You are on page 1of 16

Chapter 24

Exergy and Multiobjective Optimization


Chapter Outline
24.1 Introduction 483 24.5.2.3 Constraints 488
24.2 Optimization Formulations 484 24.5.2.4 Multiobjective Optimization 488
24.2.1 System Boundaries 484 24.6 Case Study: GT Power Generation Plant 489
24.2.2 OFs and System Criteria 484 24.6.1 Thermodynamic, Economic, and Environmental
24.2.3 Decision Variables 484 Modeling and Analysis 489
24.2.4 Constraints 484 24.6.1.1 Thermodynamic Modeling 490
24.3 Optimization Methods 485 24.6.1.2 Exergy Analysis 490
24.3.1 Classical Optimization 485 24.6.1.3 Exergoeconomic Analysis 490
24.3.2 Numerical Method Optimization 485 24.6.1.4 Exergoenvironmental Analysis 492
24.3.3 Evolutionary Algorithm 485 24.6.2 Optimization 492
24.3.3.1 GA 485 24.6.2.1 Definition of OFs 492
24.3.3.2 ANN 485 24.6.2.2 Decision Variables 492
24.3.3.3 Fuzzy Logic 486 24.6.2.3 Model Validation 493
24.4 Multiobjective Optimization 486 24.6.2.4 Optimal Solutions 493
24.5 Illustrative Example: Air Compressor Optimization 486 24.6.2.5 Optimal Relation between Total Cost
24.5.1 Thermodynamic and Economic Modeling and Rate and Exergy Efficiency 495
Analysis 486 24.6.2.6 Distribution of Design Parameters 495
24.5.1.1 Modeling 486 24.6.2.7 Use of Sensitivity Analyses to
24.5.1.2 Exergy Analysis 487 Describe Effect of Decision Variables
24.5.1.3 Exergoeconomic Analysis 487 on OFs 497
24.5.2 Optimization 488 24.7 Closing Remarks 497
24.5.2.1 OFs 488 Problems 497
24.5.2.2 Decision Variables 488

ABSTRACT 24.1 INTRODUCTION


In this chapter, the integration of exergy analysis with multi- Increasing global competition and the desire for better and
objective optimization, and the application of such a combined more efficient processes have placed expanded pressures on
method, is described. It is usually important to optimize processes
design. Interest is growing in producing higher quality prod-
so that a chosen quantity, known as the objective function (OF), is
ucts at minimal cost, while satisfying increasing concerns
maximized or minimized. Engineering design has thus become
increasingly coupled with optimization, and exergy methods can regarding environmental impact, safety, and other factors. It
enhance such approaches. Here, we describe several optimization is no longer adequate to develop a system that simply per-
methods and the formulation of an optimization problem. To forms a desired task, from the perspective of society.
improve understanding, several components are first considered For various reasons, it is important to optimize processes
individually and OFs, constraints, and optimization are developed so that a chosen quantity, known as the objective function
and applied. After, a case study is considered based on a gas (OF), is maximized or minimized. For example, the output,
turbine power generation plant in which a complete system is profit, productivity, product quality, and so forth, may be
modeled and optimized. maximized, or the cost per item, investment, energy input,
KEYWORDS and so forth, may be minimized. The success and growth of
Exergy; Optimization; Multiobjective optimization; Objective industries today is strongly based on their ability to optimize
function; Evolutionary algorithm; Economics; Environment; designs and systems. With the advent in the recent years of
Decision variable; Air compressor; Gas turbine; Hydrogen new materials, such as composites and ceramics, and new
production. manufacturing processes, several traditional industries (e.g.,

Exergy. http://dx.doi.org/10.1016/B978-0-08-097089-9.00024-3
2013 Ibrahim Dincer and Marc A. Rosen. Published by Elsevier Ltd. All rights reserved 483
484 Exergy

steel) have faced significant challenges and, in some cases, is based on the desire or purpose of the decision maker, and
diminished in size, while many new fields have emerged. It it can be either maximized or minimized.
is important to exploit new techniques for product Optimization criteria can vary widely. For instance, it
improvement and cost reduction in traditional and new can be based on efficiency (energy, exergy, or other effi-
industries. Even in an expanding area, such as consumer ciencies), economic (total capital investment, total annual
electronics, the prosperity of a company is closely connected levelized costs, cost of exergy destruction, and cost of
to its ability to apply optimization to new and existing environmental impact), technological (production rate,
process and system designs. Consequently, engineering production time, and total weight), and/or environmental
design, which has always been important, has become (rate of emitted pollutants).
increasingly coupled with optimization. Note that we can consider more than one OF to find the
Optimization is a significant tool in engineering for optimal solution for an optimization problem. This method
determining the best, or optimal, value for the decision is called multiobjective optimization.
variable of a system. Energy engineering is a field where
optimization plays a particularly important role. Engineers
24.2.3 Decision Variables
involved in thermal energy engineering, for instance, are
required to answer the questions such as: Another essential element in formulating an optimization
problem is the selection of the independent decision vari-
l What processes or equipment should be selected for
ables that adequately characterized the possible design
a system, and how should the parts be arranged for the
options. To select these decision variables, it is important
best outcome?
to: (1) include all important variables that could affect the
l What are the best characteristics for the components
performance and cost effectiveness of the system, (2) not
(e.g., size, capacity, cost)?
include variables with minor importance, and (3) distin-
l What are the best process parameters (e.g., temperature,
guish among independent variables whose values are
pressure, flow rate, and composition) of each stream
amenable to change. In each optimization problem, only
interacting with the system?
decision variables are changing. Variables whose values are
In order to answer such questions, engineers are required to calculated from the independent variables using mathe-
formulate an appropriate optimization problem. Proper matical models are dependent variables.
formulation is usually the most important and sometimes
the most difficult step in optimization. To formulate an
optimization problem, there are numerous elements that
24.2.4 Constraints
need to be defined, including system boundaries, optimi- The constraints in a given design problem arise due to
zation criteria, decision variables, and OFs. limitations on the ranges of the physical variables, basic
In this chapter, we describe several optimization methods conservation principles that must be satisfied, and other
and the formulation of an optimization problem. To improve limitations. The restrictions on the variables may arise due
understanding, several components are first considered to the space, equipment, and materials that are employed.
individually and OFs, constraints, and optimization are These may restrict, for example, the dimensions of the
developed and applied. Then, a case study is considered in system, the highest temperature that the components can
which a complete system is modeled and optimized. safely attain, the allowable pressure, the material flow rate,
the force generated, and so on. Also, minimum values of
24.2 OPTIMIZATION FORMULATIONS the temperature may be specified for thermoforming of
a plastic and for ignition to occur in an engine. Thus, both
24.2.1 System Boundaries minimum and maximum values of the design variables may
The first step in any optimization problem is to define the be involved in constraints.
system boundaries. All subsystems that affect system Many constraints in thermal systems arise because of
performance should be included. When the system is overly conservation laws, particularly those related to mass,
complex, it is often desirable to break it down into smaller momentum, and energy. For instance, under steady-state
subsystems. In this case, optimization should be done on conditions, mass inflow to the system must equal mass
each subsystem independently, that is, suboptimization of outflow. This condition gives rise to an equation that must
the subsystems is performed. be satisfied by the relevant design variables, thus restricting
the values that may be employed in the search for an
optimum. Similarly, energy balance considerations are
24.2.2 OFs and System Criteria
important in thermal systems and may limit the range of
The next step in an optimization problem is to define the temperatures, heat fluxes, dimensions, and so forth, that
system criteria, which is sometimes called the OF. The OF may be used. Several such constraints are often satisfied
Chapter | 24 Exergy and Multiobjective Optimization 485

during modeling and simulation because the governing l Dynamic programming: Studies the case in which the
equations are based on conservation principles. In this way, optimization strategy is based on splitting the problem
the OF being optimized already considers these constraints. into smaller subproblems.
In such cases, only the additional limitations that define the l Combinatorial optimization: Concerns problems
boundaries of the design domain remain to be considered. where the set of feasible solutions is discrete or can be
reduced to a discrete one.
24.3 OPTIMIZATION METHODS l Evolutionary algorithm: Involves numerical methods
based on random search.
24.3.1 Classical Optimization
Classical optimization techniques are useful in finding the
optimum solution or unconstrained maximum or minimum 24.3.3 Evolutionary Algorithm
of continuous and differentiable functions. Some specifi-
cations for classical optimization can be selected based on An evolutionary algorithm utilizes techniques inspired by
this understanding, as described below: biological evaluation reproduction, mutation, recombina-
tion, and selection. Candidate solutions to the optimization
l These are analytical methods that make use of differ- problem play the role of individuals in a population, and the
ential calculus in locating the optimum solution. fitness function determines the environment within which
l Classical methods have limited scope in practical the solutions live. Evolutionary algorithm methods
applications as some involve OFs that are not contin- include genetic algorithms (GAs), artificial neural
uous and/or differentiable. networks (ANNs), and fuzzy logic (Ghaffarizadeh, 2006).
l These methods assume that the function is differentiable These approaches are discussed further in the following
twice with respect to the design variables and that the sections. Each of the approaches is available in toolboxes
derivatives are continuous. developed by MathWorks and can thus be used easily with
l Three main types of problems can be handled by clas- MATLAB software.
sical optimization techniques:
l Single variable functions

l Multivariable functions with no constraints


24.3.3.1 GA
l Multivariable functions with both equality and A GA is a search method used for obtaining an optimal
inequality constraints solution that is based on evolutionary techniques similar to
In problems with equality constraints, the Lagrange processes in evolutionary biology, including inheritance,
multiplier method can be used. If the problem has learning, selection, and mutation. The process starts with
inequality constraints, the Kuhn-Tucker conditions can be a population of candidate solutions called individuals, and
used to identify the optimum solution. progresses through generations, with the fitness of each
individual being evaluated. Fitness is defined based on the
OF. Then multiple individuals are selected from the current
24.3.2 Numerical Method Optimization generation based on fitness and modified to form a new
This optimization technique can be categorized according population. This new population is used in the next iteration
to specific method, as follows: and the algorithm progresses toward the desired optimal
point (Schaffer, 1985; Goldberg, 1989).
l Linear programming: Studies the case in which the
OF f is linear and the set A, where A is the design
24.3.3.2 ANN
variable space, is specified using only linear equalities
and inequalities. ANNs are interconnected groups of processing elements,
l Integer programming: Studies linear programs in called artificial neurons, similar to those in the central
which some or all variables are constrained to take on nervous system of the body. The approach is thus analogous
integer values. to some elements of neuroscience. The characteristics of
l Quadratic programming: Allows the OF to have the processing elements and their interconnections deter-
quadratic terms, while the set A must be specified with mine the processing of information and the modeling of
linear equalities and inequalities. simple and complex processes. Functions are performed in
l Nonlinear programming: Studies the general case in parallel and the networks have both nonadaptive and
which the OF or the constraints or both contain adaptive elements, which change with the inputs and
nonlinear parts. outputs and the problem. The ANN approach leads to
l Stochastic programming: Studies the case in which nonlinear, distributed, parallel, local processing and adap-
some of the constraints depend on random variables. tive representations of systems (Ghaffarizadeh, 2006).
486 Exergy

FIGURE 24.1 Multiobjective optimization with


two OFs, OF1 and OF2, that are to be minimized,
showing the (a) dominant designs and (b) the Pareto
frontier.

24.3.3.3 Fuzzy Logic Figure 24.1a shows values for the two OFs at five design
points. As shown in this figure, design 2 is clearly pref-
Fuzzy logic allows us to deal with inherently imprecise erable to design 4 because both OFs are smaller for
concepts, such as cold, warm, very, and slight, and is useful design 2 compared to design 4. Similarly, design 3 is
in a wide variety of thermal systems where approximate, preferable to design 5. However, designs 1, 2, and 3 are
rather than precise, reasoning is needed. Fuzzy logic can be not preferable, or dominated, by any other designs. The
used for the control of systems and in problems where set of nondominated designs is introduced as the Pareto
a sharp cut off between two conditions does not exist. frontier, representing the best collection of design points.
This is shown in Figure 24.1b. Note that any point on the
24.4 MULTIOBJECTIVE OPTIMIZATION Pareto frontier can be considered as an optimal design
condition. The selection of a specific design from the set
Optimal conditions are generally strongly dependent on the of points constituting the Pareto frontier is at the discre-
chosen OF. However, several aspects of performance are tion of the decision maker, which may be an engineer or
often important in most practical applications. In thermal designer.
and energy systems design, efficiency (energy and/or
exergy), production rate, output, quality, and heat transfer
rate are common quantities that are to be maximized, while 24.5 ILLUSTRATIVE EXAMPLE: AIR
cost, input, environmental impact, and pressure are quan- COMPRESSOR OPTIMIZATION
tities to be minimized. Any of these can be chosen as the To enhance understanding of the application of optimiza-
OF for a problem, but it is usually more meaningful and tion, an example involving an air compressor (AC), a rela-
useful to consider more than one OF. tively simple device, is illustrated in detail. OFs, constraints,
In a single-objective optimization problem, the task is to and multiobjective optimization are involved.
find the value for the OF, which optimizes a sole OF. When An (AC) is a device that converts electrical power
more than one OF is considered in the optimization, we (usually from an electric motor, a diesel engine, or a gaso-
refer to the procedure as multiobjective optimization. line engine) into compressed air by compressing and
One of the common approaches for dealing with pressurizing air, which can be released on demand for
multiple OFs is to combine them into a single OF that is to applications. There are numerous methods of air
be minimized or maximized. For example, in the design of compression, including positive-displacement or negative-
heat exchangers and cooling systems for electronic equip- displacement types. To optimize an AC using thermody-
ment, it is desirable to maximize the heat transfer rate. namic and other factors, we first have to model the device
However, this often comes at the cost of increased fluid using energy balances and other relations. Figure 24.2
flow rates and corresponding frictional pressure losses. shows an AC as a control volume for the thermodynamic
Another approach that has attracted much attention in modeling.
recent years is multiobjective optimization. With this
approach, two or more OFs that are of interest in a given
problem are considered, and a strategy is developed to 24.5.1 Thermodynamic and Economic
balance or trade off each OF relative to the others (Deb, Modeling and Analysis
2001).
24.5.1.1 Modeling
To illustrate, we consider two OFs, OF1 and OF2.
We assume that these are to be minimized (although For the AC, we can write expressions for the outlet
maximization can be similarly handled since it is equiv- temperature T2 and the work rate requirement W_ AC ,
alent to minimization of the negative of the function). respectively, as follows:
Chapter | 24 Exergy and Multiobjective Optimization 487

T2 and where T0 and P0 denote the reference environment


temperature and pressure. These values are taken here to be
the ambient temperature and pressure, which are 20 C and
. 1 atm.
WAC The compression ratio for the AC and the specific
AC
exergy of the outlet air, respectively, can be expressed as
follows:

P2
RAC (24.7)
T1 P1
FIGURE 24.2 Model of an AC.
ex2 h2  h0  T0 s2  s0
    
T2 P2
Cp T2  T0  T0 Cp ln  R ln
T0 P0
  ga 1  (24.8)
1 g
T2 T1 1 gACa 1 (24.1)
hAC An exergy balance for the AC can be written as:
W_ AC m_ a Cpa T2  T1 (24.2)
m_ 1 ex1 W_ AC m_ 2 ex2 Ex
_ D (24.9)
where m_ a is air mass flow rate, hAC is the compressor
The exergy efficiency of the compressor can be evalu-
isentropic efficiency, ga cp =cv , and cpa is the specific
ated as follows:
heat at constant pressure of air, which can be expressed
a function of temperature as follows (Hajabdollahi et al., _ 2  Ex _ 1
Ex
2011): AC (24.10)
W_ AC
3:837  T 9:453  T 2 5:490  T 3
cpa 1:048  
104 107 1010 24.5.1.3 Exergoeconomic Analysis
7:929  T 4
An exergoeconomic analysis is provided here for the AC.
1014
(24.3) Further details on exergoeconomic analysis, cost balances,
and exergoeconomic factors are discussed earlier in this
Also, T1 is the ambient air temperature entering the AC and book and elsewhere (Kotas, 1985; Bejan et al., 1996;
T2 is the hot air temperature leaving the AC. Ahmadi and Dincer, 2010). To find the cost of exergy
destruction for the AC, a cost rate balance can be utilized.
24.5.1.2 Exergy Analysis The cost rate balance for this component can be written as
As exergy analysis is discussed extensively in other follows:
chapters, we focus here on the exergy destruction and _ 1 cw W_ AC Z_ AC c2 Ex
c1 Ex _ 2 (24.11)
exergy efficiency of the AC, which can exist in isolation
or be a component of a larger system. More details about where c1, c2, and cw are the unit costs of the inlet air, outlet
exergy analyses of compressors, applied within gas air, and work. Here, the inlet air is taken to be free, so its
turbine (GT)-based power generation, are given else- unit cost is zero, that is,
where (Ahmadi, 2006; Kanoglu et al., 2007a; Ahmadi c1 0 (24.12)
and Dincer, 2011a,b). An exergy analysis for the AC
Also, Z_ AC is the purchase cost rate of AC, which can be
follows.
expressed as follows:
The specific exergy of the air entering the compressor
can be written as follows:
ZAC  CRF  4
Z_ AC (24.13)
N  3600
ex1 h1  h0  T0 s1  s0 (24.4)
where ZAC is the purchase cost of the AC and CRF is the
where capital recovery factor, which is dependent on the interest
rate i and equipment life time n, and is determined as:
h1  h0 Cp T1  T0 (24.5)
    i1 in
T1 P1 CRF
s1  s0 Cp ln  R ln (24.6) 1 in  1
(24.14)
T0 P0
488 Exergy

Here, n denotes the annual number of operation hours for


the unit, and 4 is the maintenance factor, which is often TABLE 24.1 Physical Constraints for Optimization of
1.06 (Bejan et al., 1996). an AC
The purchase cost of the AC can be approximated as Constraints Reason
follows (Bejan et al., 1996):
RAC < 22 Material temperature limit
1
ZAC c11 m_ a RAC lnRAC (24.15) hAC < 0.92 Commercial availability
c12  hAC
where
c11 44:71$=kg s1 (24.16)
are considered, and these are given and explained in
c12 0:95 (24.17) Table 24.1.

24.5.2.4 Multiobjective Optimization


24.5.2 Optimization To determine the best among the optimal design parameters
for an AC, a modified version of a GA developed with
24.5.2.1 OFs MATLAB software is used. Figure 24.3 shows the Pareto
For each optimization problem, defining OFs is of great frontier for multiobjective optimization of an AC with the
importance. In this example multiple OFs can be consid- pressure ratio and isentropic efficiency of the compressor as
ered through multiobjective optimization. When high effi- the two main design variables. The range of values shown is
ciency is desired, it is reasonable to include the exergy limited by the problem constraints.
efficiency of the compressor as an OF. As shown in It can be seen in Figure 24.3 that the total compressor
Equation 24.10, the compressor exergy efficiency is cost increases moderately as the compressor exergy effi-
a function of the compressor pressure ratio and the ciency increases up to about 92%. Increasing the total
compressor isentropic efficiency. Another OF can be the exergy efficiency further increases the cost significantly. It
compressor cost, as expressed in Equation 24.13, which is is also shown in Figure 24.3 that the maximum exergy
a function of the compressor pressure ratio, the air mass efficiency exists at design point C (94.44%), while the
flow rate through the compressor, and the compressor compressor cost rate is the greatest at this point (33.1 $/hr).
isentropic efficiency. On the other hand, the minimum value for the compressor
Considering these two OFs, we can write: cost rate occurs at design point A and is about $5.12 per
hour. Design point C is the optimal situation when exergy
_ 2  Ex
Ex _ 1 efficiency is the sole OF, while design point A leads the
OF I AC (24.18)
W_ AC optimum design when total cost rate of product is the sole
OF. Point D is the ideal solution of the multiobjective
1
c11 m_ a RAC ln RAC CRF  4 optimization because both OFs are at their optimal values,
c12  hAC
OF II Z_ AC that is, at higher exergy efficiency and lower total cost rate.
N  3600
(24.19)

24.5.2.2 Decision Variables


The main design parameters or decision variables for
optimization of the AC are the compressor pressure ratio
and the compressor isentropic efficiency. Thus, to perform
a multi-objective optimization they are considered as our
decision variables.

24.5.2.3 Constraints
To formulate a meaningful optimization problem, cons-
traints often exist that must be satisfied while performing
the optimization. Often these constraints are used to ensure FIGURE 24.3 Pareto frontier for the optimization of an AC, highlighting
solutions are reasonable and realistic. Here, two constraints the best trade-off among values for the OFs.
Chapter | 24 Exergy and Multiobjective Optimization 489

Since this point is not located on the Pareto frontier, point B


could be selected as one of the best solutions because it is
close to the ideal solution.
The variations of compressor pressure ratio and
compressor isentropic efficiency are illustrated in Figures
24.4 and 24.5, respectively, where scatter distributions for
the populations in the Pareto frontier are shown for each of
these design parameters. The points in these figures are
obtained from the developed MATLAB code, and they
show how the design parameters change within their
allowable ranges. In a GA, a population (called chromo-
somes or the genotype of the genome), which encodes
candidate solutions (called individuals, creatures, or
phenotypes) to an optimization problem, evolves toward
better solutions. The scatter distribution of design param-
FIGURE 24.5 Scatter distribution of compressor pressure ratio with
eters are within the range exhibiting good selections of
population in Pareto frontier.
these two parameters for optimization purposes. It is noted
in Figures 24.4 and 24.5 that the points are not just near the
boundaries, they are also scattered almost randomly within
their ranges (8e22 for RAC, and 0.78e0.92 for hAC). In for this energy system, thermodynamic modeling is
real optimization, the selection of decision variables is required and is also explained.
based on the scattered distribution of the decision vari-
ables, providing an efficient search for the best optimal
solution of OF.
24.6.1 Thermodynamic, Economic, and
Environmental Modeling and Analysis
24.6 CASE STUDY: GT POWER
GT plants are widely utilized throughout the world for
GENERATION PLANT
electricity generation, and natural gas is often used as the fuel
To improve understanding of the optimization of energy in such plants. Today, many electrical generating utilities are
systems and how methods are applied, a detailed optimi- striving to improve the efficiency (or heat rate) of existing
zation case study of a complex system is considered here of thermal electric generating stations, many of which are over
in the form of a GT power generation plant. We describe the 25 years old. Often, an energy efficiency improvement of
definition of OFs and constraints and how to perform only a few percent is sought since the costs and complexity of
a sensitivity analysis. The OFs can involve efficiency, cost, the measures required to improve efficiency moderately are
environmental impact, and other factors. To determine OFs more manageable than more extensive options.
The GT power plant considered in this case study is
shown in Figure 24.6. It includes an air preheater (AP),
which is part of a heat exchanger that recovers heat from
the exhaust gas.

FIGURE 24.4 Scatter distribution of compressor isentropic efficiency


with population in Pareto frontier. FIGURE 24.6 Schematic diagram of a GT power plant.
490 Exergy

During operation of the GT power plant, air at express the pressure drop through the CC DPCC as
ambient conditions enters an AC, which is treated as follows:
adiabatic, at point 1. After compression to point 2, the P4
air enters the AP, where its temperature is increased to 1  DPcc (24.23)
P3
point 3. The hot air exiting the AP is supplied to the
combustion chamber (CC). Fuel at point 9 is injected into GT: An energy balance for a control volume around the
the CC and hot combustion gases exit (point 4) and pass GT shown in Figure 24.6 gives the following:
through a GT to produce power. The hot flue gases
leaving the GT at point 5 pass through the heat recover    1gg 
p4 g g
heat exchanger, which increases the temperature of the air T4 T3 1  hGT 1 (24.24)
p3
before it enters the CC, and then they exit the plant as flue
gases at point 6, when they are emitted to the
where T3 is the gas turbine inlet temperature (GTIT), hGT is
surroundings.
the GT isentropic efficiency and ga Cpg =Cvg . With this
equation, the GT outlet temperature can be determined and
the GT output power can be evaluated as follows:
24.6.1.1 Thermodynamic Modeling
W_ GT m_ g cpg T3  T4
Energy balances and governing equations for the compo-
nents of the GT power plant are utilized to model the
device. An energy analysis for the AC has been explained 24.6.1.2 Exergy Analysis
in the previous section. Here, the thermodynamic modeling Here we focus on the exergy destruction and exergy effi-
of the other components is explained. ciency for each component of the system. A complete
Several simplifying assumptions are made here exergy analysis is not provided; extensive details regarding
to render the modeling and analysis more tractable, exergy analyses of GT-based power generation are given
while retaining adequate accuracy to ensure meaningful elsewhere (Ahmadi, 2006; Kanoglu et al., 2007; Ahmadi
results: and Dincer, 2011c,d). Expressions for these exergy
l All processes are considered to be operating at steady parameters, which are used in the optimization procedures
state. discussed subsequently, are given in Table 24.2.
l Air and combustion products are ideal-gas mixtures.
l The fuel injected into the CC is taken to be natural gas. 24.6.1.3 Exergoeconomic Analysis
l Heat loss from the CC is 2% of the fuel energy entering
the CC, based on lower heating value (LHV), and all Exergoeconomics is the area of engineering that appro-
other components are considered adiabatic. priately combines, at the level of systems and their
l The dead state is P0 1.01 bar and T0 293.15 K. components, thermodynamic evaluations based on an
exergy analysis and economic principles, in order to
A description of the thermodynamic modeling of the provide information that is useful to the design and
components follows: operation of a cost-effective system, but that is not readily
AP: We can write the following energy balance for obtainable by conventional energy and exergy analyses
a control volume around the AP: and economic analysis (Kotas, 1985; Bejan et al., 1996).
Some suggest that, when exergy costing is not applied, the
m_ a h3  h2 m_ g h5  h6 hAP (24.20) general term thermoeconomics is more appropriate as it
where hAP is AP efficiency. A pressure drop through the characterizes any combination of thermodynamic and
preheater DPAP is considered, where: economic analysis (Kotas, 1985; Bejan et al., 1996;
Ahmadi and Dincer, 2010). In order to define a cost
P3 function that is dependent on the optimization parameter,
1  D PAP (24.21)
P2 specific component cost is expressed a function of ther-
modynamic design parameters.
CC: As for the preheater, an energy balance for When applying exergoeconomics, a parameter called
a control volume around the CC is written: flow cost rate C_ ($/hour) is defined for all flows in a system,
and a cost rate balance is written for each component as
m_ a h3 m_ f LHV m_ g h4 1  hcc m_ f LHV (24.22) follows:
Here, the LHV of the fuel, which is 50,000 kJ/kg in this X X
C_ q;k C_ i;k Z_ k C_ e;k C_ w;k (24.25)
analysis, and hcc is the CC efficiency. Also, we can i e
Chapter | 24 Exergy and Multiobjective Optimization 491

TABLE 24.2 Expressions for Exergy Destruction Rate and Exergy Efficiency for Each Component of the GT Power Plant

Component Exergy destruction rate Exergy efficiency


Compressor _ 1 W_ AC Ex
Ex _ 2 E x_ D E x_ 2  E x_ 1
AC
W_ ac
AP E x_ 2 E x_ 5 E x_ 3 E x_ 6 E x_ D E x_ 3  E x_ 2
AP
E x_ 5  E x_ 6
CC E x_ 3 E x_ 9 E x_ 4 E x_ D E x_ 4
CC
E x_ 3 E x_ 9
GT E x_ 4 E x_ 5 W_ GT Ex
_ D W_ GT
GT
E x_ 4  E x_ 5

Accordingly, the expression for the cost rate of exergy


Cost rate balances are generally written so that all terms
destruction can be expressed as follows:
are positive. Equation 24.25 can be rewritten using unit
costs as follows (Bejan et al., 1996): C_ D;k cF;k Ex
_ D;k (24.29)
X X
_ e cw;k W_ k cq;k E_ xq;k
ce Ex ci E_ x i k Z_ k Several methods have been proposed to express the
k
purchase cost of equipment in terms of design parameters in
(24.26) Equation 24.10. We convert the capital investment to a cost
per time in Equations 24.13 and 24.14. To determine the
where component exergy destruction cost rate, the exergy
C_ j cj Ex
_ j (24.27) _ D,k is evaluated using exergy balances
destruction rate Ex
from the previous section.
In this analysis, the fuel and product exergy need to be To estimate the cost rate of exergy destruction in each
defined. The product exergy is defined according to the component of the plant, we first solve the cost balances for
component under consideration, while the fuel represents each component. In the cost balance (Eq. 24.10), there is
the energy source that is consumed in generating the more than one inlet and outlet flow for some components.
product. Both product and fuel are expressed in terms of In such cases, the number of unknown cost parameters is
exergy when utilizing an exergoeconomic approach. The higher than the number of cost balances for that compo-
cost rates associated with the fuel C_ F and the product C_ p of nent. Auxiliary exergoeconomic equations are developed to
a component are obtained by replacing the exergy rates. allow such problems to be solved. Writing Equation 24.26
In the cost rate balance formulation (Eq. 24.10), there is for each component together with the auxiliary equations
no cost term directly associated with the exergy destruction forms a system of linear equations as follows:
of each component. Accordingly, the cost associated with
the exergy destruction in a component or process is _ k   ck  Z_ k 
Ex (24.30)
a hidden cost. The exergy balance in terms of fuel and where Ex_ k , ck , and Z_ k are the exergy rate matrix
product exergy rate follows: (obtained via exergy analysis), the exergetic cost vector (to
E_ xF;k E_ xP;k E_ xD;k (24.28) be evaluated), and the vector of Z_ k factors (obtained via
economic analysis), respectively. That is,

2 3 2 3 2 3
_ 1
Ex 0 0 0 0 0 0 0 0 c1 0
6 Ex
_ _ 2 _ 7 0 7 6 7 6 Z_ AC 7
6 1 Ex 0 0 0 0 Ex 0 7 6 c2 7 6 7
6 0 _
Ex2 Ex_ 3 0 _ 5
Ex _ 6
Ex 0 0 7 6
0 7 6 c3 7 7 6 Z_ AP 7
6 6 7
6 0 0 _
Ex3 Ex_ 4 0 0 0 0 _ 9 7 6 c4 7
Ex 6 Z_ CC 7
6 7 6 7 6 7
6 0 _ _ 5 _ 7 _ 8 0 7 6 7 6 _ 7
6 0 0 Ex4 Ex 0 Ex Ex 7  6 c5 7 6 Z GT 7 (24.31)
6 0 0 7 6 7 6 0 7
6 0 0 1 1 0 0 0 7 6 c6 7 6 7
6 0 1 7 6
0 7 6 c7 7 7 6 0 7
6 0 0 0 0 0 1 6 7
4 0 0 0 0 1 1 0 0 0 5 4 c8 5 4 0 5
0 0 0 0 0 0 0 0 1 c9 Fc
492 Exergy

24.6.1.4 Exergoenvironmental Analysis 24.6.2.1 Definition of OFs


A primary objective for minimizing environmental The three OFs considered here for multiobjective opti-
impact often is to increase the efficiency of energy mization are exergy efficiency (to be maximized), the
conversion processes and thereby decrease the fuel use. In total cost rate of the product and environmental impact
recent years, particular emphasis has been placed on (to be minimized), and the CO2 emission (to be mini-
releases of carbon dioxide, since it is the main greenhouse mized). The second OF expresses the environmental
gas, and optimization of thermal systems based on this impact as the total pollution damage ($/hour) due to
parameter has received much attention. A focus of the CO and NOx emissions by multiplying their respective
present case study is to consider emissions of pollutants flow rates by their corresponding unit damage costs,
(e.g., CO and NOx). Note that many approaches reported which have been reported elsewhere as CCO 0.02086
for optimizing power plants pay little attention to envi- $/kg CO and CNOx 6.853 $/kg NOx (Ahmadi and
ronmental impacts. Dincer, 2010). In the present work, the cost of pollution
The adiabatic flame temperature in the primary zone of damage is assumed to be added directly to the expen-
the CC can be expressed as follows (Gulder, 1986): ditures that must be paid, making the second OF the
   sum of thermodynamic and environmental objectives.
Tpz Asa expbs l2 px qy jz (24.32)
Consequently, the OFs in this analysis can be expressed
where p is dimensionless pressure (P/Pref), q is dimen- as follows:
sionless temperature (T/Tref), j is the H/C atomic ratio, s
f for f  1 (where f is mass or molar ratio) and s f  l GT power plant exergy efficiency:
0.7 for f  1. Moreover, x, y, and z are quadratic functions W_ Net
of s based on the following expressions: jTotal (24.38)
m_ f ;CC  LHV  x

x a1 b1 s c1 s 2
(24.33)
where W_ net , m_ f ;CC , and x denote, respectively, the GT net
y a2 b2 s c2 s2 (24.34) output electrical power, the mass flow rate of fuel injected
into the CC, and a fuel exergy/energy correlation parameter
z a3 b3 s c3 s2 (24.35) for gaseous fuels of composition (CxHy).
where A, a, b, l, ai, bi, and ci are constants. Values for these
l Total cost rate:
parameters and more details on the methodology are pre- X
sented elsewhere (Gulder, 1986). C_ Tot C_ f Zk C_ D C_ env (24.39)
The amount of CO and NOx produced in a given CC k

and combustion reaction is related to the adiabatic flame where the cost rates of environmental impact and fuel are
temperature (Gulder, 1986). The pollutant emissions (in expressed as follows:
grams per kilogram of fuel) can be determined as follows:
 
71100 C_ env CCO m_ CO CNOX m_ NOX ; C_ f cf m_ f  LHV
0:15  10 s exp
16 0:5
TPZ (24.40)
mNOx   (24.36)
DP3 _ _
Here, Z K and C D are the purchase cost rate of each
P0:05 0:5
3
P3 component and the cost rate of exergy destruction,
  respectively. In addition, m_ CO and m_ NOX are calculated with
7800
0:179  109 exp Equations 24.36 and 24.37, respectively.
T
mCO   PZ (24.37)
DP 3
P23 s 0:5
P3
24.6.2.2 Decision Variables
where s is the residence time in the combustion zone
The decision variables (design parameters) in this case
(assumed constant and equal to 0.002 s), Tpz is the primary
study are compressor pressure ratio RAC, compressor
zone combustion temperature, P3 is the combustor inlet
isentropic efficiency hAC, GT isentropic efficiency hGT, CC
pressure, and DP3 =P3 is the nondimensional pressure drop
inlet temperature T3, and GTIT. Even though the decision
in the CC.
variables may be varied in the optimization procedure, each
decision variable is normally required to be within
a reasonable range so constraints are applied. These
24.6.2 Optimization
constraints and the reasons of their applications are listed in
In this section, we consider OFs, design parameters, and Table 24.3, based on previous analyses (Ahmadi and
constraints, as well as overall optimization. Dincer, 2010, 2011a).
Chapter | 24 Exergy and Multiobjective Optimization 493

TABLE 24.3 Constraints for Optimization of GT Power TABLE 24.4 Comparison of Power Plant Data and
Plant Simulation Code Results

Constraints Reason Simulation


GTIT < 1550 K Material temperature limit Measured code Difference
Quantity Unit data results (%)
P2/P1 < 20 Commercial availability

T2 C 321.4 332.0 3.2
hGT < 0.92 Commercial availability

T6 C 500 529.0 5.5
hAC < 0.92 Commercial availability

T7 C 448 487.0 8.0
T7 > 400 K To avoid formation of sulfuric
acid in exhaust gases m_ f kg/s 8.44 8.18 3.2

j % 32.0 30.4 5.0

24.6.2.3 Model Validation


To simulate the performance of the GT power plant, a simu- exergy efficiency further, from 42.1% to 42.5%, results in
lation code developed with MATLAB software is used. To a more rapid, but still moderate, increase in the cost rate
validate the simulation code, the results of this study are of the products. Increasing the exergy efficiency above
compared with an actual operating GT power plant located in 42.5% leads to a significantly more rapid increase in the
the Shahid Salimi Power Plant, Neka, Iran. This power plant total cost rate.
is located close to Mazandaran city near the Caspian Sea, in Four points (AeD) on the Pareto frontier are shown in
one of the Northern provinces in Iran. A schematic diagram Figure 24.7, and the determined values of these points are
of this power plant is shown in Figure 24.6. shown in Figure 24.8. It can be seen that the maximum
Based on power plant data obtained in 2005, the exergy efficiency (42.9%) occurs at design point A, which
incoming air has a temperature of 20 C and a pressure of 1 also has the highest total cost rate of products. The
bar. The pressure increases to 10.1 bar in the compressor, minimum value for total cost rate of product occurs at
which has an isentropic efficiency of 82%. The GTIT is design point D. Design point A corresponds to the optimal
971 C. The turbine has an isentropic efficiency of 86%, and point when exergy efficiency is the single OF, while design
the regenerative heat exchanger has an effectiveness of 87%. point D is the optimum point when total cost rate of product
The pressure drop through the AP is considered 3% of the is a single OF. Points B and C are two different points that
inlet pressure for both flow streams and through the CC is 3% could be selected by a decision maker, but they are just
of the inlet pressure. The fuel (natural gas) is injected at 20 C selected for comparison with points A and D. Specifica-
and 30 bar. tions for sample design points A to D in the Pareto optimal
Thermodynamic properties of the power plant based on fronts are listed in Table 24.5.
modeling and simulation are listed in Table 24.4 and con-
trasted with given power plant data. The data from the
simulation code are observed to vary from those for the
actual power plant by about 3%e8%, with the maximum
difference applicable to the AP outlet temperature. This
good agreement suggests that data generated by the simu-
lation code for the GT power plant is reasonably valid.

24.6.2.4 Optimal Solutions


To optimize the system, a GA method (Deb and Goel,
2001) is used. The optimization code is developed with
MATLAB software based on the evolutionary algorithm.
Figure 24.7 shows the Pareto frontier solution for the
GT power plant, where the OFs are defined as described
in the multiobjective optimization section of this chapter.
In this figure, while the total exergy efficiency of the
process increases up to about 42.1%, the total cost rate of FIGURE 24.7 Optimal solution points in the Pareto frontier for exergy
products increases only slightly. Increasing the total efficiency and total cost rate for the GT power plant.
494 Exergy

FIGURE 24.8 Scatter of variables for the Pareto optimal frontier in Figure 24.7 for the GT power plant. (a) isentropic efficiency of air compressor versus
population, (b) isentropic efficiency of gas turbine versus population, (c) gas turbine inlet temperature versus population, (d) air compressor pressure ratio
versus population and (e) air preheater temperature versus population.
Chapter | 24 Exergy and Multiobjective Optimization 495

It is emphasized that in multiobjective optimization,


each point in the Pareto frontier can be considered an TABLE 24.6 Comparison between Actual GT Power
optimized point. Therefore, selection of a particular Plant Quantities and Optimized Data from this Case
optimum solution depends on preferences and criteria of Study
the decision maker. Hence, each decision maker may select Actual plant Optimized Difference
a different point as the optimum solution that better suits Quantity value value (%)
with his/her desires.
Exergy efficiency, j 32.0 42.8 34
(%)
24.6.2.5 Optimal Relation between Total Cost
C_ Total ($/hour) 7570 5540 37
Rate and Exergy Efficiency
C_ env ($/hour) 16.4 10.9 51
To provide a helpful tool for the optimal design points of
the GT cycle in Figure 24.6, the following equation is
derived for the Pareto frontier given in Figure 24.7 using
curve fitting: Table 24.7 lists values of several design parameters for
the optimized solution (at point B in Figure 24.7) and the
7:31j3  15:15j 3:65
12:32j2
C_ Tot  1000 actual GT power plant data.
j4 9:27j  8:65j2 1:89j 0:014
3

(24.41)
This fitted equation permits the optimal value of the total 24.6.2.6 Distribution of Design Parameters
cost rate of the power plant to be approximated conve- The distribution of optimal design parameters in the Pareto
niently as a function of exergy efficiency. frontier curve is shown in Figure 24.8 for each of the
In Table 24.6, the product cost rate, the exergy effi- decision variables in Table 24.7. The lower and upper
ciency, and the cost of environmental impacts for the bounds of these variables are represented by dashed lines.
considered power plant (the Shahid Salimi Power Plant in These distributions suggest that the GT isentropic effi-
Iran) are compared with the results from multiobjective ciency (Figure 24.8b) and air preheater temperature
optimization. (Figure 24.8e) are driven by optimization toward their
Note that the values for multiobjective optimization are maximum values. That is, the distributions indicate that
based on point B in Figure 24.7. This point is chosen increases in these two design parameters lead to improve-
because it can, in many ways, be considered the best ments in both OFs, so their maximum values are selected.
point relative to other points in the Pareto solution. This The other design variables in Figure 24.8 show the scattered
point exhibits a good balance, achieving a relatively high distribution. Therefore, it can be predicted that variations of
efficiency and low total cost rate relative to other points. these variables result in conflict between two OFs in a small
Optimization is observed in Table 24.6 to increase the total range of their variations. To better understand this trend and
exergy efficiency of the GT power plant by 34% relative to the effects of these design parameters on both OFs at
its actual value of 32%, to reduce the total cost of exergy optimal points, the variations of OF with changes in design
destruction by 37%. Also, the cost of environmental impact parameters for four different points (AeD) in the Pareto
is significantly decreased by over 50%. curve are examined in next section using sensitivity anal-
yses. Note that points other than AeD exhibit similar
variations.
TABLE 24.5 Values of Selected Design Quantities for
Selected Points on the Pareto Optimal Frontier for the
GT Power Plant TABLE 24.7 Comparison of Some Variables for the
Optimization Results with Actual Plant Data
Point on Pareto frontier
Decision variable Actual plant data Optimization results
A B C D
Quantity RAC 10.1 14.4
Exergy efficiency, 42.93 42.78 42.15 41.26
j (%) hAC 0.82 0.86
C_ D;PP ($/hour) 1463 1453 1428 1418 hGT 0.86 0.89
C_ env ($/hour) 10.89 10.91 12.28 13.75 GTIT (K) 1244 1425
C_ D; Total ($/hour) 5908 5542 5407 5365 T3 (K) 754 850
496 Exergy

FIGURE 24.9 Variation of exergy efficiency with total cost rate for five design parameters of the GT power plant, for four optimized cases (AeD).
(a) Effect of increase in AC isentropic efficiency, (b) effect of increase in GT isentropic efficiency, (c) effect of increase in GTIT, (d) effect of increase in
AC pressure ratio and (e) effect of increase in AP temperature.
Chapter | 24 Exergy and Multiobjective Optimization 497

24.6.2.7 Use of Sensitivity Analyses to Describe the impact of varying selected design parameters on the OF,
Effect of Decision Variables on OFs or all OFs when multiobjective optimization is utilized.
In the case study of the GT power plant, a multi-
Sensitivity analyses can be used to describe the effect on the objective GA is applied to optimize two important OFs: (1)
OFs of varying decision variables. The results of such sensi- the power plant exergy efficiency and (2) the total cost rate
tivity analyses are shown in Figure 24.9 for the GT power plant. of the plant (including investment cost, cost of exergy
The variations of OFs with changes in compressor destruction, and cost of environmental taxes). Multi-
isentropic efficiency are shown in Figure 24.9a. It is seen objective optimization is shown to provide significant
that an increase in this design parameter, within its allow- benefits for the GT power plant, increasing its overall
able range, increases the GT power plant exergy efficiency, exergy efficiency by 34% and decreasing associated envi-
while decreasing the total cost rate at lower exergy effi- ronmental impacts by 51%, relative to the design parame-
ciencies and increasing it for higher exergy efficiencies. A ters. Also:
conflict is thus observed between the OFs. Moreover, since
the region where improvements in both OFs are observed is l Increasing the compressor isentropic efficiency incre-
greater than the region where conflict is observed, this ases the overall exergy efficiency of the GT power plant,
design parameter must have a scattered distribution near the while decreasing the total cost below a certain exergy
maximum values. Figure 24.8a confirms this trend. In efficiency and increasing it above that exergy efficiency.
Figure 24.9b, it is indicated that an increase in GT isen- Similar qualitative behavior is observed when the GTIT
tropic efficiency leads to an increase in exergy efficiency as is increased.
well as a decrease in total cost rate. Therefore, higher l Increasing the GT isentropic efficiency improves both
values of GT isentropic efficiency can be advantageous. OFs, indicating that this parameter does not lead to
Figure 24.9c shows the variation of both OFs when GTIT a conflict between OFs. Similar qualitative behavior is
varies within its allowable range. An increase in this design observed when the AP temperature is increased. Thus
parameter is seen to increase the exergy efficiency of the maximum values are selected for these design parame-
power plant, but to decrease the total cost rate notably when ters in the optimization.
the exergy efficiency is lower and to increase the total cost
rate notably when the exergy efficiency is higher. This
behavior is because an increase in GTIT above a reasonable PROBLEMS
range results in an increase in the cost of the CC, directly 24.1 Consider a heat recovery steam generator that is used
affecting the total cost rate of the plant. to recover the exhaust energy from flue gases exiting
The effects of increasing the compressor pressure ratio a GT to produce saturated steam. How is the exergy
on both OFs are shown in Figure 24.9d. An increase in efficiency defined for this device if it is considered as
compressor pressure ratio increases the GT exergy effi- one of our OFs? What would the proper decision
ciency for all ranges, but decreases the total cost rate at variable be for this device? Explain the effect of GT
lower exergy efficiencies and increases it at higher exergy outlet temperature on the selection of heat recovery
efficiencies. This trend is mirrored for variations of GTIT steam generator decision variables.
on total cost rate in Figure 24.9c. 24.2 Identify the source of exergy destruction in a GT
In Figure 24.9e, it is seen that increasing the AP power plant and explain how this loss can be reduced.
temperature (T3) leads to improvements in both OFs. This Then optimize a power plant using appropriate
explains why the optimal points in Figure 24.8e are at the decision variables and consider the total exergy
higher values. Therefore, variations of this parameter do destruction as your OF.
not cause a conflict between the two OFs. 24.3 When an engineer or designer is deciding on the
preferred optimal point on a Pareto frontier, what
point will be selected and why?
24.7 CLOSING REMARKS 24.4 Does the Pareto frontier change with changing fuel
Optimization techniques and the benefits of each are costs? Explain.
described in this chapter. The coverage includes OFs, 24.5 For GT power plants considered in this chapter, but
constraints, and evolutionary algorithms. A focus is placed without the AP, evaluate the exergy and energy effi-
on multiobjective optimization, and validation and ciencies, the Pareto frontier, and the total exergy
comparisons between base and optimized cases are dis- destruction and compare these with the values for the
cussed. To illustrate this, an AC and a GT power plant are GT power plant with the AP. Explain why an AP is
examined. Contrasting optimization results with actual used in GT power plants.
plant data often emphasizes the benefits of employing 24.6 If we extend the GT power plant considered in this
optimization methods. A sensitivity analysis can indicate chapter as a combined heat power system by using
498 Exergy

a single pressure heat recovery heat exchanger, that 24.7 What is the effect of interest rate i given in Equation
has an economizer and an evaporator to produce 24.14 on the Pareto frontier? Draw a Pareto frontier
saturated steam at P 15 bar, define the OFs, for interest rates of 5%, 10%, and 15%.
constraints, and proper design variables. Also, opti- 24.8 For the GT considered in this chapter, perform an
mize the system and draw a Pareto frontier and optimization by considering exergy efficiency
compare it with the Pareto frontier for the simple GT and total cost rate as two OFs for multi
cycle. objective optimization and discuss the Pareto frontier.

You might also like