20 views

Uploaded by mfhaleem@pgesco.com

coal power plant

coal power plant

© All Rights Reserved

- Steam Turbine Construction
- Steam Turbines - a full range to fit your needs.pdf
- e. Steam Power Plant - Lecture
- High Pressure Boiler
- Lifecycle Optimizer
- Script MotorDesign
- Turbine Cycle Optimisation
- Why Steam Turbine Tripping Provided for Low Inlet Steam Pressure
- Prediction of Traffic Flow by an Adaptive Prediction System
- BHEL
- Mini Project
- Steam turbine and em
- syballus.pdf
- stgspecification-140417234455-phpapp01
- Thermal Power Plants Operation Detail
- BlendMaster
- Forced cooling of Turbine.
- 265
- Microsteam Brochure
- Turbine Efficiency Ch-8.1

You are on page 1of 198

vorgelegt von

Ligang Wang

geb. in Qingdao, China

der Technischen Universitt Berlin

zur Erlangung des akademischen Grades

- Dr.-Ing. -

genehmigte Dissertation

Promotionsausschuss

Berichter: Prof. Dr.-Ing. George Tsatsaronis

Berichter: Prof. Dr.-Ing. Andr Bardow

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

3.5 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.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 2.2 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

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

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 3.2 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

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

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

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

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.

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

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.

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

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

5

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

700 C

Boiler

VHPT HPT-IPT LPT

Steam turbine

Economizer

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)

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

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.

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.

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

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).

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).

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.

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.

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

exergy destruction

Variation range of

investment costs

A

Specific unavoidable

invesment cost

unit of product exergy

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))

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.

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/

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.

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

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.

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:

d, o

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

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.

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

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.

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

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.

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

Selection Recombination

(Crossover)

Evaluation Mutation

(and parent) individuals is performed, so that those individuals with larger possibility

of leading to the optimality survive and are selected as the offspring population. The

iteration continues until certain termination criterion, e. g., a limit of computation time,

fitness-evaluation number, or generation number, is reached.

Selection, crossover and mutation are three genetic operators of evolutionary algo-

rithms for maintaining local intensification and diversification of the search. Different

strategies on these three aspects lead to a variety of evolutionary algorithms. Selection

strategy mainly exerts influence on population diversity. One commonly used strategy

of selection is the ( + )-selection proposed in evolution strategies (Beyer and Schwe-

fel, 2002), where and , satisfying 1 , denote the sizes of parent and offspring

populations, respectively. Selection ranks the fitness of all + individuals and takes

the best individuals.

Depending on the search space and objective function, the crossover and/or the

mutation may or may not occur in specific instantiations of the algorithm (Beyer and

Schwefel, 2002; Glover and Kochenberger, 2003). Crossover intensifies the local search

but is not essential, as self-adaption of individuals is available due to mutation, which,

in contrast, is usually dispensable to avoid premature convergence toward sub-optimal

solutions. There are different mechanisms of crossover and mutation. For example, ge-

netic algorithm (Deb et al., 2002) usually employs bit strings to represent variables.

Typically, the crossover and mutation for integer variables can be readily and straight-

forwardly effected by gray coding, thus genetic algorithm is mainly applied for solving

combinatorial optimization problems. However, differential evolution (DE (Storn and

Price, 1997)), mentioned as the fastest evolution algorithm (Storn and Price, 1997), does

not rely on any coding but directly manipulates real-valued or discrete variables. Ba-

sically, for mutation DE adds the weighted difference between two parents variable

vectors to a third vector, thus the scheme remains completely self-organizing without

using separate probability distribution and has no limitations for implementation com-

pared to other evolutionary algorithms.

The optimization problems of thermal energy systems are usually highly constrained

and nonlinear, thus belong to (NLP) or (MINLP). The nonlinearity and integrity may

be led to by thermodynamic properties of working fluids, design and operational char-

acteristics of components, the investment cost functions of components, energy balance

20 analysis, optimization and synthesis of coal-fired power plants

equations, etc. These need to be well addressed, so that the problems, in the best case,

can be transformed to (LP) or (MILP) for deterministic optimizations.

For the properties of working fluids, particularly water and steam (IAPWS-IF97

(Wagner et al., 2000)), the highly nonlinear exact mathematical formulations can hardly

be employed. One direct means incorporates polynomial approximations of low de-

grees of nonlinearity in expense of accuracy (Savola, 2007; Jdes, 2009; Luo et al.,

2011; Manassaldi et al., 2011). However, inaccurate regressions may result frequently

in non-applicable "optimal" solutions. Moreover, integer variables may be explicitly

introduced if involving zonal or piecewise regressions.

Another approach evaluates the propertys value and associated derivatives of high

accuracy based on reformulated exact formulations or reprocessed steam tables. Exter-

nal evaluations can be fast and stable, such as in freesteam4 (based on reformulated

IAPWS-IF97), or in TILMedia Suite5 (bi-cubic spline interpolation of standard steam

tables Schulze 2014). Moreover, in these libraries, the discontinuities and even jumps

of the thermodynamic properties are smoothed, and the integer variables indicating

the state zones are encapsulated.

The nonlinear (or perhaps discrete) thermodynamic (operational) behavior of com-

ponents can be properly reformulated. For example, for modeling turbine, alternatives

include constant entropy efficiency model, Willans Line (Pachernegg, 1969), Turbine

Hardware Model (Mavromatis and Kokossis, 1998) and Stodola ellipse (Cooke, 1983).

In those models, the set of variables which the isentropic efficiency depends on dif-

fers, thus the predictions of the off-design behavior are also different in accuracy. For

heat exchangers, the logarithmic mean temperature difference can be replaced by a

refinement of the arithmetic mean (Paterson, 1984; Chen, 1987). While for mixers, the

discrete equality nonlinear relationship of the flow pressures between inlets and outlet

can be either relaxed as an inequality nonlinear constraint (Drud, 1994) or linearized

by introducing additional integer variables (FICO, 2009).

The investment cost functions are always needed if an economic objective is involved

in the optimization. A cost function links the purchased equipment cost of one com-

ponents with its key characteristic variables and associated flow parameters; thus, the

function may be of high nonlinearity. To cope with this, cost functions are usually re-

formulated with separable terms of each variable, which are subsequently piecewise

linearized with the aid of integer SOS2 variables (Tomlin, 1988).

Continuous nonconvex bilinear term (v1 v2 ) is another common source of nonlinear-

ity, e.g., the term m h involved in energy balance equations. This nonconvex nonlin-

earity is usually handled by a convex/concave McCormick relaxation (Tsoukalas and

Mitsos, 2014) or a quadratic reformulation. For the latter approach, two new variables

z1 = (v1 + v2 ) /2 and z2 = (v1 v2 ) /2 are introduced to replace the bilinear term

with z21 z22 . The quadratic term can also be further linearized by SOS2 variables.

Given a specific structure, the parametric selection and optimization of a steam cycle

becomes an easy task. Early work on optimizing steam cycles date back to 1950s and fo-

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.

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.

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.

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).

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

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.

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

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).

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.

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.

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:

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.

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)).

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.

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

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.

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

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

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)).

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.

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.

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.

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.

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.

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

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)

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

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

yD,tot = E D,tot / E F,tot , (3.3)

3.1 framework for comprehensive exergy-based evaluation 37

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.

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).

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):

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:

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:

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.

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

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.

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.

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

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

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 ) = 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

exergy destruction (or loss if emitted to the environment) can be calculated according

to Fig. 3.2b:

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

temperature. The model for calculating its endogenous exergy destruction is given

based on Fig. 3.2c:

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 .

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:

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

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.

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:

(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:

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 .

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 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.

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.

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

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

3.3 evaluation of a modern ultra-supercritical plant 47

2500 100

2000 80

Exergy efficiency / %

Exergy efficiency

Temperature / C

1000 40

Average temperature

500 20

Water (steam) temperature Air

0 0

WW SSH1 SSH2 FSH FRHC FRHH PRH2 PSH ECO1 PRH1 ECO2 APH

Figure 3.4: Exergy efficiencies and temperature profiles of hot and cold streams of all heat

exchangers in the boiler subsystem

70%. In the successive convection-dominant SSHs and FSH, the gradual decrease in

temperature differences for heat transfer generally promotes their exergy efficiencies.

Interestingly, the reheaters, FRHC, FRHH and PRH2, perform not much better than

the upstream SSHs and FSH, even with lower flue gas temperatures. This is because

of much lower average temperatures of cold-side steams in these reheaters, which may

be caused by the properties of water and steam. In addition, those heat exchangers in

the post flue gas duct, PSH, PRH1 and ECOs perform with higher exergy efficiencies.

Note that the exergy efficiency of APH is lower than all other convection heat sur-

faces in the boiler subsystem, although the heat transfer in APH occurs under almost

the lowest temperature difference among all. This is because of the temperature level of

the cold-side air, which is the lowest cold-side temperature among all heat exchangers.

A low temperature level of cold stream could partially counteract the performance en-

hancement of a heat exchanger obtained by employing a small temperature difference

for heat transfer (see section 3.3.1.3).

500

100 Average temperature

Exergy eff. / %

90 Fluid temperature

400 80

Temperature / C

70

300 60

H1 H2 H3 H4 H5 DA H6 H7 H8

200

100

0 H2 H3 H4 H5 DA FP H6 H7 H8

0 100 200 300

. 400 500 600 700 800

Q / MW

Figure 3.5: Exergy efficiencies and temperature profiles of the heat exchangers in the turbine

subsystem

48 comprehensive exergy-based assessment of coal-fired power plants

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).

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:

HE = = = , (3.62)

E hot Q HE (1 Tref /Tahot ) 1 Tref /Tahot

or

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

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

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.

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

14.1%

Total Loss 49.6%

72.6% Cooling water

Boiler subsystem 47.2%

12.5%

Exhausted fluegas

Turbine

subsystem

c) d)

Other HEs (ms) 13.4% 3.1% Pumps

5.7% FWHs 3.1% Pipes

29.2% 4.4% HEs (rh)

Water wall 3.1% APH

39.9%

Condenser

57.2% 40.5%

Coal combustion Turbines

Figure 3.7: Spatial distribution of exergy destructions and losses: a) overall distributions of ex-

ergy destructions and losses, b) distribution of exergy losses, c) exergy-destruction

distribution in the boiler subsystem, and d) exergy-destruction distribution in tur-

bine subsystem

The exergy destructions within the turbine subsystem is dominated almost equally

(40%) by all turbines and the condenser (Fig. 3.7d). The remaining exergy destructions

are contributed by feedwater preheaters (over 13%), pumps (3%) and pipes (3%). This

pinpoints the significance of advanced design of turbine blades, particularly the blades

working with wet steam, and also the topology improvement of feedwater preheating

subsystem.

The economic constants used in the economic evaluation include annual effective in-

terest rate (0.1), nominal escalation ratio for OMC (0.03), nominal escalation ratio for

FC (0.035), plant economic life (25 yr), annual operation hours (6900 h/yr), and annual

capacity factor (0.8). The average price of the coal with a LHV of 29,270 MJ/kg was

84.3 $/ton (2.88 $/GJ-LHV) in 2009, leading to a levelized FC as high as 196.5 million$.

The PEC of the entire set of PCPP, i. e., the once-through boiler (including the APH),

FWHs, STs and other separate components, could be obtained from (China Power

Engineering Consulting Group Corporation, 2009). However, the PEC of boiler and STs

needs to be allocated to their sub-components (Fig. 3.3). For the allocation of boilers

PEC, the PEC of rotary APH comprises approximately 2% of PEC of the boiler body

(the boiler subsystem excluding SAF, PAF and IDF). The allocation of the remaining

boilers PEC is based on Table B.3. For allocating the turbine PEC, weighting factors are

1, 1.5 and 2 for HPTs, IPTs and LPTs, respectively. The module factors for estimating

fixed capital investment (FCI) from the PECs of different components are referred to

(Turton et al., 2008) as Table B.4. Then, the total capital cost (TCI) is estimated as around

3.3 evaluation of a modern ultra-supercritical plant 51

828 million$ or a specific cost of 774 $/kW. The levelized CC, OMC and FC contribute

30%, 7% and 63% to total revenue requirement (TRR) (308 million$/yr). The cost of

electricity is 5.20 /kWh.

Exergoeconomic analysis can be conducted for two cases: setting Z k as zero or the

values estimated from an economic analysis. The first case enables us to understand

the formation process of the cost of the final product: how the COE is affected by the

system structure and component types. A classical conclusion from the cost formation

process is that the specific cost of exergy accumulates gradually in course of transfer

and conversion to the final product. This leads to the cost inequivalence of exergy

destructions within different components. Generally, the closer the exergy to the final

product, the higher its specific cost. For the PCPP, the total exergy destruction within

all boilers heat exchangers except those in FUR is almost two times of that within

STs (Table B.5). However, there are no big difference between the hidden costs of the

two sets of components: 1969 $/h for considered heat exchangers and 1794 $/h for all

STs. This suggests that the reduction of downstream inefficiencies closer to the final

product would be more effective than reducing upstream inefficiencies.

The second case, i. e., setting Z k as the estimated values, provides a comprehen-

sive evaluation of the component performances. The component ranking based on

the summation C D + Z could pinpoint the significance for reducing the COE. The

cost-effective design improvement should focus on those components with the high-

est C D + Z . From this perspective, the components in the boiler (particularly the

furnace), HPT1, IPT1 and EG, are those key components (Table B.5).

The components in the boiler subsystem have much larger relative cost differences rk

than the others. The ECOs achieve the greatest values (even as high as 335%). However,

the FUR and other heat exchangers have relatively small rk (from 50% to 150%). In

particular, this exergoeconomic variable is below 63% for components in the turbine

subsystem, mostly within 930%. This suggests that the added costs of exergy streams

through the boiler subsystem are much larger than those by the turbine subsystem

(due to the capital investment). Thus, future boiler design should achieve less capital

investment as well as exergy destruction.

Exergoeconomic factors can identify the major sources of the component cost. The

FUR and APH have far less exergoeconomic factor than any other component in boiler.

The remaining heat exchangers in boiler are investment-cost-dominant components

with high exergoeconomic factors. Potential measure to balance the investment and

exergy destruction costs of these heat exchangers may be: Move part of the heat ab-

sorption in these heat exchangers into the furnace and increase the air preheating tem-

perature by increasing the area of APH. In such a way, more even temperature field of

flue gas can be established in the furnace, which leads to smaller exergy destruction

within furnace, larger area of (and larger exergy destruction within) APH, and smaller

areas of (or larger exergy destructions within) the remaining heat exchangers. This

concept may result in a simultaneous reduction of both capital investment cost of and

exergy destruction within the boiler.

The STs generally have reasonable exergoeconomic factors ranging from 50% to

65% with the exception of the last two parallel STs, LPT3 and LPT6. The exergy de-

structions within these two STs have to be reduced. In addition, the first 5 feedwater

52 comprehensive exergy-based assessment of coal-fired power plants

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

The results of the advanced exergy analysis of the PCPP are listed in Table B.6, where

the endogenous exergy destructions are calculated considering fixed and varied spe-

cific exergies of streams into and out of the considered components compared with

those of real case. The removal of pressure drops in all theoretically-operated com-

ponents mainly affects the endogenous exergy destructions within heat exchangers:

The larger the pressure drop in the considered component, the greater the influence

on its endogenous exergy destruction. Given fixed temperature increases of cold-side

working fluids, endogenous exergy destructions within those fluegas-water heat ex-

changers (FUR, ECO1 and ECO2) increase. In particular, the FUR causes 20 MW more

endogenous exergy destruction. Endogenous exergy destructions within fluegas-steam

heat exchangers for main steam decrease by almost 1 MW, while those within fluegas-

steam heat exchangers for reheat steam remain nearly unchanged due to small pres-

sure drops of reheat steam. Similarly, for FWHs and APH, there are no big differences

between the values obtained by the two calculation approaches.

The coupling with the preceding theoretically-operated turbine leads to a smaller

temperature of the inlet stream into the considered component, which is common for

the turbine train in PCPP. Smaller temperatures of inlet steams reduces the endoge-

nous exergy destructions within the turbines to varying degrees (Table B.6): Those

within the HPTs before reheat are less affected, while those within the IPTs and LPTs

after reheat are reduced significantly. More importantly, the endogenous exergy de-

struction within the COND is largely reduced due to a small temperature of inlet

steam (less inlet exergy to be removed).

The splitting of exergy destructions (Table B.6) is based on the endogenous exergy

destructions and products calculated considering the removal of related pressure drops

and the possible coupling with turbines. The results show that large parts of the exergy

destructions within most components are endogenous. However, for different types of

components, the share of the endogenous part differs significantly. Almost all exergy

destruction within EG is endogenous. For fluegas-water (steam) heat exchangers and

feedwater preheaters, the shares of endogenous exergy destructions are high (7080%).

The HPTs have large shares of endogenous inefficiency (over 70%) as well, given fixed

conditions of main steam. The thermodynamic inefficiencies within these components

tend to be affected in a limited fashion by the inefficiencies of other components. The

shares of endogenous exergy destruction within each remaining STs vary around 50

60%. The performance of each turbine among these STs are strongly affected by the

inefficiencies of other turbines. Moreover, the share of endogenous exergy destruction

within APH becomes even smaller (only 45%). It suggests the performance of the APH

can also be improved largely by improving the remaining components.

The real potential for improving a component is not fully revealed by its total exergy

destruction but by its avoidable part. A large part (3050%) of the exergy destruction

within heat exchangers in the boiler subsystem except FUR is avoidable. For FUR, the

avoidable exergy destruction of the coupled combustion and radiation heat transfer

3.4 suggestions for improvement 53

comes mainly from a smaller excess air ratio. The avoidable exergy destruction within

each FWH remains below 15% of its total exergy destruction. For pumps and fans,

the ranges of avoidable shares are 5060% and 2035%. For STs, especially the IPTs

and LPTs, the share of avoidable exergy destruction varies significantly from 580%. In

addition, most heat exchangers in the boiler subsystem, IPTs, LPT1, LPT3, LPT6 have

large avoidable exergy destructions (around and over 3 MW).

For most components except FUR, the share of endogenous/avoidable part in avoid-

able exergy destruction agrees with that of endogenous part in the components exergy

destruction. Thus, it suggests that, for most boilers heat exchangers (not including

FUR and APH), FWHs and HPTs, efforts to reduce their avoidable exergy destructions

should be dedicated to improve the components themselves. More importantly, for STs,

the improvement should be focused on the whole turbine train but not only the turbine

considered itself. Over 60% of avoidable exergy destruction within FUR and over 40%

for APH is exogenous. To reduce these avoidable parts, not only the component itself

but also the remaining components should be improved simultaneously.

The splitting of investment costs and exergy destruction costs of all components are

listed in Table B.7 and Table B.8, respectively. The shares of endogenous/exogenous

parts in components total investment cost are identical to the corresponding shares of

endogenous/exogenous exergy destructions. However, the shares of avoidable invest-

ment cost of almost all components range within 8090%. The total avoidable invest-

ment costs of the system (49.9 /s, 10% out of the total investment cost) are contributed

by the heat exchangers in the boiler (28.8 /s), the STs (10.4 /s), the EG (4.6 /s), and

pumps and fans (surprisingly, 4.3 /s in total). Around 90% of EGs contribution is

endogenous, while this figure becomes 70% for boilers heat exchangers, 55% for STs.

As expected, over 70% of the avoidable investment costs from pumps and fans are

exogenous.

The overall avoidable exergy destruction costs of the system (62 /s out of 618 /s)

are contributed most by the STs (27 /s), the boiler (24 /s), pumps and fans (4.8 /s),

and EG (4 /s). Around 60% of the exergy destruction costs of both the boiler and

turbine subsystems are endogenous. Almost all avoidable exergy destruction costs of

EG are endogenous, while those of pumps and fans are mostly exogenous.

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.

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

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.

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

41 41

0.92

.4 .2

0.8 700 .8

Optimal pressure ratio of reheat 2 / -

0.8

0.9

4

39 86

40

.2

0.

39

0.7 41

6.

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

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

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.

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

IN2 OUT2 OUT2

IN3 OUT3 IN2 IN2 OUT2

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.

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.

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.

The procedure of EAs has been described in section 2.5.1.4. For solving the MINLP

problem (Eq. 4.2), a similar procedure (Fig. 4.5a) is applied: The solution randomizer

in the initialization randomly generates integer and real variables of candidate solu-

tions. Each solution is evaluated by the simulation of the specified superstructure, in

which the stream paths, starting values, and real variables are set. The solution is

added to the initial population, if the simulation is successful and extra constraints

(the four constraints on steam quality in Eq. 4.2) are satisfied; otherwise, the solution

is discarded. Once the initial population is formed, the evolution iteration starts. For

evolving each target solution (si ) in the parent population, a noisy solution (sni ) is first

generated by adding weighted difference between the two randomly selected solutions

(sai and sbi ) to another randomly selected solution (sci ):

where f is the real-valued weighting factor and Ns is the population size. The noisy

and target solutions are recombined to produce a new trial solution (sti ):

s ,

i,j r > c ,

sti,j = i = 1 . . . Ns , j = 1 . . . Nv , (4.4)

sn , r c ,

i,j

where c is the real-valued crossover constant; while r and Nv are a real-valued num-

ber (randomly chosen from 0 to 1) and the number of decision variables, respectively.

After evaluation, the better of the trial and target solutions is kept as a child solu-

tion. The evolution iterations for each solution and population continue until certain

4.4 multi-objective supported differential evolution 63

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

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

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 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.

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

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.

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

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

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

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.

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.

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

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

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

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

For current steam conditions of a temperature level below 650 C, additional poten-

tials of energy-savings is expected by optimizing the initial design of reheating and

the layout of feedwater regeneration. Therefore, an exhaustive study on case 3 is con-

ducted. The actual reference industrial design (262 bar/600 C/605 C) reaches a plant

efficiency of 44.41%, with the pressure array of extracted steams being 83.72 bar/63.60

(rh1)/24.05/11.57 (DA)/6.35/2.543/0.669/0.253. The industrial design employs only a

single reheat; however, double reheats are also considered to find the maximum energy-

saving potential under the given steam conditions. The best-known (optimal) solutions

are selected from several optimization runs for the same problem.

45.4

10

45.3 9

Plant efficiency / %

45.2

45.1

45.0

8

44.9

44.7

300 310 320 330 340

Feedwater preheating temperature / C

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

and final feedwater temperatures (solid lines: with ET; dashed lines: with OT)

4.5 design optimization for thermodynamic objective 69

Table 4.3: Comparison of the steam extractions after reheating for single-reheat case (a) and

double-reheat case (b) by keeping m ms as 861 kg/s: ppressure (bar), tsat saturation

temperature (C), ttemperature (C), tdecrease in superheating degree (C), m

mass flow rate (kg/s), OOT, EET. (The steam extractions, which has been in two-

phase zone even without an ET, are not listed)

(a) (b)

p tsat t t m p tsat t t m

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

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.

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.

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.

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

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:

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

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.

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 / -

600 C

Reheat pressures / bar

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).

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.

For mathematical optimization, the cost functions of all involved components have

been listed in App. A. With known purchased-equipment costs of all components, the

total capital investment is simply calculated by PECk (the constant is set as

4.75 (Bejan et al., 1996)) but not by a detailed calculation. Similarly, the operation and

BOILER

IPT1-4

IPT1-3

HPT1-2

/

G

EG

FWH7

OT

COND

FWH6

CP

CWP

DA

FWP

FWH5 FWH4 FWH3 FWH2 FWH1

DP

Figure 4.16: The schematic of the power plant with a fixed structure

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

maintenance cost is simply estimated by TCI (the constant is set as 0.06 (Bejan

et al., 1996)). Therefore, the COE can be readily calculated according to the economic

analysis in section 3.1.3, with the same economic constants in section 3.3.2.

A single-reheat, 1000 MW, ultra-supercritical coal-fired power plant (Fig. 4.16) is opti-

mized with respect to the two objectives. Note that the boiler is modeled in a simplified

fashion as its PEC is expressed in a condensed form (App. A). Totally 87 material and

energy streams and 26 components are involved in each simulation.

An industrial design of a Chinese ultra-supercritical plant (Wang et al., 2012a), with

steam conditions of 274 bar/605 C/603 C, is taken as a reference. The plant efficiency

and COE of the plant are 45.8% and 5.20 /kWh (with a coal price of 0.3 /MJ on a

LHV basis), respectively.

For the optimization, 38 decision variables (Table 4.5) are taken into account. The

bounds of the decision variables can be fixed (for most variables) or varied (e. g., the

outlet pressure of some turbine stages). Note that the outlet pressure of HPT1 is con-

sidered with a fixed bound, which confines the final feedwater preheating temperature

within the range from 275 C to 337 C.

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]

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

ccoal = 0.3 /MJLHV ccoal = 0.3 /MJLHV

60 5.8

5.8 tsh = trh tsh = trh

5.7 3.7

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.

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.

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

4.5 tsh = trh 1.4

Turbines

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.

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

Pressure, temperature or mass

5.5 .

mms (kg/s) 800

trh (C)

5.4 700

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.

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.

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

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.

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)

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

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

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.

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.

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

Single reheating with fixed structure

5.8 Single reheating with varied structure 300

0.65

Cost of electricity / /kWh

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).

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

With 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

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 / %

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.

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

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).

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

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

d, o

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)

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

5.1 multi-objective superstructure-free synthesis framework 89

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.

The mutation operator employs mutation rules to generate new structures. The number

of required mutation rules is minimized to a set of 6 mutation rules for structural

evolution of thermal power plants:

1. Remove one component with all of its interconnections.

4. Delete one component and insert a parallel connection of two other components.

5. Delete one component and insert a serial connection of two other components.

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

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

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.

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.

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).

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

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

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.

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

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.

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.

The presented superstructure-free approach can be coupled to any model for thermal

power plants, implemented in the lower-level deterministic optimization (Fig. 5.1). For

this illustrative study, the objective is to maximize the thermal efficiency (%):

| PST | | PFP |

max = 100 , (5.7)

Q gross

where the PST and PFP are the power generated by steam turbine and the power con-

sumed by feedwater pump, respectively. The gross heat input to the plant (Q gross ) can

be calculated based on the total heat absorbed by the working fluid (Q abs ) and the

relative heat loss rate (loss , %) of steam generation and superheating:

100 Q abs

Q gross = . (5.8)

100 loss

The parameter values and variable bounds are specified according to Table 5.1.

In general, the generic ECH (Fig. 5.2) could be applied directly. However, the presented

illustrative study does not consider all technologies in the ECH (Fig. 5.2). Thus, the

ECH is refined (Fig. C.1) by removing functions not fulfilled by any technology. For

2 This option controls the function and derivative debugger of CONOPT (Drud, 2004). The value is set as

zero so that the debugger is not used. The function values and derivatives of properties of water and

steam are computed completely from the extrinsic library (the freesteam).

3 The options, rtnwma and rtnwmi, are the maximum and minimum feasibility tolerances for constraints

(for more details, see (Drud, 2004)). Larger values of the two options increase the feasibility of the whole

model. The values are set as 1E-5 and 7E-6 for all optimization runs in the thesis.

5.3 evaluation of the superstructure-free approach 97

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.

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%.

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).

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.

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.

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.

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 (%)

Run

tot n e tot n e

min % min 1 % min % min 1 %

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

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.

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.

For the upper-level EA, the same ( + )-evolution strategy is used: = 20 and = 25.

The maximum mutation strength is set as 5. The EA will terminate once 8,000 feasi-

ble structure alternatives are generated. For the lower-level deterministic optimization,

the specifications follow Table 5.1. Note again, the total net power output is fixed

other than the mass flow rate of main steam. In addition, if thermal efficiency is the

only objective, maximum numbers of components of involved technologies are set (see

section 5.3.2): 12 for steam turbine, 8 for feedwater preheater, 2 for reheater, 2 for de-

superheater and 2 for feedwater pump.

Starting from an initial structure with 4 feedwater preheaters and a thermal efficiency

of 46.49% (Fig. 5.8a), the structural evolution finds the optimal structure with an effi-

ciency of 49.6% (Fig. 5.8b). Note that the thermal efficiency calculated in this chapter is

not directly comparable with that in chapter 4 based on Fig. 4.3 and Fig. 4.16, in which

the power consumption of auxiliary devices (e. g., the fans in the boiler subsystem and

the cooling-water pump for condenser) is considered. The flowsheets and temperature-

entropy diagrams of both initial and optimal solutions are given in Fig. C.6 and C.7.

The optimal solution features eight-stage feedwater preheating, double reheating, 2 de-

superheaters, and a secondary turbine supplying steam for one feedwater preheater.

The temperatures of the main and reheat steams reach the maximum allowed tem-

perature. The feedwater is heated up to 647 K (thermodynamically reasonable) by the

a) 500 b) 500

50 50

Pressure / bar

Pressure / bar

5 5

0.5 0.5

0.05 0.05

0.005 0.005

0 1000 2000 3000 4000 0 1000 2000 3000 4000

Enthalpy / kJ/kg Enthalpy / kJ/kg

Figure 5.8: Pressure-enthalpy diagrams of the initial and optimal solutions for maximizing the

thermal efficiency: a) initial solution ( = 46.49%), b) optimal solution ( = 49.60%)

5.4 superstructure-free synthesis of complex coal-fired power plants 103

feedwater preheater above rh1 and the successive de-superheater (see Fig. C.6b). The fi-

nal feedwater preheating temperature is constrained by a specified maximum pressure

of steam extractions (200 bar). The optimal reheating pressure ratios are 0.37 (rh1) and

0.26 (rh2), respectively. Considering the quality of the exhausted steam (0.946), these

two reheating pressure ratios agree well with Fig. 4.14 (type 1 in zone A). However, the

optimal structure (Fig. C.6b) can not be generated from the superstructure in Fig. 4.3,

because of the steam turbine with one single steam extraction and the de-superheaters.

To integrate the de-superheater into the superstructure, a vast number of possibilities

have to be considered, e. g., the positions of de-superheaters and the feedwater to be

heated by the de-superheaters. Accordingly, this requires considerable effort to revise

the superstructure and related generation algorithm for structural alternatives.

An exemplary run for the complex problem for maximizing the thermal efficiency

(Fig. 5.9) illustrates similar performance as in Fig. 5.7. The generation rates of both

structurally-feasible and feasible structures are still high, over 90%, which leads to a

fast structural evolution. Near-optimal structures are generated from around 30 gener-

ations. Surprisingly, the generated solutions after 30 generations are grouped into two

efficiency ranges according to the number of reheatings employed. Single-reheat struc-

tures are frequently generated, when the mutation operator removes one reheater from

the given preserved parent solution. It also indicates that adding one more reheating is

much more effective than adding more feedwater preheaters and de-superheaters for

increasing the thermal efficiency.

Some interesting intermediate structures are listed in Fig. C.8. After adding two

reheaters and one more feedwater preheater to the initial structure (Fig. C.6a), one

possible structure is generated as in Fig. C.8a. In this structure, the turbine train in

the initial structure now serves as a secondary turbine with steam extractions (ET).

Then, more newly-added feedwater preheaters are connected to both the ET and main

turbine (MT), as in Fig. C.8b. The mutation operator continues to change the config-

54 1000 100

rate

Accumulated time / min

1 2

Thermal eciency / %

52 800 80

3

solutions generated 4

in each generation 5

50 600 60

Rate / %

48 400 40

44 0 0

0 50 100 150 200 250 300

Generation number / -

Figure 5.9: One representative run (similar to all performed runs) for maximizing the thermal

efficiency: generation rate of structurally feasible structures ( 1 ), rate of successful

evaluation of structurally feasible structures ( 2 ), total evaluation time ( 3 ), eval-

uation time of feasible solutions + unsuccessful optimization of structurally feasi-

ble structures produced by insertion ( 4 ), evaluation time of feasible solutions ( 5 ).

( = 20, = 25, maximum mutation strength 5, Pnet = 1000 MW)

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

uration and more feedwater preheaters are connected to the MT (Fig. C.8c and C.8d).

Therefore, a complex structure can evolve gradually to completely different structures

without getting stuck. For this synthesis task, the benefit of an ET becomes weak after

the maximum number of feedwater preheaters are employed.

The superstructure-free synthesis is also performed for another five times (Table C.1).

The computation time for this complex problem is much greater than that of the illus-

trative study, as the optimization time for each structurally-feasible structure becomes

longer. This is due to that 1) more constraints and variables are involved in each eval-

uation, and 2) fixing the power output makes the GAMS model even more complex.

More importantly, not all these 6 runs identifies the optimal solution. However, the

geometric average optimality gap of all 6 runs is only 0.01%: Good near-optimal alter-

native solutions are generated, although the identification of the optimal solution is

not guaranteed.

Important information carried by all generated solutions (including near-optimal so-

lutions) are given in Fig. 5.10 and 5.11. It is confirmed that, for increasing the thermal

efficiency, adding one additional reheating is much more effective than adding sev-

eral feedwater preheaters (Fig. 5.10). The more the feedwater preheaters are employed,

the less efficiency increase can be gained by adding additional feedwater preheaters

(Fig. 5.10). The number of feedwater preheaters above rh1, in general, has limited or no

influence on the thermal efficiency (Fig. 5.11). The optimal pressure ratios for reheating

are 0.150.35 (rh1) and 0.250.45 (rh2).

For the EA, the same evolution strategy, maximum mutation strength and termination

criteria are employed. For the deterministic optimization, the cost of electricity (COE)

is calculated by the same procedure described in section 4.6.1. As only those variables,

whose bounds are listed in Table 5.1, are optimized, additional variables involved in

9 0.5 5

rh2

8

Number of FWH or reheaters / -

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

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.

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

80 2 1

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

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

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

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.

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

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

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

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

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

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

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.

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.

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

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.

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.

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.

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

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.

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.

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

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.

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 ,

thot sat tout > tup ,

cold

or tstm cold

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).

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.

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

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

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 ,

pfw fw

0

hfw

out = h p fw

out ,

out .

mixer A mixer has one incoming pin and one outgoing pin, which are of the same

stream type. The incoming pin allows for more than two streams; while the outgoing

pin only allows for one stream. The mixer can be modeled as follows:

Nin

m in,j = m out ,

j =1

Nin

(m h)in,j = (m h)out ,

j =1

splitter A splitter has one incoming pin and one outgoing pin, which are of the

same stream type. The outgoing pin allows for more than two streams; while the in-

coming pin only allows for one stream. The splitter can be modeled as follows:

Nout

m in = m out,j ,

j =1

Nout

(m h)in = (m h)out,j ,

j =1

pout,j = pin .

electronic generator A electronic generator has one incoming work pin, and

one outgoing electricity pin. The model is used for both simulation and optimization.

The electronic generator is simply modeled:

PEG = Pout = EG Pin ,

124 thermodynamic models and cost functions of the components

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.

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

water pump

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

PFP = E out

wat

E in

wat

+ E D,FP

cost balance

C in

wat

+ C in

work

+ Z FP = C out

wat

125

126 additional materials for the analysis and evaluation

air

PAF = E out,j E in

air

+ E D,AF

cost balance

C in

air

+ C in

work

+ Z AF = C out

air

feedwater preheater

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

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

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

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

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

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

128 additional materials for the analysis and evaluation

mixer

!

Nin

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

j =1

cost balance

Nin

C in,j + Z MR = C out

j =1

splitter

!

Nout

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

j =1

cost balance

Nout

C in + Z SR = C out,j

j =1

cout,i = cout,j

electronic generator

E in = E out + E D,EG

cost balance

C in + Z EG = C out

B.2 real and unavoidable conditions of involved components 129

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

130 additional materials for the analysis and evaluation

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

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

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

132 additional materials for the analysis and evaluation

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

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

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

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

Comp. Material

mmmm area/ m2 ton factor

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)

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

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

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

136 additional materials for the analysis and evaluation

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

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

138 additional materials for the analysis and evaluation

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

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

140 additional materials for the analysis and evaluation

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

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

142 additional materials for the analysis and evaluation

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

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.

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.

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.

allowed allowed allowed allowed

expander pump heater heater cooler

Function level

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

The flowsheets of the initial (Fig. C.2a) and optimal (Fig. C.2b) solutions identified in

all optimization runs are presented with their temperature-entropy diagrams (Fig. C.3).

a)

heat

condenser

steam pump

generator

b) reheater

steam turbine

condenser

feedwater

steam preheater

generator pump

Figure C.2: The flowsheets of the initial (a) and optimal (b) solutions identified in the illustra-

tive study

a) 950 b) 950

850 850

750 750

Temperature / K

Temperature / K

650 650

550 550

450 450

350 350

250 250

0 2 4 6 8 10 0 2 4 6 8 10

Entropy / kJ/(kgK) Entropy / kJ/(kgK)

Figure C.3: Temperature-entropy diagrams of the initial (a) and optimal (b) solutions identified

in the illustrative study

C.1 illustrative study for evaluating the superstructure-free approach 145

a) 500 b) 500

50 50

Pressure / bar

Pressure / bar

5 5

0.5 0.5

0.05 0.05

0.005 0.005

0 1000 2000 3000 4000 0 1000 2000 3000 4000

Enthalpy / kJ/kg Enthalpy / kJ/kg

c) 500 d) 500

50 50

Pressure / bar

Pressure / bar

5 5

0.5 0.5

0.05 0.05

0.005 0.005

0 1000 2000 3000 4000 0 1000 2000 3000 4000

Enthalpy / kJ/kg Enthalpy / kJ/kg

Figure C.4: Some intermediate solutions identified in the representative run of the illustrative

study (Fig. 5.7a): a) the 29th generation ( = 48.81%), b) the 37th generation ( =

49.04%), c) the 40th generation ( = 49.26%), d) the 52th generation ( = 49.44%)

146 additional materials for the superstructure-free synthesis

The specified energy conversion hierarchy 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.

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.

allowed allowed allowed allowed

expander pump heater heater cooler

Function level

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

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

a) 1050 b) 1050

950 950

850 850

Temperature / K

Temperature / K

750 750

650 650

550 550

450 450

350 350

250 250

0 2 4 6 8 10 0 2 4 6 8 10

Entropy / kJ/(kgK) Entropy / kJ/(kgK)

c) 1050 d) 1050

950 950

850 850

Temperature / K

650 650

550 550

450 450

350 350

250 250

0 2 4 6 8 10 0 2 4 6 8 10

Entropy / kJ/(kgK) Entropy / kJ/(kgK)

Figure C.8: Some intermediate solutions of the representative run (Fig. 5.9) for maximizing the

thermal efficiency: a) 3th generation ( = 47.45%), b) 9th generation ( = 48.30%),

c) 20th generation ( = 48.94%), d) 55th generation ( = 49.27%)

C.2 optimal synthesis of complex coal-fired power plants 149

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 (%)

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

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

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

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

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

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 %

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

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

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)

(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.)

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

page 23.)

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.)

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.)

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

153

154 bibliography

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.)

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.)

Germany. (Cited on page 98.)

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

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.)

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.)

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.)

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

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.)

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.)

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.)

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

ellipse. Technical report, Energy Incoportated PEPSE Users Group Meeting, Rich-

mond, Virginia. (Cited on pages 20 and 115.)

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

minlp synthesis of process systemsII. heat exchanger networks. Comput. Chem. Eng.,

18(8):679709. (Cited on page 24.)

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.)

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.)

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

page 14.)

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

page 22.)

2536. (Cited on pages 22 and 23.)

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

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

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.)

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.)

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.)

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.)

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

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.)

optimal synthesis-design-operation of complex systems. Internal Journal of Energy,

Environment, Economics, 4(1):267274. (Cited on page 10.)

design or improvement of complex thermal systems. PhD thesis, Georgia Institute of

Technology. (Cited on pages 9 and 10.)

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

158 bibliography

the thermoeconomic functional approach. ASME J. Eng. Gas Turb. Power, 114:707714.

(Cited on pages 9, 10, 25, and 117.)

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

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.)

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.)

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.)

models using graph theory for optimisation of thermal energy systems. Energy Con-

vers. Manage., 48(11):28182826. (Cited on page 23.)

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

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

programming for the synthesis of process systems. Latin American Applied Research,

30(4):263284. (Cited on page 22.)

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.)

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.)

a finite number of heaters. Proceedings of the Institution of Mechanical Engineers,

161(1):157164. (Cited on page 21.)

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

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.)

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

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

page 1.)

using differential evolution. In AI 2004: Advances in Artificial Intelligence, pages 861

872. Springer. (Cited on page 31.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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.)

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

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.)

page 16.)

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

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.)

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.)

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.)

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.)

energy systems. Part II: Formation structure of optimality for robust design. Appl.

Energy, 87(3):957970. (Cited on page 11.)

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.)

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

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.)

rithms and comparison of software implementations. Journal of Global Optimization,

56(3):12471293. (Cited on page 16.)

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.)

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.)

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.)

for bilinear programming problems. Journal of Global Optimization, 2(4):379410.

(Cited on page 17.)

168 bibliography

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

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.)

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.)

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.)

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.)

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.)

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.)

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

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.)

nation of exergetic and economic analysis in energy conversion processes, pages

151157. Pergamon Press, Oxford, England. (Cited on pages 8 and 9.)

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

ume 69, chapter Design Optimization Using Exergoeconomics, pages 101115.

Springer Netherlands, Netherlands. (Cited on pages 9 and 14.)

170 bibliography

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.)

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.)

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.)

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.)

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.)

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.)

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

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.)

in Chemical Engineering Research and Education, volume Volume 16, pages 499523.

Academic Press. (Cited on page 22.)

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

174 bibliography

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.)

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

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.)

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.

- Steam Turbine ConstructionUploaded byKamal Arab
- Steam Turbines - a full range to fit your needs.pdfUploaded byNitinSood
- e. Steam Power Plant - LectureUploaded byallovid
- High Pressure BoilerUploaded byjehadyam
- Lifecycle OptimizerUploaded byAruna Kari
- Script MotorDesignUploaded byKusum0
- Turbine Cycle OptimisationUploaded bysujitkgec
- Why Steam Turbine Tripping Provided for Low Inlet Steam PressureUploaded bydhinesht
- Prediction of Traffic Flow by an Adaptive Prediction SystemUploaded byAmjad Hashim
- BHELUploaded byfunkisanju1
- Mini ProjectUploaded byVenkata Sai M
- Steam turbine and emUploaded byhamidkarimi325
- syballus.pdfUploaded bykaran bansal
- stgspecification-140417234455-phpapp01Uploaded bymoe
- Thermal Power Plants Operation DetailUploaded byRAGAVENDRAN VENKATRAJ
- BlendMasterUploaded bysaeedhoseini
- Forced cooling of Turbine.Uploaded bynaba
- 265Uploaded byStefano Ramsingh
- Microsteam BrochureUploaded byviniciusafonso
- Turbine Efficiency Ch-8.1Uploaded byE.C.MADHUDUDHANA REDDY
- Jspl ReportUploaded byHimanshu Gupta
- Clay 2018Uploaded byAditya Chauhan
- Sources of vibration excitation _ The steam turbine.pdfUploaded byMuhammad Fahmi Mahmud
- 4-Chapter-4-SteamTurbine.pdfUploaded byKevin Ajat
- UntitledUploaded bymahesh_eil
- wcdmaoptimizationdrivetestanalys.pdfUploaded byMehran
- 2.1- EA- Turbine CycleUploaded bys anil kumar tatithuri
- Optimal Processes in Irreversible MicroeconomicsUploaded byAmanda Car
- Algorithmic approach to pushback design based on stochastic programming method application and comparisons.docxUploaded byyudel M ch
- 3_041Uploaded byWeins Gemerlap

- Sewage Pumping HandbookUploaded byGrundfosEgypt
- 110673739 Sewage Pumping HandbookUploaded bymfhaleem@pgesco.com
- API682 ISO21049 JOHN CRANE GENERAL COMMENTSUploaded byMixtli Garcia
- Sump Pump Manual (6!15!2016)Uploaded bymfhaleem@pgesco.com
- ReviewofWearandGallingCharacteristicsofStainlessSteel_9006_Uploaded byfahreez
- Mesh vs MicronUploaded byArif
- Turbine Overspeed Systems and Required Response Times Rev B.01Uploaded bymfhaleem@pgesco.com
- Oil and Gas Production Handbook Ed3x0_webUploaded bymfhaleem@pgesco.com
- API 610 Pump Selection and Curve EvaluationUploaded bybbmoksh
- Hayward Tyler AR2015Uploaded bymfhaleem@pgesco.com
- Cooling-Tower-Fundamentals (1).pdfUploaded byrkukg
- 3B.1Fox.pdfUploaded bymfhaleem@pgesco.com
- 247797491-Compressor-Tech-June-2014.pdfUploaded bymfhaleem@pgesco.com
- Fire Pump Testing E211 0Uploaded bymfhaleem@pgesco.com
- 1998_03_8_leibundgut_eUploaded bymfhaleem@pgesco.com
- 856_e_cr598_enUploaded bymfhaleem@pgesco.com
- Fire Pump Essentials - Derek ThompsonUploaded bymfhaleem@pgesco.com
- Thordon_PropShaftBrochure_Int.pdfUploaded bymfhaleem@pgesco.com
- Centrifugal pumpUploaded bye103050_1
- Fire Pump Essentials - Derek ThompsonUploaded byapi_fabian
- Orifice KsoUploaded bymfhaleem@pgesco.com
- 10.1115@POWER2015-49551Uploaded bymfhaleem@pgesco.com
- Ccc221coalpowerR MUploaded bymfhaleem@pgesco.com
- 3 Minchener Undp HeleUploaded bymfhaleem@pgesco.com
- 2-PGA-2008Uploaded bymfhaleem@pgesco.com
- disk couplingUploaded bymfhaleem@pgesco.com
- product.pdfUploaded bymfhaleem@pgesco.com
- Mesh vs MicronUploaded bymfhaleem@pgesco.com
- HlvUploaded bymfhaleem@pgesco.com

- Establishing a Fully Coupled CFD Analysis Tool for Floating Offshore Wind TurbinesUploaded byMayra Zezatti
- Core Physics 2014 HigherUploaded byDonte Barnett
- Boilerplate No. 15, Spacecraft Saturn Research and Development Test SA-6 SpecificationUploaded byBob Andrepont
- Client LogUploaded bySurya Saputra
- Fast-Switch Integration Scenarios to a Different SAP PI Runtime - Webinar PresentationUploaded bySwarup Sawant
- Simhydraulics Modeling Tutorial.pdfUploaded byAjit Singh
- Optimal ChoiceUploaded byBianca Bazin
- Seismic Methods and Applications: A Guide for the Detection of Geologic Structures, Earthquake Zones and Hazards, Resource Exploration, and Geotechnical EngineeringUploaded byUniversal-Publishers, Inc.
- Simulation of Air Flow Distribution in a Tray Dryer by CFDUploaded byBrayan Arias
- List of HTML TagsUploaded bydorinadid
- Efficient Text Classification of 20 Newsgroup Dataset using Classification AlgorithmUploaded byEditor IJRITCC
- FisaDisciplina (7)Uploaded byMarius Samson
- Lg 29fg2rge-Tg ChassisUploaded bybigbj
- What s New in Plts for Arcgis Nautical Solution 9 2SlidesUploaded bymichael ellis
- Upgrading the Controller Hardware on a Pair OfUploaded byPurushothama Gn
- Phase2_SlopeStabilityVerification_Part1Uploaded bydrakic
- implementing-cisco-application-centric-infrastructure-with-citrix-netscaler-application-delivery-controllers.pdfUploaded byBob Mulumba
- IT Interview QuestionsUploaded byCareerMonk Publications
- Ch022Uploaded bySusana
- Tips&Tricks JCo ProgrammingUploaded byEnzo Sellers
- Petroleum Refining Process Control and Rt OptimizationUploaded bydemos2011
- Introduction to ICC ProfilesUploaded bySappiETC
- 2Uploaded bySharma Sudhir
- Design and Performance Prediction of and Axial Flow CompressorUploaded byYusufYüksel
- Harmony and Melody in Kant's Second Analogy OfUploaded byBeatriz M
- Bulletin T101-3_Chemichal resistance chart.pdfUploaded byChristianGuerrero
- PowerPoint+Chapter+9Uploaded bySara Villarreal
- Tutorial RsamUploaded bymirko_italy
- D8CD2A4Cd01Uploaded byBelhabra Mohalhil
- DATA COLLECTION AND ROCK MASS CLASSIFICATIONUploaded byScott Foster