
FWH2 1.01 1.02 0.99 1.26 1.43 1.35 24.1 40.3 36.3 1.23 1.74 1.06 1.61 3.21 1.20 2.84 4.96 2.27 43.3 35.2 46.9
FWH3 1.01 1.02 0.99 1.25 1.29 1.29 23.3 26.6 30.3 2.62 2.55 2.10 4.96 4.47 3.51 7.58 7.02 5.62 34.6 36.4 37.5
FWH4 1.01 1.02 0.99 1.18 1.21 1.18 16.2 19.4 18.7 1.69 2.37 1.22 3.32 5.46 2.06 5.01 7.82 3.29 33.7 30.3 37.2
FWH5 1.02 1.02 0.99 1.16 1.17 1.16 13.7 14.6 17.0 2.38 3.64 3.54 4.22 6.34 7.59 6.61 9.98 11.1 36.1 36.5 31.8
FWH6 1.03 1.03 1.01 1.14 1.15 1.12 10.7 11.9 10.6 5.23 6.55 5.13 6.88 10.7 6.72 12.1 17.3 11.9 43.2 37.9 43.3
FWH7 1.03 1.03 1.01 1.10 1.10 1.07 7.28 6.33 5.83 1.87 0.72 0.41 1.45 0.44 0.18 3.32 1.16 0.58 56.2 62.2 69.9
DA 1.02 1.02 0.99 1.19 1.18 1.25 17.4 15.8 26.0 2.59 1.41 3.76 3.03 0.98 8.50 5.61 2.39 12.3 46.1 58.8 30.6
FP 1.48 1.48 1.46 1.96 1.91 1.90 32.6 29.5 30.0 10.4 7.61 6.64 4.97 7.14 6.05 15.4 14.8 12.7 67.7 51.6 52.3
OT 1.01 1.02 0.99 1.48 1.48 1.46 45.6 45.3 47.6 8.73 9.34 9.38 7.55 8.49 6.21 16.3 17.8 15.6 53.6 52.4 60.2
tot 0.39 0.39 0.39 1.45 1.42 1.39 272 266 258 610 615 578 674 620 616 1284 1235 1194 47.5 49.8 48.4
81
82 superstructure-based synthesis and optimization of coal-fired power plants
temperature / C
Industrial design 0.55
5.6
280
5.5 0.45
270
5.4
0.35
5.3 260
250 0.25
5.2
5.1 240 0.15
41 43.5 46 48.5 51 41 43.5 46 48.5 51
Plant efficiency / % Plant efficiency / %
Figure 4.24: Fronts of the four cases with Figure 4.25: Optimal pressure ratio of the single-
seven feedwater preheaters reheat case with varied structures
tions. The single-reheat case with varied structures and the double-reheat case with
OT (and varied structures) give similar dominating Pareto fronts (Fig. 4.24); mean-
while, the maximum plant efficiency reached by the three optimization with varied
structures is much larger than that reached by the optimization with a fixed structure.
Therefore, the advantage of simultaneous optimization of the structure and the param-
eters becomes clear. In addition, for improving the efficiency from 48% to 49%, there
needs just a small increase in the COE, while from 49% to 50% the increase can be
large (steeper slope).
a) 4 50 b) 1.5 50
45 45
Figure 4.26: The influences of the ET on plant efficiency and COE for double-reheat unit with
varied structures: a) the contributions of fuel costs and capital investments to COE,
b) the contribution of the boiler to COE
a) 100 b) 100
Optimal pressure ratio / %
60 60
40 40
20 20
The first reheat The first reheat
0 0
40 44 48 52 40 44 48 52
Plant efficiency / % Plant efficiency / %
Figure 4.27: Optimal pressure ratios of the two double-reheat cases with varied structures: a)
with OT, b) with ET
4.7 summary and conclusions 85
main steam are high enough for these efficiency levels. Therefore, the optimal pressure
ratio of the first reheating is quite small (around 0.1) and that of the second is large
(0.81).
For further improving the plant efficiency, the significance of the first reheating is
highlighted; meanwhile, the second reheating contributes more and more. The optimal
pressure ratio of the second reheating remains within the range 0.150.25 (with OT, the
same with that of the single-reheat case (Fig. 4.25)) and 0.10.2 (with ET). The share
of heat absorbed by the second reheating increases significantly with a decrease in the
optimal pressure ratio.
Double reheats are beneficial for the highly-efficient designs (the Pareto front). The
optimal pressure ratio for the first reheating is within the range 0.20.3 (with OT) and
0.150.25 (with ET), while that of the second reheating is recommended within the
range 0.20.4 (with OT) and 0.10.3 (with ET). The introduction of an ET causes a
slight decrease in the optimal pressure ratios of reheating.
There is a deviation between the largest-efficiency design zone and the reason-
able quality of exhausted steam in the design of double-reheat steam cycles, es-
pecially when the steam temperatures are over 600 C. For the temperature level
of 750 C/750 C/750 C, the deviation even reaches 1.23 percentage points. An
effective way to reduce the deviation is increasing the throttle pressure. There is
a transition throttle pressure for each temperature level, at which the deviation
is eliminated. However, the transition pressure seems too large (~469 bar for a
temperature level of 650 C) in practice. Therefore, in practical designs of double-
reheat units, certain amount of efficiency benefit from increasing the temperature
level must be sacrificed to achieve reasonable quality of exhausted steam.
For trade-offs between plant efficiency and cost of electricity, there is a specific
solution for each front. For the front solutions whose efficiencies are below that
of the specific solution, it is more cost-effective to adjust not the steam conditions
of main and reheated steams but other decision variables for increasing the plant
efficiency. However, for the front solutions with the efficiency over that of the
specific solution, enhancing the steam conditions is inevitable for seeking higher
efficiency.
86 superstructure-based synthesis and optimization of coal-fired power plants
Away from the economically-optimal solution, the cost of electricity of the Pareto
solutions increases sharply with increasing plant efficiency. Therefore, the solu-
tions around the economically-optimal solution are proper design alternatives
for plant enhancement. Compared with the industrial design, there is a potential
of reducing the cost of electricity by 0.1 /kWh with increasing plant efficiency
by over 2 percentage points.
Based on the given cost functions and limited structural alternatives considered,
it seems not necessary to employ a second reheat in coal-fired power plants:
Adding a second reheat increases the cost of electricity for the same plant effi-
ciency. For the economically-optimal solutions, the costs of electricity of double-
reheat cases are higher than those of single-reheat cases.
Although the optimization problem can be properly solved by the superstructure based
approach, there are still several disadvantages: 1) Only limited structural alternatives
are manually defined a priori in the superstructure, which is a time-consuming, com-
plex, and error-prone task; 2) The superstructure built is not easy to be extended; 3) The
changes of the superstructure may need corresponding adjustments of (multi-objective)
differential evolution, for manipulating the stream paths and producing feasible solu-
tions efficiently. These drawbacks, however, can be overcome by a superstructure-free
approach implemented in the next chapter.
5
SUPERSTRUCTURE-FREE SYNTHESIS
A N D O P T I M I Z AT I O N O F C O A L - F I R E D
POWER PLANTS
87
88 superstructure-free synthesis and optimization of coal-fired power plants
where the solution structure is evolved by mutation, and all structure alternatives
in the space can be possibly reached by repeated structural mutation. In contrast to
the spaces explicitly defined by superstructures, the space is not known in advance,
and is only implicitly defined by the ECH. The knowledge-integrated, generic ECH
is a hierarchically-structured graph that classifies all considered energy conversion
technologies according to their functions Voll et al. (2012). This classification enables
Evolutionary algorithm
min
min
Deterministic optimization
where the weighting factors are positive and normalized (wk [0, 1] and k wk = 1) for
normalized objective functions. The weighting factors are no longer manually specified
or adjusted by the lower-level deterministic optimization but systematically manipu-
lated by the upper-level evolutionary algorithm (Fig. 5.1). After the deterministic opti-
mization of each structural alternative based on the super-objective in Eq. 5.4, the ob-
jective function values f are employed by the upper-level evolutionary multi-objective
algorithms (EMOA) (Fig. 5.1) to rank all competing solutions by their dominance and
crowding distances (e. g., in NSGA-II (Deb et al., 2002)). The solutions dominated by
many solutions are less likely to evolve as Pareto solutions and thus more likely to be
discarded. The crowding distance of a solution indicates the solution density around
the solution. The solutions with low density are more likely to be preserved to ensure
evenly-spread Pareto solutions. Moreover, to efficiently obtain only the Pareto solutions
90 superstructure-free synthesis and optimization of coal-fired power plants
with good compromise for practical problems (the interesting regions of the Pareto
front), the solution selection can also be based on the dominated hypervolume in algo-
rithm SMS-EMOA (Beume et al., 2007). The solution selection based on the dominated
hypervolume can discard the parts of the Pareto front with bad compromise, where
one objective function value changes dramatically with the conflicting objectives.
The mutation operator employs mutation rules to generate new structures. The number
of required mutation rules is minimized to a set of 6 mutation rules for structural
evolution of thermal power plants:
1. Remove one component with all of its interconnections.
4. Delete one component and insert a parallel connection of two other components.
5. Delete one component and insert a serial connection of two other components.
Function level
Chemical Expander Compressor Heater Cooler
reactor
CO2
capture Steam Gas Water Oil Gas Steam Gas Gas Steam Water Oil
device expander expander pump pump fan cooler heater cooler heater heater cooler
Post-combustion Steam Gas Water Oil Gas Steam- Gas- Gas- Steam- Gas- Oil- Oil- Oil-
CC turbine turbine pump pump fan gas gas steam water water gas steam water
HE HE HE HE HE HE HE HE
Technology level
Figure 5.2: The energy conversion hierarchy for thermal power plants
5.1 multi-objective superstructure-free synthesis framework 91
to their main functions. The nodes on the function level connect to the nodes on the
meta level and thereby define which mutation rules are applicable. For synthesizing
thermal power plants, the function level classifies the components that perform chemi-
cal reactions (combustion, gas cleaning, etc.), heat transfer (heating and cooling), fluid
expansion, fluid compression, etc. These basic tasks are considered for various fluids
as sub-functions, i. e., gas/steam/water heater, gas/steam/thermal-oil cooler, steam/-
gas expander, water/ gas/thermal-oil compressor, etc. Thus, these sub-functions are
integrated in the ECH as an additional layer.
The classification of the conversion technologies is realized by an inheritance-relation
between the corresponding technology and function nodes: As an example, a steam-
water heat exchanger is derived from the nodes "Steam cooler" and "Water heater", (see
Fig. 5.2, highlighted nodes). To define applicable mutation rules for each technology,
the nodes on the function level are linked to the corresponding nodes on the meta
level: For instance, the node "Steam cooler" is linked to the node "Cooler" and thereby
to all the four nodes in the meta level. Thus, any steam-water heat exchanger can be
principally connected to any other steam-water heat exchanger in parallel or in serial.
However, parallel connection of components may be not desired (dashed nodes and
links in Fig. 5.2) for specific applications, in which, for example, operation optimiza-
tion is not considered. In such case, a steam-water heat exchanger can be connected to
any other steam-water heat exchanger in serial but not in parallel. Therefore, by simply
specifying the links between the meta and function levels, the generation of meaning-
less design alternatives is avoided. It is also possible to protect certain components in a
structure from being replaced by setting these components as dispensable components,
for which all the mutation rules become invalid.
An important feature of this hierarchy-supported approach is that the considered
component set and the functions can be easily extended: To integrate a new technology,
the user only needs to arrange a new node on the ECHs technology level and connect
it to appropriate (new) nodes on the function level; in particular, no technology-specific
mutation rules have to be specified. For example, different types of fuel reactors (sub-
functions of chemical reactor) and corresponding technologies, such as coal combus-
tion chamber and coal gasifier for coal reactor, and gas combustion chamber for gas
reactor, can be added for different types of applications.
components must fulfill two criteria: First, they must conform to the chosen replace-
ment rule; Second, they must conform to the selected function of the target component.
If no such component is available, the mutation step is restarted. Finally, to generate
a structurally-feasible (well-connected) flowsheet, the free nonstandalone pins, if exist,
must be removed by connecting to other suitable pins of the same type. The nonstan-
dalone pins of a component are those that must be connected to fulfill the thermody-
namic model of the component. If with free nonstandalone pins, the flowsheets become
incomplete. Example nonstandalone pins are the steam-extraction inlet pin of a feed-
water preheater and the steam inlet pin of a steam turbine. There are also standalone
pins for certain technologies, e. g., the drainage inlet pin of a feedwater preheater and
a condenser, and the steam-extraction pin of a steam turbine. If no steam is extracted
from a steam turbine, the standalone steam-extraction pin becomes a free standalone
pin. Free standalone pins (not connected by any stream) are allowed in a complete
flowsheet, if the corresponding technology models allow these pins not connected.
For illustration, a single mutation step is presented for a simple Rankine-cycle based
thermal power plant (Fig. 5.3a), consisting of a steam generator, two steam turbines, a
condenser, a feedwater pump and a feedwater preheater (steam-water heat exchanger).
Assume that the mutation operator randomly selects the feedwater preheater for this
mutation. The functions of a feedwater preheater are "Steam cooler" and "Water heater".
If the "Water heater" function is randomly selected, possible replacement rules are
rules 15. In this example, rule 5 ("delete one component and insert a serial connection
of two other components") is randomly selected. The mutation operator then iden-
tifies two candidate components to be inserted in the place of the target feedwater
preheater. The two candidates must derive from the node "Water heater" and be avail-
able for serial connection. Therefore, the candidate components can be of the same
type or different types of steam-water heat exchanger, gas-water heat exchanger and
oil-water heat exchanger. In the example, the mutation operation randomly chooses
two components of steam-water heat exchanger, and thus specifies the rule as "replace
the existing steam-water heat exchanger by a serial connection of two steam-water heat
exchangers". Finally, the target feedwater preheater is removed from the initial flow-
sheet and the serial connection of two heat exchangers are inserted (Fig. 5.3b). The
connections for the selected function can be retained (Fig. 5.3b). The mutation step is
completed by establishing the missing connection of nonstandalone pins: water pin 5
a) b) c)
steam turbine
standalone 1 2
pins heat 6
standalone
pins
heat
steam pump 34 5 7 8 9
generator
feedwater preheater
Figure 5.3: Example run of a single mutation step: a) initial flowsheet, b) flowsheet after appli-
cation of replacement rule, c) flowsheet after post-processing
5.1 multi-objective superstructure-free synthesis framework 93
and steam pin 8 (Fig. 5.3c). For pin 5, the suitable candidate pins for connection are
pin 6 and pin 7; while for pin 8, only pin 2 is suitable for connection. In this example,
it is assumed that a water stream is established between pin 5 and pin 7, and a steam
stream is established between pin 2 and pin 8. Finally, a structurally-feasible structure
is generated (Fig. 5.3c).
For efficiently finding suitable pins to remove free nonstandalone pins, the pressure
and temperature information of the pins in the initial structures, if already successfully
evaluated by the deterministic optimization, can be useful. For example, a suitable wa-
ter inlet pin for connecting outlet pin 5 must satisfy that the pressure of the candidate
pin is smaller than that of pin 1 (or pin 4); thus, from the evaluation result of the initial
structure, the inlet pins whose pressures are higher than that of pin 1 are identified
by the algorithm and not considered as candidate pins, e. g., pin 3. In this way, the
number of candidate pins can be reduced significantly, so that feasible structures can
be more easily generated. In addition, the newly-added pins by the replacement rules
are also regarded as candidate pins, even without pressure and temperature values.
This ensures that the space of feasible structural alternatives is not reduced due to the
exclusion of these pins.
The mutation rule for insertion is able to address the shortcomings of the replacement
rules (see section 5.1.2.1). The insertion rule enables the addition of any technology
into any structure. In particular, it is no longer required that another technology with
the same function has already been existing in the parent structure.
When inserting a component into a structure, the challenge is to connect the new
component to the existing structure such that a feasible structure is efficiently gen-
erated. An intuitive way of insertion would be to add a component and randomly
connect all open pins of the new component to exiting pins. Such a free insertion
scheme is not efficient: Any feasible structure must satisfy temperature and pressure
relationships imposed by the component models. For example, free insertion might in-
sert a turbine with the inlet pressure smaller than the outlet pressure, which definitely
leads to infeasible solutions. The situation becomes even worse for components with
more pins, e. g., heat exchangers. To avoid the generation of infeasible alternatives, an
inserted component is suggested to replace one or more streams associated with the
94 superstructure-free synthesis and optimization of coal-fired power plants
f)
Connect free nonstandalone pins yes
c) Split the selected stream, connect the added fail
succeed
component, remove free nonstandalone pins
Any function is with free nonstandalone pins?
condenser
no
Figure 5.4: An exemplary run for the stream- Figure 5.5: Flowchart for the generalized inser-
based insertion of a heat exchanger tion rule
components function, e. g., a steam stream is replaced by a steam expander. Note that
this stream-based insertion scheme does not reduce the search space compared to the
free insertion but the number of feasible solutions generated by insertion is higher.
For illustrating the stream-based insertion, consider the following example (Fig. 5.4):
A steam-gas heat exchanger is selected to be inserted to a given parent structure
(Fig. 5.4a). The functions of the steam-gas heat exchanger are to cool down gas and
to heat up steam (Fig. 5.4b). Thus, the heat exchanger is allowed to be "placed" on
either a gas or a steam stream. If the function of heating up steam is selected, all three
steam streams represent candidate streams for the placement of the heat exchanger.
Assuming stream 2 is selected, stream 2 is split up and connected to the cold side of
the heat exchanger. The new structure (Fig. 5.4c) is obtained assuming that the gas side
can be properly connected by replacing a gas stream in the parent structure.
This insertion rule can be generalized (Fig. 5.5): Given a parent structure, the in-
sertion rule first randomly selects a technology that allows insertion (Fig. 5.5a). The
selected technology is instantiated as a component and the technologys functions are
identified (Fig. 5.5b): Note that a single technology usually has at least two functions,
e. g., the functions of a turbine are 1) to expand a stream and 2) to generate power,
and the functions of a heat exchanger are 1) to heat up a stream and 2) to cool down a
stream; thus, a function to be selected is usually associated with one stream type and two
free pins, each of which allows multiple connections. The two free pins are of the same
stream type, if no phase change or chemical reaction is involved; otherwise, the types
of the two pins are different (Fig. 5.5c). If a function with free pins of the same type is
5.2 nlp model of thermal power plants 95
selected, one particular stream can be selected (Fig. 5.5d) and split into two streams in-
serting the instance of the selected technology in between (Fig. 5.5e). If a function with
free pins of different types is selected, new streams will be established by randomly
connecting the free pins to other suitable pins in the parent structure (Fig. 5.5f; for
details of this post-processing, see the detailed description in section 5.1.2). Moreover,
if the new component cannot be embedded into the parent structure, e. g., because
no appropriate streams are available to which the new component can be connected,
the selected component is discarded and another technology is selected for insertion.
Finally, once all functions of the new component are employed with no free nonstan-
dalone pins left for connections, a new structurally feasible (well-connected) structure
has been generated for deterministic optimization.
dTsat (v00 v0 )
= Tsat 00 (Clausius-Clapeyron relation) , (5.5)
dp h h0
dh0 dTsat
= v0 (1 0 Tsat ) + c0p , (5.6)
dp dp
where, v0 , v00 , h0 and h00 represent the specific volume and enthalpy of saturated water
and steam. The variables 0 and cp0 are isobaric expansion coefficient and specific iso-
baric heat capacity of saturated water. Note that the function h00 ( p) is multi-valued but
its derivative (not listed here) is seldom used for optimization.
Water and steam properties are calculated through the freesteam library1 . Freesteam
provides property values and derivatives of high accuracy as IAPWS-IF97 formula-
tions (Wagner et al., 2000), and thus, enables to handle the thermodynamic processes
involving a wide range of pressure and temperature, even supercritical ones. More im-
portantly, the integer variables indicating the state zones of water and steam for mathe-
matical programming as in (Tveit and Fogelholm, 2006; Jdes, 2009; Ahadi-Oskui et al.,
2010) are encapsulated in the library, and consequently, the lower-level optimization
problems can be treated as NLP.
The NLP problems are implemented in the modeling language GAMS (version
24.2.3) and solved using the NLP solver CONOPT3 (version 3.15P). All solver options
other than lkdebg2 , rtnwma and rtnwmi3 , are set as the default values.
The presented superstructure-free approach can be coupled to any model for thermal
power plants, implemented in the lower-level deterministic optimization (Fig. 5.1). For
this illustrative study, the objective is to maximize the thermal efficiency (%):
| PST | | PFP |
max = 100 , (5.7)
Q gross
where the PST and PFP are the power generated by steam turbine and the power con-
sumed by feedwater pump, respectively. The gross heat input to the plant (Q gross ) can
be calculated based on the total heat absorbed by the working fluid (Q abs ) and the
relative heat loss rate (loss , %) of steam generation and superheating:
100 Q abs
Q gross = . (5.8)
100 loss
The parameter values and variable bounds are specified according to Table 5.1.
In general, the generic ECH (Fig. 5.2) could be applied directly. However, the presented
illustrative study does not consider all technologies in the ECH (Fig. 5.2). Thus, the
ECH is refined (Fig. C.1) by removing functions not fulfilled by any technology. For
2 This option controls the function and derivative debugger of CONOPT (Drud, 2004). The value is set as
zero so that the debugger is not used. The function values and derivatives of properties of water and
steam are computed completely from the extrinsic library (the freesteam).
3 The options, rtnwma and rtnwmi, are the maximum and minimum feasibility tolerances for constraints
(for more details, see (Drud, 2004)). Larger values of the two options increase the feasibility of the whole
model. The values are set as 1E-5 and 7E-6 for all optimization runs in the thesis.
5.3 evaluation of the superstructure-free approach 97
the technology and function levels, the type of fuel supplied to the power plant is fixed.
The steam generator and condenser are treated as indispensable components. For the
meta level, parallel connections of components are not considered, as the associated
mixing process is difficult to handle by standard NLP (but by NLP with discontinuous
derivatives (McCarl et al., 2014) or MINLP). Moreover, a parallel connection of two
identical technologies is of limited importance for optimizing the thermal efficiency
at a single design point. Eventually, the allowed mutation rules include the deletion,
direct insertion and serial connection of superheaters, steam-water heat exchangers,
steam turbines and pumps (see section 5.1.1).
For all structural alternatives, at least one of each type of the technologies (steam tur-
bine, condenser and steam generator) are employed to ensure a full Rankine cycle. The
upper bounds for the number of components are needed only for the thermodynamic
objective considered. If an economic objective is considered, no upper bounds for com-
ponent numbers are required. However, in this study, only the thermal efficiency at a
single design point is considered. Without the upper bounds on the number of steam
turbines, feedwater preheaters or superheaters, the algorithm will continue introduc-
ing these components, because the optimal solution would employ an infinite number
of these components. For the demonstration purpose, the maximum numbers for each
component type are arbitrarily selected: 3 for superheater, 4 for feedwater preheater, 10
for steam turbine, and 2 for feedwater pump. In addition, the steam outlet of a steam
turbine, the drainage inlet of a feedwater preheater, and the drainage inlet and steam
inlet of a condenser, allow up to 4 connections.
The initial flowsheet (Fig. C.2a) is a simple thermal power plant consisting of only
four components: a COND, a pump, a steam generator and a steam turbine. The basic
98 superstructure-free synthesis and optimization of coal-fired power plants
a) 500 b) 500
50 50
Pressure / bar
Pressure / bar
5 5
optimal solution
initial solution
0.5 0.5
0.05 0.05
0.005 0.005
0 1000 2000 3000 4000 0 1000 2000 3000 4000
Enthalpy / kJ/kg Enthalpy / kJ/kg
Figure 5.6: Pressure-enthalpy diagrams of the initial and optimal solutions of the illustrative
study: a) the initial solution ( = 38.82%) and the optimal solution obtained by
the original superstructure-free approach ( = 40.20%), b) the optimal solution
obtained by the new superstructure-free approach ( = 49.45%)
specifications for the plant are listed in Table 5.1. The initial solution (Fig. 5.6a and
Fig. C.3a) has an efficiency of 38.82%.
When the initial flowsheet is optimized based only on the replacement rules but
without the insertion rule, the resulting optimal structure (Fig. 5.6a) has no feedwater
preheaters or reheaters, but the maximal number of turbines. The single turbine is
replaced by the max number of turbines, since the isentropic efficiency of each turbine
(expansion) is set as constant. Thus, a multi-stage expansion leads to a smaller enthalpy
or dryness of the exhausted steam, thereby a higher thermal efficiency of 40.20%.
The thermal efficiency of the optimal solution is 49.45%. All reheated streams achieve
the maximum allowed temperature (923 K). The steam for feedwater preheating is
extracted directly before each reheating, thus maximizing both steam expansion be-
fore reheating and the effect of reheating on boosting the power generation. Thus, the
benefit of increasing feedwater preheating temperature is, in fact, constrained by the
reheating pressures, which leads to a final preheating temperature of 623.1 K.
1 2 rate rate
1 2
50 400
Accumulated time / min
50 400
Accumulated time / min
80
Thermal Efficiency / %
Thermal Efficiency / %
generated
47 300 47 generated 300 60
Rate / %
solutions
solutions
3 4
44 200 44 200 40
5 3 4
5
41 100 41 100 20
accumulated time accumulated time
38 0 38 0 0
0 50 100 150 200 0 50 100 150 200
Generation number / - Generation number / -
Figure 5.7: Representative runs with different maximum mutation strengths (1 for (a) and 5
for (b)) for the illustrative study (similar to all performed tests): generation rate of
structurally feasible structures ( 1 ), rate of successful evaluation of structurally fea-
sible structures ( 2 ), total evaluation time ( 3 ), evaluation time of feasible solutions
+ unsuccessful optimization of structurally feasible structures produced by insertion
( 4 ), evaluation time of feasible solutions ( 5 ). ( = 20, = 25, m ms = 1000 kg/s)
100 superstructure-free synthesis and optimization of coal-fired power plants
Four intermediate solutions of the optimization run in Fig. 5.7a are shown in Fig. C.4.
Note that all four solutions have identified three-stage reheating and multi-stage feed-
water preheating; however, the reheating positions and the positions of steam extrac-
tions for feedwater preheating in the four structures are adjusted continuously by the
mutation operator. The first solution featuring three-stage reheating and four-stage
feedwater preheating is observed in the 29th generation (Fig. C.4a). In this solution, the
first steam extraction occurs before reheating. The best objective function value reaches
over 49% at the 37th generation (Fig. C.4b). This structure employs only three-stage
feedwater preheating but incorporates a secondary turbine supplying steam extraction
specifically to its feedwater preheater. An increase in the utilization of low-pressure
steam further slightly enhances the efficiency to 49.26% (40th generation, Fig. C.4c). In
this structure, the drainages of two feedwater preheaters enter the same downstream
preheater. In generation 52, a solution (Fig. C.4d) is identified whose objective function
value differs by only 0.02% from the optimal solution. The only difference is that the
3rd steam extraction of the intermediate solution is supplied by a secondary turbine.
Note that there are also many other intermediate solutions with their objective values
close to the best-known objective (identified in the 66th generation). The exemplary
run illustrates a typical feature of evolutionary algorithms: Near-optimal solution al-
ternatives are generated in every optimization run; thus, there is no need to implement
additional mathematical techniques (e. g., integer cuts) to identify various near-optimal
solutions, as is required in deterministic methods (Voll et al., 2015). The presented ex-
emplary run highlights the automatic identification of complex power plant cycles
without requiring the user to explicitly specify which solution alternatives are to be
considered.
Table 5.2: Computational performances and optimality gaps of 6 optimization runs with dif-
ferent maximum mutation strengths: tot total evaluation time (min), best-known
thermal efficiency (%), time
to find the best objective (%), ngeneration number to
find the best objective (1), eoptimality gap (%)
and 2) only a small number of structurally feasible structures are infeasible solutions.
Moreover, the time of unsuccessful optimization of insertion-rule generated structures
is negligible. Thus, no heuristics are required to speed up the optimization.
The superstructure-free synthesis is performed for another five times for each maxi-
mum mutation strength (Table 5.2). All the five new runs find exactly the same optimal
solution (Fig. 5.6b). The geometric mean time of these optimization runs is almost not
affected by the maximum mutation strength, due to similar generation rates of feasible
solutions and thus similar amounts of structural evaluations. However, a larger maxi-
mum mutation strength, in general, can lead to a fast structural evolution: shorter time
and fewer generations to find the optimal solution (Table 5.2). The computational time
is considerable, as the presented concept employs evolutionary algorithms to automat-
ically generate and identify the optimal plant structure. The significant computational
effort is the main disadvantage of the proposed methodology. However, evolutionary
algorithms in turn can principally find the global optimal solution together with many
near-optimal solutions.
("Steam cooler" and "Water heater"). The technologies, "Steam generator" and "Con-
denser", remain dispensable. As before, the parallel connection of two identical tech-
nologies ("Parallel connection allowed" node in Fig. C.5) is of limited importance, as
operational optimization is not considered.
Two objectives, thermal efficiency and cost of electricity, are considered separately
for single-objective optimal synthesis and simultaneously for bi-objective optimal syn-
thesis. The optimal solutions, near-optimal solutions, Pareto fronts and computational
efforts are compared and discussed in detail.
For the upper-level EA, the same ( + )-evolution strategy is used: = 20 and = 25.
The maximum mutation strength is set as 5. The EA will terminate once 8,000 feasi-
ble structure alternatives are generated. For the lower-level deterministic optimization,
the specifications follow Table 5.1. Note again, the total net power output is fixed
other than the mass flow rate of main steam. In addition, if thermal efficiency is the
only objective, maximum numbers of components of involved technologies are set (see
section 5.3.2): 12 for steam turbine, 8 for feedwater preheater, 2 for reheater, 2 for de-
superheater and 2 for feedwater pump.
Starting from an initial structure with 4 feedwater preheaters and a thermal efficiency
of 46.49% (Fig. 5.8a), the structural evolution finds the optimal structure with an effi-
ciency of 49.6% (Fig. 5.8b). Note that the thermal efficiency calculated in this chapter is
not directly comparable with that in chapter 4 based on Fig. 4.3 and Fig. 4.16, in which
the power consumption of auxiliary devices (e. g., the fans in the boiler subsystem and
the cooling-water pump for condenser) is considered. The flowsheets and temperature-
entropy diagrams of both initial and optimal solutions are given in Fig. C.6 and C.7.
The optimal solution features eight-stage feedwater preheating, double reheating, 2 de-
superheaters, and a secondary turbine supplying steam for one feedwater preheater.
The temperatures of the main and reheat steams reach the maximum allowed tem-
perature. The feedwater is heated up to 647 K (thermodynamically reasonable) by the
a) 500 b) 500
50 50
Pressure / bar
Pressure / bar
5 5
0.5 0.5
0.05 0.05
0.005 0.005
0 1000 2000 3000 4000 0 1000 2000 3000 4000
Enthalpy / kJ/kg Enthalpy / kJ/kg
Figure 5.8: Pressure-enthalpy diagrams of the initial and optimal solutions for maximizing the
thermal efficiency: a) initial solution ( = 46.49%), b) optimal solution ( = 49.60%)
5.4 superstructure-free synthesis of complex coal-fired power plants 103
feedwater preheater above rh1 and the successive de-superheater (see Fig. C.6b). The fi-
nal feedwater preheating temperature is constrained by a specified maximum pressure
of steam extractions (200 bar). The optimal reheating pressure ratios are 0.37 (rh1) and
0.26 (rh2), respectively. Considering the quality of the exhausted steam (0.946), these
two reheating pressure ratios agree well with Fig. 4.14 (type 1 in zone A). However, the
optimal structure (Fig. C.6b) can not be generated from the superstructure in Fig. 4.3,
because of the steam turbine with one single steam extraction and the de-superheaters.
To integrate the de-superheater into the superstructure, a vast number of possibilities
have to be considered, e. g., the positions of de-superheaters and the feedwater to be
heated by the de-superheaters. Accordingly, this requires considerable effort to revise
the superstructure and related generation algorithm for structural alternatives.
An exemplary run for the complex problem for maximizing the thermal efficiency
(Fig. 5.9) illustrates similar performance as in Fig. 5.7. The generation rates of both
structurally-feasible and feasible structures are still high, over 90%, which leads to a
fast structural evolution. Near-optimal structures are generated from around 30 gener-
ations. Surprisingly, the generated solutions after 30 generations are grouped into two
efficiency ranges according to the number of reheatings employed. Single-reheat struc-
tures are frequently generated, when the mutation operator removes one reheater from
the given preserved parent solution. It also indicates that adding one more reheating is
much more effective than adding more feedwater preheaters and de-superheaters for
increasing the thermal efficiency.
Some interesting intermediate structures are listed in Fig. C.8. After adding two
reheaters and one more feedwater preheater to the initial structure (Fig. C.6a), one
possible structure is generated as in Fig. C.8a. In this structure, the turbine train in
the initial structure now serves as a secondary turbine with steam extractions (ET).
Then, more newly-added feedwater preheaters are connected to both the ET and main
turbine (MT), as in Fig. C.8b. The mutation operator continues to change the config-
54 1000 100
rate
Accumulated time / min
1 2
Thermal eciency / %
52 800 80
3
solutions generated 4
in each generation 5
50 600 60
Rate / %
48 400 40
44 0 0
0 50 100 150 200 250 300
Generation number / -
Figure 5.9: One representative run (similar to all performed runs) for maximizing the thermal
efficiency: generation rate of structurally feasible structures ( 1 ), rate of successful
evaluation of structurally feasible structures ( 2 ), total evaluation time ( 3 ), eval-
uation time of feasible solutions + unsuccessful optimization of structurally feasi-
ble structures produced by insertion ( 4 ), evaluation time of feasible solutions ( 5 ).
( = 20, = 25, maximum mutation strength 5, Pnet = 1000 MW)
104 superstructure-free synthesis and optimization of coal-fired power plants
uration and more feedwater preheaters are connected to the MT (Fig. C.8c and C.8d).
Therefore, a complex structure can evolve gradually to completely different structures
without getting stuck. For this synthesis task, the benefit of an ET becomes weak after
the maximum number of feedwater preheaters are employed.
The superstructure-free synthesis is also performed for another five times (Table C.1).
The computation time for this complex problem is much greater than that of the illus-
trative study, as the optimization time for each structurally-feasible structure becomes
longer. This is due to that 1) more constraints and variables are involved in each eval-
uation, and 2) fixing the power output makes the GAMS model even more complex.
More importantly, not all these 6 runs identifies the optimal solution. However, the
geometric average optimality gap of all 6 runs is only 0.01%: Good near-optimal alter-
native solutions are generated, although the identification of the optimal solution is
not guaranteed.
Important information carried by all generated solutions (including near-optimal so-
lutions) are given in Fig. 5.10 and 5.11. It is confirmed that, for increasing the thermal
efficiency, adding one additional reheating is much more effective than adding sev-
eral feedwater preheaters (Fig. 5.10). The more the feedwater preheaters are employed,
the less efficiency increase can be gained by adding additional feedwater preheaters
(Fig. 5.10). The number of feedwater preheaters above rh1, in general, has limited or no
influence on the thermal efficiency (Fig. 5.11). The optimal pressure ratios for reheating
are 0.150.35 (rh1) and 0.250.45 (rh2).
For the EA, the same evolution strategy, maximum mutation strength and termination
criteria are employed. For the deterministic optimization, the cost of electricity (COE)
is calculated by the same procedure described in section 4.6.1. As only those variables,
whose bounds are listed in Table 5.1, are optimized, additional variables involved in
9 0.5 5
rh2
8
Number of FWH or reheaters / -
6
FWH 0.3 3
5
4
0.2 2
3 Reheater
2
0.1 1
1
0 0 0
45 46 47 48 49 50 45 46 47 48 49 50
Thermal Efficiency / % Thermal efficiency / %
Figure 5.10: Thermal efficiency versus Figure 5.11: Thermal efficiency versus optimal
numbers of feedwater pre- reheating pressure ratios and num-
heaters and reheaters ber of feedwater preheaters above
reheat 1
5.4 superstructure-free synthesis of complex coal-fired power plants 105
employed cost functions (App. A) are specified as fixed values. The net power output of
all solutions is fixed to 1000 MW. In addition, the maximum numbers of components of
involved technologies, steam turbine, feedwater preheater, de-superheater and reheater,
are no longer required (cf. section 5.3.2).
The structural evolution is initialized by the same initial structure (Fig. C.9a) as in
section 5.4.1. However, the optimization of the initial structure with respect to COE
suppresses the use of redundant turbines (the highest-pressure turbine, which is con-
nected downstream by only one turbine) and one feedwater preheater (see Fig. 5.12a
and C.10a). The initial structure reaches a COE of 5.248 /kWh with an efficiency
of 47.87%, while the optimal structure found (Fig. 5.12b) achieves much lower COE
(5.063 /kWh). The optimal solution features three-stage reheating, seven-stage feed-
water preheating, three de-superheaters and two ETs (Fig. C.9b). The main steam tem-
perature is below the maximum allowed temperature, while the reheat temperatures
remain at this highest possible temperature (Fig. C.10b). The feedwater is heated up
to 591 K, far below the value of the solution in Fig. C.7b. The positions of the three
de-superheaters (Fig. C.9b) are reasonable from the second law of thermodynamics:
The feedwater out of one high-pressure feedwater preheater is further heated by one
lower-pressure steam extraction, so that superheating degrees of the steam extractions
are used effectively to heat temperature-matched feedwater. Surprisingly, the steam en-
tering ET1, which supplies steam extractions for FWH35, is from the third reheating
other than the first two. A second ET is configured as well to supply steam for FWH1.
The layout of feedwater preheaters of the optimal solution (Fig. C.9b), i. e., the num-
ber of feedwater preheaters and connections, is similar to that of the industrial de-
sign (Fig. 4.16). However, the (near-)optimal solutions employ more reheatings and
de-superheaters than currently found in modern power plants. The reheating temper-
atures of the (near-)optimal solutions are at the upper bound (973 K in Table 5.1). The
excessive use of technologies and the optimal reheating temperatures at the upper
bound indicate the need to revise the cost functions to reflect current industrial prac-
tices. As mentioned in section 4.6.2.4, the cost functions employed in the thesis are
a) 500 b) 500
50 50
Pressure / bar
Pressure / bar
5 5
0.5 0.5
0.05 0.05
0.005 0.005
0 1000 2000 3000 4000 0 1000 2000 3000 4000
Enthalpy / kJ/kg Enthalpy / kJ/kg
Figure 5.12: Pressure-enthalpy diagrams of the initial and optimal solutions for minimizing the
cost of electricity: a) initial solution (COE = 5.248 /kWh, = 47.87%), b) optimal
solution (COE = 5.0627 /kWh, = 51.55%)
106 superstructure-free synthesis and optimization of coal-fired power plants
adjusted according to limited published data, suggesting that these cost functions are
valid only within certain scope. In addition to the use of practical cost functions, prac-
tical structural constraints can also be imposed during mutation operations to avoid
searching non-practical features. Anyway, the case study demonstrates that, if proper
cost functions and practical structural constraints are available, the superstructure-free
approach can be applicable for real industrial applications.
The representative run (Fig. 5.13) shows that, compared with previous runs (Fig. 5.7
and 5.9), the generation rate of structurally feasible structures remains almost un-
changed, while a large share (up to around 30%) of the structurally-feasible structures
are evaluated as infeasible solutions by the lower-level deterministic optimization. This
is mainly due to two reasons: 1) More structurally-feasible structures are not thermo-
dynamically reasonable. The steam pin at the hot-side of a newly-added feedwater
preheater is more frequently connected to the outlet pin of an improper steam turbine,
as those structures with redundant steam turbines are less likely preserved as parent
solutions (the redundant steam turbines lead to a larger COE). Thus, the possibility
of connecting a proper steam turbine for a newly-introduced feedwater preheater be-
comes less and the violation of pressure constraint on the steam extraction occurs more
frequently. Those proper steam turbines are mostly introduced to the same parent as
well by employing one or more insertion or parallel-connection rules of the steam
turbine. 2) The optimization of some structurally-feasible and thermodynamically rea-
sonable structures fails because the model formulation (e. g., scaling and bounds of
equations and variables) is not good enough when introducing the highly nonlinear
cost functions.
The evolution process is similar to those in Fig. 5.7 and 5.9: The COE of most struc-
tures generated in each generation is close to the best-known COE so far. However,
there are also many solutions generated in each generation that are far larger than the
best-known COE so far, referred to as bad structures/solutions. This may be because
1) truly bad structures are generated, when, for example, all the reheaters in the parent
80
Accumulated time / min
5.3 1400
1200
generated solutions 4 60
Rate / %
5 1000
5.2
800
40
600
5.1 400 20
accumulated time 200
5 0 0
0 50 100 150 200 250 300
Generation number / -
Figure 5.13: One representative run (similar to all performed runs) for minimizing the cost of
electricity: generation rate of structurally feasible structures ( 1 ), rate of successful
evaluation of structurally feasible structures ( 2 ), total evaluation time ( 3 ), evalu-
ation time of feasible solutions + unsuccessful optimization of structurally feasible
structures produced by insertion ( 4 ), evaluation time of feasible solutions ( 5 ).
( = 20, = 25, maximum mutation strength 5, Pnet = 1000 MW)
5.4 superstructure-free synthesis of complex coal-fired power plants 107
a) 7 b) 9
Number of reheating / 1
6 8
Number of FWH / 1
5 7
4 6
3 5
2 4
1 3
0 2
5 5.1 5.2 5.3 5.4 5 5.1 5.2 5.3 5.4
Cost of electricity / /kWh Cost of electricity / /kWh
Figure 5.14: Numbers of reheating (a) and feedwater preheater (b) of the solutions generated
for minimizing the cost of electricity
structure are removed by the mutation operator with a maximum mutation strength
of 5, and 2) some good structures are evaluated as bad solutions, as the optimization
is trapped at local optimums due to the high nonlinearity of involved cost functions.
In addition, some intermediate solutions (Fig. C.11) of the representative run are pre-
sented as well to show the diversity of the searched structures.
The computational effort for minimizing the COE is significantly larger (Fig. 5.13).
The total time of the representative run even reaches over 30 hr. The evaluation time of
infeasible solutions including those generated with insertion rules involved becomes
greater as well because of the decrease in generation rate of feasible solutions. Another
five optimization runs are performed as well (Table C.2). Only two of the five find the
optimal solution in Fig. 5.12b. The average time for finding each best-known solution
is around 20 hr. The average optimality gap remains a low value, only 0.013%, indi-
cating many other near-optimal solution alternatives. In addition to the near-optimal
solutions, the automatic generation of structural alternatives with no engineers atten-
dance further makes the large computation time acceptable.
Near-optimal solutions mostly employ three or four stages of reheating (Fig. 5.14a),
and 5 to 7 feedwater preheaters (Fig. 5.14b). Again, these near-optimal solutions may be
less industrially applicable due to the used cost functions. However, they do illustrate
there are diverse near-optimal choices.
For the upper-level EA, the ( + )-evolution strategy is specified as (100 + 1) with a
maximum mutation strength of 5. The SMS-EMOA is employed for the selection of so-
lutions. The lower-level deterministic optimization evaluates each individual structure
regarding a weighted super-objective, for which the weighting factor is continuously
adapted by EA. The variables, their bounds, and power outputs of all solutions are
specified the same as in section 5.4.2. The upper bound of COE is set as 7.5 /kWh.
The maximum numbers of components of involved technologies are specified for the
thermal-efficiency objective: 15 for steam turbine, 10 for feedwater preheater, 5 for re-
heater and 5 for de-superheater.
Starting from the same initial structure (Fig. C.9a), several optimization runs are per-
formed with similar performance (Fig. 5.15a). The presented performance is similar to
the synthesis task considering only the COE (Fig. 5.13): still high generation rate of
108 superstructure-free synthesis and optimization of coal-fired power plants
a) 100 8000 b) 8
5.15
7.5
Rate / % 60 6.5
3
4000 5.05
1 gen. here=100 gen. 6 51 52 53
40 4 50th
of (100+1)-ES
5.5 100th
5 2000
20 150th
5 200th
250th (Pareto front)
0 0 4.5
0 50 100 150 200 250 50 52 54 56 58
Generation number / - Thermal efficiency / %
Figure 5.15: One representative run (similar to all performed runs) for the bi-objective optimal
synthesis: a) generation rate of structurally feasible structures ( 1 ), rate of success-
ful evaluation of structurally feasible structures ( 2 ), total evaluation time ( 3 ),
evaluation time of feasible solutions + unsuccessful optimization of structurally
feasible structures produced by insertion ( 4 ), evaluation time of feasible solutions
( 5 ); b) evolution of the best-known front
a) 1050 650 b) 11
10
950 600
main steam
Number / 1
850 550 8
main steam
7 feedwater preheaters
750 500
6
reheaters
650 450
feedwater 5
550 400 4
50 52 54 56 58 50 52 54 56 58
Thermal efficiency / % Thermal efficiency / %
Figure 5.16: Variable values (a) and numbers of FWH and reheater (b) of all Pareto solutions
The insertion rule enables an efficient addition of technologies that are not al-
ready existing in parent solutions. In particular, this rule generates serial connec-
tions of units with different functions, such as heat exchangerturbine series.
Due to lack of real-world cost functions for pulverized-coal power plants, the
cost functions adapted based on limited published data are employed. These cost
functions lead to the optimal and near-optimal solutions with more units (e. g.,
reheaters and de-superheaters) than currently found in modern power plants.
Moreover, the reheating temperatures in the (near-)optimal solutions are at the
upper bound. Therefore, it is indicated that there is a need to revise the cost
functions to reflect current industrial practices. The case studies presented in
this chapter demonstrate that the superstructure-free synthesis framework can
be applicable for complex real-world applications.
6
S U M M A RY A N D C O N C L U S I O N S
111
112 summary and conclusions
Table 6.1: Comparisons between the methodologies employed in this thesis for the analysis and
synthesis of thermal power plants
Structure space
Specific structure Superstructure ECH
definition
Structural evolution DE (mutation & EA (only
algorithm crossover) mutation)
Evaluation of an Simulation &
Optimization:
individual structure Solving a linear Simulation
NLP (GAMS)
alternative equation set
Num. of
meaningless Large Small
structures?
(Near-)optimal
solution?
Expert knowledge
requirement
Multi-objective
trade-offs
Multi-objective Non-dominated Aggregation
selection technique sorting selection
Small
<1 s for each
Computational simulation Large Enormous
effort needed <10 s for ~1 day for MOO ~4 days for MOO
exergoeconomic
analysis
Flexibility and
Low High
extensibility
tion and recombination operations are activated. The building of the superstructure
needs expert knowledge to decide which structural alternatives should be included.
The superstructure may take a large number of meaningless structures into account.
The approach can easily provide many near-optimal solutions and high-quality com-
plete Pareto fronts by non-dominated sorting and crowding distance assignment. The
computational effort of exploring the superstructure is large (around 1 day) for multi-
objective trade-off. The approach turns out to be with low flexibility and extendability
due to the use of problem-specific mutation and crossover operations.
The superstructure-free optimization-based synthesis approach explores the solution
space defined by an ECH. The easy-to-extend, flexible ECH considers far more struc-
tural alternatives than the superstructure built and avoids a large number of mean-
114 summary and conclusions
The presented thesis only deals with the grassroots design of pulverized-coal power
plants but not incorporates off-design performances of all involved components. This is
reasonable for single-purpose (product), single-source power plant, as all the employed
components are set to operate under partial loads: higher efficiency at the design load
would generally lead to higher efficiency at partial loads for such a type of power plant.
However, when new power plants are synthesized for multiple products and/or mul-
tiple sources, or existing thermal power plants are expected to be further enhanced by
introducing new technologies, the operation-level synthesis must be taken into account.
For large-scale power plants, off-design models for a wide range of technologies, e. g.,
Stodola ellipse model (Cooke, 1983, 1985) for large steam turbines, Rabek method for
feedwater preheater (Rabek, 1963), etc., which are available for the design, integration
and synthesis of coal-fired power plants, can be found in the manual of Ebsilon Pro-
fessional1 . Sound mathematical models to predict off-design performances are to be
developed for different components.
The thesis has proposed a flexible system-level superstructure-free synthesis frame-
work, which is capable of coping with those new challenges (Fig. 2.4) for further en-
hancing the coal power plants. In the near future, this approach is expected to extend
the range of its application. For this, more available technologies that could be inte-
grated into the pulverized-coal power plants should be included in the energy conver-
sion hierarchy, for example, Organic Rankine Cycle (Tchanche et al., 2011; Al-Sulaiman
et al., 2013), supercritical CO2 cycle (Wright et al., 2010), CO2 capture technologies
(Harkin et al., 2012; Samanta et al., 2011), solar-thermal utilization technologies (Jamel
et al., 2013), energy storage (Morandin et al., 2012a,b), etc. Except for the grassroots de-
sign, optimal retrofits of pulverized-coal power plants considering multiple available
technologies can be solved as well by combining well-developed off-design models of
all involved technologies.
The design and retrofits of real-world pulverized-coal power plants have to consider
more realistic (structural) constraints and objective functions, which have not been
carefully considered yet. For example, the steam-extraction pressure for de-aerator
should be limited within the range from 5 to 15 bar (Hillermeier et al., 2000) due
to technical reasons; the pressures of steam extractions are not completely continuous
but are constrained by the turbine design; the secondary turbine which supplies only
one steam extractions is less likely to be implemented in real power plants; the cost
functions of all components should be developed with more available industrial data
or with the participation of industrial partners. Only when well-established objectives
are optimized under reasonable real-world constraints could the number, size and
operation mode of each component be well-constrained. The obtained optimal or near-
optimal solutions eventually can be valid for industrial applications.
In this appendix, the thermodynamic and economic models of the components used
for the simulations (chapter 3 and 4) are listed, which may need to be reformulated for
mathematical optimization (chapter 5). The cost functions used in (Uche et al., 2001;
Ameri et al., 2009; Xiong et al., 2012a) are re-regressed to reflect recently-published
economic data of coal-fired power plants (China Power Engineering Consulting Group
Corporation, 2009; National Energy Technology Laboratory, 2010).
steam turbine A steam turbine has one incoming steam (stm) pin, one incoming
work pin, one outgoing steam pin, and one outgoing work pin. The model is used
for both the simulation and optimization. The non-ideal behavior is modeled by the
isentropic efficiency:
m stm stm
in = m out ,
(m h)stm
in = ( m h)stm
out + PST ,
hstm hstm
out
s = in stm ,
hstm
in h out
pstm stm
out 6 0.9 pin ,
pstm
out 6 200 bar,
PEC = 980.088 Ft F PST
0.7
,
stm stm
t tref
Ft = 1 + 5 exp in ,
10.42
1 s,ref 3
F = 1 + ,
1 s
where all symbols are as follows: mmass flow rate (kg/s), henthalpy (kJ/kg), P
power generated (for turbine) or consumed (for pump) (kW), s isentropic efficiency
(), h out enthalpy after isentropic expansion (for turbine) or compression (for pump)
(kJ/kg), ppressure (bar), PECpurchased equipment cost ($), Ft temperature-related
factor (), F efficiency-related factor (), ttemperature (C), tref reference tempera-
ture (866 C) for temperature tin , s,ref reference isentropic efficiency (95%). The same
meaning of the symbols are used in the following.
119
120 thermodynamic models and cost functions of the components
water pump A water pump has one incoming water pin, one incoming work pin,
and one outgoing water (wat) pin. The model is used for both simulation and opti-
mization. The non-ideal behavior is modeled by the isentropic efficiency.
m wat wat
in = m out ,
(m h)wat
in + PFP = ( m h)wat
out ,
s h wat hwat
= out in
wat ,
100 hwat
out h in
pwat wat
out > 1.5 pin ,
0
hwat wat
if pwat wat
out < h pout , out < pcr ,
!
1 s,ref 3
PEC = 578 PFP
0.71
1+ ,
1 s
where term h0 is the enthalpy of saturated water (kJ/kg), while pressure pcr is the
critical pressure of water (221 bar). The values of the reference efficiency s,ref is 80.8%.
air fan (compressor) A air fan has one incoming air pin, one incoming power
pin, and one outgoing air pin. The model is used only in the simulation, and is similar
to that of the water pump:
m air air
in = m out ,
(m h)air
in + PAF = ( m h)air
out ,
s h air air
out hin
= air .
100 hout hair
in
feedwater preheater The feedwater preheater specifically stands for the surface
heater with heat source mainly from a steam extraction. A feedwater preheater has
one incoming steam pin (hot side), one standalone incoming drainage pin (hot side),
one outgoing drainage pin (hot side), one incoming feedwater pin (cold side), and
one outgoing feedwater pin (cold side). The model is used for both simulation and
optimization. The models can be switched by setting different temperature differences,
i. e., temperature difference at the pinch point tpin , or temperature difference between
the saturation temperature of steam and the cold outlet tup :
m hot,stm
in + m hot,wat
in = m hot,wat
out ,
m cold,wat
in = m cold,wat
out ,
(m h)hot,stm
in + (m h)hot,wat
in = (m h)hot,wat
out + Q FWH ,
(m h)cold,wat = (m h)cold,wat Q FWH ,
in out
m stm
in h stm
in h hot
pin = m cold
in h cold
out h cold
pin ,
0
hcold
out < h pcold
out ,
hot,stm
phot
out = pin phot ,
out = pin pcold ,
pcold cold
phot
in 6 200 bar,
0.1
1 0.08 0.04
PEC = 100 0.02 3.3 Q FWH pcold phot ,
tup + 4
where the superscripts and subscripts represent as follows: hothot side, coldcold
side, pinpinch point. The term Q is the heat transfered (kW).
m hot,stm
in + m hot,wat
in = m hot,wat
out ,
m cold,cw
in = m cold,cw
out ,
(m h)hot,stm
in + (m h)hot,wat
in = (m h)hot,wat
out + Q COND ,
(m h)cold,cw = (m h)cold,cw Q COND ,
in out
thot
in = tpin ,
cold
tout
thot hot
out = tsat pout ,
0
hhot
out = h phot
out ,
hot,stm
phot
out = pin phot ,
out = pin pcold ,
pcold cold
46.48 Q COND
PEC = + 123.5 m cw
2200 tlog
8.094 Q COND log tcw cw
a ta,ref + 25.55 QCOND ,
hcw cw
out hin
tcw
a = cw cw ,
sout sin
where new terms are expressed as follows: tlog logarithmic temperature difference
(C), cwcooling water, ta average temperature (C), ta,ref reference average tempera-
ture difference (15 C). In addition, in the optimization, the cold side is not considered,
the condenser is simplified as a heat-extraction component; thus, the terms related to
cooling water in the cost function are also neglected.
counter-current, can be flue gas-water heat exchanger, flue gas-steam heat exchanger,
flue gas-air heat exchanger and so on. It is modeled as follows:
m hot hot
in = m out ,
m cold cold
in = m out ,
(m h)hot
in = ( m h)hot
out + QHX ,
(m h)cold = (m h)cold Q HX ,
in out
tin tin = tinlet ,
hot cold
(co-current)
out tout = toutlet ,
thot cold
(co-current)
in tout = tup ,
thot cold
(counter-current)
out tin = tlow ,
thot cold
(counter-current)
out = pin + phot ,
phot hot
where the temperature differences tinlet and toutlet (C) are temperature differences
between the hot and cold streams at the inlet and outlet of the heat exchanger.
boiler A pulverized-coal fired boiler includes the combustion process, steam gen-
erator, superheaters and reheaters. A boiler has one incoming coal pin, one incoming
air pin, one outgoing slag pin, a pair of incoming feedwater pin and outgoing steam
pin for evaporation and superheating, and pairs of incoming steam pin and outgo-
ing steam pin for reheatings. This model is used in the simplified simulation and the
optimization. It can be modeled as follows:
slag fg
m coal air
in + m out + m out ,
in = m
slag fg
m coal h)coal
in LHV + ( m h)air
in + ( m h)out (m h)out = Q abs ,
in ( m
m ev sh
in = m out ,
rhi
Nout
mrhi
in = m rhi
out,k ,
k =1
Nrh
(m h)out (m h)in + (m h)out (m h)in = Q abs ,
sh
ev
rhi
rhi
in = pout + pevsh ,
pev sh
in = pout + prhi ,
prhi rhi
1 ref 7
F = 1.0 + ,
1
sh
tout tref
Ft = 1.0 + 5exp ,
75
0.0014110546 psh
out
Fp = exp ,
10
thermodynamic models and cost functions of the components 123
!
tsh tev Nrh m rhi rhi
out tout tin
rhi
Fshrh = 1.0 + out sh in + ,
tout i m sh
out trhi
out
where the meanings of the symbols are as follows: fgflue gas, Q abs absorbed heat
(kW), shsuperheater or superheated steam (also ms), evevaporator or boiling water,
rhreheater or reheated steam, evshevaporator and superheater, Fm mass flow rate
related factor (), Fp pressure-related factor (), Fshrh reheat-related factor (). The
values of the reference terms, ref , pref and tref , are 95%, 500 bar and 850 C.
de-areator A de-areator has one incoming steam pin, one standalone drainage
pin, one incoming feedwater pin, and one outgoing feedwater pin. The model is used
only in the simulation. The de-areator is simply modeled as a mixer with extra con-
straints:
m stm wat
in + m fw
in + m fw
in = m out ,
(m h)stm h)wat
in + ( m h)fw
in + ( m h)fw
in = ( m out ,
mixer A mixer has one incoming pin and one outgoing pin, which are of the same
stream type. The incoming pin allows for more than two streams; while the outgoing
pin only allows for one stream. The mixer can be modeled as follows:
Nin
m in,j = m out ,
j =1
Nin
(m h)in,j = (m h)out ,
j =1
splitter A splitter has one incoming pin and one outgoing pin, which are of the
same stream type. The outgoing pin allows for more than two streams; while the in-
coming pin only allows for one stream. The splitter can be modeled as follows:
Nout
m in = m out,j ,
j =1
Nout
(m h)in = (m h)out,j ,
j =1
pout,j = pin .
electronic generator A electronic generator has one incoming work pin, and
one outgoing electricity pin. The model is used for both simulation and optimization.
The electronic generator is simply modeled:
PEG = Pout = EG Pin ,
124 thermodynamic models and cost functions of the components
This appendix lists the equations for exergy and exergoeconomic evaluation of each
type of components (SPECO (Lazzaretto and Tsatsaronis, 2006)), and presents the spec-
ifications and simulation data under different conditions of the components. In addi-
tion, the pin information of each component has been described in App. A.
E in
stm
E out
stm out W
in + E D,ST
= W
cost balance
C in
stm
+ C in
work
+ Z ST = C out
stm
+ C out
work
F principle
cstm stm
in = cout
water pump
exergy balance (fuel exergy = product exergy + exergy destruction)
PFP = E out
wat
E in
wat
+ E D,FP
cost balance
C in
wat
+ C in
work
+ Z FP = C out
wat
125
126 additional materials for the analysis and evaluation
cost balance
C in
air
+ C in
work
+ Z AF = C out
air
feedwater preheater
cost balance
hot,stm hot,wat cold,fw hot,wat cold,fw
C in + C in + C in + Z FWH = C out + C out
F principle
(m c)hot,stm + (m c)hot,wat
chot,wat
out = in in
m hot,stm
in + m hot,wat
in
cost balance
hot,stm hot,wat hot,fw
C in + C in + C aux,dc
cw
+ Z COND = C out + C dif,dc,COND
C aux,dc
cw
= ccw cw cw
in Eout Ein
F principle
(m c)hot,stm + (m c)hot,wat
chot
out =
in
hot,stm
in
hot,wat
m in + m in
cost balance
C in
hot
+ C in
cold
+ Z HX = C out
hot
+ C out
cold
F principle
chot hot
in = cout
boiler
exergy balance (fuel exergy = product exergy + exergy destruction)
!
Nrh
+ E out
slag fg
E in
coal
+ E in
air
E out E out = E out
sh
E in
ev rhi
E in
rhi
+ E D,B
j =1
cost balance
Nrh Nrh
in + Cin + Cin + Cin + ZB = Cout + Cout + Cout + Cout
C coal air ev rhi fg slag sh rhi
j =1 j =1
F principle
fg (m c)coal
in + ( m c)air
in
cout =
m coal
in +
m air
in
slag fg
cout = cout
P principle
sh ev rhi rhi
E c out
E c in
E c out
E c in
=
E out
sh E evin E out
rhi E in
rhi
de-areator
exergy balance (fuel exergy = product exergy + exergy destruction)
fw fw fw
E in
stm
E in
wat
m stm
in +
m wat
in e out =
m in e out
E fw
in + E D,DA
cost balance
C in
stm
+ C in
wat
+ C in
fw
+ Z DA = C out
fw
mixer
cost balance
Nin
C in,j + Z MR = C out
j =1
splitter
cost balance
Nout
C in + Z SR = C out,j
j =1
cout,i = cout,j
electronic generator
E in = E out + E D,EG
cost balance
C in + Z EG = C out
Table B.2: The stream data of the design condition (100% load)
Table B.3: Weight estimation and weighted factors for heat surfaces in boiler
Table B.4: The module factors for different types of components (Turton et al., 2008)
Table B.5: Conventional exergy and exergoeconomic analysis of the real case (Table B.2)
FUR 1629 714 915 37.4 43.8 0.41 0.98 371 37.0 4.08 141 9.06
SSH2 110 89.4 20.2 0.83 81.6 0.39 0.70 7.87 19.7 0.28 79.1 71.4
FSH 98.7 84.2 14.5 0.59 85.3 0.39 0.77 5.65 26.6 0.32 98.3 82.5
FRHC 93.5 78.6 14.9 0.61 84.1 0.39 0.78 5.79 24.6 0.30 99.2 80.9
FRHH 28.2 22.6 5.64 0.23 80.0 0.39 0.97 2.20 10.9 0.13 148 83.2
PRH2 32.8 24.8 8.02 0.33 75.5 0.39 0.75 3.12 5.92 0.09 93.8 65.5
PSH 117 103 13.6 0.56 88.4 0.39 0.78 5.29 35.3 0.41 101 87.0
ECO1 60.4 51.5 8.91 0.36 85.2 0.39 0.96 3.47 26.0 0.29 147 88.2
PRH1 127 109 18.4 0.75 85.5 0.39 0.74 7.18 31.5 0.39 91.3 81.4
ECO2 29.4 26.4 3.01 0.12 89.8 0.39 1.69 1.17 33.2 0.34 335 96.6
APH 119 85.8 33.2 1.36 72.1 0.39 0.60 13.0 5.23 0.18 54.4 28.8
PAF 3.28 2.80 0.48 0.02 85.3 1.17 2.49 0.56 3.14 0.04 113 84.7
SAF 2.91 2.48 0.43 0.02 85.1 1.17 2.48 0.51 2.75 0.03 113 84.4
IDF 10.8 9.35 1.44 0.06 86.7 1.17 1.84 1.68 4.64 0.06 57.9 73.4
HPT1 278 266 12.2 0.50 95.6 1.00 1.10 12.2 14.6 0.27 10.1 54.6
HPT2 54.6 52.6 2.03 0.08 96.3 1.00 1.09 2.02 2.89 0.05 9.36 58.8
IPT1 189 180 8.40 0.34 95.6 0.95 1.08 8.00 14.9 0.23 13.3 65.0
IPT2 139 133 5.97 0.24 95.7 0.95 1.08 5.69 11.0 0.17 13.1 65.9
IPT3 90.6 89.6 0.98 0.04 98.9 0.95 1.05 0.93 7.39 0.08 9.75 88.8
LPT1 52.8 47.2 5.58 0.23 89.4 0.95 1.18 5.32 5.19 0.11 23.4 49.4
LPT2 60.1 57.1 2.98 0.12 95.0 0.95 1.11 2.84 6.28 0.09 16.8 68.8
LPT3 56.2 44.2 11.9 0.49 78.8 0.95 1.32 11.4 4.87 0.16 38.5 30.0
LPT4 54.5 51.5 3.08 0.13 94.3 0.95 1.12 2.94 5.66 0.09 17.5 65.8
LPT5 91.9 84.5 7.43 0.30 91.9 0.95 1.15 7.08 9.29 0.16 20.3 56.7
LPT6 25.5 17.0 8.54 0.35 66.5 0.95 1.54 8.14 1.86 0.10 61.9 18.6
COND - - 75.0 3.07 - 0.95 - 71.5 32.4 1.04 - 31.2
FWH1 1.14 0.75 0.39 0.02 65.5 0.95 1.49 0.38 0.03 0.00 56.2 6.44
FWH2 10.4 8.42 1.97 0.08 81.0 0.95 1.19 1.88 0.13 0.02 25.1 6.63
FWH3 11.6 10.2 1.44 0.06 87.6 0.95 1.10 1.37 0.11 0.01 15.3 7.51
FWH4 28.2 24.3 3.90 0.16 86.2 0.95 1.11 3.72 0.12 0.04 16.6 3.11
FWH5 32.0 28.8 3.18 0.13 90.1 0.95 1.06 3.03 0.13 0.03 11.5 4.00
DA 29.8 27.2 2.67 0.11 91.0 0.96 1.12 2.56 1.96 0.05 17.4 43.4
FWH6 50.7 46.4 4.22 0.17 91.7 0.96 1.14 4.05 4.33 0.08 18.8 51.7
FWH7 107 101 6.12 0.25 94.3 1.00 1.11 6.11 5.67 0.12 11.7 48.1
FWH8 42.0 40.8 1.24 0.05 97.0 1.00 1.15 1.24 5.04 0.06 15.4 80.3
OT 41.5 34.4 7.18 0.29 82.7 0.95 1.43 6.85 9.63 0.16 50.3 58.4
FP 34.3 30.6 3.77 0.15 89.0 1.43 1.96 5.41 10.7 0.16 36.8 66.4
CP 1.05 0.90 0.15 0.01 85.5 1.17 2.68 0.18 1.18 0.01 129 86.9
CWP 8.57 6.72 1.84 0.08 78.5 1.17 2.75 2.15 8.46 0.11 135 79.7
EG 1023 1013 10.7 0.44 99.0 1.11 1.17 12.0 42.5 0.54 4.8 78.0
Total 2448 986 1256 51.3 40.3 0.39 1.44 618 473 6.18 272 76.5
E L = 205.5 MW, exergy lost due to the exhausted flue gas (96.97 MW), slag (6.65 MW) and cooling water
(101.88 MW).
The total exergy destructions also include those caused by the control valve and pipes, which are not listed
separately in the table.
B.5 results of conventional and advanced analyses 137
FUR 709 915 694 714 201 904 11.2 710 194 4.19 7.05
SSH2 88.5 20.2 15.3 14.6 5.64 11.8 8.40 8.60 3.21 5.97 2.43
FSH 83.6 14.5 11.0 9.89 4.61 8.74 5.76 6.00 2.74 3.89 1.87
FRHC 79.4 14.9 11.3 11.3 3.55 7.30 7.58 5.50 1.79 5.82 1.75
FRHH 23.8 5.64 4.28 4.35 1.29 3.22 2.42 2.36 0.86 1.99 0.43
PRH2 26.1 8.02 6.09 6.13 1.89 3.84 4.19 2.78 1.06 3.35 0.83
PSH 102 13.6 10.3 8.89 4.71 9.62 3.97 6.38 3.24 2.51 1.47
ECO1 51.5 8.91 6.76 6.89 2.02 6.23 2.67 4.82 1.41 2.07 0.60
PRH1 110 18.4 14.0 13.9 4.50 13.7 4.73 10.2 3.46 3.69 1.04
ECO2 26.4 3.01 2.28 2.32 0.69 2.12 0.89 1.63 0.48 0.69 0.20
APH 92.9 33.2 15.0 15.0 18.2 24.0 9.23 10.0 14.0 5.01 4.22
PAF 2.78 0.48 0.48 0.32 0.16 0.32 0.16
SAF 2.47 0.43 0.43 0.31 0.13 0.31 0.13
IDF 9.53 1.44 1.44 1.13 0.31 1.13 0.31
HPT1 266 12.2 9.34 9.24 2.95 11.0 1.23 8.30 2.65 0.94 0.30
HPT2 52.6 2.03 1.54 1.48 0.55 1.74 0.29 1.27 0.47 0.21 0.08
IPT1 180 8.40 6.39 4.10 4.30 4.48 3.92 2.18 2.29 1.91 2.01
IPT2 133 5.97 4.54 2.89 3.08 3.16 2.81 1.53 1.63 1.36 1.45
IPT3 89.6 0.98 0.74 0.49 0.49 0.53 0.45 0.27 0.27 0.22 0.22
LPT1 47.2 5.58 4.25 3.08 2.49 0.93 4.65 0.51 0.42 2.57 2.08
LPT2 57.1 2.98 2.27 1.77 1.21 2.81 0.17 1.67 1.14 0.10 0.07
LPT3 44.2 11.9 9.13 7.51 4.41 8.29 3.64 5.22 3.06 2.29 1.35
LPT4 51.5 3.08 2.34 1.68 1.40 1.68 1.40 0.92 0.76 0.76 0.64
LPT5 84.5 7.43 5.67 4.57 2.87 6.80 0.63 4.18 2.62 0.39 0.24
LPT6 17.0 8.54 6.53 5.37 3.18 1.98 6.57 1.24 0.74 4.13 2.44
COND 75.0 69.9 30.2 44.9
FWH1 0.76 0.39 0.30 0.32 0.07 0.35 0.04 0.28 0.07 0.04 0.00
FWH2 8.43 1.97 1.50 1.50 0.47 1.73 0.24 1.31 0.42 0.19 0.06
FWH3 10.2 1.44 1.09 1.09 0.35 1.25 0.19 0.94 0.30 0.15 0.05
FWH4 24.3 3.90 3.00 2.95 0.95 3.60 0.30 2.72 0.88 0.23 0.07
FWH5 28.8 3.18 2.42 2.41 0.77 2.90 0.28 2.19 0.71 0.22 0.07
DA 27.2 2.67 2.03 2.07 0.60 2.53 0.14 1.97 0.57 0.11 0.03
FWH6 46.5 4.22 3.21 3.21 1.01 3.87 0.35 2.94 0.93 0.27 0.08
FWH7 101 6.12 4.67 4.67 1.46 5.66 0.47 4.31 1.34 0.35 0.11
FWH8 40.8 1.24 0.94 0.95 0.29 1.11 0.13 0.85 0.26 0.10 0.03
OT 37.4 7.18 5.53 4.26 2.92 4.53 2.65 2.47 2.06 1.79 0.86
FP 29.3 3.77 2.87 2.35 1.42 1.88 1.89 1.22 0.66 1.13 0.76
CP 0.89 0.15 0.12 0.08 0.07 0.07 0.08 0.04 0.03 0.04 0.04
CWP 6.65 1.84 1.84 0.71 1.13 0.71 1.13
EG 1016 10.7 10.5 10.5 0.29 7.16 3.59 6.95 0.21 3.51 0.08
Intensive parameters of streams into and out of the considered component are kept the same as in the real
case.
Intensive parameters of streams into and out of the considered component are recalculated (different from
those in the real case) after removing pressure drops in all theoretically-operated components and considering
the coupling with the preceding turbine (if exists). The subsequent splitting of exergy destruction is based on
this endogenous exergy destruction.
B.5 results of conventional and advanced analyses 139
Name E D
EN Z Z EN Z EX Z UN Z AV Z UN,EN Z UN,EX Z AV,EN Z AV,EX
MW /s /s /s /s /s /s /s /s /s
FUR 714 37.0 28.9 8.13 33.1 3.91 25.6 7.43 3.22 0.69
SSH2 14.6 19.7 14.2 5.48 17.6 2.10 12.5 5.03 1.64 0.46
FSH 9.89 26.6 18.1 8.45 23.7 2.89 16.1 7.64 2.08 0.81
FRHC 11.3 24.6 18.7 5.86 21.8 2.81 16.7 5.03 1.98 0.83
FRHH 4.35 10.9 8.39 2.48 9.17 1.70 7.44 1.72 0.95 0.76
PRH2 6.13 5.92 4.52 1.40 4.93 0.99 3.98 0.95 0.55 0.44
PSH 8.89 35.3 23.1 12.2 31.7 3.61 20.4 11.3 2.66 0.95
ECO1 6.89 26.0 20.1 5.87 23.3 2.70 18.0 5.26 2.09 0.61
PRH1 13.9 31.5 23.8 7.68 27.8 3.71 21.2 6.55 2.57 1.14
ECO2 2.32 33.2 25.6 7.56 29.7 3.45 23.0 6.77 2.66 0.79
APH 15.0 5.23 2.36 2.87 4.32 0.91 2.12 2.21 0.25 0.66
PAF 3.14 3.14 2.79 0.35 2.79 0.35
SAF 2.75 2.75 2.44 0.31 2.44 0.31
IDF 4.64 4.64 4.07 0.57 4.07 0.57
HPT1 9.24 14.6 11.1 3.55 13.1 1.52 9.94 3.18 1.15 0.37
HPT2 1.48 2.89 2.11 0.78 2.59 0.30 1.89 0.70 0.22 0.08
IPT1 4.10 14.9 7.26 7.62 13.3 1.54 6.51 6.83 0.75 0.79
IPT2 2.89 11.0 5.32 5.66 9.85 1.13 4.77 5.08 0.55 0.59
IPT3 0.49 7.39 3.69 3.70 6.62 0.77 3.30 3.32 0.38 0.38
LPT1 3.08 5.19 2.87 2.32 4.65 0.54 2.57 2.08 0.30 0.24
LPT2 1.77 6.28 3.73 2.55 5.63 0.65 3.34 2.29 0.39 0.26
LPT3 7.51 4.87 3.07 1.80 4.36 0.51 2.75 1.61 0.32 0.19
LPT4 1.68 5.66 3.09 2.57 5.07 0.59 2.77 2.30 0.32 0.27
LPT5 4.57 9.29 5.71 3.58 8.33 0.96 5.12 3.21 0.59 0.37
LPT6 5.37 1.86 1.17 0.69 1.67 0.19 1.05 0.62 0.12 0.07
COND 30.2 32.4
FWH1 0.32 0.03 0.02 0.01 0.02 0.01 0.02 0.00 0.01 0.00
FWH2 1.50 0.13 0.10 0.03 0.12 0.01 0.09 0.03 0.01 0.00
FWH3 1.09 0.11 0.08 0.03 0.10 0.01 0.08 0.02 0.01 0.00
FWH4 2.95 0.12 0.09 0.03 0.11 0.01 0.08 0.03 0.01 0.00
FWH5 2.41 0.13 0.10 0.03 0.11 0.02 0.09 0.03 0.01 0.00
DA 2.07 1.96 1.52 0.44 1.76 0.20 1.37 0.39 0.16 0.04
FWH6 3.21 4.33 3.29 1.04 3.88 0.45 2.95 0.93 0.34 0.11
FWH7 4.67 5.67 4.33 1.34 5.08 0.59 3.88 1.20 0.45 0.14
FWH8 0.95 5.04 3.86 1.18 4.51 0.53 3.46 1.05 0.40 0.12
OT 4.26 9.63 5.71 3.92 7.94 1.69 5.12 2.82 0.59 1.10
FP 2.35 10.7 6.66 4.04 9.36 1.34 5.58 3.77 1.08 0.27
Name E D
EN Z Z EN Z EX Z UN Z AV Z UN,EN Z UN,EX Z AV,EN Z AV,EX
MW /s /s /s /s /s /s /s /s /s
CP 0.08 1.18 0.62 0.56 1.00 0.18 0.52 0.48 0.10 0.08
CWP 8.46 8.46 6.89 1.57 6.89 1.57
EG 10.5 42.5 41.4 1.14 37.9 4.62 37.0 0.91 4.38 0.24
For splitting the investment costs, the endogenous exergy destructions used (column 4 in Table B.6) are
calculated after removing the pressure drops in all theoretically-operated components and coupling the
considered component with the preceding turbine (if exists).
B.5 results of conventional and advanced analyses 141
Table B.8: The splitting of exergy destruction costs in advanced exergoeconomic analysis
FUR 371 290 81.6 367 4.56 288 78.7 1.70 2.86
SSH2 7.87 5.67 2.20 4.60 3.27 3.35 1.25 2.32 0.95
FSH 5.65 3.85 1.79 3.40 2.24 2.34 1.07 1.51 0.73
FRHC 5.79 4.41 1.38 2.84 2.95 2.14 0.70 2.27 0.68
FRHH 2.20 1.70 0.50 1.26 0.94 0.92 0.33 0.78 0.17
PRH2 3.12 2.39 0.74 1.49 1.63 1.08 0.41 1.31 0.32
PSH 5.30 3.46 1.83 3.75 1.55 2.49 1.26 0.98 0.57
ECO1 3.47 2.68 0.79 2.43 1.04 1.88 0.55 0.81 0.24
PRH1 7.18 5.43 1.75 5.34 1.84 3.99 1.35 1.44 0.41
ECO2 1.17 0.91 0.27 0.82 0.35 0.64 0.19 0.27 0.08
APH 12.9 5.85 7.10 9.35 3.60 3.90 5.45 1.95 1.65
PAF 0.57 0.57 0.38 0.19 0.38 0.19
SAF 0.51 0.51 0.36 0.15 0.36 0.15
IDF 1.68 1.68 1.32 0.36 1.32 0.36
HPT1 12.2 9.24 2.95 11.0 1.23 8.30 2.65 0.94 0.30
HPT2 2.03 1.48 0.55 1.74 0.29 1.27 0.47 0.21 0.08
IPT1 8.00 3.90 4.09 4.26 3.73 2.08 2.18 1.82 1.91
IPT2 5.69 2.75 2.93 3.01 2.67 1.46 1.55 1.30 1.38
IPT3 0.93 0.46 0.47 0.51 0.42 0.25 0.25 0.21 0.21
LPT1 5.31 2.94 2.38 0.88 4.43 0.49 0.40 2.45 1.98
LPT2 2.84 1.69 1.16 2.68 0.16 1.59 1.09 0.10 0.07
LPT3 11.4 7.16 4.20 7.89 3.47 4.97 2.92 2.18 1.28
LPT4 2.94 1.60 1.33 1.60 1.33 0.88 0.73 0.73 0.61
LPT5 7.08 4.35 2.73 6.48 0.60 3.98 2.50 0.37 0.23
LPT6 8.14 5.11 3.02 1.88 6.25 1.18 0.70 3.93 2.32
COND 71.5 28.7 42.7
FWH1 0.37 0.30 0.07 0.33 0.04 0.27 0.07 0.04 0.00
FWH2 1.88 1.43 0.45 1.65 0.23 1.25 0.40 0.18 0.05
FWH3 1.37 1.04 0.33 1.19 0.19 0.90 0.29 0.14 0.04
FWH4 3.71 2.81 0.91 3.43 0.29 2.59 0.84 0.22 0.07
FWH5 3.03 2.29 0.74 2.76 0.27 2.09 0.67 0.21 0.06
DA 2.56 1.99 0.57 2.43 0.13 1.89 0.54 0.10 0.03
FWH6 4.05 3.08 0.97 3.72 0.33 2.83 0.89 0.26 0.08
FWH7 6.12 4.67 1.46 5.66 0.47 4.31 1.34 0.35 0.11
FWH8 1.24 0.95 0.29 1.11 0.13 0.85 0.26 0.10 0.03
OT 6.82 4.05 2.78 4.30 2.52 2.35 1.96 1.70 0.82
FP 5.39 3.36 2.04 2.69 2.70 1.75 0.94 1.61 1.09
This appendix presents additional materials for the cases presented in chapter 5.
In the specified hierarchy, the dashed rule in the meta level is not employed for muta-
tion, while the dashed technologies in the technology level are dispensable.
Technology level
Figure C.1: Specified energy conversion hierarchy for the illustrative study
143
144 additional materials for the superstructure-free synthesis
The flowsheets of the initial (Fig. C.2a) and optimal (Fig. C.2b) solutions identified in
all optimization runs are presented with their temperature-entropy diagrams (Fig. C.3).
a)
heat
condenser
steam pump
generator
b) reheater
steam turbine
condenser
feedwater
steam preheater
generator pump
Figure C.2: The flowsheets of the initial (a) and optimal (b) solutions identified in the illustra-
tive study
a) 950 b) 950
850 850
750 750
Temperature / K
Temperature / K
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
Figure C.3: Temperature-entropy diagrams of the initial (a) and optimal (b) solutions identified
in the illustrative study
C.1 illustrative study for evaluating the superstructure-free approach 145
a) 500 b) 500
50 50
Pressure / bar
Pressure / bar
5 5
0.5 0.5
0.05 0.05
0.005 0.005
0 1000 2000 3000 4000 0 1000 2000 3000 4000
Enthalpy / kJ/kg Enthalpy / kJ/kg
c) 500 d) 500
50 50
Pressure / bar
Pressure / bar
5 5
0.5 0.5
0.05 0.05
0.005 0.005
0 1000 2000 3000 4000 0 1000 2000 3000 4000
Enthalpy / kJ/kg Enthalpy / kJ/kg
Figure C.4: Some intermediate solutions identified in the representative run of the illustrative
study (Fig. 5.7a): a) the 29th generation ( = 48.81%), b) the 37th generation ( =
49.04%), c) the 40th generation ( = 49.26%), d) the 52th generation ( = 49.44%)
146 additional materials for the superstructure-free synthesis
The specified energy conversion hierarchy (Fig. C.5) considers a new technology, de-
superheater (a surface heat exchanger, see App. A), compared with that of the illustra-
tive study (Fig. C.1). Similarly, the dashed rule in the meta level is not employed for
mutation, while the dashed technologies in the technology level are dispensable.
Figure C.5: Specified energy conversion hierarchy for synthesizing complex PCPPs
C.2 optimal synthesis of complex coal-fired power plants 147
a)
steam turbine
condenser
steam feedwater
generator preheater
pump
b) reheater
steam
turbine condenser
steam
generator desuperheater
feedwater
preheater pump
Figure C.6: The flowsheets of the initial (a) and optimal (b) solutions identified for maximizing
the thermal efficiency
a) 1050 b) 1050
950 950
850 850
Temperature / K
Temperature / K
750 750
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
Figure C.7: Temperature-entropy diagrams of the initial (a) and optimal (b) solutions in Fig. C.6
148 additional materials for the superstructure-free synthesis
a) 1050 b) 1050
950 950
850 850
Temperature / K
Temperature / K
750 750
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
c) 1050 d) 1050
950 950
850 850
Temperature / K
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
Figure C.8: Some intermediate solutions of the representative run (Fig. 5.9) for maximizing the
thermal efficiency: a) 3th generation ( = 47.45%), b) 9th generation ( = 48.30%),
c) 20th generation ( = 48.94%), d) 55th generation ( = 49.27%)
C.2 optimal synthesis of complex coal-fired power plants 149
Table C.1: Computational performance and optimality gaps of 6 runs for maximizing the ther-
mal efficiency: tot total evaluation time (min), best-known thermal efficiency (%),
time
to find the best objective (min), ngeneration number to find the best objec-
tive (1), eoptimality gap (%)
a)
steam turbine
condenser
steam feedwater
generator preheater
pump
b) reheater
ET1
steam
turbine
steam ET2
generator condenser
desuperheater
feedwater
preheater
Figure C.9: The flowsheets of the initial (a) and optimal (b) solutions identified for minimizing
the cost of electricity
a) 1050 b) 1050
950 950
850 850
Temperature / K
Temperature / K
750 750
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
Figure C.10: Temperature-entropy diagrams of the initial (a) and optimal (b) solutions (Fig. C.9)
C.2 optimal synthesis of complex coal-fired power plants 151
a) 1050 b) 1050
950 950
850 850
Temperature / K
Temperature / K
750 750
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
c) 1050 d) 1050
950 950
850 850
Temperature / K
Temperature / K
750 750
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
Figure C.11: Some intermediate solutions of the representative run (Fig. 5.13) for minimizing
the cost of electricity: a) 23th generation (COE = 5.122 /kWh, = 50.65%), b)
58th generation (COE = 5.093 /kWh, = 51.39%), c) 105th generation (COE =
5.070 /kWh, = 51.51%), d) 141th generation (COE = 5.0633 /kWh, =
51.56%)
152 additional materials for the superstructure-free synthesis
Table C.2: Computational performance and optimality gaps of 6 runs for minimizing the cost
of electricity: tot total evaluation time (min), COEbest-known cost of electricity
(/kWh), time
to find the best objective (min), ngeneration number to find the
best objective (1), eoptimality gap (%)
tot
COE n e
Run
min /kWh min 1 %
a) 1050 b) 1050
950 950
850 850
Temperature / K
Temperature / K
750 750
650 650
550 550
450 450
350 350
250 250
0 2 4 6 8 10 0 2 4 6 8 10
Entropy / kJ/(kgK) Entropy / kJ/(kgK)
Akkalp, E., Aras, H., and Hepbasli, A. (2014a). Advanced exergoeconomic analysis
of an electricity-generating facility that operates with natural gas. Energy Convers.
Manage., 78:452460. (Cited on page 13.)
Akkalp, E., Aras, H., and Hepbasli, A. (2014b). Advanced exergy analysis of an
electricity-generating facility using natural gas. Energy Convers. Manage., 82:146153.
(Cited on page 13.)
Ahadi-Oskui, T., Alperin, H., Nowak, I., Cziesla, F., and Tsatsaronis, G. (2006). A
relaxation-based heuristic for the design of cost-effective energy conversion systems.
Energy, 31(10-11):13461357. (Cited on page 23.)
Ahadi-Oskui, T., Vigerske, S., Nowak, I., and Tsatsaronis, G. (2010). Optimizing the
design of complex energy conversion systems by branch and cut. Comput. Chem.
Eng., 34(8):12261236. (Cited on pages 23 and 96.)
Aljundi, I. H. (2009). Energy and exergy analysis of a steam power plant in Jordan.
Appl. Therm. Eng., 29(2):324328. (Cited on page 14.)
Ameri, M., Ahmadi, P., and Hamidi, A. (2009). Energy, exergy and exergoeconomic
analysis of a steam power plant: A case study. Int. J. Energy Res., 33(5):499512.
(Cited on pages 77 and 119.)
Aras, H., Hepbasli, A., et al. (2014). Advanced exergy analysis of a trigeneration system
with a diesel-gas engine operating in a refrigerator plant building. Energy Build.,
80:26275. (Cited on page 13.)
Armaroli, N. and Balzani, V. (2011). Energy for a sustainable world. Wiley-VCH, Wein-
heim. (Cited on page 1.)
153
154 bibliography
Banerjee, A., Tierney, M. J., and Thorpe, R. N. (2012). Thermoeconomics, cost bene-
fit analysis, and a novel way of dealing with revenue generating dissipative units
applied to candidate decentralised energy systems for indian rural villages. Energy,
43(1):477488. (Cited on page 11.)
Bates, E. D., Mayton, R. D., Ntai, I., and Davis, J. H. (2002). CO2 capture by a task-
specific ionic liquid. J. Am. Chem. Soc., 124(6):926927. (Cited on page 1.)
Ber, J. M. (2007). High efficiency electric power generation: The environmental role.
Prog. Energy Combust. Sci., 33(2):107134. (Cited on page 1.)
Bejan, A., Tsatsaronis, G., and Moran, M. J. (1996). Thermal design and optimization. John
Wiley & Sons, New Jersey, USA. (Cited on pages 8, 14, 38, 73, 74, 75, and 77.)
Bertok, B., Barany, M., and Friedler, F. (2012a). Generating and analyzing mathematical
programming models of conceptual process design by P-graph software. Ind. Eng.
Chem. Res., 52(1):166171. (Cited on pages 23 and 24.)
Bertok, B., Kalauz, K., Sule, Z., and Friedler, F. (2012b). Combinatorial algorithm for
synthesizing redundant structures to increase reliability of supply chains: Applica-
tion to biodiesel supply. Ind. Eng. Chem. Res., 52(1):181186. (Cited on page 24.)
Beume, N., Naujoks, B., and Emmerich, M. (2007). SMS-EMOA: Multiobjective se-
lection based on dominated hypervolume. European Journal of Operational Research,
181(3):16531669. (Cited on pages 30 and 90.)
Biegler, L. T., Grossmann, I. E., Westerberg, A. W., and Kravanja, Z. (1997). Systematic
methods of chemical process design, volume 796. Prentice Hall PTR Upper Saddle River,
NJ. (Cited on page 21.)
Blum, R., Kjr, S., and Bugge, J. (2007). Development of a PF fired high efficiency
power plant AD700. In Proceedings of Ris International Energy Conference on Energy
Solutions for Sustainable Development. (Cited on pages 1 and 7.)
bibliography 155
Bonami, P., Kilin, M., and Linderoth, J. (2012). Algorithms and software for convex
mixed integer nonlinear programs. In Mixed Integer Nonlinear Programming, pages
139. Springer. (Cited on page 17.)
Book, R. V. and Otto, F. (1993). String-rewriting systems. Springer. (Cited on page 23.)
Bugge, J., Kjr, S., and Blum, R. (2006). High-efficiency coal-fired power plants devel-
opment and perspectives. Energy, 31(10-11):14371445. (Cited on page 1.)
Burnard, K. and Bhattacharya, S. (2011). Power generation from coal: Ongoing develop-
ments and outlook. International Energy Agency. (Cited on page 1.)
Chen, F., Duic, N., Manuel Alves, L., and da Graa Carvalho, M. (2007). Renewisland
Renewable energy solutions for islands. Renewable and Sustainable Energy Reviews,
11(8):18881902. (Cited on page 22.)
Chen, K., Parmee, I., and Gane, C. (1997). Dual mutation strategies for mixed-integer
optimisation in power station design. In IEEE International Conference on Evolutionary
Computation, pages 385390. (Cited on pages 24 and 26.)
Chu, S. and Majumdar, A. (2012). Opportunities and challenges for a sustainable en-
ergy future. Nature, 488(7411):294303. (Cited on page 1.)
Cziesla, F., Tsatsaronis, G., and Gao, Z. (2006). Avoidable thermodynamic inefficiencies
and costs in an externally fired combined cycle power plant. Energy, 31(10-11):1472
1489. (Cited on pages 11 and 12.)
156 bibliography
Deb, K. (2001). Multi-objective optimization using evolutionary algorithms, volume 16. John
Wiley & Sons. (Cited on page 28.)
Deb, K., Pratap, A., Agarwal, S., and Meyarivan, T. (2002). A fast and elitist multiobjec-
tive genetic algorithm: NSGA-II. IEEE Trans. Evol. Comput., 6(2):182197. (Cited on
pages 19, 30, and 89.)
Eaves, J., Palmer, F. D., Wallace, J., and Wilson, S. (2014). The value of our existing coal
fleet: An assessment of measures to improve reliability & efficiency while reducing
emissions. Technical report, The National Coal Council. (Cited on pages 1 and 8.)
Emmerich, M., Grtzner, M., and Schtz, M. (2001). Design of graph-based evolution-
ary algorithms: A case study for chemical process networks. Evol. Comput., 9(3):329
354. (Cited on pages 23, 26, and 27.)
Erdem, H. H., Akkaya, A. V., Cetin, B., Dagdas, A., Sevilgen, S. H., Sahin, B., Teke, I.,
Gungor, C., and Atas, S. (2009). Comparative energetic and exergetic performance
analyses for coal-fired thermal power plants in turkey. International Journal of Thermal
Sciences, 48(11):21792186. (Cited on page 14.)
Espatolero, S., Corts, C., and Romeo, L. M. (2010). Optimization of boiler cold-end
and integration with the steam cycle in supercritical units. Appl. Energy, 87(5):1651
1660. (Cited on pages 1 and 8.)
Espatolero, S., Romeo, L. M., and Corts, C. (2014). Efficiency improvement strategies
for the feedwater heaters network designing in supercritical coal-fired power plants.
Appl. Therm. Eng., 73(1):449460. (Cited on pages 1 and 21.)
Fazlollahi, S., Mandel, P., Becker, G., and Marechal, F. (2012). Methods for multi-
objective investment and operating optimization of complex energy systems. Energy,
45(1):1222. (Cited on page 31.)
FICO (2009). XPRESS Optimization Suite: MIP formulations and linearizations. Tech-
nical report, Fair Isaac Corporation. (Cited on pages 20 and 118.)
Fraga, E. S. (1998). The generation and use of partial solutions in process synthesis.
Chem. Eng. Res. Des., 76(1):4554. (Cited on page 23.)
Fraga, E. S. (2009). A rewriting grammar for heat exchanger network structure evolu-
tion with stream splitting. Engineering Optimization, 41(9):813831. (Cited on pages 23
and 26.)
Frangopoulos, C. A., von Spakovsky, M. R., and Sciubba, E. (2002). A brief review of
methods for the design and synthesis optimization of energy systems. International
Journal of Thermodynamics, 5(4):151160. (Cited on pages 21, 22, and 88.)
Friedler, F., Tarjn, K., Huang, Y., and Fan, L. (1992). Graph-theoretic approach to
process synthesis: Axioms and theorems. Chem. Eng. Sci., 47(8):19731988. (Cited on
page 23.)
Friedler, F., Tarjan, K., Huang, Y., and Fan, L. (1993). Graph-theoretic approach to
process synthesis: Polynomial algorithm for maximal structure generation. Comput.
Chem. Eng., 17(9):929942. (Cited on page 24.)
Friedler, F., Varga, J., and Fan, L. (1995). Decision-mapping: A tool for consistent and
complete decisions in process synthesis. Chem. Eng. Sci., 50(11):17551768. (Cited on
pages 2 and 24.)
Gaggioli, R. and El-Sayed, Y. (1989). A critical review of second law costing methodII:
Calculus procedures. ASME J. Energy Resour. Technol., 111(1):815. (Cited on page 9.)
Gaggioli, R. and Reini, M. (2014). Panel I: Connecting 2nd law analysis with economics,
ecology and energy policy. Entropy, 16(7):39033938. (Cited on page 11.)
Graus, W., Voogt, M., and Worrell, E. (2007). International comparison of energy effi-
ciency of fossil power generation. Energy policy, 35(7):39363951. (Cited on page 1.)
Habib, M., Said, S., and Al-Zaharna, I. (1995). Optimization of reheat pressures in
thermal power plants. Energy, 20(6):555565. (Cited on pages 21 and 57.)
Hajabdollahi, F., Hajabdollahi, Z., and Hajabdollahi, H. (2012). Soft computing based
multi-objective optimization of steam cycle power plant using NSGA-II and ANN.
Applied Soft Computing, 12(11):36483655. (Cited on page 21.)
Harkin, T., Hoadley, A., and Hooper, B. (2012). Using multi-objective optimisation
in the design of CO2 capture systems for retrofit to coal power stations. Energy,
41(1):228235. (Cited on page 115.)
Hillermeier, C., Hster, S., Mrker, W., and Sturm, T. F. (2000). Optimization of power
plant design: Stochastic and adaptive solution concepts. In Parmee, I., editor, Evolu-
tionary Design and Manufacture, pages 318. Springer-Verlag London, UK. (Cited on
pages 23, 24, 26, and 116.)
Horlock, J. H., Young, J. B., and Manfrida, G. (1999). Exergy analysis of modern fossil-
fuel power plants. ASME J. Eng. Gas Turb. Power, 122(1):17. (Cited on page 14.)
Hostrup, M., Gani, R., Kravanja, Z., Sorsak, A., and Grossmann, I. (2001). Integration
of thermodynamic insights and minlp optimization for the synthesis, design and
analysis of process flowsheets. Comput. Chem. Eng., 25(1):7383. (Cited on page 23.)
Hua, B., Chen, Q., and Wang, P. (1997). A new exergoeconomic approach for analysis
and optimization of energy systems. Energy, 22(11):10711078. (Cited on page 10.)
160 bibliography
Hua, B., Yin, Q., and Wu, G. (1989). Energy optimization through exergy-economic
evaluation. ASME J. Energy Resour. Technol., 111(3):148153. (Cited on page 10.)
Huang, B., Xu, S., Gao, S., Liu, L., Tao, J., Niu, H., Cai, M., and Cheng, J. (2010). Indus-
trial test and techno-economic analysis of CO2 capture in Huaneng Beijing coal-fired
power station. Appl. Energy, 87(11):33473354. (Cited on page 1.)
Huseman, R. (2005). Advanced (700 C) PF power plant: A clean coal european tech-
nology. In Advanced Material for AD700 Boilers, Cesi Auditorium, Milano. (Cited on
page 6.)
IEA (2014). Key world energy statistics 2014. International Energy Agency. (Cited on
page 1.)
Jaksland, C. A., Gani, R., and Lien, K. M. (1995). Separation process design and syn-
thesis based on thermodynamic insights. Chem. Eng. Sci., 50(3):511530. (Cited on
pages 22 and 23.)
Jamel, M., Abd Rahman, A., and Shamsuddin, A. (2013). Advances in the integra-
tion of solar thermal energy with conventional and non-conventional power plants.
Renewable and Sustainable Energy Reviews, 20:7181. (Cited on pages 1 and 115.)
Jiang, L., Lin, R., Jin, H., Cai, R., and Liu, Z. (2002). Study on thermodynamic char-
acteristic and optimization of steam cycle system in IGCC. Energy Convers. Manage.,
43(9):13391348. (Cited on page 23.)
Jdes, M. (2009). MINLP optimization of design and steady-state operation of power plants
considering several operating points. PhD thesis, Technical University of Berlin, Berlin,
Germany. (in German). (Cited on pages 20 and 96.)
Kalinci, Y., Hepbasli, A., and Dincer, I. (2011). Exergoeconomic analysis of hydrogen
production from plasma gasification of sewage sludge using specific exergy cost
method. Int. J. Hydrogen Energy, 36(17):1140811417. (Cited on page 10.)
Karthikeyan, M., Zhonghua, W., and Mujumdar, A. S. (2009). Low-rank coal drying
technologiesCurrent status and new developments. Drying Technology, 27(3):403
415. (Cited on page 8.)
Kelly, S. (2008). Energy Systems Improvement based on Endogenous and Exogenous Exergy
Destruction. PhD thesis, Technische Universitt Berlin. (Cited on pages 11, 12, 13,
and 14.)
Kelly, S., Tsatsaronis, G., and Morosuk, T. (2009). Advanced exergetic analysis: Ap-
proaches for splitting the exergy destruction into endogenous and exogenous parts.
Energy, 34(3):384391. (Cited on pages 11, 12, and 13.)
Kirkwood, R., Locke, M. H., and Douglas, J. (1988). A prototype expert system for
synthesizing chemical process flowsheets. Comput. Chem. Eng., 12(4):329343. (Cited
on page 22.)
Kjaer, S. and Drinhaus, F. (2010). A modified double reheat cycle. ASME Conference
Proceedings, 2010(49354):285293. (Cited on page 7.)
Klatt, K.-U. and Marquardt, W. (2009). Perspectives for process systems engineering
Personal views from academia and industry. Comput. Chem. Eng., 33(3):536550.
(Cited on page 2.)
Klegman, J. (2010). Waigaoqiao: World class in clean coal. Technical report, Living
Energy. (Cited on page 5.)
Kocis, G. and Grossmann, I. (1989). A modelling and decomposition strategy for the
minlp optimization of process flowsheets. Comput. Chem. Eng., 13(7):797819. (Cited
on pages 25 and 116.)
Kossiakoff, A., Sweet, W. N., Seymour, S., and Biemer, S. M. (2011). Systems engineering
principles and practice, volume 83. John Wiley & Sons. (Cited on page 2.)
Kotas, T. (1985). The exergy method of thermal plant analysis. Butterworth Publishers,
Stoneham, MA. (Cited on pages 11 and 14.)
Kraemer, T. G., Nelson, G., Boyce, G. H., and Palmer, F. D. (2006). Coal: Americas
energy future. Technical report, The National Coal Council. (Cited on page 1.)
Krmer, K., Kossack, S., and Marquardt, W. (2009). Efficient optimization-based design
of distillation processes for homogenous azeotropic mixtures. Industrial & Engineer-
ing Chemistry Research, 48(14):67496764. (Cited on page 25.)
Lam, H. L., Kleme, J. J., Kravanja, Z., and Varbanov, P. S. (2011). Software tools
overview: Process integration, modelling and optimisation for energy saving and
pollution reduction. Asia-Pacific Journal of Chemical Engineering, 6(5):696712. (Cited
on page 23.)
Lang, Y.-D. and Biegler, L. T. (2002). Distributed stream method for tray optimization.
AIChE J., 48(3):582595. (Cited on page 25.)
Lasdon, L. S., Waren, A. D., Jain, A., and Ratner, M. (1978). Design and testing of a
generalized reduced gradient code for nonlinear programming. ACM Transactions on
Mathematical Software (TOMS), 4(1):3450. (Cited on page 16.)
Lazzaretto, A. and Toffolo, A. (2008). A method to separate the problem of heat transfer
interactions in the synthesis of thermal systems. Energy, 33(2):163170. (Cited on
page 26.)
Lazzaretto, A. and Tsatsaronis, G. (1997). On the quest for objective equations in ex-
ergy costing. Proceedings of the ASME Advanced Energy Systems Division, New York:
American Society of Mechanical Engineers, 37:197210. (Cited on page 9.)
Li, X. and Kraslawski, A. (2004). Conceptual process synthesis: Past and current trends.
Chemical Engineering and Processing: Process Intensification, 43(5):583594. (Cited on
pages 2, 21, and 22.)
Liberti, L. and Pantelides, C. (2006). An exact reformulation algorithm for large non-
convex NLPs involving bilinear terms. Journal of Global Optimization, 36(2):161189.
(Cited on page 17.)
Liu, P., Georgiadis, M. C., and Pistikopoulos, E. N. (2010a). Advances in energy systems
engineering. Ind. Eng. Chem. Res., 50(9):49154926. (Cited on page 23.)
Liu, P., Pistikopoulos, E. N., and Li, Z. (2010b). A multi-objective optimization ap-
proach to polygeneration energy systems design. AIChE journal, 56(5):12181234.
(Cited on page 31.)
bibliography 163
Lozano, M. and Valero, A. (1993). Theory of the exergetic cost. Energy, 18(9):939960.
(Cited on pages 9 and 11.)
Luo, X., Wen, Q.-Y., and Fieg, G. (2009). A hybrid genetic algorithm for synthesis of
heat exchanger networks. Comput. Chem. Eng., 33(6):11691181. (Cited on page 23.)
Luo, X., Zhang, B., Chen, Y., and Mo, S. (2011). Modeling and optimization of a util-
ity system containing multiple extractions steam turbines. Energy, 36(5):35013512.
(Cited on pages 20 and 23.)
Luo, X., Zhang, B., Chen, Y., and Mo, S. (2012). Operational planning optimization of
multiple interconnected steam power plants considering environmental costs. En-
ergy, 37(1):549561. (Cited on page 23.)
Mahalec, V. and Motard, R. (1977). Procedures for the initial design of chemical pro-
cessing systems. Comput. Chem. Eng., 1(1):5768. (Cited on page 22.)
Manassaldi, J. I., Mussati, S. F., and Scenna, N. J. (2011). Optimal synthesis and design
of heat recovery steam generation (HRSG) via mathematical programming. Energy,
36(1):475485. (Cited on pages 20 and 25.)
Manolas, D., Efthimeros, G., and Tsahalis, D. (2001). Development of an expert sys-
tem shell based on genetic algorithms for the selection of the energy best available
technologies and their optimal operating conditions for the process industry. Expert
Systems, 18(3):124130. (Cited on page 22.)
Marler, R. T. and Arora, J. S. (2010). The weighted sum method for multi-objective
optimization: New insights. Structural and multidisciplinary optimization, 41(6):853
862. (Cited on pages 28 and 89.)
Martelli, E., Amaldi, E., and Consonni, S. (2011). Numerical optimization of heat recov-
ery steam cycles: Mathematical model, two-stage algorithm and applications. Com-
put. Chem. Eng., 35(12):27992823. (Cited on page 25.)
Masuyama, F. (2004). Alloy development and material issues with increasing steam
temperature. In Advances in materials technology for fusion power plants-Proceedings
from Fourth International conference. ASM International, Cleveland. (Cited on page 6.)
Matelli, J. A., Bazzo, E., and da Silva, J. C. (2009). An expert system prototype for de-
signing natural gas cogeneration plants. Expert Systems with Applications, 36(4):8375
8384. (Cited on pages 22 and 23.)
164 bibliography
Matsuda, K., Hirochi, Y., Tatsumi, H., and Shire, T. (2009). Applying heat integration to-
tal site based pinch technology to a large industrial area in Japan to further improve
performance of highly efficient process plants. Energy, 34(10):16871692. (Cited on
page 22.)
McCarl, B. A., Meeraus, A., van der Eijk, P., Bussieck, M., Dirkse, S., Steacy, P., and
Nelissen, F. (2014). McCarl GAMS user guide. (Cited on page 97.)
Messac, A., Ismail-Yahaya, A., and Mattson, C. (2003). The normalized normal con-
straint method for generating the Pareto frontier. Structural and Multidisciplinary
Optimization, 25(2):8698. (Cited on pages 28, 29, and 89.)
Metz, B., Davidson, O., De Coninck, H., Loos, M., Meyer, L., et al. (2005). Carbon
dioxide capture and storage. Technical report, Cambridge University Press. (Cited
on page 1.)
Moran, M. J., Shapiro, H. N., Boettner, D. D., and Bailey, M. B. (2010). Fundamentals of
Engineering Thermodynamics, 7th Edition. John Wiley & Sons, New York, USA. (Cited
on pages 8 and 95.)
Morandin, M., Marchal, F., Mercangz, M., and Buchter, F. (2012a). Conceptual design
of a thermo-electrical energy storage system based on heat integration of thermody-
namic cyclePart B: Alternative system configurations. Energy, 45(1):386396. (Cited
on page 115.)
Morandin, M., Marchal, F., Mercangz, M., and Buchter, F. (2012b). Conceptual design
of a thermo-electrical energy storage system based on heat integration of thermody-
namic cyclesPart A: Methodology and base case. Energy, 45(1):375385. (Cited on
page 115.)
Morosuk, T. and Tsatsaronis, G. (2006). The "cycle method" used in the exergy analysis
of refrigeration machines: From education to research. In Proceedings of the 19th inter-
national conference on efficiency, cost, optimization, simulation and environmental impact of
energy systems, volume 1, pages 1214. (Cited on page 13.)
bibliography 165
Morosuk, T. and Tsatsaronis, G. (2008a). How to calculate the parts of exergy destruc-
tion in an advanced exergetic analysis. In Proceedings of the 21st International Confer-
ence on Efficiency, Costs, Optimization, Simulation and Environmental Impact of Energy
Systems, pages 185194. Cracow, Gliwice, Poland, June. (Cited on page 13.)
Morosuk, T. and Tsatsaronis, G. (2009b). Advanced exergy analysis for chemically re-
acting systemsApplication to a simple open gas-turbine system. Int. J. Thermophys.,
12:105111. (Cited on pages 11 and 13.)
National Energy Technology Laboratory (2010). Cost and performance baseline for fos-
sil energy plants volume 1: Bituminous coal and natural gas to electricity. Technical
report, DOE/NETL-2010/1397. (Cited on pages 76, 77, and 119.)
Nesterov, Y., Nemirovskii, A., and Ye, Y. (1994). Interior-point polynomial algorithms in
convex programming, volume 13. SIAM. (Cited on page 16.)
Pachernegg, S. (1969). A closer look at the Willans-line. Technical report, SAE. (Cited
on page 20.)
Paterson, W. (1984). A replacement for the logarithmic mean. Chem. Eng. Sci.,
39(11):16351636. (Cited on page 20.)
Penkuhn, M. (2015). Calculation method for advanced exergy analysis. Personal com-
munication. (Cited on page 40.)
Petrakopoulou, F. (2011). Comparative Evaluation of Power Plants with CO2 Capture: Ther-
modynamic, Economic and Environmental Performance. PhD thesis, Technical University
of Berlin. (Cited on pages 1, 11, and 13.)
166 bibliography
Petrakopoulou, F., Boyano, A., Cabrera, M., and Tsatsaronis, G. (2011a). Exergoeco-
nomic and exergoenvironmental analyses of a combined cycle power plant with
chemical looping technology. International Journal of Greenhouse Gas Control, 5(3):475
482. (Cited on page 13.)
Petrakopoulou, F., Tsatsaronis, G., Morosuk, T., and Carassai, A. (2011b). Advanced
exergoeconomic analysis applied to a complex energy conversion system. ASME J.
Eng. Gas Turb. Power, 134(3):031801031801. (Cited on page 13.)
Petrakopoulou, F., Tsatsaronis, G., Morosuk, T., and Carassai, A. (2012b). Conventional
and advanced exergetic analyses applied to a combined cycle power plant. Energy,
41(1):146152. (Cited on page 13.)
Popov, D. (2011). An option for solar thermal repowering of fossil fuel fired power
plants. Solar Energy, 85(2):344349. (Cited on pages 1 and 8.)
Potter, P. J. (1988). Power plant theory and design. RE Krieger. (Cited on page 21.)
Price, K., Storn, R. M., and Lampinen, J. A. (2006). Differential evolution: A practical
approach to global optimization. Springer. (Cited on pages 62 and 63.)
Ray, T. K., Datta, A., Gupta, A., and Ganguly, R. (2010). Exergy-based performance
analysis for proper O&M decisions in a steam power plant. Energy Convers. Manage.,
51(6):13331344. (Cited on page 14.)
Rukes, B. and Taud, R. (2004). Status and perspectives of fossil power generation.
Energy, 29(1215):18531874. (Cited on page 6.)
Samanta, A., Zhao, A., Shimizu, G. K., Sarkar, P., and Gupta, R. (2011). Post-combustion
CO2 capture using solid sorbents: A review. Ind. Eng. Chem. Res., 51(4):14381463.
(Cited on pages 1 and 115.)
Sanaye, S., Farshi, B., and Turk, H. (2003). Optimum turbine extraction pressures for
maximum efficiency in steam power plant cycle. In International Joint Power Gener-
ation Conference collocated with TurboExpo 2003, pages 195202. American Society of
Mechanical Engineers. (Cited on page 21.)
Sand, G., Till, J., Tometzki, T., Urselmann, M., Engell, S., and Emmerich, M. (2008). En-
gineered versus standard evolutionary algorithms: A case study in batch scheduling
with recourse. Comput. Chem. Eng., 32(11):27062722. (Cited on page 26.)
Savola, T. (2007). Modeling biomass-fuelled small-scale CHP plants for process synthesis and
optimization. PhD thesis, Helsinki University of Technology. (Cited on page 20.)
Sciubba, E. (1998). Toward automatic process simulators: Part IIAn expert system for
process synthesis. ASME J. Eng. Gas Turb. Power, 120(1):916. (Cited on page 22.)
Seferlis, P. and Grievink, J. (2001). Optimal design and sensitivity analysis of reactive
distillation units using collocation models. Ind. Eng. Chem. Res., 40(7):16731685.
(Cited on page 23.)
Sengupta, S., Datta, A., and Duttagupta, S. (2007). Exergy analysis of a coal-based 210
MW thermal power plant. International Journal of Energy Research, 31(1):1428. (Cited
on page 14.)
Seyyedi, S. M., Ajam, H., and Farahat, S. (2010a). A new approach for optimization
of thermal power plant based on the exergoeconomic analysis and structural op-
timization method: Application to the CGAM problem. Energy Convers. Manage.,
51(11):22022211. (Cited on page 10.)
Seyyedi, S. M., Ajam, H., and Farahat, S. (2010b). A new criterion for the allocation of
residues cost in exergoeconomic analysis of energy systems. Energy, 35(8):34743482.
(Cited on page 10.)
Silvestri, G. J., Bannister, R. L., Fujikawa, T., and Hizume, A. (1992). Optimization of
advanced steam condition power plants. ASME J. Eng. Gas Turb. Power, 114(4):612
620. (Cited on pages 21 and 57.)
Singh, O. K. and Kaushik, S. (2013). Energy and exergy analysis and optimization
of Kalina cycle coupled with a coal fired steam power plant. Appl. Therm. Eng.,
51(1):787800. (Cited on page 14.)
Smith, R. M. (2005). Chemical process: Design and integration. John Wiley & Sons. (Cited
on page 37.)
Solomon, S. (2007). Climate change 2007-the physical science basis: Working group I con-
tribution to the fourth assessment report of the IPCC, volume 4. Cambridge University
Press. (Cited on page 1.)
Soltani, S., Yari, M., Mahmoudi, S., Morosuk, T., and Rosen, M. (2013). Advanced
exergy analysis applied to an externally-fired combined-cycle power plant integrated
with a biomass gasification unit. Energy, 59:775780. (Cited on page 13.)
Sorgenfrei, M. (2016). Analysis of IGCC-Based Plants with Carbon Capture for an Effi-
cient and Flexible Electric Power Generation. PhD thesis, Technical University of Berlin.
(Cited on page 14.)
Spliethoff, H. (2010). Steam power stations for electricity and heat generation. In Power
Generation from Solid Fuels, Power Systems, pages 73219. Springer Berlin Heidelberg.
(Cited on pages 6, 57, and 62.)
Stepczy
nska,
K., Kowalczyk, ., Dykas, S., and Elsner, W. (2012). Calculation of a
900 MW conceptual 700/720C coal-fired power unit with an auxiliary extraction-
backpressure turbine. Journal of Power Technologies, 92(4):266273. (Cited on page 7.)
Storn, R. and Price, K. (1997). Differential evolutionA simple and efficient heuristic for
global optimization over continuous spaces. Journal of global optimization, 11(4):341
359. (Cited on pages 19 and 63.)
Suresh, M., Reddy, K., and Kolar, A. K. (2010). 4-e (energy, exergy, environment, and
economic) analysis of solar thermal aided coal-fired power plants. Energy for Sustain-
able Development, 14(4):267279. (Cited on page 14.)
Suresh, M., Reddy, K., and Kolar, A. K. (2011). ANN-GA based optimization of a high
ash coal-fired supercritical power plant. Appl. Energy, 88(12):48674873. (Cited on
page 21.)
Szargut, J., Morris, D. R., and Steward, F. R. (1987). Exergy analysis of thermal, chem-
ical, and metallurgical processes. Hemisphere Publishing, New York, NY. (Cited on
page 45.)
Tchanche, B. F., Lambrinos, G., Frangoudakis, A., and Papadakis, G. (2011). Low-grade
heat conversion into power using organic rankine cyclesA review of various applica-
tions. Renewable and Sustainable Energy Reviews, 15(8):39633979. (Cited on page 115.)
Toffolo, A. (2009). The synthesis of cost optimal heat exchanger networks with uncon-
strained topology. Appl. Therm. Eng., 29(17-18):35183528. (Cited on page 26.)
Tomlin, J. (1988). Special ordered sets and an application to gas supply operations plan-
ning. Mathematical Programming, Springer-Verlag, 42(1-3):6984. (Cited on page 20.)
Torres, C., Valero, A., Rangel, V., and Zaleta, A. (2008). On the cost formation process
of the residues. Energy, 33(2):144152. (Cited on page 10.)
Turton, R., Bailie, R. C., Whiting, W. B., and Shaeiwitz, J. A. (2008). Analysis, synthesis
and design of chemical processes. Pearson Education. (Cited on pages 37, 50, and 134.)
Tveit, T.-M. and Fogelholm, C.-J. (2006). Multi-period steam turbine network optimisa-
tion. Part II: Development of a multi-period MINLP model of a utility system. Appl.
Therm. Eng., 26(14-15):17301736. (Cited on page 96.)
Uche, J., Serra, L., and Valero, A. (2001). Thermoeconomic optimization of a dual-
purpose power and desalination plant. Desalination, 136(1-3):147158. (Cited on
pages 21, 77, and 119.)
Urselmann, M., Barkmann, S., Sand, G., and Engell, S. (2011a). A memetic algorithm
for global optimization in chemical process synthesis problems. IEEE Trans. Evol.
Comput., 15(5):659683. (Cited on page 25.)
Urselmann, M., Barkmann, S., Sand, G., and Engell, S. (2011b). Optimization-based
design of reactive distillation columns using a memetic algorithm. Comput. Chem.
Eng., 35(5):787805. (Cited on page 25.)
Urselmann, M., Emmerich, M. T., Till, J., Sand, G., and Engell, S. (2007). Design of
problem-specific evolutionary algorithm/mixed-integer programming hybrids: Two-
stage stochastic integer programming applied to chemical batch scheduling. Engi-
neering Optimization, 39(5):529549. (Cited on page 26.)
Urselmann, M. and Engell, S. (2015). Design of memetic algorithms for the efficient op-
timization of chemical process synthesis problems with structural restrictions. Com-
put. Chem. Eng., 72(0):87108. (Cited on page 25.)
Valero, A., Lerch, F., Serra, L., and Royo, J. (2002). Structural theory and thermoe-
conomic diagnosis: Part II: Application to an actual power plant. Energy Convers.
Manage., 43(9-12):15191535. (Cited on page 14.)
Valero, A., Lozano, M., and Muoz, M. (1986a). A general theory of exergy saving.
I. on the exergetic cost. Computer-Aided Engineering and Energy Systems. Second Law
Analysis and Modelling, R. Gaggioli, ed, 3:18. (Cited on page 9.)
Valero, A., Lozano, M., and Muoz, M. (1986b). A general theory of exergy saving. II.
on the thermodynamic cost. Computer-Aided Engineering and Energy Systems. Second
Law Analysis and Modelling, R. Gaggioli, ed, 3:915. (Cited on page 9.)
Valero, A., Lozano, M., and Muoz, M. (1986c). A general theory of exergy saving. III.
energy savings and thermodynamics. Computer-Aided Engineering and Energy Systems.
Second Law Analysis and Modelling, R. Gaggioli, ed, 3:1621. (Cited on page 9.)
Valero, A., Lozano, M., Serra, L., and Torres, C. (1994a). Application of the exergetic
cost theory to the CGAM problem. Energy, 19(3):365381. (Cited on page 9.)
172 bibliography
Valero, A., Lozano, M. A., Serra, L., Tsatsaronis, G., Pisa, J., Frangopoulos, C., and von
Spakovsky, M. R. (1994b). CGAM problem: Definition and conventional solution.
Energy, 19(3):279286. (Cited on page 31.)
Valero, A., Serra, L., and Lozano, M. (1993). Structural theory of thermoeconomics.
ASME, NEW YORK, NY,(USA)., 30:189198. (Cited on pages 10 and 23.)
Vance, L., Cabezas, H., Heckl, I., Bertok, B., and Friedler, F. (2012). Synthesis of sustain-
able energy supply chain by the P-graph framework. Ind. Eng. Chem. Res., 52(1):266
274. (Cited on page 24.)
Vieira, L. S., Donatelli, J. L., and Cruz, M. E. (2004). Integration of an iterative method-
ology for exergoeconomic improvement of thermal systems with a process simulator.
Energy Convers. Manage., 45(15-16):24952523. (Cited on page 117.)
Viswanathan, R., Coleman, K., and Rao, U. (2006). Materials for ultra-supercritical
coal-fired power plant boilers. International Journal of Pressure Vessels and Piping,
83(11):778783. (Cited on page 6.)
Voll, P., Jennings, M., Hennen, M., Shah, N., and Bardow, A. (2015). The optimum is
not enough: A near-optimal solution paradigm for energy systems synthesis. Energy,
82(0):446456. (Cited on page 100.)
Voll, P., Klaffke, C., Hennen, M., and Bardow, A. (2013). Automated superstructure-
based synthesis and optimization of distributed energy supply systems. Energy,
50(0):374388. (Cited on pages 2 and 24.)
Voll, P., Lampe, M., Wrobel, G., and Bardow, A. (2012). Superstructure-free synthesis
and optimization of distributed industrial energy supply systems. Energy, 45(1):424
435. (Cited on pages 2, 26, 87, 88, 89, 109, and 112.)
von Spakovsky, M. R. (1986). A practical generalized analysis approach to the optimal ther-
moeconomic design and improvement of real-world thermal systems. PhD thesis, Georgia
Institute of Technology. (Cited on pages 9 and 10.)
von Spakovsky, M. R. and Evans, R. B. (1990). The design and performance optimiza-
tion of thermal systems. ASME J. Eng. Gas Turb. Power, 112(1):8693. (Cited on
pages 9 and 10.)
Vuckovic, G. D., Stojiljkovic, M. M., Vukic, M. V., Stefanovic, G. M., and Dedeic, E. M.
(2014). Advanced exergy analysis and exergoeconomic performance evaluation of
thermal processes in an existing industrial plant. Energy Convers. Manage., 85:655
662. (Cited on page 13.)
Wagner, W., Cooper, J. R., Dittmann, A., Kijima, J., Kretzschmar, H.-J., Kruse, A., Mare,
R., Oguchi, K., Sato, H., Stcker, I., ifner, O., Takaishi, Y., Tanishita, I., Trbenbach, J.,
and Willkommen, T. (2000). The IAPWS industrial formulation 1997 for the thermo-
dynamic properties of water and steam. ASME J. Eng. Gas Turb. Power, 122(1):150184.
(Cited on pages 20 and 95.)
Wang, L., Voll, P., Lampe, M., Yang, Y., and Bardow, A. (2015). Superstructure-free
synthesis and optimization of thermal power plants. Energy, 91:700711. (Cited on
page 87.)
Wang, L., Yang, Y., Dong, C., Morosuk, T., and Tsatsaronis, G. (2014). Systematic op-
timization of the design of steam cycles using MINLP and differential evolution.
ASME J. Energy Resour. Technol., 136(3):031601. (Cited on pages 23 and 25.)
Wang, L., Yang, Y., Dong, C., Yang, Z., Xu, G., and Wu, L. (2012a). Exergoeconomic
evaluation of a modern ultra-supercritical power plant. Energies, 5:33813397. (Cited
on page 74.)
Wang, L., Yang, Y., Morosuk, T., and Tsatsaronis, G. (2012b). Advanced thermody-
namic analysis and evaluation of a supercritical power plant. Energies, 5(6):18501863.
(Cited on page 13.)
Wei, S., editor (2013). China Electric Power Yearbook 2013. China Electric Press. (Cited
on page 1.)
Weitzel, P. S. (2011). Steam generator for advanced ultra supercritical power plants
700C to 760C. In ASME 2011 Power Conference collocated with JSME ICOPE 2011,
pages 281291. American Society of Mechanical Engineers. (Cited on page 6.)
World Bank (2014). World development indicators 2014. World Bank Publications. (Cited
on page 1.)
Wright, J., Zhang, Y., Angelov, P., Hanby, V., and Buswell, R. (2008). Evolutionary syn-
thesis of HVAC system configurations: Algorithm development. HVAC&R Research,
14(1):3355. (Cited on page 26.)
Wright, S. A., Radel, R. F., Vernon, M. E., Rochau, G. E., and Pickard, P. S. (2010).
Operation and analysis of a supercritical CO2 brayton cycle. Technical report, Sandia
Report, No. SAND2010-0171. (Cited on page 115.)
Xiong, J., Zhao, H., Zhang, C., Zheng, C., and Luh, P. B. (2012a). Thermoeconomic
operation optimization of a coal-fired power plant. Energy, 42(1):486496. (Cited on
pages 14, 21, 77, and 119.)
Xiong, J., Zhao, H., and Zheng, C. (2012b). Thermoeconomic cost analysis of a 600MWe
oxy-combustion pulverized-coal-fired power plant. International Journal of Greenhouse
Gas Control, 9(0):469483. (Cited on pages 14 and 76.)
Xu, Y. C., Chen, Q., and Guo, Z. Y. (2015). Entransy dissipation-based constraint for
optimization of heat exchanger networks in thermal systems. Energy, 86(0):696708.
(Cited on page 117.)
Yan, Q., Yang, Y., Nishimura, A., Kouzani, A., and Hu, E. (2010). Multi-point and
multi-level solar integration into a conventional coal-fired power plant. Energy Fuels,
24(7):37333738. (Cited on page 1.)
Yang, H., Xu, Z., Fan, M., Gupta, R., Slimane, R. B., Bland, A. E., and Wright, I. (2008).
Progress in Carbon Dioxide separation and capture: A review. Journal of Environmen-
tal Sciences, 20(1):1427. (Cited on page 1.)
Yang, Y., Wang, L., Dong, C., Xu, G., Morosuk, T., and Tsatsaronis, G. (2013). Com-
prehensive exergy-based evaluation and parametric study of a coal-fired ultra-
supercritical power plant. Appl. Energy, 112(0):10871099. (Cited on pages 13 and 82.)
Yang, Y., Yan, Q., Zhai, R., Kouzani, A., and Hu, E. (2011). An efficient way to use
medium-or-low temperature solar heat for power generationIntegration into con-
ventional power plant. Appl. Therm. Eng., 31(2-3):157162. (Cited on page 8.)
Yee, T., Grossmann, I., and Kravanja, Z. (1990). Simultaneous optimization models
for heat integrationIII. Process and heat exchanger network optimization. Comput.
Chem. Eng., 14(11):11851200. (Cited on page 25.)
bibliography 175
Yergin, D. (2006). Ensuring energy security. Foreign affairs, 85:6982. (Cited on page 1.)
Zhang, C., Chen, S., Zheng, C., and Lou, X. (2007). Thermoeconomic diagnosis of a
coal fired power plant. Energy Convers. Manage., 48(2):405419. (Cited on page 14.)
Zhang, C., Wang, Y., Zheng, C., and Lou, X. (2006). Exergy cost analysis of a coal
fired power plant based on structural theory of thermoeconomics. Energy Convers.
Manage., 47(7):817843. (Cited on page 14.)
Journal papers:
1. Wang, L., et. al. (2016). Advanced exergoeconomic evaluation of a modern large-
scale coal-fired power plant. Applied Energy, to be submitted.
2. Wang, L., Lampe, M., Voll, P., Yang, Y., and Bardow, A. (2016). Multi-objective
superstructure-free synthesis and optimization of thermal power plants. Energy,
under review.
3. Wang, L., Voll, P., Lampe, M., Yang, Y., and Bardow, A. (2015). Superstructure-free
synthesis and optimization of thermal power plants. Energy, 91, 700711.
4. Yang, Y., Wang, L., Dong, C., Xu, G., Morosuk, T., and Tsatsaronis, G. (2013).
Comprehensive exergy-based evaluation and parametric study of a coal-fired
ultra-supercritical power plant. Applied Energy, 112, 10871099.
5. Wang, L., Yang, Y., Dong, C., Morosuk, T., and Tsatsaronis, G. (2014). Systematic
optimal design of steam cycles using MINLP and differential evolution. ASME
Journal of Energy Resources Technology, 136(3), 031601, doi: 10.1115/1.4026268.
6. Wang, L., Yang, Y., Dong, C., Morosuk, T., and Tsatsaronis, G. (2014). Multi-
objective optimization of coal-fired power plants using differential evolution. Ap-
plied Energy, 115, 254264.
7. Wang, L., Yang, Y., Dong, C., Morosuk, T., and Tsatsaronis, G. (2014). Parametric
optimization of supercritical coal-fired power plants by MINLP and differential
evolution. Energy Conversion and Management, 85, 828838.
8. Wang, L., Yang, Y., Morosuk, T., and Tsatsaronis, G. (2012). Advanced thermo-
dynamic analysis and evaluation of a supercritical power plant. Energies, 5(6),
18501863.
9. Wang, L., Yang, Y., Dong, C., Yang, Z., Xu, G., and Wu, L. (2012). Exergoeconomic
evaluation of a modern ultra-supercritical power plant. Energies, 5(9), 33813397.
10. Wang, L., Yang, Y., Dong, C., and Xu, G. (2012). Improvement and primary ap-
plication of theory of Fuel Specific Consumption. Proceedings of the CSEE, 32(11),
1621. (in Chinese)
11. Wang, L., Wu, L., Xu, G., Dong, C., and Yang, Y. (2012). Calculation and analysis
of energy consumption interactions in thermal systems of large-scale coal-fired
steam power generation units. Proceedings of the CSEE, 32(29), 914. (in Chinese)
177
178 bibliography
Conference papers:
1. Wang, L., Lampe, M., Voll, P., Yang, Y., and Bardow, A. (2016). Multi-objective
superstructure-free synthesis and optimization of thermal power plants. The 29th
International Conference on Efficiency, Cost, Optimization, Simulation and Environmen-
tal Impact of Energy Systems, June 19-23, Portoroz, Slovenia.
2. Wang, L., Huang, S., Wang, N., Dong, C., Yang, Y., and Tsatsaronis, G. (2014). A
modified specific fuel consumption analysis for predicting the rearrangement of
system structures. The 6th International Conference on Applied Energy, May 30June
2, Taibei, China.
3. Huang, S., Xu, G., Yang, Y., Zhang, C., Wang, L., Wang, N., and Yang, Z. (2014).
System integration and flowsheet optimization of 1000 MW coal-fired supercriti-
cal power generation units. The 6th International Conference on Applied Energy, May
30June 2, Taibei, China.
4. Wu, L., Wang, L., Wang, Y., Hu, X., Dong, C., Yang, Z., and Yang, Y. (2014). Com-
ponent and process based exergy evaluation of a 600MW coal-fired power plant.
The 6th International Conference on Applied Energy, May 30June 2, Taibei, China.
5. Wang, L., Yang, Y., Dong, C., Morosuk, T., and Tsatsaronis, G. (2013). Multi-
objective optimization of coal-fired power plants using differential evolution. The
26th International Conference on Efficiency, Cost, Optimization, Simulation and Envi-
ronmental Impact of Energy Systems, July 1619, Guilin, China.
6. Wang, L., Yang, Y., Morosuk, T., and Tsatsaronis, G. (2012). Conventional and
advanced exergetic evaluation of a supercritical coal-fired power plant. The 25th
International Conference on Efficiency, Cost, Optimization, Simulation and Environmen-
tal Impact of Energy Systems, June 2629, Perugia, Italy.
7. Yang, Y., Wang, L., Dong, C., and Xu, G. (2012). Exergy analysis and paramet-
ric study of a large-scale coal-fired supercritical power generation unit. The 4th
International Conference on Applied Energy, July 58, Suzhou, China.