You are on page 1of 198

Thermo-economic Evaluation, Optimization and

Synthesis of Large-scale Coal-fired Power Plants

vorgelegt von
Ligang Wang
geb. in Qingdao, China

von der Fakultt III Prozesswissenschaften


der Technischen Universitt Berlin
zur Erlangung des akademischen Grades

Doktor der Ingenieurwissenschaften


- Dr.-Ing. -

genehmigte Dissertation

Promotionsausschuss

Vorsitzender: Prof. Dr. Tatiana Morosuk


Berichter: Prof. Dr.-Ing. George Tsatsaronis
Berichter: Prof. Dr.-Ing. Andr Bardow

Tag der wissenschaftlichen Aussprache: 01.07.2016

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

3.4 Suggestions for improvement . . . . . . . . . . . . . . . . . . . . . . . . . . 53


3.5 Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4 superstructure-based synthesis and optimization of coal-fired
power plants 57
4.1 Sensitivity analysis of key variables on the cycle performance . . . . . . . 57
4.2 Graphical superstructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3 Statement of the optimization problem . . . . . . . . . . . . . . . . . . . . 61
4.4 Multi-objective supported differential evolution . . . . . . . . . . . . . . . 62
4.4.1 Basic DE/MODE and the enhancement . . . . . . . . . . . . . . . . 62
4.4.2 Issues in the implementation of DE/MODE . . . . . . . . . . . . . 64
4.5 Design optimization for thermodynamic objective . . . . . . . . . . . . . . 66
4.5.1 Algorithm evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.5.2 Optimization for current steam conditions . . . . . . . . . . . . . . 68
4.5.3 Optimization for higher steam conditions . . . . . . . . . . . . . . . 70
4.6 Design optimization for thermodynamic and economic objectives . . . . . 73
4.6.1 Calculation of the cost of electricity for mathematical optimization 73
4.6.2 On the fixed structure . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.6.3 On the predefined superstructure . . . . . . . . . . . . . . . . . . . 82
4.7 Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5 superstructure-free synthesis and optimization of coal-fired
power plants 87
5.1 Multi-objective superstructure-free synthesis framework . . . . . . . . . . 87
5.1.1 Knowledge-integrated mutation operator . . . . . . . . . . . . . . . 90
5.1.2 Replacement rules for structural mutation . . . . . . . . . . . . . . 91
5.1.3 Insertion rule for structural mutation . . . . . . . . . . . . . . . . . 93
5.2 NLP model of thermal power plants . . . . . . . . . . . . . . . . . . . . . . 95
5.3 Evaluation of the superstructure-free approach . . . . . . . . . . . . . . . . 96
5.3.1 Thermodynamic objective function . . . . . . . . . . . . . . . . . . 96
5.3.2 Specified energy conversion hierarchy . . . . . . . . . . . . . . . . . 96
5.3.3 Initial solution and optimal solution by only replacement rules . . 97
5.3.4 Evaluation of the approach . . . . . . . . . . . . . . . . . . . . . . . 98
5.4 Superstructure-free synthesis of complex coal-fired power plants . . . . . 101
5.4.1 Optimal synthesis for maximizing thermal efficiency . . . . . . . . 102
5.4.2 Optimal synthesis for minimizing cost of electricity . . . . . . . . . 104
5.4.3 Bi-objective optimal synthesis . . . . . . . . . . . . . . . . . . . . . . 107
5.5 Summary and conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6 summary and conclusions 111
6.1 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
6.1.1 Real-world optimal designs and retrofits . . . . . . . . . . . . . . . 115
6.1.2 Evaluation and synthesis methodologies . . . . . . . . . . . . . . . 116
6.1.3 Mathematical modeling practice . . . . . . . . . . . . . . . . . . . . 118
a thermodynamic models and cost functions of the components 119
b additional materials for the analysis and evaluation 125
b.1 Exergy and exergoeconomic analysis of each type of component . . . . . 125
b.2 Real and unavoidable conditions of involved components . . . . . . . . . 129
b.3 Stream data of the simulation under real conditions . . . . . . . . . . . . . 131
contents vii

b.4 Additional materials for economic evaluation . . . . . . . . . . . . . . . . . 134


b.5 Results of conventional and advanced analyses . . . . . . . . . . . . . . . . 135
c additional materials for the superstructure-free synthesis 143
c.1 Illustrative study for evaluating the superstructure-free approach . . . . . 143
c.1.1 Specified energy conversion hierarchy . . . . . . . . . . . . . . . . . 143
c.1.2 Initial and optimal solutions . . . . . . . . . . . . . . . . . . . . . . 144
c.1.3 Some intermediate solutions of one representative run . . . . . . . 145
c.2 Optimal synthesis of complex coal-fired power plants . . . . . . . . . . . . 146
c.2.1 Specified energy conversion hierarchy . . . . . . . . . . . . . . . . . 146
c.2.2 Single-objective synthesis for maximizing thermal efficiency . . . . 147
c.2.3 Single-objective synthesis for minimizing cost of electricity . . . . 150
c.2.4 Bi-objective optimal synthesis . . . . . . . . . . . . . . . . . . . . . . 152

bibliography 153
Publications 177
LIST OF FIGURES

Figure 2.1 Technology development of pulverized coal power plants . . . . 5


Figure 2.2 Schematics of the advanced ultra-supercritical coal power plant
under development . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Figure 2.3 Key idea of the Master Cycle . . . . . . . . . . . . . . . . . . . . . 7
Figure 2.4 Fundamental considerations and new challenges of the design
of pulverized coal-fired power plants . . . . . . . . . . . . . . . . . 7
Figure 2.5 Definition of specific unavoidable exergy destruction and spe-
cific unavoidable investment cost . . . . . . . . . . . . . . . . . . . 12
Figure 2.6 Complete splitting of the exergy destruction in an advanced ex-
ergetic analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Figure 2.7 Flowchart of an evolutionary algorithm . . . . . . . . . . . . . . . 19
Figure 2.8 Graphical representation of the e-constraint method . . . . . . . . 29
Figure 2.9 Graphical representation of the normal constraint method for
bi-objective problems . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Figure 2.10 Solution selection in S-merit multi-objective evolution algorithm 30
Figure 3.1 Flowsheet of an ideal coal-fired power plant . . . . . . . . . . . . 41
Figure 3.2 Systems I for calculating the endogenous exergy destructions
within different components . . . . . . . . . . . . . . . . . . . . . . 42
Figure 3.3 Schematic of the ultra-supercritical pulverized coal power plant . 46
Figure 3.4 Exergy efficiencies and temperature profiles of hot and cold
streams of all heat exchangers in the boiler subsystem . . . . . . . 47
Figure 3.5 Exergy efficiencies and temperature profiles of the heat exchang-
ers in the turbine subsystem . . . . . . . . . . . . . . . . . . . . . . 47
Figure 3.6 Performance of heat exchangers on the temperature levels and
differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Figure 3.7 Spatial distribution of exergy destructions and losses . . . . . . . 50
Figure 4.1 Simple plant of the preliminary study . . . . . . . . . . . . . . . . 58
Figure 4.2 Effects of reheat parameters on plant efficiency and quality of
the exhausted steam . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Figure 4.3 The superstructure considering up to two-stage reheating, up
to 10 feedwater preheaters and a secondary turbine with steam
extractions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Figure 4.4 Possible stream paths controlled by switches SW1SW4 . . . . . . 60
Figure 4.5 Multi-objective differential evolution and the implementation . . 63
Figure 4.6 Mutation of reheat positions . . . . . . . . . . . . . . . . . . . . . . 65
Figure 4.7 Generation of pressure arrays for successful simulations during
the initialization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
Figure 4.8 Evolution details of the three cases . . . . . . . . . . . . . . . . . . 67
Figure 4.9 Plant efficiency versus solution number in generation 0 (initial
generation) and 50 for case 1 . . . . . . . . . . . . . . . . . . . . . 67
Figure 4.10 One representation run of the DE/MODE algorithm . . . . . . . 68

viii
List of Figures ix

Figure 4.11 The best-known plant efficiencies under different numbers of


feedwater preheaters and final feedwater temperatures . . . . . . 68
Figure 4.12 Optimal efficiency for different steam conditions under con-
straints type1 and type2 . . . . . . . . . . . . . . . . . . . . . . . . 71
Figure 4.13 The efficiency penalty and the optimal reheat pressure under
the constraint type2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Figure 4.14 Statistical optimal reheat pressure ratios of zone A and C . . . . . 72
Figure 4.15 Optimal reheat pressures of constraint type1 varying with the
throttle pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Figure 4.16 The schematic of the power plant with a fixed structure . . . . . . 73
Figure 4.17 Cost of electricity versus plant efficiency at different generation
numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
Figure 4.18 Contribution of fuel cost and capital investment cost to cost of
electricity of the front solutions . . . . . . . . . . . . . . . . . . . . 75
Figure 4.19 Share of the investment cost of each component for the econom-
ically optimal solution with equal temperatures of main and re-
heated steams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Figure 4.20 Distributions of the investment costs of all front solutions . . . . 76
Figure 4.21 Variables related to purchased equipment costs of the boiler and
the condenser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
Figure 4.22 The front and important decision variables of the case with
nonequal temperatures of main and reheated steams . . . . . . . 78
Figure 4.23 The impact of the temperature-related factor on the front of the
case with identical temperatures of main and reheated steams . . 79
Figure 4.24 Fronts of the four cases with seven feedwater preheaters . . . . . 83
Figure 4.25 Optimal pressure ratio of single-reheat case with varied structures 83
Figure 4.26 The influences of the secondary turbine with steam extractions
on plant efficiency and cost of electricity for the double-reheat
unit with varied structures . . . . . . . . . . . . . . . . . . . . . . . 84
Figure 4.27 Optimal pressure ratios of the two double-reheat cases with var-
ied structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Figure 5.1 Multi-objective superstructure-free optimization approach . . . . 88
Figure 5.2 The energy conversion hierarchy for thermal power plants . . . . 90
Figure 5.3 Example run of a single mutation step . . . . . . . . . . . . . . . . 92
Figure 5.4 An exemplary run for stream-based insertion of a heat exchanger 94
Figure 5.5 Flowchart for the generalized insertion rule . . . . . . . . . . . . . 94
Figure 5.6 Pressure-enthalpy diagrams of the initial and optimal solutions
of the illustrative study . . . . . . . . . . . . . . . . . . . . . . . . . 98
Figure 5.7 Representative runs for the illustrative study . . . . . . . . . . . . 99
Figure 5.8 Pressure-enthalpy diagrams of the initial and optimal solutions
for maximizing the thermal efficiency . . . . . . . . . . . . . . . . 102
Figure 5.9 One representative run for maximizing the thermal efficiency . . 103
Figure 5.10 Thermal efficiency versus numbers of feedwater preheaters and
reheaters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
Figure 5.11 Thermal efficiency versus optimal reheating pressure ratios and
number of feedwater preheaters above reheat 1 . . . . . . . . . . . 104
x List of Figures

Figure 5.12 Pressure-enthalpy diagrams of the initial and optimal solutions


for minimizing the cost of electricity . . . . . . . . . . . . . . . . . 105
Figure 5.13 One representative run for minimizing the cost of electricity . . . 106
Figure 5.14 Numbers of reheating and feedwater preheater of solutions gen-
erated for minimizing the cost of electricity . . . . . . . . . . . . . 107
Figure 5.15 One representative run for the bi-objective optimal synthesis . . . 108
Figure 5.16 Variable values and numbers of feedwater preheaters and re-
heaters of all Pareto solutions . . . . . . . . . . . . . . . . . . . . . 109
Figure C.1 Specified energy conversion hierarchy for the illustrative study . 143
Figure C.2 The flowsheets of the initial and optimal solutions identified in
the illustrative study . . . . . . . . . . . . . . . . . . . . . . . . . . 144
Figure C.3 Temperature-entropy diagrams of the initial and optimal solu-
tions identified in the illustrative study . . . . . . . . . . . . . . . 144
Figure C.4 Some intermediate solutions identified in the representative run
of the illustrative study . . . . . . . . . . . . . . . . . . . . . . . . . 145
Figure C.5 Specified energy conversion hierarchy for synthesizing complex
coal-fired power plants . . . . . . . . . . . . . . . . . . . . . . . . . 146
Figure C.6 The flowsheets of the initial and optimal solutions identified for
maximizing the thermal efficiency . . . . . . . . . . . . . . . . . . 147
Figure C.7 Temperature-entropy diagrams of the initial and optimal solu-
tions for maximizing the thermal efficiency . . . . . . . . . . . . . 147
Figure C.8 Some intermediate solutions of the representative run for maxi-
mizing thermal efficiency . . . . . . . . . . . . . . . . . . . . . . . 148
Figure C.9 The flowsheets of the initial and optimal solutions identified for
minimizing the cost of electricity . . . . . . . . . . . . . . . . . . . 150
Figure C.10 Temperature-entropy diagrams of the initial and optimal solu-
tions for minimizing the cost of electricity . . . . . . . . . . . . . . 150
Figure C.11 Some intermediate solutions of the representative run for mini-
mizing the cost of electricity . . . . . . . . . . . . . . . . . . . . . . 151
Figure C.12 Economically- and thermodynamically-optimal solutions of the
representative run for the bi-objective optimal synthesis . . . . . 152
L I S T O F TA B L E S

Table 3.1 Framework of comprehensive exergy-based evaluation . . . . . . 36


Table 3.2 Simulations needed for calculating all terms . . . . . . . . . . . . 36
Table 3.3 Basic specification of the pulverized-coal fired power plant un-
der real conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
Table 4.1 Specifications of cases 1, 2, 3 . . . . . . . . . . . . . . . . . . . . . . 66
Table 4.2 The performance variables of each type of component embed-
ded in the superstructure for thermodynamical optimization . . . 66
Table 4.3 Comparisons of the steam extractions after reheating . . . . . . . 69
Table 4.4 The transition throttle pressures corresponding to each temper-
ature level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
Table 4.5 Decision variables and the bounds . . . . . . . . . . . . . . . . . . 74
Table 4.6 Comparison of the decision variables between the industrial de-
sign and the economically optimal solutions . . . . . . . . . . . . 80
Table 4.7 Exergoeconomic comparisons between the industrial design and
the economically optimal solutions . . . . . . . . . . . . . . . . . . 81
Table 5.1 Basic specifications of the thermal power plant . . . . . . . . . . . 97
Table 5.2 Computational performances and optimality gaps of 6 optimiza-
tion runs with different maximum mutation strengths . . . . . . . 101
Table 6.1 Comparisons between the methodologies employed in this the-
sis for the analysis and synthesis of thermal power plants . . . . 113
Table B.1 Real and unavoidable conditions of all involved components . . . 129
Table B.2 The stream data of the design condition . . . . . . . . . . . . . . . 131
Table B.3 Weight estimation and weighted factors for heat surfaces in boiler134
Table B.4 The module factors for different types of components . . . . . . . 134
Table B.5 Conventional exergy and exergoeconomic analysis of the real case135
Table B.6 Advanced exergy analysis of the coal-fired power plant . . . . . . 137
Table B.7 The splitting of investment costs . . . . . . . . . . . . . . . . . . . 139
Table B.8 The splitting of exergy destruction costs . . . . . . . . . . . . . . . 141
Table C.1 Computational performance and optimality gaps of 6 runs for
maximizing the thermal efficiency . . . . . . . . . . . . . . . . . . 149
Table C.2 Computational performance and optimality gaps of 6 runs for
minimizing the cost of electricity . . . . . . . . . . . . . . . . . . . 152

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

IPT intermediate-pressure turbine


IP integer programming
LHV lower heating value
LPT low-pressure turbine
LP (continuous) linear programming
L loss
MILP mixed integer linear programming
MINLP mixed integer nonlinear programming
MODE multi-objective differential evolution
MOO multi-objective optimization
MT main turbine
MX mexogenous
NLP (continuous) nonlinear programming
NP non-deterministic polynomial-time
NSGA-II non-dominated sorting genetic algorithm II
OMC operation and maintenance cost
OT ordinary secondary turbine
PAF primary air fan
PCPP pulverized-coal power plant
PEC purchased-equipment cost
PP percentage point
PRH primary reheater
PSH primary superheater
P product
QP quadratic programming
rh reheating
SAF secondary air fan
SMS S-merit selection
SOS2 special ordered set of type 2
SPECO specific exergy costing
SQP successive quadratic programming
SSH screen-type superheater
ST steam turbine
SW path switch
TCI total capital cost
TRR total revenue requirement
UN unavoidable
WW waterwall
Greek Symbols
, real-valued constants [0,1] 1
difference
crowding distance
ti ,pi efficiency penalty at throttle pressure ti and throttle temperature pi PP
e interval of objective value in e-constraint method
e optimality gap %
efficiency %
xiv Nomenclature

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

Fm mass-flow-rate related factor


Fp pressure-related factor
Fshrh reheat-related factor
Ft temperature-related factor
FC fuel cost $/yr
g global integer variables for identifying structural alternatives
g inequality constraint
h enthalpy kJ/kg
h equality constraint
h0 specific enthalpy of saturated water kJ/kg
h00 specific enthalpy of saturated steam kJ/kg
k, n number 1
l lower bound of a branch
LHV lower heating value kJ/kg
Naoh annual operation hours h/yr
Nc number of components 1
Ns number of exergy streams 1
Ns population size in evolutionary optimization 1
Nv number of variables 1
O operational variable space
o operational variable
OMC operation and maintenance cost $/yr
p, p pressure bar
pt transition throttle pressure bar
r non-domination rank in multi-objective optimization 1
r relative cost difference in exergoeconomic analysis %
S structural variable space, solution space or a set of solutions
s structural variable
t temperature C
u objective value of a feasible solution
v0 specific volume of saturated water kg/m 3

v00 specific volume of saturated steam kg/m3


v, z variable
w weighting factor
X feasible solution space
x indenpendent decision variable
x quality of the exhausted steam 1
yD exergy destruction ratio %
Subscripts and Superscripts
abs absorbed
aux auxiliary
dc dissipative component
dif difference
ex exhausted
fw feedwater
in incoming
xvi Nomenclature

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

require cost-effective integration to maximize their benefits. Therefore, it is necessary


to re-consider the system-level evaluation, optimization and synthesis of PCPPs.
The methodologies for evaluation, optimization and synthesis of energy systems
have also been upgraded significantly during last thirty years (Tsatsaronis, 2008; Gross-
mann and Guilln-Goslbez, 2010). There are even many automatic computer-aided
approaches for conceptual synthesis of chemical processes, e. g., Friedler et al., 1995,
or energy systems, e. g., Voll et al., 2013. Unfortunately, engineering expertise and ex-
perience, or simply simulation-based sensitivity analysis, still guide the generation of
alternatives for system configurations and the selections of the devices and key pa-
rameters in engineering design practice (Klatt and Marquardt, 2009; Kossiakoff et al.,
2011). The involved sequential and iterative design adjustments and decision-making
limit the number of solution alternatives considered, thus generally result in only sub-
optimal designs (Li and Kraslawski, 2004). From such perspectives, it is advantageous
to solve the optimal design of conventional thermal power plants by implementing the
state-of-the-art methodologies.
The thesis, therefore, focuses on enhancing the thermodynamic and economic per-
formance of pulverized-coal power plants by comprehensively analyzing or, more im-
portantly, automatically and evolutionarily optimizing the structural configurations.

1.1 Structure of the thesis


In chapter 2, the development trends of pulverized-coal power plants and potential
challenges for the technology improvement are first introduced. Afterwards, state-of-
the-art methodologies for analysis, optimization and synthesis of energy systems are
reviewed with multi-objective optimization techniques. Finally, scientific contributions
of the thesis are specified.
In chapter 3, a framework for comprehensive evaluation of energy systems is pre-
sented, which is based on conventional and advanced exergy (exergoeconomic) analy-
sis. The conventional analysis reveals only the spatial distribution of exergy dissipation
(formation process of product cost), while the advanced analysis splits the exergy dissi-
pation (cost) into endogenous/exogenous and avoidable/unavoidable parts to identify
the interactions among components and the sources for energy(cost)-saving potential
of each component. A new approach for quantitatively calculating endogenous exergy
destructions is introduced and applied to thermal power plants in detail. The frame-
work is then employed to evaluate a typical modern ultra-supercritical pulverized-coal
power plant to identify energy- and cost-saving potentials and improvement strategies.
In chapter 4, superstructure-based synthesis and optimization approach is applied to
handle the optimal design of current and future plants. A graphic superstructure, con-
sidering single or double reheats, up to 10 feedwater preheaters (plus one de-aerator)
and the existence of a secondary steam turbine with stream extractions, has been de-
veloped with the help of a professional process simulator. The optimization problem
related to traditional design considerations, e. g., the matches of parameters of main
and reheat steam for a reasonable dryness of exhausted steam and optimal reheating
ratios, is handled by an evolutionary algorithm using bi-objective optimization.
In chapter 5, a generic superstructure-free synthesis framework proposed originally
by Voll et al. (2012) for distributed energy supply systems is generalized and extended
1.1 structure of the thesis 3

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

In this chapter, traditional considerations in the system-level design of coal-fired power


plants are introduced with new challenges lying ahead (section 2.1). Then, various
methodologies for evaluating energy systems are reviewed (section 2.2); while the fol-
lowing two sections (section 2.3 and 2.4) present widely used optimization and syn-
thesis approaches. Section 2.5 includes multi-objective decision-making techniques. Fi-
nally, the contributions of the thesis are highlighted in section 2.6.

2.1 Trends and challenges of the design of coal-fired power plants


Coal-fired power plants have went through nearly one hundred years of development.
Technology progress originates mainly from the milestones of material improvement
(Fig. 2.1). The technology trend is toward higher steam parameters and larger gen-
erating capacity. Current best practice, the Waigaoqiao III power plant (21000 MW,
280 bar 605/603 C) in Shanghai, has reached an operation efficiency of up to 46%
(Klegman, 2010) by employing specific operational strategies (Feng, 2008). The next
generation technology, advanced ultra-supercritical power plants (Fig. 2.2), aiming at
steam temperatures over 700 C and pressures over 350 bar, has been under intensive
R&D since the mid-1990s and promises to constitute a benchmark plant with a design

250/540/560 280/600/620 350/700/720 bar/C/C 380/700/720


Ferritic Ferritic
Ferritic R&D ongoing
Efficiency improvement

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

Figure 2.1: Technology development of pulverized coal power plants

5
6 analysis, optimization and synthesis of coal-fired power plants

35 MPa 720 C 720 C


700 C

Boiler
VHPT HPT-IPT LPT
Steam turbine
Economizer

Conventional materials Material under development Material under development


a
Ferrite Ni-based Ferrite Austenitea
Austenite Ni-based
a
Composition of Ferrite and Austenite are adjusted for particular applications.

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)

efficiency of approximately 50% (Masuyama, 2004; Huseman, 2005; Viswanathan et al.,


2006; Weitzel, 2011).
Pulverized-coal power plants are based on the classical Rankine cycle. The efficiency
of an ideal Rankine cycle (ideal ) is determined by average temperatures of heat absorp-
tion (Ta,in ) and heat release (Ta,out ) of the working fluid:
Ta,out
ideal = 1 . (2.1)
Ta,in
The higher the average temperature of heat absorption and the lower the average tem-
perature of heat release, the greater the cycle efficiency can be achieved. For condensing
power plants, the average temperature of heat release depends on local ambient condi-
tions. Thus, to achieve a higher cycle efficiency, the major means is to increase the aver-
age temperature of heat absorption, which can be achieved by increasing the tempera-
tures of main and reheated streams, increasing the final feedwater preheating tempera-
ture, adding more feedwater preheaters and employing multiple reheating (Rukes and
Taud, 2004; Spliethoff, 2010). For real-world Rankine-cycle-based coal power plants, the
increase of the pressure level of main steam and the reduction of thermodynamic inef-
ficiencies occurring in real components (e. g., friction loss and steam leakage in steam
turbines) can improve the plant efficiency as well. These design options for efficiency
improvement have been considered during the development of future coal-fired power
plants (Fig. 2.2).
Although the temperature increase of main and reheated steams can improve the
plant efficiency, it may lead to overheat crisis of feedwater preheaters, especially those
that extract superheated steam from the turbines after reheating. In addition, the super-
heat degree of steam extractions indicates incomplete steam expansion (i. e., the loss of
2.1 trends and challenges of the design of coal-fired power plants 7

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

able technologies or concepts, which deliver significant improvement of overall plant


efficiency, becomes possible. The options include topping or bottoming cycles (such
as the CO2 -based closed Brayton cycle or the organic Rankine cycle Eaves et al. 2014),
low-grade waste heat recovery from flue gas (Espatolero et al., 2010), low-rank coal
predrying (Karthikeyan et al., 2009), multiple heat sources (especially solar thermal,
Turchi et al. 2011; Popov 2011; Yang et al. 2011), etc. In addition, pollutant-removal
technologies, particularly for CO2 capture, should be considered as well.
Therefore, except for those fundamental considerations for the design of a power
plant itself, such as employing more stages of reheating, increasing feedwater preheat-
ing temperature and implementing more feedwater preheaters, future design concept
of this conventional power generation system emphasizes system-level synthesis for in-
tegrating many available advantageous technologies (Fig. 2.4). The question is then to
find the best integration of multiple technologies considered by a systematic, effective
synthesis and optimization method. However, to the authors best knowledge, there
has been no approach that could perfectly solve this complex synthesis problem.

2.2 Analysis of energy systems


The analysis of energy systems is a prerequisite for identifying the design imperfec-
tions and promoting improvement strategies. The analysis can be based on energy or
exergy. An energy analysis accounts only for the quantity of energy, identifies only
the energy transfers to the environment as thermodynamic inefficiencies, thus fails to
identify any inefficiency in an adiabatic process (Tsatsaronis, 2011). While combing the
concept of exergy, which corrects these misconceptions, an exergy analysis considers
also the quality of energy, thus complements and enhances the energy-based analysis.

2.2.1 Exergy analysis

All real processes are irreversible as their occurrence is driven by non-equilibrium


forces, leading to thermodynamic inefficiencies inside the process boundaries (destruc-
tion (D) of exergy) and those across the process boundaries (loss (L) of exergy). An ex-
ergy analysis identifies the spatial distribution of thermodynamic inefficiencies within
an energy system, pinpoints the components and processes with high irreversibilities,
thus highlights the areas of improvement of the system.
The formulation of an exergy analysis usually includes exergy balance equations
of the total system, a subsystem or a single component, which can be based on the
incoming and outgoing exergy flows or the fuel (F) and product (P) definitions, key
concepts introduced by Tsatsaronis (1984) to define the exergetic efficiency ( = E P / E F )
for an evaluation purpose. In addition, by properly selecting the system boundaries,
exergy losses occur only at the system level.
For the calculation of exergy, detailed methods for physical and chemical exergies of
different types of material flows, work and heat flows have been discussed in (Bejan
et al., 1996; Moran et al., 2010). The handling of exergy losses and dissipative com-
ponents, such as condenser, gas cleaning units and throttle valves, can be found in
(Tsatsaronis and Park, 2002; Lazzaretto and Tsatsaronis, 2006).
2.2 analysis of energy systems 9

2.2.2 Exergoeconomic analysis

Exergoeconomic analysis provides a deep understanding of costs related to equip-


ments and thermodynamic inefficiencies as well as their interconnection and also takes
into account the interaction between the components and the whole system by unit
costs of exergy flows and those of exergy destructions, thus tells us how we could iter-
atively improve the efficiency and cost effectiveness of the system (Tsatsaronis, 2011).
More importantly, in an exergoeconomic optimization1 , individual optimization of sys-
tem components decomposed from the whole optimization problem is made possible.
This decomposition relies on the statement that exergy is the only rational basis for the
costs of energy flows and the inefficiencies within a system (Tsatsaronis, 2011).
Major theoretical fundamentals of exergoeconomics have been established during
1980s and 1990s. The term exergoeconomics was coined by Tsatsaronis (1984), referred
to as an exergy-aided cost-reduction method (Tsatsaronis, 1999a). Key contributions of
exergoeconomics came from a number of scientists, such as Tsatsaronis and Winhold
(1985b,a), Tsatsaronis (1993), Tsatsaronis and Pisa (1994), Tsatsaronis et al. (1993), Laz-
zaretto and Tsatsaronis (1997, 1999), Valero et al. (1986a,b,c), Valero and Torres (1988),
Valero et al. (1994a), Lozano and Valero (1993), Frangopoulos (1983, 1988, 1992, 1994),
von Spakovsky (1986), von Spakovsky and Evans (1990), von Spakovsky (1994), etc.
These works can be classified as accounting and calculus methods (Gaggioli and El-
Sayed, 1989).

2.2.2.1 The accounting methods


The accounting methods aim at understanding the formation of product costs, evaluat-
ing performance of components and the system, and improving the system iteratively.
To obtain unknown costs of all exergy flows, a set of algebraic equations is built. The
equation set consists of cost balance equations associated with each unit (a component
or a set of components of the system) and auxiliary cost equations that are needed
for the units, of which the number of output streams is larger than the number of
input streams. Evaluation of the equation set starts from the known costs of all input
resources. With the costs of all exergy flows known, several exergoeconomic variables
associated with each unit are calculated for performance evaluation and system im-
provement (Tsatsaronis, 1993; Tsatsaronis and Pisa, 1994).
The allocation of costs to internal flows and products are mostly performed on mon-
etary basis (sometimes on exergetic cost basis (Valero et al., 1986a)). The monetary cost
of an exergy flow usually is accounted by the average cost associated with different
exergy forms (thermal, mechanical and chemical) (Tsatsaronis et al., 1993; Lazzaretto
and Tsatsaronis, 1997). A systematic, generic and easy-to-use methodology, the specific
exergy costing (SPECO) method, has been proposed by Lazzaretto and Tsatsaronis
(2006), which has been the milestone of the accounting methods. In SPECO method,
cost balance equations of each unit include the cost flow rates associated with capital
amortization from an economic accounting, while fuel and product definitions and
auxiliary cost equations are developed at the component level and also in the most

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).

2.2.2.2 The calculus methods


The calculus methods serve directly for mathematical cost minimization. The central
idea is to closely approach thermoeconomic isolation, by means of thermoeconomic de-
composition, for quickly and accurately assessing the effect of a certain parameter
on the system performance without optimizing the whole problem (local optimiza-
tion) (Frangopoulos, 1983). Different decomposition approaches, i. e., the thermoeco-
nomic functional analysis (Frangopoulos, 1983, 1991, 1992, 1994), Engineering Func-
tional Analysis (von Spakovsky, 1986; von Spakovsky and Evans, 1990; von Spakovsky,
1994) and Three-Link Approach (Hua et al., 1989, 1997), have been developed for en-
ergy systems of different levels of design detail.
When the method of Lagrange multipliers is applied as the optimization algorithm,
such as in the thermoeconomic functional analysis, the system is first decomposed by
a functional analysis2 into units (the functional diagram (Frangopoulos, 1983), which
is, in fact, the productive structure), each one of which has one particular function with
a single exergy product. Then, the cost objective function is reformulated by adding a
summation of Lagrange multipliers-weighted exergy products of all units. Thus, the
multipliers do have their physical meaning: marginal costs of the exergy flows in the
functional diagram. Introducing the marginal costs makes the problem readily solved
by sequential algorithms.
However, the marginal costs are difficult to interpret regarding the process of cost
formation (Valero et al., 1993), thus these methods are unable to reveal the physical
and economic interrelationships among the components (Frangopoulos, 1994). In ad-
dition, thermoeconomics decomposition becomes limited when complex systems are
considered and also less necessary due to the rapid developments of direct mathemat-
ical optimization tools and the computation ability. Therefore, there have been no new
developments or interesting applications of these calculus methods in recent years.

2.2.2.3 Recent development


In general, the maturity of exergoeconomics is marked by the SPECO method (Laz-
zaretto and Tsatsaronis, 2006); however, methodological and fundamental discussions
have still been continued. One recent focus is the cost accounting associated with dissi-
pative components, i. e., those whose productive purpose is neither intuitive nor easy
to define. Torres et al. (2008) and Seyyedi et al. (2010b) discussed the mathematical ba-
sis and different criteria for cost assessment and formation process of the residues, and
suggested costs entering a dissipative component should be charged to the productive
component responsible for the residue. Piacentino and Cardona (2010b) introduced

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.

2.2.3 Advanced exergy-based analysis

When attempting to reduce thermodynamic inefficiencies within a system, additional


factors must be taken into account: (a) Not all inefficiencies can be avoided (Kotas,
1985), due to physical and economic constraints. The technical possibilities of exergy
savings (i. e., the avoidable inefficiencies) of a component or system are always lower
than the corresponding theoretical limit of thermodynamic exergy savings (Lozano
and Valero, 1993). (b) Components in an energy system are not isolated whereas inter-
actions among them always exist. Thus, part of exergy destruction within a component
is, in general, caused by the inefficiencies of the remaining components of the system
(Cziesla et al., 2006). (c) The same amount of exergy destruction within different com-
ponents is not equivalent (Kotas, 1985), because of different fundamental mechanisms
of irreversibility and the component-system interactions. In other words, the same
amount of decrease in exergy destruction within two different components has differ-
ent impacts on the overall fuel consumption of the system (Lozano and Valero, 1993).
These issues, however, cannot be addressed by the conventional exergy-based analysis.
Thus, as one solution, an advanced exergy (exergoeconomic) analysis has been de-
veloped continuously for last ten years by Tsatsaronis and his coworkers (Tsatsaronis,
1999b; Tsatsaronis and Park, 2002; Cziesla et al., 2006; Kelly, 2008; Tsatsaronis, 2008;
Kelly et al., 2009; Morosuk and Tsatsaronis, 2009a,b; Tsatsaronis and Morosuk, 2010;
Petrakopoulou, 2011), in which the exergy destruction (and cost) within a system com-
ponent are further split: the avoidable (AV) and unavoidable (UN) parts, the endogenous
(EN) and exogenous (EX) parts, and their combinations.
12 analysis, optimization and synthesis of coal-fired power plants

Specific unavoidable

unit of product exergy


exergy destruction

Investment cost per


Variation range of
investment costs
A

Specific unavoidable
invesment cost

Exergy destruction per


unit of product exergy

Figure 2.5: Definition of specific unavoidable exergy destruction ( E D / E P )UN


k and specific un-
UN
avoidable investment cost ( Z/ EP )k based on the expected relationship between
investment cost and exergy destruction (or exergetic efficiency) for the kth compo-
nent. (Reproduced from Tsatsaronis and Park (2002))

2.2.3.1 Avoidable/unavoidable exergy destruction and cost


By employing technically feasible designs and/or operational enhancement, part of
exergy destruction and costs associated with a system or component can be avoided,
thus this part is considered as avoidable.
The estimation procedure has been initially discussed in (Tsatsaronis, 1999b; Tsatsa-
ronis and Park, 2002; Cziesla et al., 2006). Practically, the cost behavior exhibited by
most components is that the investment cost (Z) per unit of product exergy increases
with decreasing exergy destruction (E D ) per unit of product exergy or with increasing
efficiency (Tsatsaronis and Park, 2002). Thus, for the kth system component, which is
considered in isolation, if two limit states (Fig. 2.5), one with extremely large invest-
ment cost and one with extremely high thermodynamic inefficiency, can be estimated
with reasonably, then the unavoidable exergy destruction ratio ( E D / E P )UN and the un-
avoidable investment cost ratio ( Z/ E P )UN with respect to per unit of product exergy
k
could be determined: E D,k
UN = E UN
P,k ( ED / EP )k and Zk
UN = E P,k ( Z/
E P )UN . Once the ex-
k
ergy destruction E D,k
UN and the cost Z UN are known, the avoidable parts can be obtained:
k
E D,k
AV = E UN AV
D,k ED,k and Zk = Zk Zk .
UN
In general, both extreme states for the ratios ( E D / E P )UN
k and ( Z/ E P )UN are not indus-
k
trially achievable; however, they can be simulated by adjusting a set of thermodynamic
parameters associated with the considered component, including the parameters of in-
coming and outgoing streams, and the key design parameters of the component itself.

2.2.3.2 Endogenous/exogenous exergy destruction and cost


The endogenous exergy destruction within the kth component (E D,k
EN ) is that part of the

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/

nous part is obtained: Z kEX = Z k Z kEN .

2.2.3.3 Combination of the two splittings


All the exergy destruction rates and cost rates can be further split into avoidable/en-
dogenous, unavoidable/endogenous, avoidable/exogenous and unavoidable/exoge-
nous parts. The complete list of all possible terms involved in an advanced exergoeco-
nomic evaluation has been illustrated in Fig. 2.6.
An evaluation should consider all available data and be conducted in a comprehen-
sive way. In general, improvement efforts should be made to those components with
relatively high avoidable exergy destructions or costs. However, it is also possible these
avoidable parts are mainly exogenous, for which another component or a group of ad-
jacent components are responsible. Therefore, the sources of the avoidability are more
reasonably identified and the improvement or optimization will not be misguided.

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.

2.2.4 Analysis of coal-fired power plants

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

However, these applications donot focus on modern large-scale (ultra-)supercritical


coal-fired power plants. Only rough distributions of exergy dissipation and costs, and
formation process of electricity cost are presented, as almost all of them consider the
steam generator (boiler) as a black box and with no insights into it. More importantly,
these works have not quantified the avoidable exergy destructions and costs within
each component and the interactions among different components.

2.3 Optimization of energy systems


A general optimization problem consists of an objective function to be minimized (or
maximized) subject to (s.t.) equality and/or inequality constraints, and the considered
independent decision variables. For energy systems, there are usually three types of
decision variables (Frangopoulos, 2003), i. e., binary structural variables (s) associated
with the structure of the system, continuous or discrete design variables (d) related to
nominal characteristics and sizes of the system and the components, and continuous or
discrete operational variables (o) determining operation strategies at the system and/or
component levels. The degrees of freedom in the system structure lead to optimal
synthesis problems (section 2.4). Thus, the "optimization" referred to in this section is
for a given system structure and can be formulated as follows:

min f (d, o ), "objective function",


d, o

s.t. h(d, o ) = 0, "equality constraints", (2.2)


g(d, o ) 0, "inequality constraints".

The optimization of energy systems are usually solved by mathematical approaches


with or without the aid of special speed-up techniques associated with thermodynam-
ics, e. g., (Uhlenbruck and Lucas, 2000), or thermoeconomics (see section 2.2.2.2). Thus,
in the following, mathematical optimization is briefly introduced at first. Then, the non-
linearity and integrity in energy system formulations is discussed. Finally, the selection
and optimization of design parameters of steam cycles are reviewed. Note the "selec-
tion" and "optimization" are distinguished: the optimal selection of decision variables
is usually based on the sensitivity analysis of a set of variables, thus it usually cannot
lead to an optimum; while the optimization is capable of finding the mathematically
optimal combinations of all independent decision variables.

2.3.1 Mathematical optimization

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

methods in continuous optimization (Nocedal and Wright, 2006b), an extensive review


on mathematical optimization for process engineering (Biegler and Grossmann, 2004;
Grossmann and Biegler, 2004), recent advances in global optimization (Floudas and
Gounaris, 2009), derivative-free algorithms for bound-constrained optimization prob-
lems (Mor and Wild, 2009; Rios and Sahinidis, 2013), and a broad coverage of the
concepts, themes and instrumentalities of metaheuristics (Glover and Kochenberger,
2003). According to these, the basis of commonly used deterministic and metaheuristic
optimization algorithms associated with the thesis scope are briefly introduced below.

2.3.1.1 Deterministic algorithms


Given a particular input, a deterministic algorithm always passes through the same
sequence of search pattern and converges potentially fast to the same result. The algo-
rithms usually take advantage of analytical properties of the optimization problems;
thus, the problems need to be well formulated to avoid misguiding the search. How-
ever, for good formulations, particularly of complex problems, the user may have to
manually address some trivial issues (Drud, 2004), e. g., scaling of (intermediate) vari-
ables and functions. In addition, the search may end up with bad local optimal solu-
tions for complex problems.
The optimization of (LP), if no global solution algorithm is used, is a relatively ma-
ture field. For a well-conditioned linear problem with a bounded objective function, the
feasible region is geometrically a convex polyhedron, which implies a local extremum is
always globally optimal. The optimal solution, possibly not unique, is always attained
at the boundary of the feasible region. The optimality can be reached with a finite
steps, from any feasible solution either at the boundary (primal-dual simplex algorithms
(Nocedal and Wright, 2006a)) or at the interior (interior point algorithms (Nesterov et al.,
1994)) of the feasible region. Several modern solvers, e. g., XPRESS, CPLEX and IPOPT,
are capable of handling (LP) with an unlimited number of variables and constraints3 ,
subject to available time and memory.
For NLP problems, the optimal solution can basically occur anywhere of the feasible
region. Global optimality can be guaranteed for convex (NLP). Most NLP algorithms
require derivative information of the objective function and constraints for efficiently
determining effective searching directions. Commonly used solvers are usually based
on successive quadratic programming (SQP), e. g., IPOPT, KNITRO and SNOPT, which
generate Newton-like steps and need the fewest function evaluations, or generalized
reduced gradient (GRG (Lasdon et al., 1978)), e. g., GRG2 and CONOPT, which work
efficiently when function evaluations are relatively cheap. However, in case of non-
convexity, the search depends on the problem initialization and may get stuck at lo-
cal optima; thus, for global optimality, other optimization techniques, e. g., branch and
bound and branch and cut (see below), are necessary.
Integer (linear) programming problems have a combinatorial feature and are usually
NP-hard (Kppe, 2012). The basic idea to solve (MILP) is to generate and refine the
lower and upper bounds of the optimal objective value ( L and U), which are given by
solving LP relaxations and by the objective values of feasible solutions, respectively.
The solving algorithms are mostly based on a branch-and-bound idea, which incorpo-

3 www.solver.com/linear-quadratic-technology
2.3 optimization of energy systems 17

rates a systematic rooted-tree enumeration of candidate solutions by "branch" and effi-


cient eliminations of non-promising solutions (for integer optimality) by "bound". The
problem is first completely relaxed as a continuous root LP problem with the same
bounds for all variables. The root problem yields an initialization for bound L, while
the bound U is initially specified as infinity or the objective value of any integer fea-
sible solution (u). Starting from the root problem, the integer space is successively
partitioned (i. e., branched) into a set of relaxed LP subproblems, each one of which
yields a lower bound of the branch it represents (l). A qualified subproblem (satisfy-
ing the condition l < U) is further branched, if it, as is often the case, exhibits one or
more relaxed integer variables with fractional values; otherwise, it becomes a feasible
solution with a new bound u. The bounds L and U are updated conditionally and
tightened: If the condition l > L is satisfied, then the bound L is set identically as
bound l; if the condition u < U is satisfied, then the bound U is set the same as bound
u. The searching procedure continues until the optimality gap e (e = 100 (U L )/U),
is
reduced to zero (the global optimality guaranteed) or below a given value (e-optimal).
The algorithm can be further enhanced, as branch and cut, by introducing cutting planes
(linear inequities) to tighten the lower bound of LP relaxations. The best known MILP
solvers include CPLEX, XPRESS.
Mixed integer nonlinear problems are also NP-hard. The solving idea is similar by
generating and tightening the bounds of the optimal solution value. The algorithms,
generally branch-and-bound or branch-and-cut like, rely on relaxations of the integrity
to yield NLP subproblems and (linear) relaxations of the nonlinearity, which overesti-
mate (outer approximate) the feasible region and underestimate the objective function
(Bonami et al., 2012). Accordingly, commonly used solvers are mainly based on branch
and bound (e. g., SBB and Bonmin) and outer approximation (Duran and Grossmann,
1986) (e. g., DICOPT, and Bonmin). These solvers are efficient but limited to problems
of medium size, i. e., few hundreds of binary variables and several thousands of con-
tinuous variables and constraints (Grossmann and Guilln-Goslbez, 2010).
There is another problem of the above mentioned MINLP methods: when fixing cer-
tain discrete variables as zero for branching or approximation, the redundant equations
and intermediate variables may cause singularities and poor numerical performance
(Grossmann and Guilln-Goslbez, 2010). To circumvent this, GDP methods have been
developed as an alternative and receive increasing attention (see Grossmann and Ruiz,
2012). In GDP, the combination of algebraic and logical equations is allowed, thus the
representation of discrete decisions is simplified. However, the algorithms for GDP are
mostly under development (see Grossmann and Trespalacios, 2013) and currently only
the LOGMIP software (Vecchietti and Grossmann, 1999) is available.
The global optimality could be achieved for nonconvex (NLP) or (MINLP) by branch-
and-bound and cutting-plane techniques, e. g., in BARON. The idea is to remove the
nonconvexity by decomposing and refining the decision space with additional linear
inequalities, and by tightening the objective function value. In addition, reformula-
tion and approximation of original problems are also capable of reducing nonlinearity
or even generating convex formulations: Nonlinear integrity may be linearized, e. g.,
(Balas, 1985; Sherali and Alameddine, 1992; Liberti and Pantelides, 2006), while highly
nonlinear terms with continuous variables may be approximated by separable program-
ming or even linearizations. However, for real-world problems, whose global optimal
18 analysis, optimization and synthesis of coal-fired power plants

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.

2.3.1.2 Metaheuristic algorithms


Metaheuristic optimization algorithms are referred to as all nature-inspired algorithms,
in which the pattern of improving a single candidate solution is usually approximate,
non-deterministic and diversified, because of an "intelligent" local sampling or a com-
plex learning mechanism; hence, these algorithms are capable of escaping from lo-
cal optima and robustly exploring a decision space. Although the metaheuristics are
still not able to guarantee the global optimality for some classes of problems, e. g.,
(MILP) and (MINLP), they can generally find sufficiently good solutions. Moreover,
metaheuristic algorithms require no derivative information and donot analyze the al-
gebra of an optimization problem; thus, they can be applied to highly nonlinear (even
ill-conditioned) or blackbox problems. The major disadvantages, however, include po-
tential slow speed of convergence, unclear termination criterion, incapability of certi-
fying the optimality of the solutions, and potential need of designing problem-specific
searching strategies.
A wide variety of metaheuristics have been developed with claims of novelty and
practical efficacy (Blum and Roli, 2003). Commonly used algorithms mainly include
those single-solution based, e. g., simulated annealing, tabu search, and those population
based, e. g., evolutionary algorithms, ant colony optimization and particle swarm optimization.
In the following, the basis of population-based evolutionary algorithms, which are
used in the thesis, is briefly introduced.

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

Termination check Mating selection

Selection Recombination
(Crossover)

Evaluation Mutation

Figure 2.7: Flowchart of an evolutionary algorithm (adapted from (Voll, 2014))

(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.

2.3.2 Nonlinearity and integrity in energy system optimization

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.

2.3.3 Parametric selection and optimization of steam cycles

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-

4 freesteam.sourceforge.net, accessed on 8 June 2015.


5 www.tlk-thermo.com/en/software-products/tilmedia.html, accessed on 8 June 2015.
2.4 synthesis of energy systems 21

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.

2.4 Synthesis of energy systems


Process synthesis, namely complete flowsheet synthesis when performed at an overall sys-
tem level, deals with the selection of process structure (topology), i. e., the set of tech-
nical component employed and their interconnections. The optimal synthesis phase
usually contributes a major part to achieving the predefined goal or finding the glob-
ally optimal design option (Biegler et al., 1997). However, optimal synthesis tends to be
a tough task compared to a simple design or operation optimization: It normally takes
the design and/or operation optimization into account in a sequential or simultaneous
fashion; moreover, the design space of structural alternative is basically not known a
priori for a complex system, thus a complete, exact mathematical formulation of the
synthesis problem seems not possible (Westerberg, 2004). To systematically address
the synthesis of energy and process systems, a vast number of research has been con-
ducted in this field and methodologically reviewed by many authors, e. g., (Frangopou-
los et al., 2002; Kaibel and Schoenmakers, 2002; Li and Kraslawski, 2004; Westerberg,
2004; Barnicki and Siirola, 2004; Grossmann and Guilln-Goslbez, 2010). Accordingly,
the synthesis methodologies can be basically categorized into three groups, which are
complementary to each other: a) heuristic methods, b) targeting or task-oriented meth-
ods, and c) mathematical optimization based methods.
22 analysis, optimization and synthesis of coal-fired power plants

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.

2.4.1 Superstructure-based synthesis

The superstructure explicitly defines a priori structural space to mathematically formu-


late the synthesis problems. The superstructure concept was first proposed by Du-
ran and Grossmann (1986) to describe the outer approximation algorithm for solving
(MINLP), and was initially illustrated for addressing process synthesis issues in HEN
(Yee and Grossmann, 1990). Later, the synthesis concept was generalized as a system-
atical superstructure-based synthesis method (Grossmann et al., 2000; Westerberg, 1991;
2.4 synthesis of energy systems 23

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.

2.4.1.1 Superstructure representation


A (super)structure can be presented in forms of string, connectivity matrix or graph,
such as digraph, signal-flow graph, P-graph (for these three types, see Friedler et al. 1992)
and S-graph (Emmerich et al., 2001). The string-based representation is favorable for ap-
plying replacement rules (grammars), such as in a string rewriting system (Book and
Otto, 1993) for HEN (Fraga, 2009); however, the grammars tend to be too complicated
for presenting detailed flowsheets. The connectivity matrix, digraph and signal-flow
graph are only suitable for process analysis, e. g., the matrix representation in struc-
tural theory of thermoeconomics (Valero et al., 1993), but may become ambiguous for
variable structures. P-graph (Friedler et al., 1992) represents the structure of a process
system in a unique and mathematically rigorous form, while S-graph (Emmerich et al.,
2001) is more suitable for representing a detailed flowsheet. Current software status
(see Lam et al., 2011) allows for modular graphical representation of a flowsheet, e. g.,
(Wang et al., 2014; Bertok et al., 2012a).

2.4.1.2 Superstructure generation


For most applications, the superstructure considers only a limited number of promis-
ing alternatives, which may be generated by knowledge-based methods, such as heuris-
tic rules (Douglas, 1995; Kravanja and Grossmann, 1997) and thermodynamic insights
(Jaksland et al., 1995; Hostrup et al., 2001). Great efforts, e. g., (Fraga, 1998), have been
made to enhance user-friendly generation. However, the generation procedure usually
requires trivial manual interactions and specifications. More importantly, many good
alternatives may be left out of the solution space spanned by the superstructure.
In principle, an excessively large superstructure can include as many good alter-
native as possible. However, it may encompass also a large number of meaningless or
even infeasible alternatives, which potentially lead to forbiddingly large computational
effort, as the computation complexity and difficulty of the optimal synthesis problems
almost always increase exponentially with the number of components considered in
the superstructure (Friedler et al., 1992; Seferlis and Grievink, 2001; Matelli et al., 2009).
In addition, for realistic problems, the number of structural alternatives tends to be
very large, e. g., over 109 structural enumerations for the feed-water preheating train
of thermal power plants (Hillermeier et al., 2000). Considering current computation ca-
24 analysis, optimization and synthesis of coal-fired power plants

pability of mathematical programming, it is basically not possible to take all possible


alternatives into account.
To cope with some of these fundamental problems, many systematic or even algo-
rithmic generation of superstructure have been developed. Toward systematic genera-
tion, there are stage-wise synheat superstructure for HEN (Daichendt and Grossmann,
1994), multi-level hierarchical aggregation (Kravanja and Grossmann, 1997; Daichendt
and Grossmann, 1998; Manninen and Zhu, 2001), state-task and state-equipment network
(Yeomans and Grossmann, 1999) for process systems, or decision tree (Chen et al., 1997;
Hillermeier et al., 2000) for power plants.
The algorithmic generation of superstructure automatically and systematically ig-
nores structurally infeasible structures. The most prominent algorithms (Friedler et al.,
1993, 1995) are based on the P-graph representation. The P-graph framework explores
the combinatorial nature of considered technical components and minimizes the num-
ber of components in the maximal structure (Friedler et al., 1993). Therefore, the com-
plexity of the superstructure is reduced. The P-graph framework was originally pro-
posed for synthesizing chemical processes and has been deployed to a wide range of
synthesis problems (see Vance et al., 2012). Detailed implementation of the framework
is introduced by Bertok et al. (2012a). The disadvantage of the original framework,
however, is that multiple redundant instances of one type of technical components are
not considered. Recently, a combinatorial algorithm was proposed to add necessary
redundancy of supply chains (Bertok et al., 2012b).
To enable the automated synthesis of distributed energy supply systems, Voll et al.
(2013) proposed a superstructure generation algorithm based on the P-graph frame-
work. The algorithm first generates a maximal structure considering all feasible types
of components. Then, the maximal structure is successively expanded by adding multi-
ple redundant components, which is achieved by manipulating the connectivity matrix.
However, limited by the matrix representation, the connections of the newly added re-
dundant component are identical to the already existing component of the same type.
Although these methods make superstructure generation an easy task for certain
processes, there are more challenges for complex energy systems, particularly thermal
power plants: A complex flowsheet comprises only several types of components, which
indicates that multiple redundant components are always involved with different con-
nections. Additionally, one task may be fulfilled by several sequentially or parallelly
connected components of the same or different types. Thus, it seems these generation
methods are not adequate for such applications.

2.4.1.3 Superstructure-based modeling and solving


The superstructure is usually modeled by introducing binary selection variables to
allow the activation/deactivation of each considered component, as reviewed in (West-
erberg, 2004; Biegler and Grossmann, 2004; Grossmann and Guilln-Goslbez, 2010).
Following section 2.3, such superstructure-based problems are generalized as (MILP),
(MINLP) or (GDP):
min f (s, d, o ), "objective function",
s, d, o

s.t. h(s, d, o ) = 0, "equality constraints", (2.3)


g(s, d, o ) 0, "inequality constraints",
2.4 synthesis of energy systems 25

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).

2.4.2 Superstructure-free synthesis

As mentioned, the fundamental problems of superstructure-based optimization re-


mains: On the one hand, good alternatives (in particular, the optimal solution) might
be excluded from the superstructure, while on the other hand a large number of mean-
ingless or even infeasible alternatives may be considered. To overcome these problems,
superstructure-free approaches apply metaheuristic algorithms to explore a practically
unconstrained solution space, which is not limited a priori by a superstructure model.
In fact, back to 1970s, Stephanopoulos and Westerberg (1976) have outlined a crucial
view of the evolutionary synthesis: Given an initial structure and rules to systemati-
cally adjust the structure with small changes, an effective strategy applying the rules
produces neighbor structures and thus "enumerates" all feasible structures, in which
the optimal structure lies. Based on this idea, Seader and Westerberg (1977) synthe-
sized a simple separation sequence. Modern superstructure-free approaches, however,
apply metaheuristic algorithms, which perform "intelligently" and stochastically, thus
a large number of unpromising structures are not taken into account. Two-level hybrid
algorithms are always involved: the upper level manipulates the structural represen-
26 analysis, optimization and synthesis of coal-fired power plants

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.

2.4.3 Synthesis of steam cycles

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.

2.5 Multi-objective optimization


Real-world optimization and synthesis of energy systems always involves more than
one criterion (objective), e. g., efficiency, cost and environmental impact. These crite-
ria usually become, for a set of solutions, conflicting with each other, i. e., improving
one objective worsens at least one other objective. This solution set is named Pareto set
or Pareto front, while each solution on the front is called Pareto solution. Pareto solu-
tions present trade-offs among different objectives, thus enable decision making more
comprehensively.
The techniques to obtain such Pareto front, multi-objective (or multi-criterial) opti-
mization (MOO) techniques, have been detailed reviewed elsewhere, e. g., (Marler and
Arora, 2004; Deb, 2014). A multi-objective optimization problem is formulated:

min f ( x ) = ( f 1 ( x ), f 2 ( x ), . . . , f k ( x ))T , (2.4)


x
s.t. x X,

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.

2.5.1 Multi-objective optimization techniques

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)).

2.5.1.1 The weighted sum method


The weighing method reformulates multiple objectives into a single super-objective F:
k
min
x
F(x) = wi f i ( x ), (2.5)
i =1
s.t. x X,

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.

2.5.1.2 The e-constraint method


The e-constraint method splits the objective function space into many sub-spaces by
introducing upper or lower bounds for all objective functions but one. Each sub-
space represents a single-objective optimization problem. Thus, this method trans-
forms MOO into a set of single-objective problems. The optimal solution of each sub-
problem is a Pareto solution. The generic e-constraint method is formulated as follows:

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

Figure 2.8: Graphical representation of the e-constraint method

2.5.1.3 The normalized normal constraint method


The normalized normal constraint method (Messac et al., 2003) is for generating a set of
evenly-spaced Pareto solutions. The method introduces normal lines (Fig. 2.9) to divide
the objective space instead of the horizontal or vertical lines in the e-constraint method
(Fig. 2.8). After obtaining the anchor points 1 and 2 , an Utopia line is established and
divided by evenly-spread points, i. e., 1 5 . For each of the points, 1 5 , a normal line
of the Utopia line is added, e. g., NU1 for 1 and NU2 for 2 . Thus, the objective space
is divided into many sub-spaces, 5 for the example in Fig. 2.9. For each subproblem, the
space above the corresponding normal line is the feasible region; while the remaining
objective space becomes the infeasible region. The optimal solution of each subproblem,
e. g., s1 s5 , is a Pareto solution. The advantage of the method is its ability of generating
a set of well-distributed Pareto solutions, even those on the non-convex regions of the
Pareto frontier.

2.5.1.4 Evolutionary multi-objective algorithms


The population-based feature alone endows evolutionary multi-objective algorithms a
tremendous advantage in solving MOO, i. e., generating a population of Pareto solu-
tions in one single run of an evolution algorithm. In addition, good spread of Pareto

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

solutions may be easily achieved, if non-dominated solutions can be identified and


preserved with diversity in each generation. There are many algorithmic variants (see
Deb, 2014). In the following, the well-established non-dominated sorting genetic algorithm
II (NSGA-II (Deb et al., 2002)) and recently proposed S-merit selection EMOA (SMS-
EMOA (Beume et al., 2007)) are introduced with respect to the strategies of finding
non-dominated fronts and selecting proper solutions.
The NSGA-II applies a fast non-dominated sorting to assign non-domination ranks.
For each solution i in a population, the number of solutions that dominate solution i
(ni ) and the set of solutions dominated by solution i (Si ) are calculated. The solutions
not dominated by any other solution (ni = 0) are identified as the best non-dominated
front F0 . Then, a loop is started to classify all solutions into different non-dominated
fronts: For each solution i in a known front Fr with non-domination rank r, each
dominated solution j in the solution Si is manipulated by n j = n j 1; meanwhile, if
the number n j becomes 0, the solution j is picked as a member of the neighboring
non-dominated front Fr+1 .
The NSGA-II assigns crowding distance () to a set of solutions for estimating the solu-
tion density. For each objective f m , the solutions in a front are sorted in a descending or-
der, among which the two with the largest and smallest objective f m are specified with
an infinity distance; while, for each remaining solution i, its distance with respect to
the objective f m is defined by its two neighbor solutions: m ( xi ) = f m ( xi1 ) f m ( xi+1 ).
Therefore, after looping all k objectives, the crowding distances of solutions in all non-
dominated fronts are known: ( xi ) = km=1 m ( xi ).
The NSGA-II merges the offspring and parent populations first and assigns non-
domination ranks and crowding distances to all solutions. For selection, the solutions
closer to the best non-dominated front (with lower non-domination ranks) and those
less densely spread (with the same non-domination rank but larger crowding distance)
are preferred. Thus, the NSGA-II aims at generating uniformly distributed solutions
on the whole Pareto front.
In comparison with NSGA-II, SMS-EMOA aims at generating only the part of the
complete Pareto front, which includes good compromise Pareto solutions (Fig. 2.10).
This method is expected to be more computationally efficient for practical problems.

Reference point
Objective function 2

Hypervolume of
solution

Complete Pareto front


Pareto solutions with (NSGA-II)
better compromise (SMS-EMOA)
Objective function 1

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)).

2.5.2 Multi-objective optimization of coal-fired power plants

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.

2.6 Contribution of the thesis


Synthesis and evaluation are at the heart of overall system design of thermal sys-
tems. The synthesis methods enable the engineers to create novel conceptual system
designs, which are then evaluated with respect to various criteria for suggesting fur-
ther improvements. The review in this chapter reveals the shortcomings of available
methods for the evaluation and optimization-base synthesis of pulverized-coal power
plants, and the insufficiency in the literary applications. In this thesis, a comprehensive
exergy-based evaluation method and two optimization-based synthesis methodologies
are introduced, further developed or extended, and applied for the improvement of
pulverized-coal power plants. Thus, the scope of the thesis goes beyond the evaluation
and synthesis applications by addressing also methodology development.

comprehensive evaluation method for coal-fired power plants and


its application In this thesis, a state-of-the-art large-scale ultra-supercritical pul-
verized coal power plant is evaluated in a deep and complete fashion, based on the
conventional and advanced exergy-based evaluation methods. The whole system of
32 analysis, optimization and synthesis of coal-fired power plants

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.

automated generation of structural alternatives In this thesis, two


approaches for automated optimization-based synthesis are further developed for syn-
thesizing pulverized-coal power plants: the graphically-superstructure-based and the
superstructure-free synthesis approach.
The superstructure-based synthesis approach generates structural alternatives based
on a graphical superstructure, which considers available features that can enhance the
performance of a pulverized-coal power plant. Each structural alternative is mapped
to an integer-variable representation. Problem-specific mutation and crossover opera-
tors are developed for automatically generating new structural alternatives. Combined
with additional algorithms to generate proper boundary conditions for each structural
alternative, the differential evolution realizes a high generation rate of feasible solution
structures and identifies the optimal solution along with many near-optimal solutions.
The superstructure-free synthesis approach generates structural alternatives based
on a hierarchically-structured energy conversion hierarchy and a minimum set of mu-
tation rules. The approach is originally developed for distributed energy supply sys-
tem. However, for the synthesis of more complex and difficult thermal power plants,
the superstructure-free approach is further extended by a new insertion mutation rule.
The insertion rule allows for the addition of technology that are not already exist-
ing in parent solutions. Particularly, the insertion rule generates serial connections of
components with different functions. Combined with the energy conversion hierarchy
developed for thermal power plants in this work, the approach is capable of automati-
cally generating solution structures of pulverized-coal power plant, but avoids a priori
definition of a superstructure and technology-specific replacement rules.

multi-objective decision support The system-level design of thermal sys-


tems confronts usually not a single objective but multiple (contradicting) objectives.
Traditional multi-criterial trade-offs (a set of Pareto optimal solutions) enable the de-
signer identify the most preferable solution; however, other fronts (not Pareto front)
contain useful information as well for the design of thermal systems. In this thesis, tra-
ditional multi-objective selection method, NSGA-II is further extended for generating
2.6 contribution of the thesis 33

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.

3.1 Framework for comprehensive exergy-based evaluation


The framework (Table 3.1) is based on recent developments in exergy-based analy-
sis. The aims are to evaluate the exergetic and exergoeconomic performances of each
component and the whole system, the (endogenous or exogenous) avoidable exergy
destruction within each component, the interactions among different components, and
the avoidable (endogenous or exogenous) investment cost (or exergy destruction cost).
Therefore, it provides not only the locations and magnitudes (the spatial distribution)
of thermodynamic inefficiencies, but also their causes, sources and avoidability. There-
fore, this framework provides more valuable information than conventional exergy-
based analysis on how to improve the design of thermal systems in reasonable and
effective ways.
As discussed in section 3.1.2, the key idea of the framework is the reasonable split-
ting of exergy destruction within each component and the investment costs of different
components, according to their sources (endogenous and exogenous) and avoidabil-
ity (avoidable and unavoidable). The splitting of endogenous and exogenous parts
relies on the nature of thermal systems: The components in a system are not isolated

35
36 comprehensive exergy-based assessment of coal-fired power plants

Table 3.1: Framework of comprehensive exergy-based evaluation

Approach (the corresponding section) Key terms

Exergy analysis (section 3.1.1) E L,tot , E D,tot , tot , E D,k , k


Advanced exergy analysis (section 3.1.2) E EN , E EX , E AV , E AV,EN , E AV,EX
D,k D,k D,k D,k D,k
Exergoeconomic analysis (section 3.1.3) Z k , cF,k , cP,k , C D,k , rk , f k
Advanced exergoeconomic analysis (section 3.1.4) Z EN , Z EX , Z AV , C EN , C EX , C AV
k k k D,k D,k D,k

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)

Type and description Aim Time

0: All components under real conditions E D,k , E P,k , Z k , cF,k , cP,k 1


I: Component k under real condition, the
E D,k
EN , E EN
P,k Nc
remaining under their theoretical conditions
II: Component k under unavoidable condition UN UN
E D / E P k
E P
, Z/ k
Nc
and in isolation (isolated from the system)
III: Components k and r under real conditions, EN,k +r EN,k +r
E D,k , EP,k 2 N
CN c
others under their theoretical conditions c

3.1.1 Exergy analysis

Conventional exergy analysis identifies the location and magnitude of thermodynamic


inefficiencies in thermal systems. Thus, the components with large exergy destructions
or low exergy efficiencies can be pinpointed. The equations and variables involved in
an exergy analysis include

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

tot = E P,tot / E F,tot , (3.2)


yD,tot = E D,tot / E F,tot , (3.3)
3.1 framework for comprehensive exergy-based evaluation 37

yL,tot = E L,tot / E F,tot ; (3.4)


for the kth component E F,k E P,k = E D,k , (3.5)
k = E P,k / E F,k , (3.6)
yD,k = E D,k / E F,tot , (3.7)

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.

3.1.2 Advanced exergy analysis

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).

3.1.3 Exergoeconomic analysis

A complete exergoeconomic analysis consists of an exergetic analysis (section 3.1.1), an


economic analysis and an exergoeconomic evaluation. The economic analysis estimates
the major costs associated with the plant construction and operation to calculate cost
of electricity (COE) and the levelized cost rates assigned to each component (Z k ). The
cost estimation starts with the calculation of purchased-equipment cost (PEC) of each
plant component (Tsatsaronis and Park, 2002), for which detailed estimation methods
are presented in (Smith, 2005; Turton et al., 2008). With calculated PECs, the levelized
38 comprehensive exergy-based assessment of coal-fired power plants

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):

FCL + CCL + OMCL


COE = , (3.19)
Pnet Naoh
CCL + OMCL PECk
Z k = (3.20)
Naoh PECk
where the terms Pnet , Naoh and are net power output, annual operation hours and
annual capacity factor.
The exergoeconomic evaluation first calculates the specific costs of all exergy flows,
based on cost balance equations of each component and the associated auxiliary equa-
tions. The cost balance of a productive component k can be generally given as:
Nin Nout
E c + Z k = E c
 
in,i out,j
, (3.21)
i =1 j =1

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:

cF,k = C F,k / E F,k , (3.24)


cP,k = C P,k / E P,k , (3.25)
C D,k = cF,k E D,k , (3.26)
rk = (cP,k cF,k ) /cF,k , (3.27)
f k = Z k / Z + C D .

k
(3.28)

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:

C P,tot + C L,tot + C dif,dc,tot


COE = , (3.29)
Pnet

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.

3.1.4 Advanced 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

Z kEX = Z kEX,r + Z kMX , (3.38)


r
Z kEX,r = EEN,k+r E P Z kEN ,

P,k Z/ k
(3.39)
Z kUN,EX,r = EEN,k+r Z/ E P UN Z kUN,EN ,

P,k k
(3.40)
Z kAV,EX,r = EX,r
Zk Zk UN,EX,r
. (3.41)

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.

3.2 Calculation of endogenous and unavoidable exergy destructions


The determination of endogenous and unavoidable exergy destructions within each
component is the key issue in an advanced exergy analysis. The calculation of un-
avoidable part is much easier, as the component in regard is isolated from the system.
The product exergy of each considered component is specified as the same as that of
the real case. For the unavoidable condition, the component requires less fuel exergy,
which leads to a change of its investment cost as well.
For the endogenous/exogenous splitting, one simulation of type I (Table 3.2) for
each component, in which the considered component works with the exergy efficiency
under its real condition while the remaining components are under theoretical con-
ditions, has to be constructed. The approaches previously developed for calculating
endogenous exergy destructions (cf. section 2.2.3.2) encounter conceptual problems
when analyzing specific components. For instance, the thermodynamic cycle approach
cannot handle chemical reactors and becomes not flexible for open systems; the engi-
neering approach, however, is not available for dissipative components. Moreover, the
handling of a series of heat exchangers which directly connects to each other is rather
tricky, since setting a heat exchanger to its theoretical condition may defect its succeed-
ing heat exchanger: The temperature of an outgoing hotside stream of a heat exchanger
working theoretically may be below the temperature of the cold-side streams of its suc-
ceeding component. In other words, the ideally operated productive heat exchangers
or chemical reactors cannot be built in a complete flowsheet simulation. Therefore,
more flexible approaches which avoid the simulations of ideal heat exchangers and
chemical reactors are needed.

3.2.1 A new approach for calculating endogenous exergy destruction

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

Boiler subsystem Turbine subsystem


slag
10 2 4
coal .
coal Pnet
reheat steam 3 turbine
8 combustion 5
9 6
feedwater
air
air feedwater

pump
preheater 1
preheater
11 exhausted fluegas 7

Figure 3.1: Flowsheet of an ideal coal-fired power plant

thermodynamically-reversible operated blackbox, which makes the determination of


each endogenous exergy destruction fairly easy. Note that the ideal operation of the
blackbox scales the mass flow rates of all streams and may change the thermodynamic
properties of streams flowing into and out of the considered component. Based on this
general concept, this section focuses on a more complex system, which involves both
an open subsystem and a thermodynamic cycle.
A condensing coal-fired power plant comprises an open boiler subsystem for heating
feedwater and reheating steam, and a Rankine cycle for generating power, compressing
and regenerating feedwater. These two subsystems are linked only by the feedwater
and reheat steam. Thus, the ideal plant is given as Fig. 3.1, with the gross exergy
balance equations as follows:

m 2 (e2 e1 ) + m 4 (e4 e3 ) = m 8 e8 + m 9 e9 m 10 e10 m 11 e11 , (3.42)


m 2 (e2 e1 ) + m 4 (e4 e3 ) = Pnet . (3.43)

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

A dissipative condenser is necessary to ensure a full Rankine cycle. The endogenous


exergy destruction (or loss if emitted to the environment) can be calculated according
to Fig. 3.2b:

(e2 e1 ) m 2 + (e4 e3 ) m 4 (e33 e32 ) m 32 E D,COND


EN
= Pnet , (3.47)
e15 m 15 + e16 m 16 e17 m 17 (e33 e32 ) m 32 E D,COND
EN
= 0, (3.48)
m 15 h15 + m 16 h16 e17 m 17 (h33 h32 ) m 32 = 0. (3.49)

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

The productive purpose of a feedwater preheater is to heat the feedwater to a specific


temperature. The model for calculating its endogenous exergy destruction is given
based on Fig. 3.2c:

(e2 e1 ) m 2 + (e4 e3 ) m 4 E D,FWH


EN
= Pnet , (3.50)
(e22 e21 ) m 22 + E F,FWH
EN
E D,FWH
EN
= 0, (3.51)
(e22 e21 ) m 22 + FWH E F,FWH
EN
= 0, (3.52)

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 .

Pump, compressor, and fan

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:

(e2 e1 ) m 2 + (e4 e3 ) m 4 E D,PUMP


EN
= Pnet , (3.53)
(e19 e18 ) m 18 + E F,PUMP
EN
E D,PUMP
EN
= 0, (3.54)
(e19 e18 ) m 18 + PUMP E F,PUMP
EN
= 0, (3.55)

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

exergies e1 , e2 , e3 and e4 can be readily determined by removing all pressure losses in


the boiler subsystem. The specific exergy e19 , however, has to be derived, based on the
pump model in App. A, from the new pressure level p19 after removing all successive
pressure drops in the real case. Finally, the above equation set is closed for m 2 , E D,PUMP
EN

and E F,PUMP
EN . The same considerations are taken also for compressors.

Fluegas-water/steam heat exchanger

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:

(e2 e1 ) m 2 + (e4 e3 ) m 4 = Pnet , (3.56)


(e27 e26 ) m 27 + E F,HE
EN
E D,HE
EN
= 0, (3.57)
(e27 e26 ) m 27 + HE E F,HE
EN
= 0. (3.58)

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:

e8 m 8 + e9 m 9 e10 m 10 e11 m 11 E D,APH


EN
= Pnet , (3.59)
(e31 e9 ) m 9 + E F,APH
EN
E D,APH
EN
= 0, (3.60)
(e31 e9 ) m 9 + APH E F,APH
EN
= 0, (3.61)

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 .

3.3 Evaluation of a modern ultra-supercritical plant


The plant selected for the evaluation (Fig. 3.3) is an existing 1000 MW-level ultra-
supercritical, single-reheat, condensing PCPP with key design specifications listed in
Table 3.3. A bituminous coal is fired with the ultimate analysis (wt% as received basis)
of C (57.37), H(4.19), O (7.57), S (0.87), N (1.40), moisture (21.30), ash (7.30) and the
3.3 evaluation of a modern ultra-supercritical plant 45

Table 3.3: Basic specification of the pulverized-coal fired power plant under real conditions

Terms Value

Conditions of main steam 274.3 bar/605 C


Conditions of steam to be reheated 61.8 bar/369 C
Conditions of reheated steam 53.1 bar/600 C
Condenser pressure 0.062 bar
Final feedwater preheating temperature 295.6 C
Temperature of exhausted flue gas 129.5 C
Gross/net power output 1012.6/985.9 MW
Net energy/exergy efficiency 42.43%/40.28%

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.

3.3.1 Exergy analysis

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.

3.3.1.1 Boiler subsystem


The furnace, where the fierce chemical reaction and radiation heat transfer occur, dom-
inates the total irreversibility in boiler subsystem with its exergetic efficiency as low as
44% (Table B.5). Almost 1/4 of the fuel exergy is destructed in the combustion process.
The exergy efficiencies of heat exchangers depend on the temperature levels of cold
and hot streams (Fig. 3.4), i. e., the thermodynamic average temperatures of heat ab-
sorption and release: Ta = h/s. The water wall, in which radiation heat transfer
is prevailing, performs badly, compared with the convection heating surfaces, such as
FSH and PSH. The radiation heat transfer plays the leading role only when the flue gas
temperature is extremely high, meanwhile causing a large temperature difference be-
tween flue gas and working fluid. The exergy efficiency of the water wall achieves only
1 Ebsilon: www.steag-systemtechnologies.com/ebsilon_professional.html, accessed on 29 May 2015.
2 Engineering Equation Solver: www.fchart.com/ees, accessed on 29 May 2015.
46
comprehensive exergy-based assessment of coal-fired power plants
53

CV
8

12 FSH 26
52 FRHH 11

SSH2 7 127 128


HPT1-2 LPT1-3 LPT4-6
124
51 FRHC IPT1-3
1012.645 MW
15 19
8 9 10 11 14 18
F 2 3 4 13 17
6 12 16
U SSH1 129 166
156 159 162 165
R G
N 130 133 PRH2 EG
5
A
50 81 82
C
E
49 PSH
54 59 64 69 76 83 87 15 91 19
PRH1

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

Figure 3.3: Schematic of the ultra-supercritical pulverized coal power plant


3.3 evaluation of a modern ultra-supercritical plant 47

2500 100

2000 80

Exergy efficiency / %
Exergy efficiency
Temperature / C

1500 Fluegas temperature 60

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

3.3.1.2 Turbine subsystem


The turbines working with superheated steam perform with little thermodynamic in-
efficiencies, leading to exergetic efficiencies over 90%. However, those driven by wet
steam (LPT3 and LPT6) endure worse working conditions with exergy efficiencies be-
low 80%. This is mainly due to the wetness loss, the tip clearance loss as well as the
residual speed loss of the last turbine stage.
The exergetic performances of regenerative preheaters improve steadily along the
feedwater flow (Fig. 3.5), because of the increase in feedwater temperature. However,
FWH47 present lower exergy efficiencies than expected, due to larger temperature dif-
ferences for heat transfer. In the 4 feedwater preheaters, the benefit from the increase in
feedwater temperature is partially counteracted by increased temperature differences.
More insights on the performances of heat exchangers are given in the following (sec-
tion 3.3.1.3).

3.3.1.3 Further discussion on exergetic performances of heat exchangers


As mentioned above, the temperature level of cold fluid and the temperature differ-
ence for heat transfer are two key factors characterizing the exergetic performances of
different heaters. The exergy efficiency of a productive heat exchanger can be generally
written as follows:

E cold Q HE (1 Tref /Tacold ) 1 Tref /Tacold


HE = = = , (3.62)
E hot Q HE (1 Tref /Tahot ) 1 Tref /Tahot
or

E cold E cold / Q HE m cold ecold / m cold hcold



HE = = =
E hot E cold / Q HE m hot ehot / (m hot hhot )
1 Tref (s/h)cold
p 1 Tref /Tacold
= = . (3.63)
1 Tref (s/h)hot
p
1 Tref /Tahot

100

0.5 A Average temperature difference Ta / C


90
Exergy efficiency / %

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

Substituting Tahot by Tacold + Ta , Eq. 3.62 or 3.63 can be rewritten as follows:

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.

3.3.1.4 Distribution of exergy destruction and losses


The exergy destruction within the boiler subsystem (73%) dominates the overall exergy
dissipation (Fig. 3.7a), followed by that within the turbine subsystem (over 14%) and
the total exergy loss (over 12%). Therefore, the boiler subsystem may be where the
largest energy-savings potential exists.
The cold ends of both boiler (exhausted flue gas) and turbine (cooling water) subsys-
tems contribute almost equally to the overall exergy losses (Fig. 3.7b), 47% and 49%,
respectively. The remaining exergy loss is contributed by boiler slag. The efficient uti-
lization of the large amount of exergy from waste flue gas and cooling water should
be further investigated for the energy-savings potential.
The largest part of exergy destruction within boiler subsystem comes from coal com-
bustion (57%). Another important source (around 30%) is from the water wall with
radiation heat transfer (Fig. 3.7c). The exergy destructions within other heat exchang-
ers are much smaller: The remaining heat exchangers for generating main steam lead
to nearly 6% of boilers total exergy destructions, while this figure reduces to 4.4% for
reheaters. Interestingly, the remaining 3.2% of boilers exergy destructions are caused
within the single APH. Therefore, much more attention should be paid to the improve-
ment of exergy transportation process from the fuel exergy to the physical exergy of
steam. Novel ways of heating combustion air may be favorable for reducing the ineffi-
ciency within the APH.
50 comprehensive exergy-based assessment of coal-fired power plants

a) 0.7% EG b) 3.2% Slag

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.

3.3.2 Exergoeconomic evaluation

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

preheaters are far less exergy-destruction-cost-dominant with exergoeconomic factors


(38%), while those of FWH68 are within 4080%.

3.3.3 Advanced exergy analysis

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.

3.3.4 Advanced exergoeconomic analysis

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.

3.4 Suggestions for improvement


The comprehensive evaluation of PCPP not only demonstrates where and to what ex-
tend the system and the components can be enhanced, but also suggests favorable
sequence of improving an individual component or subsystem: The enhancements of
EG must be given a high priority. The fluids pressure drops should be kept as low as
possible. Only when the mechanical energy is saved as much as possible can the im-
provements of other components be more efficient and practical. Then, the STs should
be improved at the subsystem level, due to high interactions among them and also
their large contribution to the overall avoidable exergy destruction costs. Subsequently,
for the investment-cost-dominant boiler subsystem, the effort should be dedicated to
54 comprehensive exergy-based assessment of coal-fired power plants

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.

3.5 Summary and conclusions


In this chapter, one of the most advanced PCPPs has been analyzed comprehensively
by both conventional and advanced exergy-based methods. In contrast to those previ-
ous analyses of PCPPs, the boiler of the PCPP is simulated in detail by a coal combustor
and a series of heat exchangers. The exergoeconomic performance of each component,
and the dissipation of exergy destructions and costs within the whole system are eval-
uated. For the advanced analysis, a new method to calculate the endogenous exergy
destructions and costs is introduced and applied to the PCPP. Thus, exergy-saving
and cost-saving potentials of each component are identified into endogenous and ex-
ogenous parts. Accordingly, suggestions are given for improving the system. Main
conclusions include:
Three performance ranges for all heat exchangers are distinguished according
to the (thermodynamically average) temperature levels of hot and cold fluids.
The heat transfer occurring under low temperature levels or large temperature
difference leads to low exergy efficiency of the heat exchanger, e. g., the first low-
pressure FWH, APH and WW. Other heat exchangers perform much better with
exergy efficiencies over 80%.
The overall exergy dissipation come mainly from the exergy destruction of the
boiler subsystem (over 70%), especially coal combustion and WW. The contribu-
tions of exergy destructions within the turbine subsystem and the total losses
are similar to each other. Exergy is mainly lost in the exhausted flue gas and the
cooling water.
For the boiler subsystem, the FUR is with extremely high exergy destruction cost,
while the remaining heat exchangers are with high investment cost. Future design
of the boiler should achieve less capital investment and also exergy destruction.
For STs, the turbines working with wet steam are exergy-destruction-cost domi-
nant, thus inefficiencies within these STs should be reduced more than those in
other STs.
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. More than 70% of the exergy destruction within each boilers
heat exchangers (except APH) and STs is endogenous, while this figure reduces
to around 50% for APH and STs after reheat. Most boilers heat exchangers, IPTs
and LPTs have large avoidable exergy destructions.
Around 10% of both total investment and exergy destruction costs of the system
are avoidable. The boiler contributes the most avoidable investment cost, while
STs contribute the most avoidable exergy destruction cost. For boilers heat ex-
changers, STs and EG, most (over 60%) of the avoidable costs are endogenous,
while for pumps and fans the most parts are exogenous.
3.5 summary and conclusions 55

Analysis methods can evaluate thermodynamic inefficiencies of a specific system and


potentially guide the parametric optimization of the analyzed system. These methods
can assist the improvement of system flowsheet if combining with engineers expe-
rience and judgments. However, they cannot, at least until now, generate structural
alternatives automatically. Thus, in the following, the superstructure-based (chapter 4)
and superstructure-free (chapter 5) approaches for automatic generation of structural
alternatives are investigated for synthesis and optimization of PCPPs.
4
SUPERSTRUCTURE-BASED 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

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.

4.1 Sensitivity analysis of key variables on the cycle performance


The temperatures and pressures of main and reheat steams significantly affect the effi-
cient and safe operation of thermal power plants. Certain matches of these parameters
are necessary for a reasonable quality (x 0.9, Spliethoff 2010) of the exhausted steam
(i. e., steam to be condensed in the condenser). In particular, a reheat temperature that
is too high can lead for given pressures to overheating of the low-pressure turbine,
especially at partial load, while too low reheat temperatures result in condensed water
droplets in the exhaust steam, leading to blade erosion.
Although the calculation of the thermodynamically optimal combination of reheat
pressures and temperatures is a classical engineering task and has been discussed else-
where, for example, in (Silvestri et al., 1992; Habib et al., 1995). However, the constraints
for steam quality at the turbine exhaust were not described. Thus, a sensitivity of key
variables related to the steam cycle design on both cycle efficiency and exhaust steam
quality of a simple cycle (Fig. 4.1) was performed.

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

Temperature of steam out of reheat 2 / C

41 41
0.92

.4 .2
0.8 700 .8
Optimal pressure ratio of reheat 2 / -

0.8

40. 0.99 0.99


0.9

4
39 86
40

650 0.98 0.98


.2
0.
39

0.7 41
6.

600 0.96 0.96


0.6 Second-reheat zone
41
.4
40
.8 0. 550 0.94 40 40 0.94
0.5 88 39
0.99

.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

4.1 sensitivity analysis of key variables on the cycle performance


1 6
Block5 a f k Block7
(To Block6) (To Block8)
2 7 b g l

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

1 Block1 (to Block2): possible stream (1-11) of rh1


SW1 2 E1
3

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.

4.2 Graphical superstructure


A superstructure for the steam cycle with highly nonlinear features (see section 2.3.2) is
difficult to solve by deterministic MINLP algorithms: The branch-and-bound-like algo-
rithms require an optimal root solution (cf. section 2.3.1.1), which is hardly guaranteed
in case of highly nonlinear properties of water and steam. More importantly, the wide
temperature and pressure ranges of water and steam involved in modern PCPPs make
the commonly-used property approximations (section 2.3.2) rather complex, unstable
as well as inaccurate. Therefore, the superstructure of PCPP used in this chapter is not
built in algebraic modeling environments, e. g., GAMS, but by a professional simulator,
Ebsilon. The superstructure is then solved by an efficient heuristic algorithm, DE.
The graphical superstructure (Fig. 4.3) incorporates two-stage reheating, ten-stage
feedwater preheating (not including DA) and an ET (in this thesis a non-condensing
turbine, of which the exhausted steam flows directly to a FWH). All components in-
volved in the superstructure are well defined and properly connected to realize any
anticipated and featured system layout.
One specific structural alternative included in the superstructure can be active by
opening (or closing) certain stream paths and defecting (or activating) certain com-
ponents. In the superstructure, four stream-path switches SW1 to SW4 (Fig. 4.3) are
developed, which allow only one active path in each specific structural alternative
(Fig. 4.4). By proper control of these switches, the components not included in a spe-
cific structure are by-passed. In addition, once a stream path is open or closed, relevant
boundary conditions (temperature and pressure) need to be properly handled.
As an example, detailed performances of SW1 to SW3 are presented to activate a
reheating path: from stream1 in block1 to MT2. If the outlet steam of MT1 is selected
for reheating, the upper outlet of MT1 will be closed; meanwhile, the lower outlet of
MT1 (stream1 in block1) is active and the remaining streams (stream211) in block1
are closed. Afterwards, the cold reheat steam (stream1) flows through block2 to steam
generator. In block2, properties of the outlet stream are specified the same as those of
stream1 from block1. If an ET exists, part of the steam out of block2 may flow to the

SW1 SW2 SW3 SW4

OUT IN1 IN OUT1 IN1 IN1 OUT1


IN2 OUT2 OUT2
IN3 OUT3 IN2 IN2 OUT2

IN6 IN5 IN4 OUT6 OUT5 OUT4

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.

4.3 Statement of the optimization problem


For the synthesis and optimization at single design point, the values of objective func-
tions for a specific structural alternative () depend also on real-valued decision vari-
ables, i. e., free variables associated with material flows and component performance.
Therefore, for the superstructure (Fig. 4.3), the MINLP problem can be stated as finding
the best (optimal) structure , the best values and combination of all decision variables
to achieved the best objective values, i. e., plant efficiency and COE in this thesis:
min/max f (, tms , pms , pex,MT , trh1 , trh2 , , pMT , pET ), (4.2)
s.t. tmax min
ms tms tms ,
pmax min
ms pms pms ,
pmax min
ex pex,MT pex ,
tmax min
rh1 trh1 trh1 ,
tmax min
rh2 trh2 trh2 ,
max
i i min
i ,
pms pMT1 . . . pMTi1 pMTi pMTi+1 . . . pMT20 pex,MT ,
prh1 pET1 . . . pETi1 pETi pETi+1 . . . pET9 pMT19 ,
0.95 xex,MT 0.9, (type1),
xex,MT 0.9, (type2),
xex,ET 0.9,
xex,OT 0.9,
where the subscripts ms and ex stands for the main and exhausted steams; while the
vector contains the performance variables associated with all components. If only
62 superstructure-based synthesis and optimization of coal-fired power plants

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.

4.4 Multi-objective supported differential evolution


As mentioned in section 4.2, the superstructure is not solved by deterministic MINLP
algorithms but heuristic algorithms. Structural alternatives and continuous variables
are manipulated and optimized continuously by mutation and crossover operations,
while each generated alternative is simulated for solution evaluation.
Compared with genetic algorithm, DE is a simple and efficient algorithm (Price et al.,
2006). Extended by multi-objective selection techniques in NSGA-II (see section 2.5.1.4),
the algorithm is flexible for supporting multi-objective decision-makings.

4.4.1 Basic DE/MODE and the enhancement

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 ):

sni = sci + f (sai sbi ), i = 1 . . . Ns , (4.3)

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

a) Solution randomizer infeasible c) Merged pops (size: )


solution Ebsilon: evaluate solution
integer & real variables find
Evaluation The superstructure
feasible solution Split the pop to two sub-pops

Initial pop set stream path
set starting values
Subpop1 ( ) Subpop2 ( )
A specified structure
a b c Parent pop set real variables /2 ?
+ + Simulation fail
no

target succeed
solution
Crossover Extra constraints not fulfill a b c Sub-pop

Selection Evaluation fulfill +


not selected +

selected Feasible Infeasible random
solution
Crossover
Child pop
Selection Evaluation
not selected
next target solution
selected
Merge parent and offspring pops
Sub-pop
COE b)
Multi-objective selections F3
non-dominated sorting
crowding-distance assignment objective space Select /2 solutions separately
rank all solutions and do selection
Pareto front
from each sub-pop:
F0 non-dominated sorting
F1
new parent pop crowding-distance assignment
no
Termination? rank all solutions and do selection
yes
Desired solutions New parent population

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

4.4.2 Issues in the implementation of DE/MODE

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.

initialization and mutation of the feedwater-preheater train The


initialization for the feedwater-preheater train randomly employs g2 feedwater pre-
heaters, while for mutation the on/off condition of a FWH is replaced by that of a
randomly-selected preheater in the same solution:

1, i = int(floor( (imax imin + 1) + 1)), for initialization,
r
bi = (4.5)
b j , i 6= j, for mutation.

After mutation, if the number of employed feedwater preheaters is no longer equal to


g2 , extra feedwater preheaters are randomly selected to be employed or unemployed:

bi = 0 or 1, i = int(floor(r (imax imin + 1) + 1)). (4.6)

initialization and mutation of the reheat positions The initialization


and mutation for the reheat positions can be done as follows:

int(floor( cmax )), for initialization,
r rhi
ci = (4.7)
int(ccrhi + f (carhi cbrhi )), for mutation.

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.

4.4.2.1 Generating pressure array for turbine trains


The generation of reasonable pressure arrays for the turbine trains, which satisfy the
constraints in Eq. 4.2, is the greatest challenge for successful simulations of the super-
structure. Once the generated outlet pressure of certain turbine is smaller than the in-
let pressure (bound violation), the simulation will definitely fail. The situation always
4.4 multi-objective supported differential evolution 65

, and identical? A valid layout of feedwater preheaters and reheaters

New layout of feedwater preheaters a


yes Set outlet pressures of turbines with in- The turbines without
1/3
1/3
1/3 no service feedwater preheaters (incl. DA) in-service feedwater
Move rhi upstream Move rhi downstream preheaters

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

pnFWHi = pcFWHi + f ( paFWHi pbFWHi ), (4.8)


66 superstructure-based synthesis and optimization of coal-fired power plants

pnrhi = pcrhi + f ( parhi pbrhi ). (4.9)

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.

4.5 Design optimization for thermodynamic objective


The matches of the parameters of the main and reheat steams, the optimal pressure ra-
tios of reheating, the layout of feedwater regeneration and the thermodynamic benefit
of an ET are investigated at a single design point with regard to only the thermody-
namic objective, i. e., the plant efficiency.

4.5.1 Algorithm evaluation

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 4.1: Specifications of cases 1, 2, 3. (The brackets mean range)

Case pms /bar tms /C trh1 /C tfw /C ET Number of FWHs

1 (230, 500) 600 (550, 650) 300


7 FWHs + 1 DA
2 (230, 500) 600 605 300 7 FWHs + 1 DA
3 262 600 605 300 7 FWHs + 1 DA

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

Variable Value Variable Value Variable Value Variable Value

p,ms , p,rhi 13.0% s,ET 85.0% B 93.0% tup,hFWH 0.00 C


s,OT , s,FP 86.0% s,CWP 80.0% pFWH 0.00 bar tup,lFWH 2.78 C
s,CP , s,IDF 87.0% m 99.8% pCWP 3.33 bar tlow,FWH 5.56 C
s,MT 90.0% EG 98.6% pCOND 0.052 bar tup,COND 5.00 C
This value is 80% for the last stage (MT21 in Fig. 4.3).
This value is 1.7 C for FWH10.
4.5 design optimization for thermodynamic objective 67

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

Generation rate of feasible


80 20

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

4.5.2 Optimization for current steam conditions

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

44.8 Number of feedwater preheaters: 7

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

63.60 (rh1) 370 712.0 83.7 (rh1) 415 775


23.84 (O) 221 471 44.7 47.4 (O) 261 529 46.4
218 194
23.84 (E) 221 253 55.5 47.4 (E) 261 335 57.1
10.93 (O) 183 362 34.7 26.8 (rh2, O) 442 671
179
10.93 (E) 183 183 41.0 26.8 (rh2, E) 442 565
4.57 (O) 149 257 95.4 26.8 (O) 228 442 57.2
108 175
4.57 (E) 149 149 47.4 26.8 (E) 228 267 67.2
This value includes the mass flow rate 10.45 (O) 181 479 31.1
297
of the steam entering OT (53.0 kg/s). 10.45 (E) 181 182 39.7
4.411 (O) 147 359 85.5
This value includes the mass flow rate 212
of the steam entering OT (46.9 kg/s). 4.411 (E) 147 147 45.6

4.5.2.1 Single reheat


Compared with the reference plant efficiency of 44.41%, an additional 0.4 percentage
point (PP) increase in plant efficiency (case 3 in Fig. 4.8) is achieved by adjusting pres-
sures of steam extractions, the reheat position and the layout of feedwater preheaters.
The resulting optimal pressures of steam extractions are 83.72 bar/63.30 (rh1)/23.84/
10.93/4.57 (DA)/1.331/0.439/0.154. The usage of low-quality steam to heat the feedwa-
ter increases, resulting in the efficiency increase. The qualities of the exhausted steams
of MT and OT remain reasonable: 0.9 and 0.92, respectively. The optimal reheat pres-
sure is almost the same as that of the reference industrial design.
Replacing the OT with an ET (s,ET = 80%) for the above optimal solution leads to an
extra efficiency improvement of 0.15 PP (Fig. 4.11). A detailed comparison (Table 4.3a)
show the temperatures of the three steam extractions after reheating are reduced by
218, 179, and 108 C, respectively. The benefit of reducing the superheating degrees of
steam extractions after reheating overwhelms the efficiency decay caused by the low-
efficient steam expansion in the ET. The overall efficiency enhancement could be even
larger if employing a high-efficiency ET.
Increasing feedwater preheating temperature and adding more feedwater preheaters
are classical approaches to enhance the overall performance (Fig. 4.11): below 0.2 PP by
increasing the final feedwater temperature from 300 to 340 C under a given number of
feedwater preheaters, and less than 0.3 PP by adding three more feedwater preheaters
at a given feedwater preheating temperature. The efficiency increase by introducing
an ET is similar (around 0.15 PP) for different cases (Fig. 4.11) with different FWH
numbers and feedwater preheating temperatures.
70 superstructure-based synthesis and optimization of coal-fired power plants

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.

4.5.2.2 Double reheats


Case 3 (Table 4.1) is also optimized with double reheats under constraint type1 (see
Eq. 4.2). The optimal pressures of steam extractions and layout of feedwater preheaters
are 83.7 bar (rh1)/47.4/26.8(rh2)/10.45/4.411(DA)/1.322/0.445/0.153. The overall effi-
ciency reaches 45.08% without an ET and 45.36% with an ET: the efficiency increases
by 0.3 PP because of introducing the ET (see detailed comparison in Table 4.3b). Feed-
water preheating under large temperature differences is avoided. The superheating
degree of the first steam extraction after reheat rh2 is reduced by 300 C. Addition-
ally, the mass flow rate of the steam for the second reheat is reduced by almost 16%,
which would significantly reduce the boiler investment. Therefore, more economical
and secure benefits from employing an ET can be obtained for double-reheat units.
If 10 feedwater preheaters are employed and the final feedwater temperature reaches
340 C, the plant efficiency will achieve 45.48% even without an ET: The efficiency
is enhanced by 0.4 PP, compared to the optimal efficiency under 7 feedwater pre-
heaters (plus 1 DA) and a feedwater preheating temperature of 300 C. The pressures
of steam extractions becomes 146 bar/98.5(rh1)/46.1/26.7(rh2)/11.0/5.594(DA)/2.953/
1.626/0.735/0.344/0.150.

4.5.3 Optimization for higher 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.

4.5.3.1 Match of the conditions of main and reheated steams


The match of the conditions of main and reheated steams is discussed only for double-
reheat units by keeping 9 feedwater preheaters, 1 DA and a fixed feedwater preheating
temperature of 320 C. The MINLP problem is solved considering constraint type1 and
type2 separately (Fig. 4.12). As expected, the higher the temperature and pressure lev-
els of the main steam are, the greater the plant efficiency is. With steam conditions of
500 bar/750 C/750 C/750 C, the unit can reach an efficiency of 49.46%. More impor-
tantly, for constraint type1, the larger the throttle pressure, the greater the benefit that
can be obtained from the same increment in the temperature level. However, for con-
straint type2, the benefits gained from the same temperature-level increment remain
almost unchanged under different throttle pressures: given throttle and condenser pres-
sures, the higher the temperature level, the larger the deviation of the largest efficiency
zone from the reasonable range of the exhausted steams will be (Fig. 4.2a). Note that,
given a throttle pressure, the efficiency penalty caused by the upper limit of xex,MT
becomes larger when the steam temperature increases over 600 C (Fig. 4.12). The
main reason for the efficiency penalty is that, once the exhausted steam becomes su-
perheated, the temperature of the exhausted steam increases significantly. For the unit
4.5 design optimization for thermodynamic objective 71

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

Reheat pressures / bar


48 0.9

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:

ti ,pi = ati ln( pi ) + bti , (4.10)

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

Temperature level ti /C ati / bti / pt,ti /bar

750/750/750 -0.96 6.135 596


700/700/700 -0.73 4.576 527
650/650/650 -0.44 2.707 469

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.

4.5.3.2 Selection of reheat pressures


The reheat pressures are important for the design of steam cycles. Interestingly, when
considering a wide range of pressure and temperature, there are three different de-
sign zones for reheat pressures according to the pressures pt,ti and pMT1 (Fig. 4.13):
(A) If the pressure pms is far smaller than the corresponding pressure pt,ti , the optimal
pressure prh1 will be less than pressure pMT1 and the pressure prh2 is between 10 and
20 bar. In zone A, there would be a FWH above the first reheat point. (B) Further in-
creasing the pressure pms narrows variation range of pressure prh1 and the pressure
prh1 finally reaches the pressure pMT1 ; however, the pressure prh2 remains almost un-
changed (around 20 bar) for each temperature level. (C) If the pressure pms increases

0.4 300
type 1 (9 FWHs, 1 DA and tfw = 320 C) 650 C
Pressure ratio of the 2rd reheat / -

0.35 type 2 250


600 C
Reheat pressures / bar

0.3 tms = trh1 = trh2= 750 C


C 200
0.25 550 C C
A
0.2 150 700 C

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).

4.6 Design optimization for thermodynamic and economic objectives


The multi-objective optimization considering plant efficiency and COE is investigated
at single design point for both a fixed structure and the developed superstructure
(Fig. 4.3). In these optimization runs, the constraints on the quality of exhausted steam
(both constraint type1 and type2 in Eq. 4.2) are no longer considered.

4.6.1 Calculation of the cost of electricity for mathematical optimization

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.

4.6.2 On the fixed structure

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.

Table 4.5: Decision variables and the bounds

Decision Variables Symbol Bounds Unit

Temperatures of main and reheat steam tms , trh [550,700] C


Pressure of main steam pms [230,350] bar
Isentropic efficiency of turbines (including OT) s [0.75,0.95]
Outlet pressure of HPT1 pHPT1 [60,140] bar
Outlet pressure of HPT2, IPT13, LPT13 pTi [pTi+1 ,pTi1 ] bar
Outlet pressure of LPT4 pLPT4 [0.03,0.08] bar
Pinch temperature difference of COND tpin [2,6] K
Temperature difference of FWH (Table 4.2) tup [-3,3] K
Temperature difference of FWH (Table 4.2) tlow [0,7] K
Isentropic efficiency of pumps s [0.75,0.87]

4.6.2.1 Validation of the enhanced MODE


The fronts desired (not only the Pareto front) are smooth and continuous (Fig. 4.17):
The randomly generated initial solutions are dispersed almost uniformly within a wide
range. The dominated solutions are less likely to be preserved, thus the front gradually
gets populated more and more densely with increasing generation number. Moreover,
the advancement of the front is quite fast in the first tens of generations and tends
to be slow afterwards. The front is preliminarily visualized at the 30th generation
yet not smooth enough. With continued evolution, the solutions on the front become
4.6 design optimization for thermodynamic and economic objectives 75

5.9 5.9 4.2


ccoal = 0.3 /MJLHV ccoal = 0.3 /MJLHV
60 5.8
5.8 tsh = trh tsh = trh

Fuel and capital costs / /kWh


5.7 3.7
Cost of electricity / /kWh

Cost of electricity / /kWh


5.7
Initial generation 5.6 Fuel cost
5.6 150 5.5 3.2
5.5 5.4
1 Industrial design
5.4 5.3 (45.8, 5.2) 2.7
5.2
5.3
5.1 2.2
Capital cost
5.2 5
200 (only the front)
5.1 4.9 1.7
39 42 45 48 51 41 43.5 46 48.5 51
Plant efficiency / % Plant efficiency / %

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.

4.6.2.2 Contribution of fuel cost and capital investment cost to COE


The front obtained (Fig. 4.18) is quite classical (Bejan et al., 1996). The economically
optimal solution when the temperatures of main and reheated steams are set to be
equal to each other (s1 ) has a plant efficiency of 47.8% and a COE of 5.11 /kWh. This
solution splits the front to the left and right parts. In the left part, the COE decreases
significantly with increasing the plant efficiency. In the right part, the two objectives
oppose each other sharply (Pareto front). However, around the solution s1 , the two
objectives interact mildly: There are many design alternatives with similar thermody-
namical and economical performances, which introduces a certain freedom for the
designer.
The fuel cost and the capital investment cost contribute differently to the COE
(Fig. 4.18). The FC depends on the coal consumption and in turn on the plant effi-
ciency. For a plant with a single fuel, the relationship between FC and plant efficiency
is approximately linear (consistent with the mathematical feature of an inverse func-
tion). The decrease in fuel consumption contributes to a reduction of COE: the share
of FC in the COE declines from 66% to 55%, when the efficiency increases from 40%
to 50%. Thus, given a design of a PCPP, increasing the coal price by 5% can lead to
a increase in COE by 3%. The contribution of CC remains almost unchanged at plant
efficiency values below 44%. However, it increases gradually when the plant efficiency
increases above 44%, since more advanced materials for improving the component
performances are required.

4.6.2.3 Distribution of investment costs


The investment-cost shares of all components of the economically-optimal solution
with identical temperatures of main and reheated steams are different from each other
(Fig. 4.19). The boiler contributes more than half of the total investment costs, followed
76 superstructure-based synthesis and optimization of coal-fired power plants

5 1.6
3% ccoal = 0.3 /MJLHV

Levelized cost rate of the boiler / $/s


4.5 tsh = trh 1.4
Turbines

Levelized cost rates of other


2% 9% 4
1.2
3.5

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.

4.6.2.4 On the uncertainty of cost functions


For mathematical optimization, the quality and practical feasibility of an optimal solu-
tion highly depend on the accuracy of the modeling, especially the economic modeling,
because it is not always reasonable or even possible to express the PEC as a function
of associated variables (Bejan et al., 1996). For the cost functions in App. A, although
the coefficients have been adjusted according to (China Power Engineering Consulting
Group Corporation, 2009; National Energy Technology Laboratory, 2010), the limited
number of cases provided there makes it not possible to verify the validity and accu-
racy of these cost functions within the entire decision space.
As aforementioned, the boiler contributes more than half to the total PEC (Fig. 4.19).
Thus, it is reasonable to investigate the front uncertainty caused by the boiler. The cost
function of the boiler used here is expressed as PECB = a Fm F Ft Fp Fshrh , where a
is a constant. The factor F becomes a constant due to a constant boiler efficiency, while
the coefficients associated with the mass flow rate and pressure of main and reheated
steams, Fm and Fp , have been widely used (Uche et al., 2001; Xiong et al., 2012a; Ameri
et al., 2009) and thus are considered as reasonable. As a consequence, the uncertainty
caused by the remaining factors, the reheat-related and temperature-related coefficients
Fshrh and Ft , is investigated:

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

flow rate of main (reheated) steam


Pressure, temperature or mass
5.5 .
mms (kg/s) 800

Cost of electricity / /kWh


trh (C)
5.4 700

5.3 tsh (C) 600

5.2 tsh = trh 500

5.1 tsh trh 400

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.

Uncertainty caused by the reheat-related factor


In the current industrial designs (see chapter 3), the temperatures of main and re-
heated steams are usually kept nearly the same. However, the outlet temperature of
the reheated steam can also be a free decision variable (i. e., the temperatures of main
and reheated steams are not equal to each other). The temperatures of main and re-
heated steams determine the reheat-related factor, Fshrh (see equations for BOILER in
App. A), and affect the investment cost of the boiler.
The front obtained and major associated variables for the case with nonequal tem-
peratures of main and reheated steams are investigated (Fig. 4.22): The COE of the
front for the case with nonequal temperatures of main and reheated steams is much
lower than that with equal temperatures of main and reheated steams for the most
part but increases more sharply after the economically-optimal solution of the case
with nonequal temperatures of main and reheated steams (s2 : 48%, 5.0 /kWh). Based
on the given cost function of the boiler (see equations for BOILER in App. A), the re-
heat temperature should be kept as high as possible (700 C for all front solutions in
Fig. 4.22), while the temperature of the main steam is as low as possible for reaching a
given plant efficiency.

Uncertainty caused by the temperature-related factor


The uncertainty caused by the temperature-related factor Ft is because the prices of
new advanced materials can hardly be predicted. The temperature-related factor is
associated with three coefficients b, c and tref (see equations for BOILER in App. A).
Each coefficient determines a certain property of the front shape and position of the
economically-optimal solution (Fig. 4.23). For each optimization run of the case with
identical temperatures of main and reheated steams, only one of the three coefficients
is changed. Meanwhile, to ensure the COE of the given industrial design is kept as
4.6 design optimization for thermodynamic and economic objectives 79

a) 5.6 b) 5.6
a = 247583, c = 50, tref = 850
5.5 5.5
Cost of electricity / /kWh

Cost of electricity / /kWh


5.4 a = 156908, b = 10 5.4 a = 179397, c = 100, tref = 850

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.

4.6.2.5 Comparisons between the industrial design and economically-optimal designs


The real industrial design (ID) and the two economically-optimal designs, s1 (temper-
atures of main and reheated steams equal to each other) and s2 (temperatures of main
and reheated steams not equal to each other) are compared by an exergoeconomic
analysis to show how the system has been improved.
All decision variables of the three designs are listed in Table 4.6. The variations of
the variables tsh , trh and psh have been discussed in previous sections. For HPT1, a
significant increase in the isentropic efficiency is needed for the cost-effective design.
Additionally, determined by the outlet pressure of HPT1, the final feedwater preheat-
ing temperature of the design s1 remains almost the same as in the ID; however, that
of the design s2 (even with a larger plant efficiency) is smaller by around 6 C. For
HPT2, both the pressure ratio and isentropic efficiency are reduced in the two optimal
designs. This may be due to the thermodynamic interaction between the HPT2 and
the boiler, as a higher inlet temperature of reheat steam leads to a decrease in PECB
(see Fshrh in App. A). For the remaining turbines, the outlet pressures in the optimal
80 superstructure-based synthesis and optimization of coal-fired power plants

Table 4.6: Comparison of the decision variables between the industrial design (ID) and the
economically optimal solutions (t /C, p /bar)

Comp. Variable ID Solution s1 Solution s2

tsh trh 605.0 603.0 608.1 608.1 564.3 700.0


Boiler
psh m sh 274.2 856.0 295.0 836.7 268.6 778.3
HPT1 s pTi 0.869 85.70 0.915 87.88 0.914 78.53
HPT2 s pTi 0.925 65.56 0.905 80.41 0.882 74.72
IPT1 s pTi 0.894 24.79 0.921 22.14 0.911 25.55
IPT2 s pTi 0.915 12.18 0.926 7.691 0.915 9.180
IPT3 s pTi 0.929 6.490 0.881 5.461 0.927 2.049
LPT1 s pTi 0.912 2.677 0.919 1.327 0.898 0.778
LPT2 s pTi 0.919 0.704 0.921 0.346 0.932 0.172
LPT3 s pTi 0.914 0.266 0.940 0.0549 0.912 0.058
LPT4 s pTi 0.835 0.052 0.845 0.036 0.838 0.037
COND tup 5.0 2.027 2.029
FWH1 tup tlow 2.8 5.6 0.075 5.604 0.801 6.016
FWH2 tup tlow 2.8 5.6 1.904 5.919 0.034 5.156
FWH3 tup tlow 2.8 5.6 1.113 4.376 0.454 4.062
FWH4 tup tlow 2.8 5.6 1.952 6.713 1.775 5.082
FWH5 tup tlow 0.2 5.6 -1.197 5.609 -1.635 4.206
FWH6 tup tlow 0.2 5.6 1.349 5.808 0.948 5.234
FWH7 tup tlow -1.7 5.6 -0.553 6.546 -0.591 4.331
FP s 0.840 0.799 0.795
OT s 0.810 0.810 0.829

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

cF,k /MJ1 cP,k /MJ1 rk /% Zk /s1 C D,k /s1 C D,k + Zk /s1 f k /%

4.6 design optimization for thermodynamic and economic objectives


Comp
ID s1 s2 ID s1 s2 ID s1 s2 ID s1 s2 ID s1 s2 ID s1 s2 ID s1 s2

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

turbines is significantly reduced. In addition, the relative cost differences of feedwater


preheaters gradually decrease owing to a continuous performance improvement with
increasing temperature level for heat transfer (Yang et al., 2013).
For the cost of exergy destruction C D and the investment cost Z, both costs of the
boiler are reduced in the two economically optimal solutions. The reduction in the cost
C D,B is almost the same for both optimal solutions, while the solution s2 has a larger
decrease in the cost Z B due to the formulation of factor Fshrh (see equations for BOILER
in App. A). The turbines cannot be compared separately, since the work generated
by each turbine varies. However, the exergy destruction cost of all turbines falls from
0.72 $/s in the ID to 0.57 $/s in the optimal designs. Considering that the capital
investment is increased slightly, this finally contributes to the slight reduction in the
cost C D + Z of all turbines. A smaller temperature difference in the condenser requires
more heat transfer area and, thus, more monetary investment, while the corresponding
exergy destruction cost is significantly reduced, leading to a drop of the cost C D + Z of
the condenser. Although both the temperature differences tup and tlow of feedwater
preheaters are smaller in the optimal solutions, the feedwater preheaters present quite
different performances. The cost C D + Z of feedwater preheaters (including DA) in the
design s1 increases due to the increase in exergy destruction; however, in the solution
s2 , it remains almost unchanged compared with the ID. Since the capital investment
of all feedwater preheaters is rather small compared to other components, it might be
beneficial to add more feedwater preheaters.
For the exergoeconomic factors, those of the turbines are reasonable within the range
60%75% in the two optimal solutions: Those of HPT2, IPT3 and LPT1 are reduced
but those of HPT1, IPT1, LPT2 and LPT4 are increased significantly. Additionally, the
exergoeconomic factors of feedwater preheaters in the optimal solutions are within the
range 30%45% with the exception of FWH1 and FWH7 (over 60%). However, for these
two preheaters, both the costs of the exergy destruction and investment are reduced
significantly to small values.

4.6.3 On the predefined superstructure

Based on the superstructure, different single-reheat and double-reheat cases with an


OT or ET were optimized without considering constraints type1 and type2 in Eq. 4.2.
All cases consider 5 to 10 feedwater preheaters. However, the cases with 7 feedwater
preheaters are discussed in detail in the following.

4.6.3.1 Comparison with current industrial design


The fronts of four different cases are compared with the same industrial design with
the plant efficiency of 45.8% and the COE of 5.21 /kWh (Fig. 4.24). For the efficiency
of the industrial design, the optimal COE of both single-reheat cases are lower, while
those of double-reheat cases are slightly higher. Introducing a second reheat into a
single-reheat unit leads to an increase in both plant efficiency and the COE. In addition,
considering all economically-optimal solutions (Fig. 4.24), there is still potential for
enhancing both thermodynamic and economic performances of real designs.
The Pareto fronts (the increasing parts from the plant efficiency of around 48% and
onwards) indicate the current industrial design is dominated by all the four optimiza-
4.6 design optimization for thermodynamic and economic objectives 83

5.9 310 0.75


Single reheating with fixed structure

Optimal reheat-pressure ratio / -


5.8 Single reheating with varied structure 300

Final feedwater preheating


0.65
Cost of electricity / /kWh

Double reheating with OT


5.7
Double reheating with ET 290

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).

4.6.3.2 Single-reheat cases


The front of the single-reheat case with a fixed structure (Fig. 4.24) is, in a wide range,
higher than that obtained from the superstructure; however, the economically-optimal
design points of these two cases are quite close to each other. A fixed system structure
constrains further decrease in COE for a given plant efficiency.
The economically-optimal solution has almost the same structure with that of the
industrial design but reaches a larger plant efficiency, because more variables, e. g.,
steam conditions and isentropic efficiency of the turbines, are set as decision variables.
For single-reheat cases, the reheat position is always recommended at the extrac-
tion point of the highest- or second-highest-pressure steam for feedwater preheating.
Particularly, the reheat of solutions at the flat zone around the economically-optimal
solution occurs at the point of highest-pressure steam extraction, which is different
from the given industrial design (Fig. 4.16).
The profile of the optimal reheat-pressure ratio is similar with that of feedwater
preheating temperature (Fig. 4.25). At the left part of the front, the optimal pressure
ratio of reheat is within 0.150.25; while at the Pareto front, the feedwater preheating
temperature increases largely and the pressure of the main steam reaches (and remains
at) its upper bound (350 bar). The optimal pressure ratio for future plant design (the
Pareto front) can be further narrowed within the range 0.20.25.
84 superstructure-based synthesis and optimization of coal-fired power plants

a) 4 50 b) 1.5 50
45 45

Shares of FC/CC to COE / /kWh

Shares of PECB to COE / /kWh


With OT

Power output ratio (ET/OT) / %


Power output ratio (ET/OT) / %
3.5 With ET 40 With OT 40
1.4 With ET
Share of FC 35 35
3 30 30
With OT 25 1.3 25
With ET
2.5 20 20
Share of CC 15 15
1.2
2 10 10
5 5
1.5 0 1.1 0
41 43 45 47 49 51 41 43 45 47 49 51
Plant efficiency / % Plant efficiency / %

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

4.6.3.3 Double-reheat cases


For double-reheat cases with ET/OT and varied structures (Fig. 4.24), when the plant
efficiency is lower than 45%, there are no significant differences between the two fronts.
Then, the COE of the front with an ET becomes greater than that with an OT at each
efficiency level. The reasons are investigated in Fig. 4.26: The difference between the
two double-reheat fronts (Fig. 4.24) is lead to by the increase in the capital investment
costs (CC) after introducing an ET (Fig. 4.26a). More importantly, the increase in CC is
due to the PEC increase in the boiler (Fig. 4.26b): Introducing an ET causes a decrease
in the mass flow rate of reheat steam; thus, to reach the same plant efficiency, the
temperature and pressure of main and reheat steams have to be improved. Considering
the cost function of the boiler, a small increase in the PEC of boiler is caused for keeping
the same plant efficiency.
The profiles of the optimal pressure ratios of the two double-reheat cases are almost
the same (Fig. 4.27): When the plant efficiency is quite low (around 40%), there seems
no need to use reheating, as the lower bounds of the temperature and pressure of the

a) 100 b) 100
Optimal pressure ratio / %

Optimal pressure ratio / %

80 The second reheat 80 The second reheat

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.

4.7 Summary and conclusions


In this chapter, the superstructure-based synthesis and optimization is performed to
investigate the classical fundamental options for the design of coal-fired power plants
(Fig 2.4), such as adding additional reheatings and feedwater preheaters. A graphical
superstructure is built based on a professional simulator. Specially-designed problem-
specific structural mutation and crossover operators are developed to automatically
generate solution structures from the superstructure. Additional algorithms to enhance
DE/MODE are developed to increase the generation rate of feasible structure alterna-
tives and to find the front desired of the objective space. The superstructure is then
optimized for investigating the optimal matches of temperature and pressure of main
and reheat steam, optimal reheating pressure ratios, optimal layout of feedwater pre-
heating system, and the bi-objective trade-offs. Main conclusions include:

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

In the superstructure-based approach (chapter 4), the structure evolution is based on


a manually-defined superstructure and a problem-specific EA. The superstructure def-
inition is usually difficult and error-prone; more importantly, an inappropriate super-
structure may leave out good structural alternatives but considers a large number of
meaningless or even infeasible alternatives. In addition, the EA requires a large set of
manual specifications for structural mutation and recombination.
The superstructure-free approach overcomes the drawbacks of superstructure-based
approach by using a knowledge-integrated rule-based mutation operator. However,
the most widely used superstructure-free approaches require a large set of technology-
specific mutation rules that have to be manually set by the user (e. g., section 4.4.2).
Therefore, a generic concept avoiding any manual input for the definition of technology-
specific mutation rules has been developed by Voll et al. (2012) for synthesizing dis-
tributed energy supply systems. In the concept, all technologies are placed into an
energy conversion hierarchy (ECH), for which one set of generic mutation rules was de-
signed once and for all. The superstructure-free approach has been further extended
by the author (Wang et al., 2015) to synthesize thermal power plants.
In this chapter, the extended superstructure-free framework is generalized for syn-
thesizing thermal power plants (section 5.1) and illustrated by the synthesis of a simple
Rankine-cycle based thermal power plant (section 5.3). The proposed approach is then
applied for synthesizing complex pulverized-coal power plants, considering thermo-
dynamic and economic objectives separately and also simultaneously for bi-objective
trade-offs (section 5.4). Finally, the concept is summarized and conclusions are drawn
(section 5.5).

5.1 Multi-objective superstructure-free synthesis framework


The general multi-objective optimization-based synthesis problem for energy systems
is given:
T
min f ( x ) = ( f 1 ( x ), . . . , f k ( x )) , (5.1)
x
s.t. x = (s, d, o ), s S, d D, o O.

87
88 superstructure-free synthesis and optimization of coal-fired power plants

In this formulation, the vector f represents k usually conflicting objective functions


f k . Depending on the scope of the synthesis and optimization problems of energy
systems, a solution x in the objective function space may comprise of three independent
decision-variable vectors s, d, and o, which belong to the continuous and/or integer
variable spaces S, D, and O for the synthesis, design, and operation of the considered
energy systems, respectively. On the synthesis level, the system structure is considered,
i. e., which units are connected in which way; on the design level, the units sizing
is determined; and finally, on the operation level, the operational status (on/off) and
operational loads are specified for each installed unit. The three levels correspond to
an inherent hierarchical structure of energy systems (Frangopoulos et al., 2002). Thus,
the problem formulation (Eq. 5.1) is decomposed into two levels: the upper level deals
with the synthesis, while the lower level copes with the design and operation,
T
min f (s, d, o ) = ( f 1 (s, d, o ), . . . , f k (s, d, o )) , (5.2)
s

s.t. min f (s, d, o ) = ( f 1 (s, d, o ), . . . , f k (s, d, o ))T .


d, o

To exploit the bi-level formulation, the superstructure-free optimization employs a


hybrid algorithm combining an evolutionary algorithm for the upper level with de-
terministic optimization for the lower level (Fig. 5.1). The upper-level evolutionary
algorithm generates structural alternatives s, i. e., units selection and interconnections
among the employed units, while each alternative generated by the upper level is then
optimized deterministically in the lower level, i. e., identification of optimal sizing d
and operation o of the employed units. The structural decisions s (Eq. 5.2) are not
explicitly modeled in a superstructure, but the structures are evolved with the new
structural alternatives generated by an evolutionary algorithm. Consequently, the
formulation of the multi-objective superstructure-free synthesis problem solved by the
hybrid decomposition becomes
T
min f (s(), d, o ) = ( f 1 (s(), d, o ), . . . , f k (s(), d, o )) , (5.3)

s.t. min f (s(), d, o ) = ( f 1 (s(), d, o ), . . . , f k (s(), d, o ))T ,


d, o

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

Figure 5.1: Multi-objective superstructure-free optimization approach


5.1 multi-objective superstructure-free synthesis framework 89

an efficient definition of all reasonable connections between the considered energy


conversion technologies. Thereby, a minimal set of generic replacement and insertion
rules suffices to generate all feasible solution structures by structural mutations. More
importantly, the manual definition of technology-specific replacement and insertion
rules is avoided. For the upper-level evolutionary algorithm, a mutation operator has
been designed in (Voll et al., 2012). The mutation operator either randomly replaces
units in a candidate structure by alternative designs or randomly inserts units into a
candidate structure based on the given ECH, so that meaningful structural alternatives
are continuously generated.
A straightforward idea to solve the multi-objective problem described by Eq. 5.3 is
as follows: 1) The upper-level evolutionary algorithm repeatedly modifies and gener-
ates structural alternatives to richly explore the space of structural alternatives S. 2)
For each single feasible structural alternative s, the lower-level deterministic optimiza-
tion tunes the design and operation variables (d and o) to perform a multi-objective
optimization with any available mathematical technique, e. g., weighted sum method
Marler and Arora (2010), the e-constraint method Mavrotas (2009) or the normalized
normal constraint method Messac et al. (2003). Such that, a set of Pareto solutions are
obtained for each specific structural alternative. 3) Then, the upper-level evolutionary
algorithm performs a multi-objective selection considering both the Pareto solutions
of newly-evaluated structural alternatives and the solutions already kept in the up-
per level. Consequently, the Pareto front of all structural alternatives evaluated so far
can be generated, and eventually, the Pareto front of the whole solution space can
be obtained after adequately exploring the space of structural alternatives. This idea,
however, is not efficient since the number of structures explored by the superstructure-
free optimization is enormous. Hence, we utilize the population-based nature of the
upper-level evolutionary algorithm to efficiently generate the Pareto front of the whole
problem in a single run. The lower-level multi-objective optimization problem (Eq. 5.3)
is reformulated to a single-objective optimization problem (Fig. 5.1) with an aggregated
objective formulated by the weighted sum method (Marler and Arora, 2010), such that
it can be handled efficiently by the low-level deterministic optimization:
T
min f (s(), d, o ) = ( f 1 (s(), d, o ), . . . , f k (s(), d, o )) , (5.4)

s.t. min
d, o
F (s(), d, o ) = (wk f k (s(), d, o)) ,
k

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.

5.1.1 Knowledge-integrated mutation operator

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.

2. Remove one component and short-circuit all of its interconnections.

3. Delete one component and insert another component.

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.

6. Insert one component by replacing the technology-related streams (for details,


see section 5.1.3).
The first 5 mutation rules are replacement rules, while the last is an insertion rule.
During optimization, these generic mutation rules are used to mutate given solution
structures to alternative solution structures. For this purpose, the mutation operator
consults the energy conversion hierarchy.
An energy conversion hierarchy consists of the three levels: meta, function, and tech-
nology levels (Fig. 5.2). Nodes on the meta level represent the mutation rules. Nodes
on the technology level represent specific energy conversion technologies. The con-
necting nodes on the function level classify energy conversion technologies according

Meta level Meta

Deletion Insertion Parallel connection Serial connection


allowed allowed allowed allowed

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

HE: heat exchanger

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.

5.1.2 Replacement rules for structural mutation

The purpose of the replacement rules is to remove a component in given structures


or replace it with other component(s). By one or few replacement operations on given
structures, similar structural alternatives (i. e., neighboring structures) are generated.
The replacement rules work as follows: The mutation operator starts by randomly se-
lecting one component (the target component) to be removed from the initial flowsheet.
Then, a function of the target component is randomly chosen, according to which pos-
sible replacement rules are identified. The possible replacement rules are rules 15
(listed in section 5.1.1) and linked to the selected function. Afterwards, one of these
possible replacement rules is randomly selected and implemented. Subsequently, the
mutation operator consults the ECH to identify the candidate components of specific
technologies that may be inserted in place of the target component. Feasible candidate
92 superstructure-free synthesis and optimization of coal-fired power plants

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.

5.1.2.1 Limitations of the replacement rules


The replacement mutation rules (rule 15 in section 5.1.1) are not sufficient for the
synthesis of thermal power plants, since these replacement rules lack the possibility
to add technologies with functions that have not been existing in any parent structure.
The mutation operator only replaces parts of a parent solution by other components
but does not add new functions. For example, a steam-water heat exchanger cannot be
introduced into a parent structure, if the functions of "Steam cooler" and "Water heater"
have not been employed in the parent structure. This shortcoming prevents the gener-
ation of serial connections of units with different functions, such as heat exchanger -
turbine - heat exchanger - turbine, etc. This shortcoming, however, is overcome by the
insertion mutation rule (rule 6 in section 5.1.1).

5.1.3 Insertion rule for structural mutation

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

a) Technology selected for insertion Parent structure


condenser steam
gas
water a)
Randomly select one
heat
technology allowing insertion
steam heat gas
Part of the parent structure b)
Instantiate the technology and select
one ECH function of the technology
b) Select a function of the technology and
one particular stream c)
no Free pins related to the selected
gas to cool gas to heat steam function are of the same stream type?
yes
steam d)
Randomly select one stream of fail
the function-required stream type
succeed
1 2 3 e)
Split the selected stream into two and
connect the technology instance in between
heat

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

New structurally feasible structure


heat

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.

5.2 NLP model of thermal power plants


The thermodynamic models of the considered components can be found in usual ther-
modynamic textbooks, e. g., (Moran et al., 2010). For details on the component models
(including cost functions), the reader is kindly referred to App. A. In this thesis, the
considered technologies include steam generator, steam turbine, water pumps, feedwa-
ter preheater (steam-water heat exchanger), steam de-superheater (DSH) (steam-water
heat exchanger), reheaters, condenser. The nonideal behavior of turbines and pumps
is modeled by the isentropic efficiency. For feedwater preheaters (FWHs), the temper-
ature difference at pinch point (Tpin ) and that between the hot outlet and the cold
inlet (Tlow ) are specified. Steam generator, reheaters and condenser are considered as
heat-injection or heat-extraction components, respectively.
For calculating function values and derivatives of necessary properties of water and
steam, the pair ( p, h) is preferred as input variable set, since the relationship of pressure
and temperature is singular in two-phase zone. Therefore, the necessary properties are
expressed as T = T ( p, h), s = s ( p, h), h0 = h0 ( p), h00 = h00 ( p) and Tsat = Tsat ( p). The
partial or full derivatives required for the optimization, (T/p)h , (T/h)p , (s/p)h ,
(s/h)p , dTsat /dp and dh0 /dp, are also calculated: The first four terms can be easily
derived by Maxwell and Jacobin relations (Moran et al., 2010), while the last two terms
are calculated by:

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

1 Freesteam 2.0: freesteam.sourceforge.net.


96 superstructure-free synthesis and optimization of coal-fired power plants

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.

5.3 Evaluation of the superstructure-free approach


In this section, the superstructure-free approach is illustrated for the optimal synthesis
of a thermal power plant. The goal is to maximize the plant efficiency (or the ther-
mal efficiency). The synthesis process is initialized with the simple, classic Rankine
cycle. It is shown that the proposed approach identifies complex cycle configurations
employing technologies that have not been included in the initial structure.

5.3.1 Thermodynamic objective function

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.

5.3.2 Specified energy conversion hierarchy

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

Table 5.1: Basic specifications of the thermal power plant

Parameters/Variables Setting Bounds

Maximum pressure of water and steam 300 bar [230,500]


Minimum pressure of water and steam 0.05 bar [0.03,0.08]
Maximum temperature of water and steam 923 K [823,973]
Maximum mass flow rate of water and steam 1000 kg/s [300,1500]
Isentropic efficiency of steam turbines 90% [85,96]
Isentropic efficiency of feedwater pumps 82% [75,87]
Temperature difference at pinch point of FWHs 2K [1,4]
Temp. diff. between hot outlet and cold inlet of FWHs 6K [4.5,8]
Temp. diff. between hot outlet and cold inlet of DSHs 20 K [15,25]
Energy loss rate (loss in Eq. 5.8) 8% [5,15]
Parameter values are employed when considering only thermal efficiency as the objective.
Variable bounds are employed when an economic objective is involved.

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.

5.3.3 Initial solution and optimal solution by only replacement rules

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%.

5.3.4 Evaluation of the approach

For the upper-level evolutionary algorithm (Fig. 5.1), a ( + )-evolution strategy is


employed (Beyer, 2001), where is the size of the parent population (20 for this study)
and is the number of the offspring solutions generated from the parent population
in each generation (25 for this study). For each mutation, only one mutation rule is
selected to modify the parent structures and all mutation rules have equal probability
to be selected. Meanwhile, to avoid the preserve of solutions with identical flowsheets,
the mutation on a given parent solution will terminate if a new feasible solution is
generated with the relative objective variation (|parent new |/parent ) larger than 1E-5.
The computations are performed on an Intel Core2 Duo CPU P8600 2.4 GHz with 4
GB RAM. The operating system is Windows 8.1 Enterprise (64-bit).

5.3.4.1 Optimal solution


The best-known solution, in the following referred to as the optimal solution (Fig. 5.6b
and C.3b), features ten turbines with multi-stage reheating and feedwater preheating
(Fig. C.2b). As expected, when seeking a maximum thermal efficiency regarding the
design conditions, the numbers of reheating and feedwater preheating reach the maxi-
mum numbers. However, the actual configuration of the units is far from being trivial.
5.3 evaluation of the superstructure-free approach 99

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.

5.3.4.2 Evolution details of exemplary runs


Figure 5.7 illustrate two exemplary optimization runs with different maximum mu-
tation strengths, which represent the similar behavior of all performed optimization
runs. When only one mutation rule is allowed for generating each structural alternative
(Fig. 5.7a), the objective value of the best solution identified so far increases sharply in
the first 50 generations and improves slowly afterwards. Moreover, the objective func-
tion values of solutions produced in each generation are within a limited range. After
50th generation, the objective range remains almost unchanged, which indicates the
preserved solutions in the parent generation possess similar objective values.
When performing the optimization with a maximum mutation strength over 1, the
number of mutation rules successfully employed to given parent structures will be
automatically adjusted by EA. The larger the maximum mutation strength, the farther
the neighboring structures can be generated by one single mutation and the faster
the structural evolution will be. This is confirmed by the comparison between the
representative runs in Fig. 5.7a (with a maximum mutation strength of 1) and Fig. 5.7b
(with a maximum mutation strength of 5). In in Fig. 5.7b, the structural evolution
advances faster; near-optimal structures have even been reached within 25 generations;
and more importantly, the same optimal solution is found with less generations and
less computation time.

a) 53 500 b) 53 500 100

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.

5.3.4.3 Computational performance


To evaluate the efficiency of the superstructure-free approach, the computation time
and the generation rate of feasible solutions, i. e., solutions that have been successfully
optimized by the lower-level deterministic optimization, are analyzed (Fig. 5.7). For
both representative runs, the generation rates of structurally-feasible solutions (well-
connected structures) achieve high values, over 90% of all mutation tries. However,
this number is relatively low for the first several generations, in which the parent struc-
tures have no available steam pins to remove those free steam pins at the hot side
of newly-added feedwater preheaters. Another exciting conclusion is that more than
90% of structurally-feasible structures (well-connected structures with no free nonstan-
dalone pins) generated are evaluated as feasible structures by successful deterministic
optimization, although this number decreases slightly with larger maximum mutation
strengths. Failure in evaluating a structurally feasible solution is mainly caused by the
violation of the pressure constraint on the steam extraction. Most importantly, the gen-
eration rates of structurally-feasible alternatives and feasible solutions remain stable in
course of structural evolution.
The accumulated computation time with respect to the generation number is pre-
sented (Fig. 5.7). The total computation time is almost the same as the total evaluation
time, since the mutation time can be neglected. Unsuccessful structure optimization
consumes little time as well compared with the total time, as 1) GAMS evaluation
always terminates early if the constraints of component models cannot be satisfied,
5.4 superstructure-free synthesis of complex coal-fired power plants 101

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 (%)

Max. mutation strength 1 Max. mutation strength 5


Run
tot n e tot n e
min % min 1 % min % min 1 %

1 464 49.45 397 176 0 382 49.45 95 58 0


2 482 49.45 136 70 0 449 49.45 140 84 0
3 393 49.45 132 75 0 456 49.45 168 86 0
4 396 49.45 108 66 0 390 49.45 137 77 0
5 423 49.45 211 111 0 467 49.45 196 98 0
6 435 49.45 257 131 0 449 49.45 149 85 0

Geometric mean 431 49.45 186 98 0 431 49.45 144 80 0


Termination at 200th generation with the settings: = 20, = 25, m ms = 1000 kg/s
The relative difference between the objective values of the best-known solution from all
runs (all ) and the best solution identified in each run i ( i ): e = 100 |all i |/all

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.

5.4 Superstructure-free synthesis of complex coal-fired power plants


In this section, the superstructure-free synthesis framework is applied for synthesiz-
ing complex pulverized-coal power plants with a fixed capacity of 1000 MW. A new
energy conversion technology, de-superheater (a surface heat exchanger, see App. A),
is considered. Therefore, the specified ECH in Fig. C.1 is further extended (Fig. C.5):
The de-superheater node in the technology level is linked by two sub-function nodes
102 superstructure-free synthesis and optimization of coal-fired power plants

("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.

5.4.1 Optimal synthesis for maximizing thermal efficiency

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

46 accumulated time 200 20

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).

5.4.2 Optimal synthesis for minimizing cost of electricity

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 / -

rh1 Number of FWH above rh1 / -


7 0.4 4
Optimal pressure ratio / -

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

5.4 1800 100


1600
1 2 rate 3
Cost of electricity / /kWh

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.

5.4.3 Bi-objective optimal synthesis

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

Cost of electricity / /kWh


80 2 1

Accumulated time / min


6000
7 5.1

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

structurally feasible structure but a decaying generation rate of feasible solutions. In


addition, since there is no clear termination criteria for multi-objective optimization,
the total number of evaluation is set as 30,000, which leads to more considerable com-
putation time. However, the average computation time of one single structure is almost
the same with that in the case of only COE.
The evolution of the best-known front is converged at around 25,000 evaluations
(Fig. 5.15b). As expected, the presented Pareto front dominates all fronts obtained
from the superstructure-based approach (Fig. 4.24), as larger decision/objective space
are investigated by the superstructure-free approach. The flat part of the front with
those solutions near the economically optimal solution has been almost converged
within 15,000 evaluations. Afterwards, the structural evolution preserves more solu-
tions achieving higher thermal efficiencies. The front evolution from generation 50 to
generation 250 suggests that the structures of Pareto solutions at the right part of the
Pareto front, i. e., those structures close to the thermodynamically optimal solution
(Fig. C.12b), are much more complex than those close to the economically optimal so-
lution (Fig. C.12a). However, the economically optimal structures of all performed runs
are not identical with the optimal solution obtained when minimizing only the COE
(Fig. 5.12b), with an average optimality gap of 0.03%. The structural evolution becomes
more difficult when multiple objectives are involved.
The flowsheet and component variables, and solution structure vary among all
Pareto solutions (Fig. 5.16). Temperatures of main steam and feedwater should be kept
at lower values for economically optimal design but increase gradually to reach higher
thermal efficiencies (Fig. 5.16a). Particularly, when the efficiency achieves near 56.3%,
both temperatures increase up to their upper bounds. For those solutions with an effi-
ciency over 56.3%, a further increase in efficiency comes mainly from adapting compo-
nent variables and the solution structure. Surprisingly, almost all the Pareto solutions,
even those near the economically-optimal solutions, employ the maximum number of
reheating. The number of feedwater preheater among all these solutions varies from 6
to 10 and, more importantly, a smaller number is favorable for cost-effective design.
5.5 summary and conclusions 109

a) 1050 650 b) 11
10
950 600

Mass flow rate / kg/s


9
Termperature / K

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

5.5 Summary and conclusions


In this chapter, an extension of the generic framework for superstructure-free opti-
mal synthesis of energy systems proposed recently by Voll et al. (2012) is introduced.
The framework employs a hybrid algorithm, combining evolutionary algorithm with
deterministic optimization. At the upper level, the knowledge-integrated mutation op-
erator employs mutation rules to adapt parts of given structures for generating new
structural alternatives, which are then optimized by the lower-level deterministic opti-
mization to identify optimal units sizing and operation. Both the number of mutation
rules and meaningless structural alternatives are minimized by classifying all involved
energy conversion technologies into a hierarchically-structured ECH. The ECH fulfills
an efficient definition of all reasonable connections between components in regard by
a minimum set of generic mutation rules. The easy-to-extend feature of the ECH facil-
itates the mutation operator to integrate more available technologies.
The original superstructure-free framework for synthesizing distributed energy sup-
ply systems is generalized and extended to synthesize thermal power plants. The muta-
tion rules are enriched by a insertion mechanism, which allows for an efficient stream-
based insertion of any technology and thus the addition of the promising design alter-
natives which can hardly be introduced by only the replacement mutation rules. Then,
the extended superstructure-free synthesis framework is evaluated by an illustrative
case study, and applied for both single- and bi-objective syntheses of complex coal-
fired power plants, based on an ECH particularly developed for thermal power plants.
The framework automatically identifies complex structures with new features, such
as multi-stage reheating, multi-stage feedwater preheating and steam desuperheating.
Main conclusions include

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.

The generation rate of structurally-feasible structures remains stable and high,


while the generation rate of feasible solutions depends highly on the model for-
mulation for the deterministic optimization. Highly-nonlinear cost functions, if
involved, lead to a decrease in the generation rate of feasible solutions.
110 superstructure-free synthesis and optimization of coal-fired power plants

The computational time is generally enormous due to the employment of EAs.


However, high-quality near-optimal solutions are generated as well in course of
searching the optimal solution. In addition, the feature of automatic structural
generation without engineers attendance further makes the large computation
effort acceptable.

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

Highly energy-efficient, cost-effective design and retrofits of pulverized-coal power


plants have become urgent challenges. In the future, the features of coal-fired power
plants will be large-scale, higher temperature and pressure levels, multiple heat sources
and products, and the integration of many available technologies to utilize different-
grade heat efficiently. From this perspective, it is crucial to further improve the sys-
tem configurations, particularly to propose rational conceptual synthesis for integrat-
ing those energy-saving technologies. Exergy-based evaluation of existing designs and
optimization-based synthesis methods for complex integration tasks offer great po-
tential for the improvement or optimization of the system structures. However, for
complex power plants, on the one hand, there is a lack of sufficient and user-friendly
methods for automated conceptual synthesis; on the other hand, there is a lack of
complete and systematic applications for the system analysis and synthesis. Thus, the
thesis concerns not only the applications of different methodologies but also certain
degree of methodology development.
In previous research, conventional exergy-based evaluation method, superstructure-
based and superstructure-free synthesis approaches are most commonly applied to
the analysis and synthesis of thermal power plants. Traditional exergy and exergoe-
conomic analyses aim at assessing the thermodynamic and exergoeconomic perfor-
mances of each component, and spatial distribution of exergy dissipation and their
associated costs. However, the nature of thermal power plants, i. e., the interactions
among different components and the avoidability of involved exergy dissipations (de-
structions and losses) and costs, cannot be revealed by these conventional methods but
by recently-developed advanced exergy-based analysis. In the advanced analysis, the
exergy destruction and costs associated with each component are split into endoge-
nous/exogenous and avoidable/unavoidable parts. The exogenous part pinpoints the
interaction between the considered component and the remaining system, while each
avoidable part highlights true energy- or cost-savings potential.
The superstructure-based synthesis approach explores a set of structural alternatives
embedded in a user-defined superstructure and identifies the optimal solution among
all. The implementation of this approach has been largely facilitated by various soft-
ware platforms, which enable efficient, or even graphical, building of a superstructure.
However, the fundamental problems of the superstructure-based approach remains:
on the one hand, a priori manual definition of a superstructure runs the risk of leaving
out good promising structural alternatives; on the other hand, an excessive large su-
perstructure reduces this risk but includes a large number of meaningless alternatives,
thus leading to extremely great computational time. To circumvent these fundamen-

111
112 summary and conclusions

tal problems, superstructure-free synthesis approaches, which enables simultaneous


generation and evaluation of structural alternatives, have been developed for specific
energy systems. Most of these superstructure-free approaches, however, involve ei-
ther manual definitions of technology-specific replacement rules or complex structure-
coding. Fortunately, a generic, easily extensible and user-friendly superstructure-free
synthesis concept has been proposed by Voll et al. (2012). The approach allows for
an automatic generation of structural alternatives, a systematic structural evolution,
based on a hierarchically-structured energy conversion hierarchy, which classifies all
available technologies with respect to their functions, and a minimum number of re-
placement rules. Each generated structural alternative is evaluated by deterministic
algorithms for identifying the optimal one. So far, this superstructure-free concept has
only been applied to the synthesis of distributed energy supply systems.
In this thesis, progressive studies on the structural improvement and optimization of
coal-fired power plants are performed: 1) For the advanced analysis (chapter 3), a ratio-
nal method for calculating endogenous exergy destructions, which avoids the model-
ing of theoretical chemical reactors and heat exchangers, is introduced and developed
for coal power plants. An existing detail-modeled ultra-supercritical pulverized-coal
power plant is evaluated comprehensively for proposing possible suggestions on struc-
tural improvements. 2) The superstructure-based approach is applied to solve some
classical design problems of coal-fired power plants (chapter 4), for which promising
structural alternatives are less likely excluded from a well-defined superstructure. As
highly-nonlinear properties of water and steam are difficult to handle in deterministic
MINLP algorithms, the superstructure, which considers limited types and numbers of
technologies, is graphically built with the aid of a simulator, properly represented and
efficiently solved by problem-specific differential evolution. 3) The superstructure-free
concept proposed by Voll et al. (2012) is extended and re-implemented to cope with
complex synthesis problems of coal-fired power plants, which involve a large number
of different available technologies. A new energy conversion hierarchy categorizes the
available technologies for synthesizing coal-fired power plants. The set of mutation
rules is enriched by a new insertion rule, which allows for efficient stream-based inser-
tion of any technology; thus, new features that have not been existing in given parent
structures can be readily introduced. The optimization of each structural alternative is
fulfilled by NLP but not MINLP solvers. Moreover, in both superstructure-based and
-free approaches, a set of near-optimal and Pareto solutions generated enable rational
decision making.
The features of the mentioned three studies for pulverized-coal power plants are
compared in Table 6.1. The exergy-based analysis aims at the evaluation of a specific
system design. It cannot directly propose new structural alternatives but only suggests
effective measures for the performance improvement with the aid of expert judgment.
The computational effort of the analysis is associated with a number of simulations for
advanced analysis (Table 3.2) and solving a linear sparse equation set for exergoeco-
nomic analysis. For the coal-fired power plant analyzed in the thesis, the computational
time for each simulation is mostly below 1 s, while that for solving a linear equation
set with a dimension of 200 200 is less than 10 s.
The superstructure- and simulation-based synthesis approach explores the super-
structure by a problem-specific differential evolution algorithm, in which both muta-
summary and conclusions 113

Table 6.1: Comparisons between the methodologies employed in this thesis for the analysis and
synthesis of thermal power plants

Exergy-based Superstructure- Superstructure-


Terms
analysis based free

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

ingless structures. The solution space is explored by an evolution algorithm, in which


only the mutation operation is activated. By using aggregation selection, a set of Pareto
solutions with good compromise and near-optimal solutions are obtained even with-
out expert knowledge. The disadvantage of the approach, however, is the enormous
computational effort, around 4 days for a multi-objective optimization.
For the exergy-based evaluation, the exergetic performances of heat exchangers of
the plant analyzed are grouped into three ranges. The heat exchangers in furnace,
air preheater and first low-pressure feedwater preheater perform with low exergetic
efficiencies. Exergy destructions of the whole system are mainly from the boiler sub-
system, while exergy losses are mainly caused by exhausted flue gas and cooling wa-
ter. The furnace and steam turbines are exergy-destruction-cost leading, while the rest
components are mostly investment-cost leading. Most boilers heat exchangers and
steam turbines have large avoidable exergy destructions, mostly endogenous. The fur-
nace and air preheater have large portions of exogenous/avoidable exergy destruction.
More importantly, only around 10% of the investment and exergy destruction costs of
the whole system can be avoided. The improvement strategy of an individual compo-
nent is different with respect the different portions of endogenous/exogenous parts in
avoidable thermodynamic inefficiencies and costs. It is also concluded that limited ben-
efits would be obtained by only improving design parameters but not the structures.
Suggestions on the arrangement of heating surfaces of the boiler are proposed as well.
Both the superstructure-based and -free synthesis approach identify complex opti-
mal and near-optimal solutions. When considering only thermodynamic objective and
keep reasonable exhausted steam quality at the same time, the benefits from increasing
temperature levels of the main and reheated steams can be fully obtained only if the
throttle pressure reaches a value specifically to each temperature level, namely transi-
tion throttle pressure. The larger the temperature level, the greater the transition throt-
tle pressure. Unfortunately, the transition throttle pressures for the temperature levels
over 600 C turn out to be too large for practical applications. It is thus concluded that
in future design of coal-fired power plants, particularly 700 C plants, certain amount
of benefits from the increase in temperature level have to be discarded.
Both superstructure-based and -free synthesis approach identify high-quality Pareto
fronts for the objective functions, plant efficiency and the cost of electricity. The part of
each front close to the economically-optimal solution is flat, while the cost of electricity
rises sharply for the rest of the front. Many good solution alternatives around the
economically-optimal solution are available for decision makers. It is concluded as well,
based on given cost functions, it is more cost-effective not to improve pressure and
temperature levels of the main and reheated steam ahead of the remaining decision
variables. Introducing a second reheating is not necessary when considering limited
structural alternatives, but becomes quite beneficial if the superstructure-free approach
is employed.
Due to lack of real-world cost functions, cost functions from studies on parametric
optimization of thermal power plants with a fixed structure are adopted. Employing
these cost functions, the superstructure-free synthesis and optimization leads to op-
timal and near-optimal solutions with more units than currently found in modern
power plants indicating the need to revise the cost functions to reflect current indus-
trial practices. The capability and effectiveness of the superstructure-free approach are
6.1 future work 115

evaluated and demonstrated. It is concluded that this approach is promising to cope


with complex synthesis problems of pulverized-coal power plants, which considers
more available energy- and cost-saving technologies.
In the following section, some future perspectives are discussed considering partic-
ularly the drawbacks and insufficiency of the presented research.

6.1 Future work


In the following, three direction of future research on evaluation and optimization-
based synthesis of pulverized-coal power plants, or more generally, thermal power
plants, are recommended: real-world designs and retrofits, evaluation and synthesis
methodologies, and good modeling practice.

6.1.1 Real-world optimal designs and retrofits

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

1 Ebsilon: www.steag-systemtechnologies.com/ebsilon_professional.html, accessed on 29 May 2015.


116 summary and conclusions

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.

6.1.2 Evaluation and synthesis methodologies

Future perspectives on the methodology development concern mainly reasonable es-


timation approaches for endogenous exergy destructions, effective utilization of valu-
able results from the exergy-based analysis, possible combination of analysis methods
with automated synthesis approaches, and further algorithm enhancements of the au-
tomatic structural evolutions.

6.1.2.1 More reasonable way of estimating endogenous exergy destructions


There are still some open fundamental problems for calculating endogenous exergy
destructions. A drawback of the presented new approach (section 3.2.1) is that, for
complex systems, mass flow rates of the streams entering the considered component,
whose endogenous exergy destruction is to be calculated, are difficult to determine,
since the theoretically-operated components is treated as a reversible blackbox. In fact,
the mass-flow relationships between the streams of the considered component and the
blackbox are hard to establish reasonably, especially when the splitting and mixing of
streams are presented in a flowsheet. Moreover, the fundamentals on the means of han-
dling specific exergies of those streams flowing into the considered component have
not been clearly described, particularly for complex systems with, e. g., heat-exchanger
and steam-turbine trains.

6.1.2.2 Adaptive structural evolution strategies


These strategies are expected to help further avoid the generation and evaluation of a
number of meaningless structure alternatives. In the presented superstructure-based
and superstructure-free approaches, there are no fundamental differences on mutating
which part of the structure. It is almost equivalently assumed that, before each evalua-
tion of the generated structure, the change of any part of the structure would lead to
the same effect on the objectives. However, different components, or more clearly, sub-
systems, have different impacts on the overall system performance. The subsystems,
which present larger potentials for improving the objectives, should be given priority
to be adjusted. Therefore, proper decomposition methods, e. g., (Kocis and Grossmann,
1989; Daichendt and Grossmann, 1998), and adaptive evolution strategies should be
developed or coupled to efficiently evolve the structure: 1) decompose the whole sys-
tem into several subsystems ranked with regard to their influences on the performance
of the whole system, and 2) evolve higher-rank subsystems and estimate their effects
6.1 future work 117

on the overall performance of the system. Consequently, frequent adjustments of the


parts of the system structure that lead to only limited improvement of overall perfor-
mance can be, to a large extent, suppressed. In addition, the decomposed subsystems,
which are expected to be smaller, easy-to-solve subproblems out of the whole problem,
require much less computation efforts to be optimized.

6.1.2.3 Efficient identification of duplicate structures searched


Although a large number of meaningless structural alternatives can be avoided in the
superstructure-free approach, duplicate meaningful structures are frequently gener-
ated. The duplicate structures may lead to not only huge extra computational time to
find the same optimal structure, but also a decrease in the diversity of the preserved so-
lution structures. For complex problems, e. g., cost-effective synthesis of thermal power
plants (section 5.4.2), the average evaluation time of an individual structure can be as
long as over 20 s. In the thesis, the preserve of identical flowsheets are suppressed
in each mutation by discarding those offspring solutions with very similar objective
values as the parent solution but generating an offspring solution with a different ob-
jective value. However, on the one hand, duplicate structures are distinguished only
after the evaluation of the structures, thus it does not help reduce the total computation
time; on the other hand, duplicate structural alternatives can still be frequently gener-
ated from two independent mutations. Effective algorithms are expected to efficiently
identify whether the generated solution has been evaluated or not in the history of
the current structural evolution. These algorithms would be quite favorable to further
enhance the superstructure-free synthesis approach for synthesizing thermal power
plants.

6.1.2.4 Integration of analysis methods into system synthesis


There is still a large gap between different analysis methods and optimization-based
synthesis of thermal systems. It is expected that the coupling of proper analysis meth-
ods would further improve the performance of automated synthesis approaches, par-
ticularly the optimization of structural alternatives. For example, the thermoeconomic
functional analysis (Frangopoulos, 1992) and the entransy dissipation theory (Xu et al.,
2015) could formulate the objective functions in explicit relations with decision vari-
ables by the insights into the considered energy system. These reformulations of ob-
jective functions could lead to proper system decompositions and the removals of sur-
plus intermediate variables and equations. Accordingly, the computational effort for
optimizing the same structural alternative can be reduced. However, currently, these
approaches are hardly applicable for complex problems and, most importantly, funda-
mentally, they can only support the optimal synthesis based on a predefined super-
structure so far. There are challenges to automatically and properly reformulate the
objective functions with respect to different system structures.
For the most widely used accounting methods, e. g., SPECO, few references, e. g.,
(Uhlenbruck and Lucas, 2000; Cziesla and Tsatsaronis, 2002; Vieira et al., 2004), have
been published on employing these analysis information for the parametric optimiza-
tion; while for structural improvements engineers expertise usually have to be inte-
grated to judge which parts of the analyzed structure should be modified and how to
118 summary and conclusions

modify. Thus, for automated optimization-based synthesis, these accounting methods


should probably be employed to rank different subsystems for modification.
A far long way is ahead as well for the developing advanced exergy-based analysis to
be a supportive method of synthesis approach. There have been no available references
yet on how to reasonably use the information from the splitting of exergy destructions
and costs for parametric and structural optimization.

6.1.3 Mathematical modeling practice

Bad mathematical formulation may evaluate feasible structures as infeasible solutions,


good structures as bad solutions as the search procedure may be trapped at local opti-
mums. For example, in the superstructure-free application in section 5.4.2, many good
structural alternatives are evaluated with quite large cost of electricity, which is not fa-
vorable for the structural evolution. Good modeling practice, particularly for NLP and
MINLP, could help build efficient and sound formulations that can be solved much
faster, easier with less possibilities of being trapped locally. Various techniques are
available for good model formulations, such as setting good initial values and bounds
of all variables (including intermediate variables), properly scaling variables and equa-
tions, reformulations (piecewise/polynomial/separable approximations or even lin-
earization) of nonlinear formulations, convexification of nonconvex formulations, etc.
Some of these techniques that have been applied for the thesis have been introduced
in section 2.3.2. Many techniques for good modeling practice can be found in detail
in (Floudas, 1995; Drud, 2004; FICO, 2009; Williams, 2013). Note the reformulations of
nonlinear or nonconvex equations may lead to the lose of certain degree of accuracy
or even unacceptable optimal solutions, thus they have to be carefully developed and
checked before replacing the original formulations.
In the thesis, the properties of water and steam are provided by the external library,
which forces the GAMS models for the superstructure-free synthesis to be NLP. In this
way, the linearization of those involved nonlinear and nonconvex cost functions may
be not necessarily good, as the MINLP could require larger computational effort than
the original NLP. Thus, these cost functions may be reasonably reformulated by, e. g.,
separable programming techniques. Moreover, the difficulty of modeling (even not
good modeling) would increase significantly, when off-design performances of each
component are involved as well.
A
THERMODYNAMIC MODELS AND COST
FUNCTIONS OF THE COMPONENTS

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 ,

pin tpin > tpin ,


thot sat tout > tup ,
cold
or tstm cold

out tin > tlow ,


thot cold
 
0
hhot
out < h p hot
out ,
thermodynamic models and cost functions of the components 121

 
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).

condenser (dissipative component) A condenser has one incoming steam


pin (hot), one standalone incoming drainage pin (hot), one outgoing condensated-
water pin (hot), one incoming cooling-water pin (cold), and one outgoing cooling-water
pin (cold). The model is used for both simulation and optimization. The temperature
difference of the pinch point (or the saturation temperature minus the cold outlet) is
set for modeling the nonideal behavior:

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.

surface heat exchanger (productive component) A heat exchanger has


one incoming hot pin, one outgoing hot pin, one incoming cold pin, and one outgoing
cold pin. This model is used only in the simulation. The heat exchanger, co-current or
122 thermodynamic models and cost functions of the components

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

out = pin + pcold ,


pcold cold

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

PEC = 215679 Fm F Ft Fp Fshrh ,


 0.7718795
Fm = m sh
out ,

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 ,

out = pin pDA ,


pfw fw
 
0
hfw
out = h p fw
out ,

PEC = 74788 m 0.7


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

pout = min ( pin ) .

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

PEC = 31.758 PEG


0.95
.
B
A D D I T I O N A L M AT E R I A L S F O R T H E
A N A LY S I S A N D E VA L U AT I O N O F
THERMAL POWER PLANTS

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.

b.1 Exergy and exergoeconomic analysis of each type of component


steam turbine
exergy balance (fuel exergy = product exergy + exergy destruction)

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

P principle (only one exergy product, no auxiliary equations are needed)

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

F principle (no auxiliary equations are needed)

P principle (no auxiliary equations are needed)

125
126 additional materials for the analysis and evaluation

air fan (compressor)

exergy balance (fuel exergy = product exergy + exergy destruction)


  air 
PAF = E out,j E in
air
+ E D,AF

cost balance

C in
air
+ C in
work
+ Z AF = C out
air

F principle (no auxiliary equations are needed)

P principle (no auxiliary equations are needed)

feedwater preheater

exergy balance (fuel exergy = product exergy + exergy destruction)


   
E hot,stm + E hot,wat E out
in in
hot,wat cold,fw
= E out E cold,fw + E D,FWH
in

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

P principle (no auxiliary equations are needed)

condenser (dissipative component)

exergy balance (inlet exergy = outlet exergy + exergy destruction)


   
hot,stm hot,wat cold,cw hot,fw cold,cw
E in + E in + E in = E out + E out + E D,COND

cost balance
hot,stm hot,wat hot,fw
C in + C in + C aux,dc
cw
+ Z COND = C out + C dif,dc,COND

where the cost rate of the cooling water C aux,dc


cw is calculated:

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

P principle (not available)


B.1 exergy and exergoeconomic analysis of each type of component 127

surface heat exchanger (productive component)


exergy balance (fuel exergy = product exergy + exergy destruction)
   
E in
hot
E out
hot
= E out
cold
E in
cold
+ E D,HX

cost balance

C in
hot
+ C in
cold
+ Z HX = C out
hot
+ C out
cold

F principle

chot hot
in = cout

P principle (no auxiliary equations are needed)

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

F principle (no auxiliary equations are needed)

P principle (no auxiliary equations are needed)


128 additional materials for the analysis and evaluation

mixer

exergy balance (inlet exergy = outlet exergy + exergy destruction)


!
Nin
E in,j = E out + E D,MR

j =1

cost balance
Nin
C in,j + Z MR = C out
j =1

F principle (no auxiliary equations are needed)

P principle (no auxiliary equations are needed)

splitter

exergy balance (inlet exergy = outlet exergy + exergy destruction)


!
Nout
E in = E out,j + E D,SR

j =1

cost balance
Nout
C in + Z SR = C out,j
j =1

Auxiliary equations (not based on F or P principles)

cout,i = cout,j

electronic generator

exergy balance (fuel exergy = product exergy + exergy destruction)

E in = E out + E D,EG

cost balance

C in + Z EG = C out

F principle (no auxiliary equations are needed)

P principle (no auxiliary equations are needed)


B.2 real and unavoidable conditions of involved components 129

b.2 Real and unavoidable conditions of involved components

Table B.1: Real and unavoidable conditions of all involved components

Comp. Variables Unit Real Unavoidable

FUR , pwf 1, bar 1.10, 25.1 1.03, 17.0


SSH2 tup , pwf C, bar 581, 4.8 300, 2.0
FSH tup , pwf C, bar 428, 5.0 250, 2.0
FRHC tup , pwf C, bar 358, 1.7 150, 1.0
FRHH tup , pwf C, bar 239, 2.0 150, 1.0
PRH2 tup , pwf C, bar 308, 2.5 150, 1.0
PSH tup , pwf C, bar 300, 3.8 250, 2.0
ECO1 tup , pwf C, bar 193, 1.0 150, 0.5
PRH1 tup , pwf C, bar 299, 2.5 275, 1.0
ECO2 tup , pwf C, bar 104, 1.0 80, 0.5
APH tlow , pair C, bar 101, 0.011 65, 0.050
PAF s , m 1, 1 0.880, 0.990 0.920, 0.999
SAF s , m 1, 1 0.875, 0.990 0.910, 0.999
IDF s , m 1, 1 0.870, 0.990 0.900, 0.999
HPT1 s , m 1, 1 0.898, 0.998 0.910, 0.999
HPT2 s , m 1, 1 0.921, 0.998 0.930, 0.999
IPT1 s , m 1, 1 0.893, 0.998 0.940, 0.999
IPT2 s , m 1, 1 0.910, 0.998 0.950, 0.999
IPT3 s , m 1, 1 0.982, 0.998 0.990, 0.999
LPT1 s , m 1, 1 0.837, 0.998 0.970, 0.999
LPT2 s , m 1, 1 0.938, 0.998 0.940, 0.999
LPT3 s , m 1, 1 0.760, 0.998 0.820, 0.999
LPT4 s , m 1, 1 0.912, 0.998 0.950, 0.999
LPT5 s , m 1, 1 0.903, 0.998 0.910, 0.999
LPT6 s , m 1, 1 0.629, 0.998 0.880, 0.999
COND tup C 5.0 3.0
FWH1 tup , tlow , pwf C, C, bar , 5.6, 0.54 , 2.0, 0.3
FWH2 tup , tlow , pwf C, C, bar 2.8, , 0.54 1.0, , 0.3
FWH3 tup , tlow , pwf C, C, bar 2.8, , 0.54 1.0, , 0.3
FWH4 tup , tlow , pwf C, C, bar 2.8, , 0.54 1.0, , 0.3
FWH5 tup , tlow , pwf C, C, bar 2.8, 5.6, 0.54 1.0, 3.0, 0.3

continued on the next page


130 additional materials for the analysis and evaluation

(Continued) Real and unavoidable conditions of all involved components

Comp. Variables Unit Real Unavoidable

DA pwf C 0.07 0.03


FWH6 tup , tlow , pwf C, C, bar 0, 5.6, 1.09 -2.0, 3.0, 0.75
FWH7 tup , tlow , pwf C, C, bar 0, 5.6, 1.12 -2.0, 3.0, 0.75
FWH8 tup , tlow , pwf C, C, bar -1.7, 5.6, 1.11 -3.0, 3.0, 0.75
OT s , m 1, 1 0.810, 0.998 0.880, 0.999
FP s , m 1, 1 0.818, 0.998 0.900, 0.999
CP s , m 1, 1 0.870, 0.998 0.950, 0.999
CWP s , m 1, 1 0.800, 0.998 0.930, 0.999
EG m 1 0.9895 0.9930
excess air ratio/ 1
tup : temperature difference between hot inlet and cold outlet for counter-current
heat exchangers, between hot inlet and cold inlet for co-current heat exchangers
tlow : temperature difference between hot outlet and cold inlet for counter-current
heat exchangers, between hot outlet and cold outlet for co-current heat exchangers
B.3 stream data of the simulation under real conditions 131

b.3 Stream data of the simulation under real conditions

Table B.2: The stream data of the design condition (100% load)

Stream No. m / kgs1 p / bar t / C E / MW c / MJ1

2 815.7 262.5 601.4 1434 1.00


3 774.1 81.13 408.2 1097 1.00
4 679.5 61.78 368.8 914.7 1.00
5 679.5 59.28 475.0 1023 0.97
6 679.5 56.78 501.0 1048 0.97
7 679.5 55.08 577.0 1127 0.95
8 679.5 53.08 600.0 1149 0.95
9 644.5 23.21 470.1 911.3 0.95
10 573.0 11.14 366.6 686.4 0.95
11 538.7 6.190 287.1 560.2 0.95
12 269.3 6.190 287.1 280.1 0.95
13 250.4 2.475 195.3 211.3 0.95
14 229.1 0.657 88.27 138.4 0.95
15 229.1 0.118 49.08 82.20 0.95
16 269.3 6.190 287.1 280.1 0.95
17 250.4 2.475 187.6 209.7 0.95
18 226.4 0.274 67.03 106.5 0.95
19 226.4 0.118 49.08 80.99 0.95
24 546.4 0.062 36.76 6.509 0.95
25 546.4 16.08 36.86 7.405 1.16
26 546.4 15.54 39.64 8.154 1.19
27 546.4 15.00 63.07 16.57 1.19
28 546.4 14.46 84.14 26.77 1.16
29 546.4 13.92 122.6 51.03 1.13
30 618.7 13.92 123.0 58.06 1.12
32 618.7 13.38 155.2 86.88 1.10
34 815.7 10.44 181.8 150.4 1.08
35 815.7 317.3 187.8 180.9 1.23
37 815.7 316.2 218.5 227.4 1.21
39 815.7 315.1 275.5 328.4 1.18
41 815.7 314.0 295.6 369.2 1.18
42 815.7 314.0 295.6 369.2 1.18

continued on the next page


132 additional materials for the analysis and evaluation

(Continued) The stream data of the design condition (100% load)

Stream No. m / kgs1 p / bar t / C E / MW c / MJ1

43 300.6 314.0 295.6 136.1 1.18


44 300.6 313.0 327.0 162.4 1.26
45 515.1 314.0 295.6 233.1 1.18
46 515.1 313.0 331.0 284.6 1.14
47 815.7 313.0 329.5 447.0 1.18
49 815.7 292.1 441.0 1069 1.06
50 815.7 288.3 474.0 1172 1.04
51 815.7 284.1 513.0 1265 1.03
52 815.7 279.3 558.0 1354 1.01
53 815.7 274.3 605.0 1438 1.00
54 41.65 81.13 408.2 59.01 1.00
55 41.65 78.70 406.5 58.87 1.00
57 41.65 78.70 281.1 16.85 1.00
59 94.60 61.78 368.8 127.4 1.00
60 94.60 59.93 367.2 127.0 1.00
62 136.2 59.93 224.1 36.72 1.00
64 34.93 23.21 470.1 49.39 0.95
65 34.93 22.51 469.7 49.26 0.95
67 171.2 22.51 193.4 35.32 0.99
69 71.58 11.14 366.6 85.75 0.95
70 71.58 10.58 366.1 85.29 0.95
71 45.70 10.58 366.1 54.46 0.95
72 45.70 0.062 36.76 12.93 0.95
75 25.88 10.58 366.1 30.84 0.95
76 34.30 6.190 287.1 35.67 0.95
77 34.30 5.881 286.7 35.45 0.95
79 34.30 5.881 128.6 3.450 0.95
81 18.97 2.475 195.3 16.01 0.95
82 18.97 2.475 187.6 15.89 0.95
83 37.94 2.475 191.5 31.90 0.95
84 37.94 2.351 191.2 31.65 0.95
85 72.24 2.351 125.4 6.936 0.95
86 72.24 13.92 125.6 7.032 1.05
87 21.27 0.657 88.27 12.84 0.95

continued on the next page


B.3 stream data of the simulation under real conditions 133

(Continued) The stream data of the design condition (100% load)

Stream No. m / kgs1 p / bar t / C E / MW c / MJ1

88 21.27 0.624 86.94 12.71 0.95


89 21.27 0.624 86.94 1.073 0.95
91 23.98 0.274 67.03 11.28 0.95
92 23.98 0.260 65.87 11.13 0.95
93 23.98 0.260 65.87 0.746 0.95
94 45.24 0.260 42.46 0.676 0.95
97 19917 4.330 18.02 104.0 0.18
102 19917 4.280 31.76 199.2 0.18
104 230.1 1.144 32.05 3.810 1.61
113 711.5 1.058 24.40 6.927 1.01
118 893.1 1.031 302.5 96.52 0.67
119 104.7 1.020 25.00 2442 0.39
121 22.31 1.012 600.0 6.650 0.39
124 1008 0.998 1139 903.6 0.39
125 1008 0.998 1033 794.0 0.39
126 1008 0.998 935.4 695.3 0.39
127 1008 0.998 839.0 601.8 0.39
128 1008 0.998 809.0 573.6 0.39
129 1008 0.998 773.7 540.8 0.39
130 553.7 0.998 773.7 296.9 0.39
131 553.7 0.998 523.7 179.8 0.39
132 553.7 0.983 372.0 119.5 0.39
133 454.8 0.998 773.7 243.9 0.39
134 454.8 0.998 430.9 116.7 0.39
135 454.8 0.983 334.3 87.33 0.39
136 1008 0.983 355.1 206.6 0.39
145 1057 0.972 120.3 87.62 0.39
146 1057 1.049 129.5 96.97 0.53
134 additional materials for the analysis and evaluation

b.4 Additional materials for economic evaluation

Table B.3: Weight estimation and weighted factors for heat surfaces in boiler

Tube size ro Tube surface Weight Weighted


Comp. Material
mmmm area/ m2 ton factor

WWS 15CrMoG 38.1 7 8,880 406.77 1.0


WWV 12Cr1MoVG 31.8 7 4,603 207.24 1.0
SSH1 T91, S304H 41.3 7 3,308 143.66 3.0
SSH2 S304H 47.7 10 2,880 180.44 3.0
FSH S304H, HR3C 44.5 10 4,105 252.27 3.0
FRHC S304H, HR3C 57 6 5,916 233.11 3.0
FRHH S304H, HR3C 57 6 2,843 112.02 3.0
PRH2 T91 63.5 5 3,305 109.57 1.5
PSH 12Cr1MoVG, T91 46.1 6 25,106 1038.82 1.0
ECO1 SA-210C 50.8 9 14,190 833.25 0.8
PRH1 12Cr1MoVG, T91 63.5 5 25,137 833.38 1.0
ECO2 SA-210C 50.8 9 20,112 1180.99 0.8
pipe thickness (mm); ro pipe radius (mm).
Weight of heat exchanger is estimated by [0.02491 (ro ) L] /1000 ton, where units of , ro ,
and L (total pipe length) should be m.
WWSspiral part of WW; WWVvertical part of WW.

Table B.4: The module factors for different types of components (Turton et al., 2008)

Comp. name Factor Comp. name Factor

Fired heaters and furnace 2.8 Other heaters 1.7


Main turbine 6 Drive turbine for pumps 3.7
Fans 2.8 Pumps 1.3
Electric generator 1.7
B.5 results of conventional and advanced analyses 135

b.5 Results of conventional and advanced analyses

Table B.5: Conventional exergy and exergoeconomic analysis of the real case (Table B.2)

Exergy analysis Exergoeconomic analysis


Name
E F E P E D yD cF cP C D Z C D +Z r f
MW MW MW % % /MJ /MJ /s /s $/s % %

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

continued on the next page


136 additional materials for the analysis and evaluation

(Continued) Conventional exergy and exergoeconomic analysis of the real case

Exergy analysis Exergoeconomic analysis


Name
E F E P E D yD cF cP C D Z C D +Z r f
MW MW MW % % /MJ /MJ /s /s $/s % %

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

Table B.6: Advanced exergy analysis of the coal-fired power plant

UN,EN UN,EX AV,EN AV,EX


Name E PUN E D E D
EN E EN E EX
D D E D
UN E D
AV E D E D E D E D
MW MW MW MW MW MW MW MW MW MW MW

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

continued on the next page


138 additional materials for the analysis and evaluation

(Continued) Advanced exergy analysis of the coal-fired power plant

UN,EN UN,EX AV,EN AV,EX


Name E PUN E D E D
EN E EN E EX
D D E D
UN E D
AV E D E D E D E D
MW MW MW MW MW MW MW MW MW MW MW

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

Table B.7: The splitting of investment costs in advanced exergoeconomic analysis

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

continued on the next page


140 additional materials for the analysis and evaluation

(Continued) The splitting of investment costs in advanced exergoeconomic analysis

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

UN,EN UN,EX AV,EN AV,EX


Name C D C D
EN C D
EX C D
UN C D
AV C D C D C D C D
/s /s /s /s /s /s /s /s /s

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

continued on the next page


142 additional materials for the analysis and evaluation

(Continued) The splitting of exergy destruction costs in advanced exergoeconomic analysis

UN,EN UN,EX AV,EN AV,EX


Name C D C D
EN C D
EX C D
UN C D
AV C D C D C D C D
/s /s /s /s /s /s /s /s /s

CP 0.18 0.09 0.08 0.08 0.09 0.04 0.04 0.05 0.04


CWP 2.16 2.16 0.83 1.33 0.83 1.33
EG 12.0 11.6 0.32 7.98 3.99 7.74 0.24 3.91 0.08
For splitting the exergy destruction 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).
C
A D D I T I O N A L M AT E R I A L S F O R T H E
SUPERSTRUCTURE-FREE SYNTHESIS OF
THERMAL POWER PLANTS

This appendix presents additional materials for the cases presented in chapter 5.

c.1 Illustrative study for evaluating the superstructure-free approach


The specified energy conversion hierarchy for evaluating the superstructure-free ap-
proach (section 5.3) is presented in Fig. C.1. Some intermediate best-known solutions
identified at certain specific generations (Fig. C.4) are also presented to give an intuitive
understanding of the structural evolution.

c.1.1 Specified energy conversion hierarchy

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.

Meta level Meta

Deletion Insertion Parallel connection Serial connection


allowed allowed allowed allowed

Expander Compressor Heater Cooler

Steam Water Steam Water Steam


expander pump heater heater cooler
Function level

Steam Water Superheater Steam Feedwater


turbine pump generator preheater Condenser

Technology level

Figure C.1: Specified energy conversion hierarchy for the illustrative study

143
144 additional materials for the superstructure-free synthesis

c.1.2 Initial and optimal solutions

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)

steam turbine heat

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

c.1.3 Some intermediate solutions of one representative run

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

c.2 Optimal synthesis of complex coal-fired power plants


The specified energy conversion hierarchy for synthesizing the complex PCPPs is pre-
sented in Fig. C.5. The synthesis is performed with respect to the thermal efficiency,
the COE, and both of them, respectively.

c.2.1 Specified energy conversion hierarchy

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.

Meta level Meta

Deletion Insertion Parallel connection Serial connection


allowed allowed allowed allowed

Expander Compressor Heater Cooler

Steam Water Steam Water Steam


expander pump heater heater cooler
Function level

Steam Water Superheater Steam Feedwater Desuper-


turbine pump preheater heater Condenser
generator
Technology level

Figure C.5: Specified energy conversion hierarchy for synthesizing complex PCPPs
C.2 optimal synthesis of complex coal-fired power plants 147

c.2.2 Single-objective synthesis for maximizing thermal efficiency

c.2.2.1 Initial and optimal solutions


The flowsheets of the initial (Fig. C.6a) and optimal (Fig. C.6b) solutions identified in all
optimization runs are presented, with their temperature-entropy diagrams (Fig. C.7).

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

c.2.2.2 Some intermediate solutions of the representative run

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

750 Temperature / K 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.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

c.2.2.3 Computational efforts


The computation runs consider a maximum mutation strength of 5. Note the net power
output of the plant is fixed (1000 MW) but no longer the mass flow rate of the main
steam (as in the illustrative study in section 5.3 and App. C.1).

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 (%)

Run tot / min / % / min n / 1 e/ %

1 987 49.600 615 217 0


2 851 49.600 538 220 0
3 954 49.594 913 308 0.011
4 1060 49.590 884 281 0.020
5 1002 49.600 811 267 0
6 862 49.585 532 211 0.031

Geometric mean 950 49.595 697 248 0.010


Termination after 8000 successful evaluations with settings: = 20, = 25, Pnet =
1000 MW, and a maximum mutation strength of 5
The relative difference between the objective values of the best-known solution from all
runs (all ) and the best solution identified in each run i ( i ): e = 100 |all i |/all
150 additional materials for the superstructure-free synthesis

c.2.3 Single-objective synthesis for minimizing cost of electricity

c.2.3.1 Initial and optimal solutions


The flowsheets of the initial (Fig. C.9a) and optimal (Fig. C.9b) solutions identified in all
optimization runs are presented, with their temperature-entropy diagrams (Fig. C.10).

a)

steam turbine
condenser

steam feedwater
generator preheater
pump

b) reheater

ET1
steam
turbine
steam ET2
generator condenser

desuperheater
feedwater
preheater

FWH3 FWH1 pump


FWH5

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

c.2.3.2 Some intermediate solutions of the representative run

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

c.2.3.3 Computational efforts


All computation runs consider a maximum mutation strength of 5 and a fixed net
power output of 1000 MW.

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 %

1 1977 5.0635 1130 202 0.015


2 1765 5.0636 1345 282 0.016
3 1895 5.0627 1345 259 0.000
4 1785 5.0627 1173 232 0.000
5 1553 5.0635 1126 268 0.015
6 1690 5.0643 1413 303 0.031

Geometric mean 1772 5.0634 1250 255 0.013


Termination after 8000 successful evaluations with settings: = 20, = 25, Pnet =
1000 MW, and a maximum mutation strength of 5
The relative difference between the objective values of the best-known solution from all
all ) and the best solution identified in each run i (COE
runs (COE i ):
all COE
e = 100 |COE i |/COE
all

c.2.4 Bi-objective optimal 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)

Figure C.12: Economically- and thermodynamically-optimal solutions of the representative run


(Fig. 5.15b) for bi-objective optimal synthesis: a) economically-optimal solution
(COE = 5.065 /kWh, = 51.92%), b) thermodynamically-optimal solution
(COE = 7.480 /kWh, = 57.10%)
BIBLIOGRAPHY

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.)

Ahmetovic, E. and Grossmann, I. E. (2011). Global superstructure optimization for


the design of integrated process water networks. AIChE J., 57(2):434457. (Cited on
page 23.)

Al-Sulaiman, F. A., Dincer, I., and Hamdullahpur, F. (2013). Thermoeconomic optimiza-


tion of three trigeneration systems using organic rankine cycles: Part IFormulations.
Energy Convers. Manage., 69:199208. (Cited on pages 10 and 115.)

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.)

Alkan, M. A., Keebas, A., and Yamankaradeniz, N. (2013). Exergoeconomic analysis


of a district heating system for geothermal energy using specific exergy cost method.
Energy, 60:426434. (Cited on page 10.)

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.)

Bagajewicz, M. J. and Manousiouthakis, V. (1992). Mass/heat-exchange network repre-


sentation of distillation networks. AIChE J., 38(11):17691800. (Cited on page 25.)

153
154 bibliography

Balas, E. (1985). Disjunctive programming and a hierarchy of relaxations for discrete


optimization problems. SIAM Journal on Algebraic Discrete Methods, 6(3):466486.
(Cited on page 17.)

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.)

Barnicki, S. D. and Siirola, J. J. (2004). Process synthesis prospective. Comput. Chem.


Eng., 28(4):441446. (Cited on page 21.)

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.)

Beyer, H. G. (2001). The Theory of Evolution Strategies. Springer Berlin Heidelberg,


Germany. (Cited on page 98.)

Beyer, H. G. and Schwefel, H. P. (2002). Evolution strategiesA comprehensive intro-


duction. Natural computing, 1(1):352. (Cited on page 19.)

Biegler, L. T. and Grossmann, I. E. (2004). Retrospective on optimization. Comput. Chem.


Eng., 28(8):11691192. (Cited on pages 16 and 24.)

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, C. and Roli, A. (2003). Metaheuristics in combinatorial optimization: Overview


and conceptual comparison. ACM Computing Surveys (CSUR), 35(3):268308. (Cited
on page 18.)

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.)

BP (2014). Energy Outlook 2035. BP. (Cited on page 1.)

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.)

Campbell, R. J. (2013). Increasing the efficiency of existing coal-fired power plants.


Technical report, Congressional Research Service. (Cited on page 1.)

Cardona, E. and Piacentino, A. (2006). A new approach to exergoeconomic analysis


and design of variable demand energy systems. Energy, 31(4):490515. (Cited on
page 11.)

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, J. (1987). Comments on improvements on a replacement for the logarithmic


mean. Chem. Eng. Sci., 42(10):24882489. (Cited on page 20.)

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.)

China Power Engineering Consulting Group Corporation (2009). Reference construction


cost index for the quota design of thermal power plants. China Electric Press: Peking,
China. (Cited on pages 50, 76, 77, and 119.)

Chu, S. and Majumdar, A. (2012). Opportunities and challenges for a sustainable en-
ergy future. Nature, 488(7411):294303. (Cited on page 1.)

Cooke, D. (1985). On prediction of off-design multistage turbine pressures by stodols


ellipse. ASME J. Eng. Gas Turb. Power, 107(3):596606. (Cited on page 115.)

Cooke, D. H. (1983). Modeling of off-design multi-stage turbine pressures by stodolas


ellipse. Technical report, Energy Incoportated PEPSE Users Group Meeting, Rich-
mond, Virginia. (Cited on pages 20 and 115.)

Cziesla, F. and Tsatsaronis, G. (2002). Iterative exergoeconomic evaluation and im-


provement of thermal power plants using fuzzy inference systems. Energy Convers.
Manage., 43(9-12):15371548. (Cited on pages 10 and 117.)

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

Daichendt, M. and Grossmann, I. (1994). Preliminary screening procedure for the


minlp synthesis of process systemsII. heat exchanger networks. Comput. Chem. Eng.,
18(8):679709. (Cited on page 24.)

Daichendt, M. M. and Grossmann, I. E. (1998). Integration of hierarchical decomposi-


tion and mathematical programming for the synthesis of process flowsheets. Comput.
Chem. Eng., 22(12):147175. (Cited on pages 24, 25, and 116.)

Deb, K. (2001). Multi-objective optimization using evolutionary algorithms, volume 16. John
Wiley & Sons. (Cited on page 28.)

Deb, K. (2014). Multi-objective optimization. In Search methodologies, pages 403449.


Springer. (Cited on pages 27 and 30.)

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.)

Dincer, I. and Al-Muslim, H. (2001). Thermodynamic analysis of reheat cycle steam


power plants. International Journal of Energy Research, 25(8):727739. (Cited on
page 14.)

Douglas, J. M. (1985). A hierarchical decision procedure for process synthesis. AIChE


J., 31(3):353362. (Cited on page 22.)

Douglas, J. M. (1988). Conceptual design of chemical processes. McGraw-Hill. (Cited on


page 22.)

Douglas, J. M. (1995). Synthesis of separation system flowsheets. AIChE J., 41(12):2522


2536. (Cited on pages 22 and 23.)

Drud, A. (2004). CONOPT documentation. Technical report, ARKI Consulting and


Development A/S, Bagsvaerd, Denmark. (Cited on pages 16, 96, and 118.)

Drud, A. S. (1994). CONOPTA large-scale GRG code. ORSA Journal on Computing,


6(2):207216. (Cited on page 20.)

Duran, M. and Grossmann, I. (1986). An outer-approximation algorithm for a class of


mixed-integer nonlinear programs. Mathematical Programming, 36(3):307339. (Cited
on pages 17 and 22.)

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.)

Emmerich, M. T. (2002). Optimisation of thermal power plant designs: A graph-


based adaptive search approach. In Parmee, I., editor, Adaptive Computing in Design
and Manufacture V, pages 8798. Springer-Verlag London, UK. (Cited on pages 26
and 27.)
bibliography 157

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.)

Escobar, M., Trierweiler, J. O., and Grossmann, I. E. (2014). A heuristic Lagrangean


approach for the synthesis of multiperiod heat exchanger networks. Appl. Therm.
Eng., 63(1):177191. (Cited on page 23.)

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.)

Feng, W. (2008). Technical innovation of energy-saving and emission-reducing during


debugging of 1000 MW ultra-supercritical units of waigaoqiao IIIstage engineering
project. East China Electric Power, 6. (Cited on page 5.)

FICO (2009). XPRESS Optimization Suite: MIP formulations and linearizations. Tech-
nical report, Fair Isaac Corporation. (Cited on pages 20 and 118.)

Floudas, C. and Gounaris, C. (2009). A review of recent advances in global optimiza-


tion. Journal of Global Optimization, 45(1):338. (Cited on page 16.)

Floudas, C. A. (1995). Nonlinear and mixed-integer optimization: Fundamentals and applica-


tions. Oxford University Press. (Cited on page 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. (1991). Intelligent functional approach: A method for analysis and


optimal synthesis-design-operation of complex systems. Internal Journal of Energy,
Environment, Economics, 4(1):267274. (Cited on page 10.)

Frangopoulos, C. A. (1983). Thermoeconomic functional analysis: A method for optimal


design or improvement of complex thermal systems. PhD thesis, Georgia Institute of
Technology. (Cited on pages 9 and 10.)

Frangopoulos, C. A. (1988). Functional decomposition for optimal design of complex


thermal systems. Energy, 13(3):239244. (Cited on page 9.)
158 bibliography

Frangopoulos, C. A. (1992). Optimal synthesis and operation of thermal systems by


the thermoeconomic functional approach. ASME J. Eng. Gas Turb. Power, 114:707714.
(Cited on pages 9, 10, 25, and 117.)

Frangopoulos, C. A. (1994). Application of the thermoeconomic functional approach


to the CGAM problem. Energy, 19(3):323342. (Cited on pages 9 and 10.)

Frangopoulos, C. A. (2003). Methods of energy systems optimization. In Summer School:


Optimization of Energy Systems and Processes. (Cited on page 15.)

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.)

Fukuda, Y. (2010). Development of advanced ultra supercritical fossil power plants


in Japan: Materials and high temperature corrosion properties. Mater. Sci. Forum,
696:236241. (Cited on page 6.)

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.)

Glover, F. and Kochenberger, G. A. (2003). Handbook of metaheuristics. Springer. (Cited


on pages 15, 16, and 19.)

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.)

Grekas, D. N. and Frangopoulos, C. A. (2007). Automatic synthesis of mathematical


models using graph theory for optimisation of thermal energy systems. Energy Con-
vers. Manage., 48(11):28182826. (Cited on page 23.)

Grossmann, I. and Ruiz, J. (2012). Generalized disjunctive programming: A framework


for formulation and alternative algorithms for MINLP optimization. In Lee, J. and
Leyffer, S., editors, The IMA Volumes in Mathematics and its Applications, volume 154,
pages 93115. Springer New York. (Cited on pages 17 and 18.)
bibliography 159

Grossmann, I. E. and Biegler, L. T. (2004). Part II. Future perspective on optimization.


Comput. Chem. Eng., 28(8):11931218. (Cited on page 16.)

Grossmann, I. E., Caballero, J. A., and Yeomans, H. (2000). Advances in mathematical


programming for the synthesis of process systems. Latin American Applied Research,
30(4):263284. (Cited on page 22.)

Grossmann, I. E. and Guilln-Goslbez, G. (2010). Scope for the application of math-


ematical programming techniques in the synthesis and planning of sustainable pro-
cesses. Comput. Chem. Eng., 34(9):13651376. (Cited on pages 2, 17, 21, 22, 23, 24,
and 25.)

Grossmann, I. E. and Trespalacios, F. (2013). Systematic modeling of discrete-


continuous optimization models through generalized disjunctive programming.
AIChE J., 59(9):32763295. (Cited on page 17.)

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.)

Haywood, R. W. (1949). A generalized analysis of the regenerative steam cycle for


a finite number of heaters. Proceedings of the Institution of Mechanical Engineers,
161(1):157164. (Cited on page 21.)

Henao, C. A. and Maravelias, C. T. (2011). Surrogate-based superstructure optimization


framework. AIChE J., 57(5):12161232. (Cited on page 25.)

Hepbasli, A. and Keebas, A. (2013). A comparative study on conventional and ad-


vanced exergetic analyses of geothermal district heating systems based on actual
operational data. Energy Build., 61:193201. (Cited on page 13.)

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 (2012). Technology roadmap: High-efficiency, low-emissions coal-fired power generation.


International Energy Agency. (Cited on pages 1 and 6.)

IEA (2014). Key world energy statistics 2014. International Energy Agency. (Cited on
page 1.)

Iorio, A. W. and Li, X. (2005). Solving rotated multi-objective optimization problems


using differential evolution. In AI 2004: Advances in Artificial Intelligence, pages 861
872. Springer. (Cited on page 31.)

Jacobson, M. Z. (2009). Review of solutions to global warming, air pollution, and


energy security. Energy & Environmental Science, 2(2):148173. (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.)

Kaibel, G. and Schoenmakers, H. (2002). Process synthesis and design in industrial


practice. Computer Aided Chemical Engineering, 10:922. (Cited on page 21.)

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.)

Kanoglu, M., Ayanoglu, A., and Abusoglu, A. (2011). Exergoeconomic assessment of a


geothermal assisted high temperature steam electrolysis system. Energy, 36(7):4422
4433. (Cited on page 10.)
bibliography 161

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.)

Kavvadias, K. and Maroulis, Z. (2010). Multi-objective optimization of a trigeneration


plant. Energy Policy, 38(2):945954. (Cited on page 31.)

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.)

Kppe, M. (2012). On the complexity of nonlinear mixed-integer optimization. In


Mixed Integer Nonlinear Programming, pages 533557. Springer. (Cited on page 16.)

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.)

Kravanja, Z. and Grossmann, I. E. (1997). Multilevel-hierarchical MINLP synthesis of


process flowsheets. Comput. Chem. Eng., 21, Supplement(0):S421S426. (Cited on
pages 22, 23, and 24.)
162 bibliography

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. (2004). Energy, economy and environment as objectives


in multi-criterion optimization of thermal systems design. Energy, 29(8):11391157.
(Cited on page 31.)

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.)

Lazzaretto, A. and Tsatsaronis, G. (1999). On the calculation of efficiencies and costs


in thermal systems. In Proceedings of the ASME advanced energy systems division,
volume 39, pages 421430. American Society of Mechanical Engineers. (Cited on
page 9.)

Lazzaretto, A. and Tsatsaronis, G. (2006). SPECO: A systematic and general method-


ology for calculating efficiencies and costs in thermal systems. Energy, 31(8-9):1257
1289. (Cited on pages 8, 9, 10, 38, and 125.)

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.)

Linnhoff, B. (1993). Pinch analysis: A state-of-the-art overview: Techno-economic anal-


ysis. Chemical engineering research & design, 71(5):503522. (Cited on page 22.)

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.)

Madavan, N. K. (2002). Multiobjective optimization using a Pareto differential evolu-


tion approach. In Computational Intelligence, Proceedings of the World on Congress on,
volume 2, pages 11451150. IEEE. (Cited on page 31.)

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.)

Manninen, J. and Zhu, X. X. (2001). Level-by-level flowsheet synthesis methodology


for thermal system design. AIChE J., 47(1):142159. (Cited on page 24.)

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. (2004). Survey of multi-objective optimization methods


for engineering. Structural and multidisciplinary optimization, 26(6):369395. (Cited on
page 27.)

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.)

Mavromatis, S. and Kokossis, A. (1998). Hardware composites: A new conceptual tool


for the analysis and optimisation of steam turbine networks in chemical process
industries: Part I principles and construction procedure. Chem. Eng. Sci., 53(7):1405
1434. (Cited on pages 20 and 22.)

Mavrotas, G. (2009). Effective implementation of the epsilon-constraint method in


multi-objective mathematical programming problems. Applied Mathematics and Com-
putation, 213(2):455465. (Cited on pages 28 and 89.)

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. and Ismail-Yahaya, A. (2002). Multiobjective robust design using physical


programming. Structural and multidisciplinary optimization, 23(5):357371. (Cited on
page 28.)

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.)

Mor, J. and Wild, S. (2009). Benchmarking derivative-free optimization algorithms.


SIAM J. Optim., 20(1):172191. (Cited on page 16.)

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. (2008b). A new approach to the exergy analysis of


absorption refrigeration machines. Energy, 33(6):890907. (Cited on page 13.)

Morosuk, T. and Tsatsaronis, G. (2009a). Advanced exergetic evaluation of refrigera-


tion machines using different working fluids. Energy, 34(12):22482258. (Cited on
pages 11 and 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.)

Nocedal, J. and Wright, S. (2006a). Conjugate gradient methods. Springer. (Cited on


page 16.)

Nocedal, J. and Wright, S. (2006b). Numerical Optimization, 2rd Edition. Springer-Verlag


New York, USA. (Cited on page 16.)

Omer, A. M. (2008). Energy, environment and sustainable development. Renewable and


sustainable energy reviews, 12(9):22652300. (Cited on page 1.)

Pachernegg, S. (1969). A closer look at the Willans-line. Technical report, SAE. (Cited
on page 20.)

Papalexandri, K. P. and Pistikopoulos, E. N. (1996). Generalized modular representa-


tion framework for process synthesis. AIChE J., 42(4):10101032. (Cited on page 25.)

Paterson, W. (1984). A replacement for the logarithmic mean. Chem. Eng. Sci.,
39(11):16351636. (Cited on page 20.)

Paulus, D. M. and Tsatsaronis, G. (2006). Auxiliary equations for the determination of


specific exergy revenues. Energy, 31(15):32353247. (Cited on page 11.)

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., and Morosuk, T. (2012a). Advanced exergoenviron-


mental analysis of a near-zero emission power plant with chemical looping combus-
tion. Environ. Sci. Technol., 46(5):30013007. (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.)

Piacentino, A. and Cardona, E. (2010a). Scope-oriented thermoeconomic analysis of


energy systems. Part II: Formation structure of optimality for robust design. Appl.
Energy, 87(3):957970. (Cited on page 11.)

Piacentino, A. and Cardona, F. (2010b). Scope-oriented thermoeconomic analysis of en-


ergy systems. Part I: Looking for a non-postulated cost accounting for the dissipative
devices of a vapour compression chiller. is it feasible? Appl. Energy, 87(3):943956.
(Cited on pages 10 and 11.)

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.)

Rabek, G. (1963). Die ermittlung der betriebsverhltnisse von speisewasservorwrmern


bei verschiedenen belastungen. Energie und Technik. (Cited on page 115.)

Rao, A. B. and Rubin, E. S. (2002). A technical, economic, and environmental assess-


ment of amine-based CO2 capture technology for power plant greenhouse gas con-
trol. Environmental Science & Technology, 36(20):44674475. (Cited on page 1.)

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.)

Rios, L. M. and Sahinidis, N. V. (2013). Derivative-free optimization: A review of algo-


rithms and comparison of software implementations. Journal of Global Optimization,
56(3):12471293. (Cited on page 16.)

Rosen, M. A. and Dincer, I. (2003). Thermoeconomic analysis of power plants: An


application to a coal fired electrical generating station. Energy Convers. Manage.,
44(17):27432761. (Cited on page 14.)
bibliography 167

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.)

Schulze, C. W. (2014). A Contribution to Numerically Efficient Modelling of Thermodynamic


Systems. PhD thesis, Technischen Universitt Carolo-Wilhelmina zu Braunschweig.
(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.)

Seader, J. D. and Westerberg, A. W. (1977). A combined heuristic and evolutionary


strategy for synthesis of simple separation sequences. AIChE J., 23(6):951954. (Cited
on page 25.)

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.)

Sherali, H. and Alameddine, A. (1992). A new reformulation-linearization technique


for bilinear programming problems. Journal of Global Optimization, 2(4):379410.
(Cited on page 17.)
168 bibliography

Siirola, J. J. (1996). Strategic process synthesis: Advances in the hierarchical approach.


Comput. Chem. Eng., 20, Supplement 2(0):S1637S1643. (Cited on page 22.)

Siirola, J. J. and Rudd, D. F. (1971). Computer-aided synthesis of chemical process


designs: From reaction path data to the process task network. Ind. Eng. Chem. Fund.,
10(3):353362. (Cited on page 22.)

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.)

Silvestri Jr, G. J. (1995). Boiler feedpump turbine drive/feedwater train arrangement.


US Patent 5,404,724. (Cited on page 7.)

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.)

Skiborowski, M., Harwardt, A., and Marquardt, W. (2012). Conceptual design of


distillation-based hybrid separation processes. Annual review of chemical and biomolec-
ular engineering, 4:4568. (Cited on page 23.)

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.)

Stein, O., Oldenburg, J., and Marquardt, W. (2004). Continuous reformulations of


discrete-continuous optimization problems. Comput. Chem. Eng., 28(10):19511966.
(Cited on page 25.)

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.)

Stephanopoulos, G. and Westerberg, A. W. (1976). Studies in process synthesis-II: Evo-


lutionary synthesis of optimal process flowsheets. Chem. Eng. Sci., 31(3):195204.
(Cited on page 25.)
bibliography 169

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.)

Surma, J. and Braunschweig, B. (1996). Case-base retrieval in process engineering: Sup-


porting design by reusing flowsheets. Engineering Applications of Artificial Intelligence,
9(4):385391. (Cited on page 22.)

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.)

Toffolo, A. (2014). A synthesis/design optimization algorithm for rankine cycle based


energy systems. Energy, 66(0):115127. (Cited on pages 26 and 27.)

Toffolo, A. and Lazzaretto, A. (2002). Evolutionary algorithms for multi-objective en-


ergetic and economic optimization in thermal system design. Energy, 27(6):549567.
(Cited on page 31.)

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.)

Tsatsaronis, G. (1984). Energy Economics and Management in Industry, chapter Combi-


nation of exergetic and economic analysis in energy conversion processes, pages
151157. Pergamon Press, Oxford, England. (Cited on pages 8 and 9.)

Tsatsaronis, G. (1993). Thermoeconomic analysis and optimization of energy systems.


Prog. Energy Combust. Sci., 19(3):227257. (Cited on page 9.)

Tsatsaronis, G. (1999a). Thermodynamic Optimization of Complex Energy Systems, vol-


ume 69, chapter Design Optimization Using Exergoeconomics, pages 101115.
Springer Netherlands, Netherlands. (Cited on pages 9 and 14.)
170 bibliography

Tsatsaronis, G. (1999b). Thermodynamic Optimization of Complex Energy Systems, chapter


Strengths and limitations of exergy analysis, pages 93100. Springer Netherlands,
Netherlands. (Cited on pages 11 and 12.)
Tsatsaronis, G. (2008). Recent developments in exergy analysis and exergoeconomics.
International Journal of Exergy, 5(5):489499. (Cited on pages 2 and 11.)
Tsatsaronis, G. (2011). Thermodynamics and the Destruction of Resources, chapter Exergoe-
conomics and Exergoenvironmental Analysis, pages 377401. Cambridge University
Press New York. (Cited on pages 8 and 9.)
Tsatsaronis, G., Kelly, S. O., and Morosuk, T. V. (2006a). Endogenous and exogenous
exergy destruction in thermal systems. In ASME 2006 International Mechanical En-
gineering Congress and Exposition, pages 311317. American Society of Mechanical
Engineers. (Cited on page 13.)
Tsatsaronis, G., Lin, L., and Pisa, J. (1993). Exergy costing in exergoeconomics. ASME
J. Energy Resour. Technol., 115(1):916. (Cited on page 9.)
Tsatsaronis, G. and Moran, M. J. (1997). Exergy-aided cost minimization. Energy Con-
vers. Manage., 38(15):15351542. (Cited on page 10.)
Tsatsaronis, G. and Morosuk, T. (2010). Advanced exergetic analysis of a novel system
for generating electricity and vaporizing liquefied natural gas. Energy, 35(2):820829.
(Cited on pages 11 and 13.)
Tsatsaronis, G., Morosuk, T., and Kelly, S. (2006b). Approaches for splitting the exergy
destruction into endogenous and exogenous parts. In The 5-th Workshop "Advances
in Energy Studies", Porto Venere, Italy, pages 1216. (Cited on page 13.)
Tsatsaronis, G. and Park, M.-H. (2002). On avoidable and unavoidable exergy de-
structions and investment costs in thermal systems. Energy Convers. Manage., 43(9-
12):12591270. (Cited on pages 8, 11, 12, and 37.)
Tsatsaronis, G. and Pisa, J. (1994). Exergoeconomic evaluation and optimization of
energy systemsApplication to the CGAM problem. Energy, 19(3):287321. (Cited
on page 9.)
Tsatsaronis, G. and Winhold, M. (1985a). Exergoeconomic analysis and evaluation
of energy-conversion plantsI. Analysis of a coal-fired steam power plant. Energy,
10(1):8194. (Cited on pages 9 and 14.)
Tsatsaronis, G. and Winhold, M. (1985b). Exergoeconomic analysis and evaluation of
energy-conversion plantsII. A new general methodology. Energy, 10(1):6980. (Cited
on page 9.)
Tsoukalas, A. and Mitsos, A. (2014). Multivariate McCormick relaxations. Journal of
Global Optimization, 59(2-3):633662. (Cited on page 20.)
Turchi, C. S., Langle, N., Bedilion, R., and Libby, C. (2011). Solar-augment potential
of US fossil-fired power plants. In ASME 2011 5th International Conference on Energy
Sustainability, pages 641651. American Society of Mechanical Engineers. (Cited on
page 8.)
bibliography 171

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.)

Uhlenbruck, S. and Lucas, K. (2000). Exergy-aided cost optimization using evolu-


tionary algorithm. Int.J. Applied Thermodynamics, 3(3):121127. (Cited on pages 15
and 117.)

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.)

Valero, A. and Torres, C. (1988). Algebraic thermodynamic analysis of energy systems.


Proceedings of approaches to the design and optimization of thermal systems AES, 7:1323.
(Cited on page 9.)

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.)

Vecchietti, A. and Grossmann, I. E. (1999). LOGMIP: A disjunctive 01 non-linear


optimizer for process system models. Comput. Chem. Eng., 23(4):555565. (Cited on
page 17.)

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. (2014). Automated Optimization-Based Synthesis of Distributed Energy Supply Sys-


tems. PhD thesis, RWTH Aachen Univeristy, Aachen, Germany. (Cited on page 19.)

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. (1994). Application of engineering functional analysis to the


analysis and optimization of the CGAM problem. Energy, 19(3):343364. (Cited on
pages 9 and 10.)
bibliography 173

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.)

Weir, C. (1964). A generalized analysis of regenerative feedheating trains. Proceedings


of the Institution of Mechanical Engineers, 179(1):477494. (Cited on page 21.)

Weir, C. D. (1960). Optimization of heater enthalpy rises in feed-heating trains. Proceed-


ings of the Institution of Mechanical Engineers, 174(1):769796. (Cited on page 21.)

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.)

Westerberg, A. W. (1991). Process engineering. In Colton, C. K., editor, Perspectives


in Chemical Engineering Research and Education, volume Volume 16, pages 499523.
Academic Press. (Cited on page 22.)

Westerberg, A. W. (2004). A retrospective on design and process synthesis. Comput.


Chem. Eng., 28(4):447458. (Cited on pages 21 and 24.)
174 bibliography

Williams, P. H. (2013). Model Building in Mathematical Programming, 5th Edition. John


Wiley & Sons. (Cited on page 118.)

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. and Grossmann, I. (1990). Simultaneous optimization models for heat


integrationII. Heat exchanger network synthesis. Comput. Chem. Eng., 14(10):1165
1184. (Cited on page 22.)

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

Yeomans, H. and Grossmann, I. E. (1999). A systematic modeling framework of su-


perstructure optimization in process synthesis. Comput. Chem. Eng., 23(6):709731.
(Cited on page 24.)

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.)

Zitzler, E. and Thiele, L. (1999). Multiobjective evolutionary algorithms: A comparative


case study and the strength Pareto approach. IEEE Trans. Evol. Comput., 3(4):257271.
(Cited on page 28.)
P U B L I C AT I O N S

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.