Energy Conversion and Management 43 (2002) 1165–1185

www.elsevier.com/locate/enconman

Keynote paper

Application of exergy to design

Yehia M. El-Sayed *

Advanced Energy System Analysis, 41658 Higgins Way, Fremont, CA 94539, USA

Abstract
The design of an energy-intensive system for lower cost and higher efficiency is one of the essential
approaches to sustainable development. A principle is presented whereby exergy is elevated from a tool that
reveals opportunities of higher efficiency to a tool that reveals opportunities of both lower cost and higher
efficiency.
The principle states that minimizing the individual costs of the energy conversion devices of a system
with respect to their efficiency decision variables leads to a minimized cost of the system and an optimal
overall system efficiency. The cost of a device is the sum of its fixed charges and its running expenses. The
cost of exergy destruction of a device is its major running expense.
The premises on which the principle is established are explained. Second-law analysis and design models
of energy conversion devices play a central role in establishing these premises. Three useful design model
manipulations seeking equations for the costing and for the performance of energy conversion devices are
presented. Applications to the optimal design of systems and extensions to their optimal operation are
given elsewhere.
2002 Elsevier Science Ltd. All rights reserved.

Keywords: Exergy for higher efficiency and lower cost; Optimal design; Second-law-based decomposition; Enhanced
optimization

1. Introduction

Three factors drive the need for higher efficiency at lower cost for systems that use or produce
power and heat. These are fossil fuel scarcity, environmental concerns and increased world
population at rising standard of living. Even with fossil fuel abundance, the other two factors
maintain the drive. Higher efficiency at lower cost calls for more intensive and extensive system
analysis while the system is still in its design phase. Because of cost consideration, the analysis
becomes multi-disciplinary.

*
Tel./fax: +1-510-656-2783.
E-mail address: aesa87@aol.com (Y.M. El-Sayed).

0196-8904/02/$ - see front matter
2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 1 9 6 - 8 9 0 4 ( 0 2 ) 0 0 0 0 6 - 7
1166 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

Nomenclature

A heat exchange surface, flow passage surface, characterizing dimension of a device,

constant
c unit price: cF of fuel, cP of electricity, cf , cp per unit exergy, cd of dissipation per unit
exergy destruction, cz of capital cost, ca of a characterizing surface, cs social price
d infinitesimal change
do tube outside diameter
D exergy destruction in a device
E exergy: Ef of fuel, Ep of product, Ej of dumped exergy
g change of objective function with a decision
h film coefficient of heat transfer
H enthalpy, enthalpy per unit mass
J an objective function: J2 by second-law transformation, JR remaining objective
k constant coefficient: kd of exergy destruction
L Lagrangian, Ltube tube length
M mass, mass rate: mh of heating stream, mc of heated stream
N number of units: Ntube of tubes, Npas of tube passes in a heat exchanger
n exponent: nd of exergy destruction
P pressure: P0 for dead state pressure, power
PR pressure ratio
Q heat rate: Qf by fuel
R gas constant
rp pressure loss ratio DP =Pin : rph hot stream, rpc heated stream
S entropy, entropy per unit mass
T temperature, absolute temperature: T0 for dead state temperature
U a decision design variable, fU g a decision vector, an overall heat transfer coefficient
V specific volume, a variable (dependent or decision): VT of thermodynamic, VD of de-
sign, VM of manufacture, VT;D thermodynamic input to design, VD;M design input to
manufacture
W work, width of a duct
X a dependent variable: XT thermodynamic, XD design, XM manufacture, fX g state
vector
x species fxg composition vector
Y a decision variable: YT thermodynamic, YD design, YM manufacture, YL local, YG
global, fY g a decision vector
Z an equipment capital cost
Greek symbols
d a small change
D a difference: DT a temperature difference, DTm logarithmic mean temperature differ-
ence, DP a pressure loss, DPh loss of a heating stream, DPc of a heated stream; DPs shell
side, DPt tube side
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1167

g an efficiency variable, adiabatic efficiency, heat exchange effectiveness

k
P an internal price,
P LagrangeP multiplier P P
summation: f over feeds, i over devices, j over reject streams, p over products
t value: tAD relative value of A to D

Thermoeconomics, developed during the 1960s, views this multi-disciplinary analysis from the
discipline of thermodynamics. The name was coined in 1962 by Professor Tribus [1]. Seawater
desalination processes were of prime concern to gain insight into the interaction between the
surface of separation and the energy requirement. Even though at that time oil prices were 0.1 to
0.2 today’s prices, the impact on the price of water was significant compared to conventional
water prices. The publications [2,3] in 1970 were perhaps two of earliest on the subject matter.
Later, in the early 1980s, the interest in further development of thermoeconomics to handle en-
ergy-intensive systems in general was initiated by Professor Gaggioli [4,5]. Many researchers re-
sponded positively to the initiation. In the last 25 years, the development of thermoeconomics has
been impressive in more than one direction. The recent developments by Valero et al. [6,7], the
author [8] and Lazzaretto and Tsatsaronis [9] may adequately represent the different directions of
development. These directions are not yet free from inconsistencies [10].
In this paper, the communication/optimization strategy proposed by the author [8] to manage
the optimization complexity of an energy-intensive system given a cost objective function is re-
viewed. The strategy then focuses on ways by which designers can handle the optimization of the
system’s devices in conformity with the system’s objective function. Two ways are described that
have the advantage of enhanced system optimization while gaining the expertise and judgment of
the designers of energy conversion devices. The first requires the designers to design for minimum
material requirements given the exergy destructions of the devices, in other words, given the input
and output states of the devices. The second requires designers to select, from a series of available
designs, the design that minimizes exergy destructions given what defines only the essential input
parameters of the devices. The first is a tailored design procedure that does not require designers
to use exergy analysis and is the one used in previous applications. The second is a selection
procedure presented here for the first time. It exports exergy to the discipline of design and makes
it part of the design practice of the energy conversion devices. The design model of a heat ex-
changer is used to explain the two ways. Applications revealing higher efficiency at lower cost,
predicting the performance equation of a system and scheduling the optimal operation of a group
of systems can be found in [8,11–14].

2. The essentials of optimal system design

Any attempt claiming design improvement of a complex energy system should be explicitly
characterized by an objective function and a selected set of decision variables that are the degrees
of improvement freedom. Because the decision variables are diverse and their number is large, an
approach to system decomposition becomes most desirable. The objective, the decisions variables
and the approach to decomposition are essentials to the optimal design of complex systems. They
1168 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

are not independent of each other and they have to be matched to each other to establish an
efficient communication and optimization strategy.

2.1. The objective function

A cost objective function, in monetary units, suitable for the design phase of an energy-
intensive system of a specified production rate is the production cost J :
X
f X
i
J¼ cF F þ czi Zi þ CR : ð1aÞ
cF is the unit cost of a feeding resource F as occurring in the market place and czi is the capital
discount rate of a device i of capital cost Zi . CR is a constant remainder cost as far as the system
design is concerned. When a design becomes a project, CR may become a variable with respect to
other non-system-design decisions.
Eq. (1a) applies to single- and multi-product systems. For the case of multi-products, the
production rate is often identified by the rate of one product, and the rest of the products follow
as dependent variables. Also, an allocation assumption of the production cost among the products
is usually introduced. The objective function
X
f X
i X
p
J¼ cF F þ czi Zi  c P P þ CR ð1bÞ
is a net cost that may be used for multi-product systems. The negative of J of Eq. (1b) is prof-
itability. cP is the unit cost of a product P as occurring in the market place. Eq. (1b) reflects the
cost variation by the dependent products. The minimization of net cost is a maximization of
profitability. However the minimum by Eq. (1a) does not necessarily coincide with the minimum
by Eq. (1b).

2.2. Communication

The cost minimization of an energy system of interconnected devices interfaces at least four
disciplines of knowledge: thermodynamics fF ; P g, design and manufacture fZg and economics
fcF ; cP ; cz g. Each discipline has its own practice expressed by models having their own input
decision variables. Communication is needed to deal with the various practices of the involved
disciplines. Fig. 1 sets the stage for optimal system design. It illustrates a direct communication
strategy among the four disciplines triggered by a thermodynamic model and aiming at optimal
system design for minimum cost. The figure shows also the principle of generating costing
equations for indirect communication to be discussed later on.

2.3. Decomposition

The strategy proposed by the author [8] decomposes an energy-intensive system at two
levels; at the discipline level and at the device level. Decomposition at the discipline
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1169

Fig. 1. Direct and indirect interdisciplinary exchange of information.

level manages the large number of variables involved in the optimization process. Decom-
position at the device level manages the interactions among the system processes due to their
1170 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

complex connectivity. In simple terms the two levels of decomposition can be established as
follows:

• Decomposition at the discipline level: Select one of the participating disciplines as the active dis-
cipline. The flow of information from the other disciplines is expressed by costing equations in
terms of the variables of the active discipline. From the viewpoint of the active discipline, the
system is decomposed. Two active disciplines of strong interaction may be merged together.
• Decomposition at the device level: Divide the decision variables into local decisions of devices
and global decisions of the system. Approximate the major effect of a local decision variable
of a device as localized, then later correct for the relatively minor interactions among local de-
cisions. Devices can then be optimized one by one with respect to their respective local deci-
sions. The devices are not decomposed with respect to the global decision variables since
they strongly affect several devices simultaneously.

3. The two-level decomposition used

3.1. The active discipline

The fact that the creation of a system occurs in the discipline of thermodynamics and the fact
that other disciplines see only individual parts of the system and not all the system make the
selection of thermodynamics as the active discipline the most suitable selection. Thermodynamic
decision variables are identified from conventional thermodynamic computations. The compu-
tations obtain solutions in terms of a production rate, efficiency parameters (local decisions) and a
few levels of pressure, temperature and composition (global decisions) along with property re-
lations and mass and energy balance equations. Efficiency parameters constitute most of the
thermodynamic decision variables. Other decision variables are processed within their disciplines
and communicated to the thermodynamic discipline as shown in Fig. 1.

3.2. Decomposition at the discipline level

P
Design and manufacture decide the capital cost Z of a device. Express Z as ca A or ca A þ C,
where A is the characterizing dimension (length, area, volume or mass), ca is its unit cost and C is a
constant. Quite often, the characterizing dimension is a surface area shaped as a flow passage or as
a surface accommodating a chemical reaction or exchanges of heat and/or mass. Design practice
decides A and manufacturing practice decides ca . Let the manufacture unit costs fca g and the
economic ones fcF ; cP ; cz g be expressed as constants or discrete variables in a tabulated form. The
unit costs fca g would depend on the material, shape and severity of operation of the device.
The unit costs fcF ; cP ; cz g would depend on the economic environment of the system, which in
turn is time and location dependent.
Now the characterizing dimension A of a device needs to be communicated between the dis-
ciplines of thermodynamics and design. There are two well-known mutually inclusive practices of
deciding A: ‘‘design the device’’ and ‘‘select a device’’. In design, minimum A is sought for a given
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1171

performance (exergy destruction). In selection, A that minimizes exergy destruction (maximizes

performance) is selected from an available series of designs.
In the design practice, designers supply, over a range of interest, the minimum characterizing
A’s as function of the efficiency variables fgg and the essential (and not all) thermodynamic
loading variables fVT in g at the device boundaries such as mass rates, pressures and temperatures
set by a system solution. The variables fgg and the essential fVT in g mean prescribed exergy de-
structions. The generation of the minimized surface Amin ðfgg and fVT in gÞ is illustrated in Fig. 1
and may be called the capital (or material) costing equation.
In the selection practice, designers select the characterizing A’s that minimize exergy destruc-
tions for a range of essential input thermodynamic variables fVT in g set by a system solution. The
generation of the minimized destruction Dmin ðfAg and fVT in gÞ is similarly illustrated in Fig. 1 and
may be called the fuel (or exergy destruction) costing equation.

3.2.1. Capital costing equations

The augmented thermodynamic model decides all efficiency decision variables fgg of its devices
over a range of interest to the system and sends them along with the corresponding essential input
thermodynamic variables fVT in g to the respective design models. Each design model responds
with the smallest characterizing dimension fAmin g expressed as function of the sets of inputs
fVT in ; fggg. The resulting equation
Amin ¼ AðVT in ; fggÞ ð2Þ
decomposes the system at the discipline level. A capital cost ca AðVT in ; fggÞ is defined as a costing
equation and implies the correlativity of Amin in terms of ðVT in ; fggÞ. The derivation of capital
costing equations is described later under the manipulation of the design models.
A few of fVT in g and most of fgg are system decision variables. The rest are dependent vari-
ables. Efficiency variables fgg of the system devices are local decisions to each device. They
emerge naturally as decision variables when computing the system’s overall efficiency. Adiabatic
efficiency, heat exchange effectiveness, pressure loss ratios, heat loss ratios, temperature differ-
ences, extent of reaction, etc., are examples of the efficiency variables of devices. The thermo-
dynamic loading variables fVT in g are mainly pressures, temperatures, compositions, power, mass
rate and/or heat rate. A few of the fVT in g may be decision variables belonging to the system as a
whole, such as variables at the boundary of the system or pressure and temperature operating
levels within the system.

3.2.2. Fuel costing equations

The designer sets the candidate types and sizes of the device that can be considered for em-
bedding in the system. The augmented thermodynamic model sends to the designer the essential
input variables fVT in g, without the efficiency parameters, over the range of variation of interest to
the system. The designer responds with the minimum exergy destructions Dmin as a function of the
characterizing dimension A and the input variables fVT in g. The resulting equation
Dmin ¼ DðA; VT in Þ decomposes the system at the discipline level. From a selection viewpoint,
expressing D as a function of A tends to occur naturally:
Dmin ¼ kAnd ; ð3aÞ
1172 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

where
k ¼ k 0 ðVT in Þ: ð3bÞ
The derivation of the fuel costing equations is also described later under the manipulation of
design models.

3.2.3. Market place costing equations

In the market place, capital costs are often over-simplified by the unit cost of a single char-
acterizing parameter for a collection of devices. The parameter may be a loading parameter such
as mass rate, heat rate or power or a physical parameter such as mass, length, area or volume.
These costs are not helpful to system improvement. They already lost a great deal of the useful
information that was there during system design in forms similar to Eqs. (2) and (3a).

3.3. Decomposition at the device level

Decomposition at the device level takes off smoothly from second-law analysis when the effi-
ciency variables are selected as decision variables. The take-off is as follows:
The overall system exergy balance is
X X X X
Ep þ Dþ Ej  Ef ¼ 0; ð4Þ
where fEp ; Ef g are the exergies of feeds and products, fDg exergy destructions by the devices and
fEj g exergy of wasted streams. Convert fcF ; cP g of Eq. (1a) or (1b) to be per unit corresponding
exergies fcf ; cp g. Add Eq. (4) as a constraint priced at an exergy destruction price cd , where cd is
an undetermined Lagrange multiplier. Augment the objective to a Lagrangian to obtain for Eq.
(1a)
X
i X
j X
f X
p
L¼ ðcd Di þ czi cai Ai Þ þ cd Ej þ ðcf  cd ÞEf þ cd Ep ¼ J2 þ JR ð5aÞ
and for Eq. (1b)
X
i X
j X
f X
p
L¼ ðcd Di þ czi cai Ai Þ þ cd Ej þ ðcf  cd ÞEf  ðcp  cd ÞEp ¼ J2 þ JRa ; ð5bÞ
where
X
i
J2 ¼ ðcd Di þ czi cai Ai Þ; ð6Þ

X
j X
f X
p X
f X
i
JR ¼ cd Ej þ ðcf  cd ÞEf þ cd Ep ¼ cf Ef  c d Di ; ð7aÞ

X
j X
f X
p X
f X
p X
i
JRa ¼ cd Ej þ ðcf  cd ÞEf  ðcp  cd ÞEp ¼ cf Ef  c p Ep  cd Di : ð7bÞ
J2 is a second-law-based objective function pairing the cost of devices (dissipaters) and the cost of
their dissipations and JR is a remainder objective, both to be minimized. Eq. (7a) gives JR of
objective equation (1a) and (7b) gives JRa of (1b).
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1173

The objective J2 involves strong trade-off between the exergy destruction Di and the charac-
terizing dimension Ai of a device with respect to its efficiency decisions. The trade-offs between
Di and the Pi Ak6¼i ’s of other devices are weak. Idealizing the weak trade-off as non-existent, then
J2 min ¼ Ji min and the devices can be optimized individually. If the weak trade-offs are ac-
tually absent, the optimum is obtained in one system computation using the form of Eq. (2) or
(3a). Since efficiency decisions affect other parts of the system mainly through changes in flow
rates, fast convergence to an optimum is expected. Note that any positive dissipation price cd
minimizes each pair of J2 . The one to use is that which also minimizes JR . The average exergy
destruction price cd a gives a reasonable starting exergy destruction price and is sometimes suf-
ficient:
X X . X X 
cd a ¼ cF F þ cp P Ep þ Ef : ð8Þ

Fig. 2 illustrates qualitatively the various interactions involved in Eqs. (2)–(7b). Fig. 2a shows
the interaction between the minimized characterizing dimension Amin of a device and its exergy
destruction D with respect to an efficiency parameter and with respect to a loading parameter.
Similar interaction occurs at the system level. Fig. 2b shows the trend of variation of the mini-
mized exergy destruction Dmin with the extent of the characterizing dimension A and illustrates the
broader sense of cost as being the relative values of the making and the operating resources of a
device. Fig. 2c shows the trend of variation of J2 , JR and J in monetary units with the exergy
destruction price cd . The first increases, the second decreases and their sum passes through a
relatively flat minimum. Fig. 2c is based on the quantitative results of a simple combined cycle for
the production cost of a unit power by Eq. (1a).

3.3.1. The price of exergy destruction

It is important to note that the above derivation assumes that one effective exergy destruction
price can be applied throughout a system. The flatness of the optimum with respect to the exergy
destruction price cd of Fig. 2c supports this assumption. Recently, however, a simple method has
been developed to assign an exergy destruction price cdi to each Di depending on the device lo-
cation in the system. The method does not violate the use of one effective exergy destruction price.
Thus the option of distributed prices is now available beside the search for one effective price.
Only the method is presented in Appendix A. A future publication will be devoted to its appli-
cation.

3.3.2. Global decisions

Devices can be treated as decomposed from each other with respect to their efficiency decisions
on the assumption that their influence is mostly localized and any global effect converges rapidly
to the system’s optimum. There are, however, few global decisions that belong to the system as a
whole. Operating pressure and temperature levels of a system are examples of global decisions.
Devices are not decomposed with respect to these decisions. A nonlinear programming algorithm
must be invoked. Ref. [8] gives a simplified gradient-based method that ignores cross second
derivatives at the expense of slow convergence.
1174 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

Fig. 2. Illustration of key interactions and their quantization.

 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1175

3.4. The needed feasibility proofs of the decomposition used

Two proofs are needed to support the two levels of decomposition discussed in this paper:

• the feasibility of correlations expressing a physical design dimension A in terms of thermody-

namic variables or expressing exergy destruction D in terms of physical dimensions along with
the quality of the correlations;
• the convergence of the individual optimizations of devices to a system optimum and the speed
of convergence as well as the existence of an exergy destruction price that minimizes
J ¼ J2 þ JR .

To establish both proofs, design models have to be created to encode examples of design
practices for sufficient types of energy conversion devices that help analyze a number of power
and process systems. The design models created to establish the proof are described in [8,13] along
with their references. No claim is made that these are the best design models. In the same ref-
erences, about 20 capital costing equations generated by the design models are summarized. Again
no claim of universality is given to the listed costing equations.
The generated capital costing equations showed scatter of the correlations averaged 10% with
an occasional maximum of 15–20%. Improved correlations are obtained for narrower range of
applicability. Only one fuel costing equation is generated so far. The generated equation is given
under the manipulation of design models. The correlation showed the same trend of scatter. Note
that, unlike property models, improving, updating and reviewing design models for various ap-
plications is an ongoing process even if they were the best available.
To establish the second proof, local optimization applied to various systems [8,12,14] showed
reliable convergence within 4–6 system computations irrespective of the number of the efficiency
decision variables. Fluctuations around a convergence indicate a tendency of global effect. In a
few instances, a temperature difference decision may show this tendency that may require treating
the decision as global. As for the existence of an exergy destruction price that minimizes the
objective J ¼ J2 þ JR , the results of local optimization confirm the trend indicated by Fig. 2c.

4. The principle of common objective

The proof of converging local optimizations with respect to the respective efficiency decision
variables allows the piecewise optimal design of energy conversion devices in conformity to
the system’s optimum while the designers are able to exercise their expertise and judgment. The
converging local optimization establishes the powerful principle of the common objective. The
principle states that minimizing the individual costs of the energy conversion devices of a system
as being made up of fixed capital charges and running exergy destruction expenses with respect to
their efficiency decision variables leads to the minimum cost of the system and an optimal overall
system efficiency. In simple terms, the principle says that what is good for a process in a system is
1176 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

Fig. 3. Indirect interdisciplinary exchange of information via capital costing equations.

good for the system. Exergy is thus elevated from a tool of revealing opportunities of higher
efficiency to a tool of revealing opportunities of higher efficiency and lower cost.

4.1. The principle using capital costing equations

Fig. 3 illustrates the decomposition obtaining Amin by design for a given D, i.e. when capital
costing equations are communicated. The design details of the devices of the improved system are
recoverable by the design models that generated the costing equations. The objective function of a
device i becomes
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1177

Minimize ji ¼ cz ca Ai min þ cd Di ; ð9Þ

fgi g

where
Ai min ¼ kz gni z and Di ¼ ke gni e :
Designers, however, can go about their work without the need for exergy computations as far as
the determination of Amin . The analytical solution of Eq. (9) for optimum is
gi opt ¼ ½ðkz nz Þ=ðke ne Þ 1=ðne nz Þ : ð10Þ
Substituting Ai and Di for kz and ke , the updating equation for convergence becomes
1=ðne nz Þ
gi new ¼ gi old ½ðnz =ne Þðcz ca Ai min Þ=ðcd Di Þ : ð11Þ

4.2. The principle using fuel costing equations

Fig. 4 illustrates the decomposition obtaining A by selection for Dmin , i.e. by fuel costing
equations. Designers in this case have to include exergy analysis in their design process. Exergy
destruction by design and by thermodynamic computation may serve as a validating check. When
the fuel costing equations are communicated, the objective function of a device i becomes
Minimize ji ¼ cz ca Ai þ cd kd Ani d : ð12Þ
Ai

The analytical solution of Eq. (12) for optimum is

Ai opt ¼ ½ðcz ca Þ=ðcd nd kd Þ 1=ðnd 1Þ : ð13Þ
Substituting Di min for kd , the updating equation for convergence becomes
1=ðnd 1Þ
Ai new ¼ Ai old ½cz ca ðAi Þ=ðcd nd Di min Þ ð14Þ

4.3. Direct and indirect communication

Both Figs. 3 and 4 represent an indirect communication among the four disciplines using
costing equations in contrast to the direct communication of Fig. 1. Both the direct and the in-
direct communications handle essentially the same information.

4.4. The broader sense of cost

In both Eqs. (13) and (14), the prices are grouped as cz ca =cd . Dividing the local objective by cd
gives
ji =cd ¼ ðcz ca =cd ÞAi þ Di ; ð15Þ
where
tAD ¼ ðcz ca =cd Þ: ð16Þ
All items of Eq. (15) are measured in exergy units, and tAD of Eq. (16) is a complex techno-
economic function that expresses the relative value of the resources of creating a device to the
1178 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

Fig. 4. Indirect interdisciplinary exchange of information via fuel costing equations.

resources of running the device. Eq. (16) may be extended to a socio-techno-economic function by
including a social price cs , where tAD ¼ ðcz ca Þ=ðcd cs Þ.
It may be noted that in both cases of costing, it is sufficient to communicate either Ai min or Di min
to minimize the local objective function ji for any set of prices since the set of prices emerge as one
value parameter.

5. The manipulation of the design models of devices

Design practices and more importantly design innovations by designers can be encoded as
design models. The models can be manipulated in various ways to reprocess design information
by correlations useful to system optimality when thermodynamics is the active discipline. The
manipulation is simply running a design model several times to generate tables of related ther-
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1179

modynamic and design parameters computed to obtain a required correlation that is meant to
condense information. Three useful correlations can be generated:

• the minimized characterizing dimension of a device that satisfies prescribed exergy destructions
in terms of its loading and efficiency variables ðVT in ; gÞ, i.e. a capital costing equation;
• the minimized exergy destruction in terms of the device characterizing dimension A and the es-
sential input thermodynamic variables ðVT in Þ, i.e. a fuel costing equation;
• the off-design efficiency variables of a device as function of its loading variables arising from the
part-load performance of the system, i.e. performance equation.

The first two are two alternative ways to deal with optimal design. The third deals with optimal
operation.
For illustration, let the device be a forced convection heat exchanger. It is assumed to be
the superheater, component 7 of the heat recovery steam generator of the simple combined cycle
analyzed in [8]. A duct shell-and-finned tube type is assumed. The fins are assumed circular on the
outside, which is the gas side. The design model of heat exchangers described in [8] is used. The
model is basically for forced convection heat transfer and pressure drops for single-phase and
two-phase (liquid–vapor) fluids. It contains more than 50 main equations for film coefficients and
friction factors. A heat exchanger can be composed of four generic geometries: double-tube, fin–
plate, shell-and-tube, plain or outside-finned tubes. The shell may be cylindrical or duct-type. The
flow may be pure counter or cross-counter.

5.1. The required correlation is a device capital costing equation

The boundary parameters P , T , fxg, M at inlets and exits of the exchanger as embedded in the
system at a design point for the system are used. The exchanger physical surface and its geometry
are defined by length, diameter, pitches, number, material, thickness and fin geometry of the
tubes. These parameters are usually more than needed to adjust in order to match the computed
surface and pressure drops by film coefficients and friction factors for the given heat load and its
temperature profile. Any extra design degrees of freedom are used to minimize the surface and/or
to satisfy reliable design practices. The design process is thus a matching/minimizing process.
The minimized surface is generated by repeating this design process for different bound-
ary parameters within a range relevant to the optimization of the system. A specific geometry of
minimized surface is obtained for each set of boundary parameters. The surface is then expressed
by an appropriate set of performance parameters such as heat loads, mass rates, heat exchange
temperature differences, effectiveness and pressure losses. In this paper, the surface as fins and
tubes is expressed in terms of heat load, the logarithmic mean temperature difference and pressure
losses on the shell side and on the tube side. The following form is used:

Amin ¼ kQn1 DTmn2 DPtn3 DPsn4 ; ð17Þ

along with the cost conversion equation

Z ¼ ca Amin : ð18Þ
1180 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

Table 1
The superheater minimized surfaces: surface vs. thermodynamic and geometrical parameters
Run Thermodynamic parameters Geometrical parameters
Atube Q DTlm g Dpt Dps Ltube do Wsh Pitch Pitch Afin =Ai Ntube Npas
(m2 ) (MW) (C) (kPa) (kPa) (m) (cm) (m) 1 (cm) 2 (cm)
1 486 15.76 66 0.921 42 0.462 20.4 2.5 11.9 5 4.52 11.8 364 1
2 915 66.80 128 0.609 41 0.475 5.8 2.5 52.1 5 4.52 11.8 2397 2
3 620 17.32 49 0.883 42 0.544 29.6 2.5 8.8 5 4.52 11.8 321 1
4 897 31.50 66 0.921 48 0.627 20.4 2.5 20.4 5 4.52 11.8 673 1
5 856 34.66 66 0.921 37 1.192 16.8 2.5 20.1 5 4.52 11.8 776 1
6 976 34.28 39 0.921 82 0.903 12.2 5 15.5 10 9.04 19.7 1258 8
7 188 7.88 66 0.921 90 0.834 85.3 7.6 0.91 15 13.6 27.8 10 1
8 276 8.67 66 0.921 90 0.227 45.7 3.8 3.7 7.6 6.78 15.7 57 1
9 355 9.52 66 0.919 21 0.234 29.6 3.8 6.4 7.6 6.78 15.7 114 1
10 112 9.52 126 0.400 83 0.965 34.1 7.6 2.1 15 13.6 27.8 15 1

Scatter of the correlating costing equation

Run 1 2 3 4 5 6 7 8 9 10
Aeqn = 0.965 1.10 1.08 0.98 1.06 0.92 1.02 0.92 0.976 1.08
Atable

The unit cost ca is a function of material, manufacture and severity of operation. It is time and
location dependent.
In this example, ca is assumed per unit total surface of fins and tubes. Ten minimized surfaces
were generated by changing inlet P , T , M, the allowed pressure losses and effectiveness. Heat load,
exit conditions, and logarithmic mean temperature difference are recorded. The parameters kept
fixed are the fin geometry, tube thickness, tube arrangement (staggered), fouling factors, flow
directions (gas horizontal, steam with gravity). In this particular example, the effect of gravity on
pressure losses is negligible. Table 1 shows the recorded parameters of the 10 minimized surfaces
and the quality of the correlation.
The constant k and the exponents n1 , n2 , n3 and n4 of Eq. (17) are computed by using the
surfaces of five cases simultaneously. These five cases are selected randomly from the total number
of cases. The computed constant and exponents that best fit the surfaces of all the cases are se-
lected. The simultaneous solution involves the inverse of a 4 4 matrix. When the matrix de-
terminant is relatively too small, unreasonable exponents are obtained and have to be rejected.
Also some selections may give rise to singular solutions and fail to give any values altogether.
There are, however, many sets that give solutions. There is also a room to round off the best-fit
exponents along with a modified value of the constant k such that the quality of the fit is not
changed. The best fit is identified by comparing the fits by the various sets. No further im-
provement of the best fit is made by applying a multiple regression approach.
The obtained constant and exponents were k ¼ 30:71, n1 ¼ 1, n2 ¼ 1, n3 ¼ 0:15 and
n4 ¼ 0:14, applicable for Q ¼ 8–66 MW, DTm ¼ 38–130 C, DPt ¼ 20–90 kPa, and DPs ¼ 0.2–1.2
kPa, with average scatter 8%, maximum þ10%. Inside tube surfaces covered the range 110–
975 m2 .
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1181

5.2. The required correlation is a device fuel costing equation

The minimized exergy destruction equation as a function of the characteristic dimension Ai for
a suitable range of input thermodynamic variables has to be communicated back to the ther-
modynamic model. Exergy destructions are also calculated by the thermodynamic model. They
may serve as a validity check when compared to their corresponding design values. The local
objective function, Eq. (12), gives
Dmin ¼ kd And : ð19Þ
The constant kd and the exponent nd are functions of the essential input thermodynamic variables
P , T , and mass rates at the inlets of the superheater. The maximum temperature range Tmax and
the mass rates of the heating and heated fluids are selected to describe the input thermodynamic
variables. The correlation for kd and nd is assumed as
kd ¼ kmnh1 mnc 2 DTmax
n3
; ð20aÞ

nd ¼ k 0 mnh4 mnc 5 DTmax

n6
; ð20bÞ
where k, n1 , n2 , n3 , k 0 , n4 , n5 , n6 depend on the range of variation of the input thermodynamic
variables.
The exergy destruction is the sum of one thermal destruction arising from temperature dif-
ference and two mechanical destructions arising from pressure losses tube side and shell side.
Formulating the mechanical destructions by similar equations as (19), (20a) and (20b) allows the
determination of the three components of exergy destruction and hence the three efficiency
variables of the superheater.
Table 2 shows six variations of thermodynamic input variables, the resulting correlations and
their qualities. Each set of the six inputs assumes five sizes and four types that may differ in tube
length, duct width, tube diameter and number of tube passes (120 design choices). For each set of

Table 2
Coefficient kd and index nd of the fuel costing equation
Input state kd nd mh (Mlb/h) mc (Mlb/h) DTmax (F)
1 371 )0.412 4.0 0.722 314
2 559 )0.440 3.0 0.678 314
3 1719 )0.404 4.5 1.422 418
4 459 )0.369 3.5 0.741 383
5 323 )0.389 3.7 0.479 383
6 1310 )0.434 5.0 1.792 314
Correlation
nd ¼ constant ¼ 0:04; kd ¼ kmnh1 mnc 2 DTmax
n3
; k ¼ 0:76; n1 ¼ 1:3; n2 ¼ 1:6; n3 ¼ 1:5
Quality of correlation
1 2 3 4 5 6
nd =nd eqn 1.03 1.10 1.01 0.90 0.97 1.09
kd =kd eqn 0.954 0.835 0.80 1.29 0.850 0.80
1182 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

Table 3
Input Case 1 to compute its kd and nd
Input variables
m (106 lb/h) hot ¼ 4 cold ¼ 0.722
T (F) hot ¼ 800 cold ¼ 486 sat. vapor
P (psia) hot ¼ 14.8 cold ¼ 600

Exergy destruction vs. surface

Run Thermodynamic parameters Geometrical parameters
2
A (ft ) D (Mbtu/h) Q g Dph (psi) Dpc (psi) L (ft) W (ft) do (in.) Npas
1 3495 17.0 101 0.637 0.476 7.55 15 69 1 2
2 3495 13.4 92 0.570 0.356 2.00 21 49 2 2
3a 3495 12.9 78 0.463 0.358 0.33 11 98 2 2
4 3495 24.1 109 0.698 0.734 16.8 15 68 0.5 1
5 6990 9.1 115 0.739 0.145 5.54 22 98 1 2
6 6990 7.6 100 0.630 0.108 0.565 22 98 2 2
7a 6990 7.6 100 0.630 0.108 0.565 22 98 2 2
8 6990 11.7 122 0.794 0.225 12.7 22 98 0.5 1
9 13979 6.4 126 0.827 0.044 4.12 31 139 1 2
10 13979 5.9 108 0.685 0.033 0.162 22 196 2 2
11a 13979 5.9 120 0.781 0.033 1.04 43 98 2 2
12 13979 7.6 132 0.871 0.069 9.62 31 139 0.5 1
13 20969 5.7 132 0.870 0.022 3.48 37 170 1 2
14 20969 5.5 112 0.716 0.016 0.079 22 294 2 2
15a 20969 5.4 129 0.851 0.016 1.52 65 98 2 2
16 20969 6.5 136 0.907 0.035 8.18 37 170 0.5 1
17 34949 5.1 137 0.914 0.009 2.82 48 219 1 2
18 34949 5.2 117 0.753 0.007 0.032 22 490 2 2
19a 34949 5.0 137 0.917 0.007 2.48 108 98 2 2
a
Least exergy destruction of a surface to compute kd and nd .

inputs the coefficient kd and the exponent nd are computed considering D to be the sum of three
components of exergy destruction. Table 3 shows kd and nd for the first as an example. Table 4
shows the correlation of kd and nd of the mechanical components of the exergy destruction and its
quality for the same six inputs.

5.3. The required correlation is the off-design performance equation of a device

This can be generated using the same design model in a different mode of computation from
that of the costing equation or the exergy destruction equation. The geometrical parameters of a
design case are kept constant at their design point while the boundary parameters are varied.
Selected boundary parameters are varied from their design values to cover the expected changes
in the boundary conditions of the device. Generally, changes in P , T , fxg and M at the inlets of
the heating and heated streams are sufficient to determine all off-design performance parameters
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1183

Table 4
The shell and tube sides nd and kd of exergy destruction by pressure losses
Input state Shell-side Tube-side
6
kd =10 nd kd 106 nd
1 12 )1.74 25.25 0.908
2 36 )1.73 3.56 0.950
3 2553 )1.78 6.60 0.883
4 136 )1.72 2.22 0.930
5 17 )1.73 4.79 0.903
6 227 )1.78 42.85 0.883

Correlation
kd ¼ kmnh1 mnc 2 DTmax
n3

Shell-side: nd ¼ constant ¼ 1:75; k ¼ 5:3 1014 ; n1 ¼ 4:9; n2 ¼ 4:7; n3 ¼ 9:6

Tube-side: nd ¼ constant ¼ 0:93; k ¼ 5000; n1 ¼ 5:4; n2 ¼ 0:3; n3 ¼ 4:7
Quality
Run 1 2 3 4 5 6
Shell-side nd =nd eqn 0.994 0.989 1.017 0.983 0.989 1.017
Shell-side kd =kd eqn 0.986 1.00 0.996 1.29 1.020 0.997
Tube-side nd =nd eqn 0.976 1.022 0.95 1.00 0.971 0.950
Tube-side kd =kd eqn 0.716 1.090 1.10 1.50 1.100 1.090

of interest. In this study only the effects of mass rates are assumed significant. The following
correlations for computing off-design effectiveness and the pressure losses are used:
n n
g=gd ¼ ðMs =Msd Þ 1 ðMt =Mtd Þ 2 ; ð21Þ

DPs =DPsd ¼ ðMs =Msd Þn ; ð22Þ

DPt =DPtd ¼ ðMt =Mtd Þn : ð23Þ

With the surface dimensions and geometry of the first case of generating costing equations held
fixed, 10 off-design cases were generated. Output parameters of the exchanger were recorded for
each case. The exponents of the best fits are decided by comparing different fits. No further im-
provement of the fits by a regression approach were made. The exponents obtained were n1 ¼ 0:2,
n2 ¼ 0:15 for efficiency (Eq. (21)), n ¼ 1:75 for shell-side pressure loss (Eq. (22)) and n ¼ 1:8 for
tube-side pressure loss (Eq. (23)), all applicable for mass rate changes in t=h 540–770 shell side and
45–365 tube side, within average scatter 8%, maximum 25%. Computer inputs and outputs
along with a graph for the distribution of scatter are detailed in [11].

6. Concluding remarks

• Exergy is elevated from a tool that reveals opportunities of higher efficiency to a tool that
reveals opportunities of both lower cost and higher efficiency.
1184 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185

• Design practices and design innovations in the form of design models of the devices of a system
are rich resources for predicting the cost and the performance of a system while still in its design
phase.
• Condensing and recovering information is an effective approach to manage the complexity of
high dimensionality problems and to enhance system analysis for improvements.
• The concept of costing equations and the principle of common objective are useful decompo-
sition techniques for the optimization of complex energy systems given a cost objective func-
tion.

Appendix A. On exergy destruction prices

Consider the objective function of Eq. (1a). Consider for simplicity fuel as the only feeding
resource.
X
i
J ¼ cF F þ czi cai Ai þ CR : ðA:1Þ
Let by definition
X
i
J ðcdi Di þ czi cai Ai Þ þ CR : ðA:2Þ
Then
X
i
c F F ¼ c f Ef ¼ ðcdi Di Þ: ðA:3Þ
For a local decision Yi
cf oEf ðYi Þ=oYi ¼ cdi oDi ðYi Þ=oYi ; ðA:4Þ

cdi ¼ cf ½oEf ðYi Þ=oYi =½oDi ðYi Þ=oYi ¼ cf ½oEf =oDi ðA:5Þ
by small changes in Yi .
Thus an exergy destruction price is the fuel price modified by the sensitivity of fuel consumption
to a change in exergy destruction induced by change in an efficiency decision variable and is
always a positive quantity. Experimenting with Eq. (A.5) as applied to a simple combined cycle
always gave non-negative prices somewhere between the fuel price per unit exergy and the market
value price of the power product. The search for one effective exergy destruction price is thus
justified and at the same time the option of distributed exergy destruction prices is created.

References

[1] Tribus M, Evans R. The thermoeconomics of seawater conversion. UCLA No. 62-63, 1962.
[2] El-Sayed Y, Evans R. Thermoeconomics and the design of heat systems. J Eng Power 1970;(January):27–35.
[3] El-Sayed Y, Aplenc A. Application of thermoeconomic approach to the analysis and optimization of vapor
compression desalting system. J Eng Power 1970;(January):17–26.
[4] Gaggioli R, editor. Thermodynamics: second-law analysis. ACS symposium series, No. 122. 1980.
[5] Gaggioli R, editor. Efficiency and costing. ACS symposium series, No. 235. 1983.
 Y.M. El-Sayed / Energy Conversion and Management 43 (2002) 1165–1185 1185

[6] Valero A, Tsatsaronis G, von Spakovski M, Frangopoulos C, Lozano M, Serra L, Piza J. CGAM problem:
definition and conventional solution. Int J Energy 1994;19(3).
[7] Valero A et al. In: Structural theory of thermoeconomics. ASME AES, vol. 30. 1993. p. 189–98, 3.
[8] El-Sayed Y. A second-law based optimization, parts 1 and 2. J Eng Gas Turbine Power 1996;118:693–703.
[9] Lazzaretto A, Tsatsaronis G. On the quest for objective equations in exergy costing. In: Proceedings of ASME
Advanced Energy Systems Division. AES, vol. 37. 1997. p. 197–210.
[10] Cerqueira S, Nebra S. Cost attribution methodologies in cogeneration systems. In: ECOS’98; Nancy, France, vol. I.
1998. p. 255–62.
[11] El-Sayed Y. Predicting part-load performance of an energy system concept. In: TAIES’97, Thermodynamic
Analysis and Improvement of Energy Systems Proceedings, 1997; Beijing, China.
[12] El-Sayed Y. Revealing the cost–efficiency trends of the design concepts of energy-intensive systems. Energy
Conserv Manage 1999;40:1599–615.
[13] El-Sayed Y. The thermoeconomics of desalination; cost vs. efficiency. In: Third International Water Technology
Conference IWTC, 1998 March; Alexandria, Egypt.
[14] El-Sayed Y. The thermoeconomics of seawater desalination systems. In: IDA World Congress on Desalination and
Water Reuse; Madrid, Spain, vol. IV. 1997. p. 149–66.

Yehia M. El-Sayed has a B.Sc. in Mechanical Engineering in 1950 from Alexandria University, Egypt and Ph.D. in
1954 from Manchester University, England. A teaching career that spanned over more than 30 years (1955–1987) has
covered four educational institutes: Assiut University, Egypt, Tripoli University, Libya, Glasgow University, Scotland,
and Massachusetts Institute of Technology (MIT), Cambridge, MA, US. Since 1987, El-Sayed is President, Advanced
Energy Systems Analysis, Fremont, CA. Honours include the IDA Best Paper Award in 1997, Madrid, Spain, the
ASME Edward Obert International Award in 1996, Atlanta, GA, and being elected a Fellow of the American Society of
Mechanical Engineers in 1990. El-Sayed has more than 60 publications in energy analysis and desalination.