Clean Techn Environ Policy (2009) 11:329338

DOI 10.1007/s10098-008-0186-z

ORIGINAL PAPER

An algorithmic approach to the optimization of process

cogeneration
Nasser Al-Azri Musaed Al-Thubaiti
Mahmoud El-Halwagi

Received: 21 June 2008 / Accepted: 24 September 2008 / Published online: 31 October 2008
Springer-Verlag 2008

Abstract In most industrial processes, there is a significant

need for electric power and for heating. Process cogeneration
is aimed at the simultaneous provision of combined heat and
power. The net result is usually a reduction in the overall cost
and emissions of greenhouse gases. Therefore, there is a
significant need for the optimal design of process cogeneration systems. This objective of this paper is to introduce an
algorithmic approach to the optimal design of process
cogeneration systems. Focus is given to the interaction of the
power cycle with the process heat requirements. Because of
the need for explicit thermodynamic expressions, a new set
of thermodynamic correlations of steam properties is
developed for proper inclusion within a mathematical-programming approach. An optimization formulation is
developed to provide a generally applicable tool for integrating the process and the power cycle and for identifying
the optimum equipment size, operating parameters (such as
boiler pressure, superheat temperature and steam load). The
objective can be chosen as minimizing the cost, satisfying the
heat requirement of the process, or producing the maximum
possible of power. A case study is solved to illustrate the
applicability of the devised approach and associated thermodynamic correlations.

N. Al-Azri
Department of Mechanical and Industrial Engineering,
Sultan Qaboos University, Muscat, Oman
M. Al-Thubaiti
Research and Development Center, Saudi Aramco,
Dharan, Saudi Arabia
M. El-Halwagi (&)
Department of Chemical Engineering,
Texas A&M University, College Station, TX 77843, USA
e-mail: El-Halwagi@tamu.edu

Introduction
The escalating usage of energy along with the associated
environmental emissions in industrial processes are causing
industry to identify various options for cleaner production and
less energy usage while satisfying the technical and business
objectives of the industrial processes. Much attention is being
devoted to the reduction of energy usage and environmental
emissions. In this respect, process optimization and integration have proven indispensable (e.g., El-Halwagi 1997, 2006).
In this paper, focus is given to process cogeneration which
seeks to integrating the process demands for heating and
power. The chemical process is integrated with the power
cycle to save on the heating requirement while simultaneously
producing power. The net result is higher efficiency and lower
level of environmental emissions.
The optimization of process utility systems has been the
subject of several research efforts. Papandreou and Shang
(2008) used a multi-objective mixed integer linear programming approach to optimize the chemical plant by
setting up an economic and environmental criteria in their
objectives. They included binary variables to select the
different design and operating conditions. Al-Thubaiti et al.
(2008) developed a procedure for the simultaneous heat
integration and process debottlenecking. El-Halwagi et al.
(2008) developed a graphical procedure for identifying
targets of process cogeneration and managing steam systems. Rodrguez-Toral et al. (2001) used a sequential
quadratic programming to solve the same problem. They
used an equation-oriented optimization that involves the
steam thermodynamic properties. Wilkendorf et al. (1998)
introduced a synthesis approach of the utility system by
satisfying the process requirement at the minimal possible
cost. Maia and Qassim (1997) used simulated annealing
which is a metaheuristic method to optimize a multi-period

123

330

N. Al-Azri et al.

utility system. Hul and Natori (1996) introduced a mixedinteger approach for multi-period utility system. Dhole and
Linnhoff (1993) investigated the interaction between process heat integration and power. Grossmann (1985) and
Papoulias and Grossmann (1983) introduced an MINLP
approach for the superstructure utility system optimization.
This paper introduces a mathematical-programming
approach to the integration of process heating requirements
with power cogeneration. A new set of correlations are
developed to explicitly predict the steam properties in a
way which is easily programmable and solvable using
optimization software. An optimization formulation is
developed to identify the optimal design and operating
parameters of the cogeneration system. A case study is
solved to illustrate the introduced approach.

Fig. 1 Rankine power cycle

Problem statement
Consider a process with a given heat requirement to be
fulfilled by steam. Recognizing that steam can be produced
at a higher pressure than needed, then let down through a
turbine, power can be generated and the exhaust steam can
still be used to satisfy the process heating requirement.
Consider a process that requires heat to be delivered at
pressure P2 with the minimum temperature being slightly
above the saturation temperature at P2. Available to the
process, a boiler that can operate at a maximum pressure
P1max and also a steam turbine to be purchased assuming
that it operates at or below a maximum load Mmax and has a
maximum inlet temperature T1max.
It is desired to develop a mathematical-programming
approach that enables the optimal design of the process
cogeneration system for combined heat and power (CHP).

Approach
Before proceeding to the design approach, it is useful to
first consider power generation through a Rankine cycle
using a heat engine such as the one shown by Fig. 1.
The basic Rankine cycle is composed of the following
thermodynamic processes:

1-2
2-3
3-4
4-1

Constant pressure heat addition in a boiler

Isentropic expansion in a turbine
Constant pressure heat rejection in a condenser
Isentropic compression in a pump

These four processes are better described with respect to

Mollier (T-S) diagram (Fig. 2).
Fresh treated water is first fed to the boiler which is fueled
by combustions gases or nuclear reactor and operates at a
constant pressure. The boiler section with the superheated

123

Fig. 2 Mollier (T-S) diagram

section is termed as steam generator (Process 1-2). The high

pressure steam is then fed to the turbine (expander), the high
pressure energy stored as high enthalpy is used to rotate the
turbine blades and hence produce power (Process 2-3). This
process is ideally isentropic, but due to the mechanical
deficiencies, the process line deviates toward the left on the
T-S diagram causing less enthalpy difference than the idea
process. At the exit of the turbine the working fluid can
either superheated steam (back-pressure turbine) or saturated steam (condensing turbine) with quality of 90% or
more to avoid erosion of the turbine hardware.
The steam at the exit is then fed to a condenser (Process
3-4). In order to pump the steam back to the boiler, the
steam is condensed and heat is discharged to the ambient in
a condenser. At the exit of the condenser, the liquid is fed
to a water pump (4-1) to elevate the pressure to the boiler
level in order to start the process over.
By investigating the Rankine cycle, in addition to the
mechanical losses, a major loss in the overall cycle is the
heat discharged to the ambient from the condenser. Work is
input to the cycle in the pump and the boiler. Power is
produced from the turbine and the rest is losses in the forms

Optimization of process cogeneration

of heat in the condenser and mechanical losses in the turbine. In order to improve the overall efficiency of the cycle,
the heat discharge to the ambient can instead be utilized by
the chemical process, Fig. 3.
Looking at Fig. 3 from an economic viewpoint, the heat
discharged at the condenser is no longer a loss. This
approach is not only viable to utilize that heat, a reverse
approach can also be implemented, i.e., determining the
cycle parameters as dictated by the process requirement.
This approach is practical when producing steam at high
pressure and cogenerating power is more economical than
producing steam at low pressure.
In this work, it is assumed that the boiler operates at
relatively high pressure levels. The pressure of the boiler is
set in conjugation with the process need in such a way that
will optimize the tradeoff between the cost and the output
of the process.
Instead of sticking to a specific equipment model, the
approach is made open to be utilized by any real case
problem. The cost and performance estimations can be
incorporated either as a continuous function in terms of
other parameters, e.g., cost of turbine as a function of its
load, or as discrete integers, e.g., a set of possibilities of
different turbines at different costs.

Optimization formulation
The optimization formulation will encompass three
important elements:
1.
2.

Cost estimation to be used in the objective function

Performance model for the power cycle and process
heating system

331

3.

Thermodynamic properties of steam to be developed

and used properly in an optimization formulation

Figure 4 is a schematic representation of these elements.

The overall optimization formulation of the problem is
given by:
Objective function:
Min cost Turbine cost Fuel cost Operation cost

Power price
Subject to:
Turbine cost = f(power)
Fuel cost = f(Qboiler, gboiler)
Power price = f(shaft work, ggenerator)
Shaft work = f(h1, h2a, m)
The inlet pressure and temperature should be in the
range of those values of the process and the maximum
values recommended by the turbine and boiler
manufacturers.
P2 \P1  P1 max
Tsat;1 \T1  T1 max
The saturation temperatures are expressed as functions
of the pressure values.
Tsat2 f P1
Tsat;1 \T1  T1 max
The inlet entropy needs two parameters for its
expression. For convenience, it is expressed in terms of
the inlet temperature and saturation temperature at that
pressure level.
s1 f Tsat1 ; T1
The inlet enthalpy is expressed in terms of the in let
saturation temperature and entropy.

Fig. 3 Integrating the chemical process with the heat engine

Fig. 4 The optimum utility design consideration

123

332

N. Al-Azri et al.

h1 f Tsat1 ; s1
The efficiency is expressed in terms of the inlet pressure
and the load.
gturbine f P1 ; load
The outlet isentropic enthalpy is expressed in terms of
inlet entropy and outlet saturation temperature using the
same expression for the inlet enthalpy.
h2;is f s1 ; Tsat2
The outlet actual enthalpy is expressed in terms of the
isentropic efficiency of the turbine, inlet enthalpy and
outlet isentropic enthalpy.
h2;a f gturbine ; h1 ; h2;is
For a back-pressure turbine, it should be stipulated that
the outlet enthalpy is greater than the saturated vapor
enthalpy.
h2a [ hg2
The outlet entropy is expressed in terms of the actual
enthalpy and saturation temperature.
s2 f h2a ; Tsat2
The outlet temperature is expressed in terms of the
actual enthalpy and the outlet entropy.
T2 f s2 ; h2a
The mass flow rate is chosen so as to deliver the required
heat and be less than the maximum allowed by the turbine.
m

Qprocess
 mmax
h2a
hf 2

The heat delivered to the boiler is expressed in terms of

the mass flowrate, the inlet enthalpy at the turbine inlet and
the saturated liquid enthalpy at the turbine outlet pressure
level.
Qboiler f h1 ; hf 2 ; m
The efficiency of the boiler can be expressed in terms of
the pressure level, load and fuel type.
gtboiler f fuel; P1 ; load
The above formulation is the most general one. The
following sections provide details on the various
constraints.
Turbine efficiency
The steam turbine has a mechanical efficiency as well as
isentropic efficiency. The overall efficiency is the product
of both. The mechanical efficiency is highly dependent on

123

the turbine structure and manufacturing and it can be

higher than 95% for modern steam turbines. The isentropic
efficiency, however, is more dependent on the operation
parameters. It exhibits a high sensitivity to the inlet pressure and steam load for example.
The turbine efficiency can be expressed in terms of the
inlet pressure as well as the load (Evans Jr 1971; Fronseca
Jr 2003). The efficiency is given as a function of the mass
or power load and produced for a family of inlet pressure
values.
These efficiency correlates well with the following
empirical equation:
g

aP
bP Pmax

where P is the power, Pmax is the maximum power at that

inlet pressure and, a and b are functions of the inlet
pressure.
Mavromatis and Kokossis (1998) established another
correlation by modeling the steam turbine using Willans
line and came up with the following correlation:



6
3:41443  106 A
M max
gis
1

1


2
5B
Dhis M max
6M
whereas M and Mmax (in lb/h) are the mass flow and the
maximum mass flow respectively, Dhis is the isentropic
enthalpy difference (in Btu/lb) and, A and B are functions
of the saturation temperature of the inlet pressure. It should
be noted that the coefficient of this equation is changed to
the American customary system here while it is used in
different units in the original reference.
The A and B factors in Eq. 2 correlation change with the
power output as explained in the appendix. This expression
very closely matches the one given by Eq. 1. In this works,
Mavromatis and Kokossis correlation is used without
losing generality.
Most models suggest a generic form regardless of the
manufacturer but there is a model that involves a parameter
that requires to be estimated using experimental runs of a
specific turbine (Varbanov et al. 2004). This model is given
by the following equation:
gis

n:m
WINT
m:gmech :DHis

whereas m, gmech and DHis are the mass flow rate,

mechanical efficiency and isentropic enthalpy difference
respectively. WINT and n are two parameters, given in the
appendix, that are calculated in terms of the mass flow rate,
isentropic enthalpy difference, saturation temperature difference, type of turbine (backpressure versus condensing)
and a parameter L that needs to be defined by correlating
the reading of the specific turbine.

Optimization of process cogeneration

333

Boiler efficiency

The boiler efficiency is basically the heat delivered to the

steam to the heat contained in the fuel. The typical boiler
efficiency is somewhere between 75 and 80% and can
reach to over 90%. The boiler efficiency depends on the
material from which it is constructed and the fuel type in
addition to the operating and ambient conditions. Due to
the fuel cost, that is, with the maintenance cost, is much
higher than the capital investment, a slight improvement in
the boiler efficiency may result in substantial savings. The
evaluation of the boiler efficiency is more complicated than
that of the steam turbine and the cost is more sensitive to
the boiler efficiency. A common way to predict the boiler
efficiency is the Power Test Code for Steam Generating
Units developed by the American Society of Mechanical
Engineers (ASME 1998).
The cost function
The generation cost is the sum of all costs of operations and
materials during production. For the optimization problem,
the generation cost of steam will be calculated using the
following equation (Kumana and Associates 2003):
Cg Cf Cw CBFW Cp Ca Cb Cd Ce Cm

Cf : fuel cost ($/klb)

Cw: raw water supply cost($/klb)
CBFW: boiler feed water treatment cost including
clarification, softening and demineralization ($/klb)
Cp: feed water pumping power cost ($/klb)
Ca: combustion air fan power cost ($/klb)
Cb: sewer charges for boiler blowdown ($/klb)
Cd: ash disposal cost ($/klb)
Ce: environmental emission control cost ($/klb)
Cm: maintenance materials and labor ($/klb)

In this work however, to maintain a traceable formulation, a consistent set of units will be used. Instead of $/klb
for the operation cost, $/MMBtu of heat delivered by the
boiler will be used.
The fuel cost contributes substantially to the generation
cost. It contributes with over 90% of the generation cost.
The fuel cost is calculated as:
Cf $=h af Qb 10
6 =gb

h1: enthalpy of steam at turbine inlet (Btu/lb)

hf: enthalpy of boiler feedwater assumed at the
saturated liquid point (Btu/lb)
m: steam flow rate (lb/h)
gb: overall boiler enthalpy

The maintenance cost of the boiler is about 30% of the

fuel cost (Kumana and Associates 2003). In order to
account for the increase in pressure, a flexible factor is
introduced.
Cg 1:3Fp Cf

Fp: a factor accounting for the increase in pressure. This

factor can be unity if the increase in pressure is insensitive to the operation cost but sometimes there are
additional costs incorporated with higher pressure values
such as the minimum allowed concentrations of the water
composition in the boiler which gets less and more costly
to remove as the pressure values get higher (Branan
2002).
The electric power price ($/kWh) is:
Pe $=h ae Pt gg =3;413

ae: electrical power price ($/KWh)

Pt: the turbine shaft power output (Btu/h); Pt = m
(h1 - h2a)
h1 : enthalpy of steam at turbine inlet (Btu/lb)
h2a: actual enthalpy at steam outlet (Btu/lb)
gg: generator efficiency

The capital costs of the equipment are the most unstable

element of the formulation. This cost can be introduced as
a function of other operation variables or as integers of a
set of possible selections with their operation capacities.
For the illustration purposes in this work, cost equations are
improvised by comparing different cost estimations (Seider
et al. 2004), moreover, the most significant cost is assumed
contributed by the turbine and the boiler. The variation in
the pump cost between the different operation variables is
insignificant.
The boiler is assumed a water-tube boiler fueled with oil
or gas with the following cost estimation:
Cboiler $ 3 Np NT Q0:77
b

af : fuel cost ($/MMBtu)

Qb: the amount of heat transferred to the steam at the
boiler (Btu/h)
Qb = m (h1 - h1f)

Qb: the amount of heat transferred to the steam (Btu/h)

Np : a factor accounting for the operation pressure and
is given by:
Np = 7 9 10-4 P1g ? 0.6; P1g is the gauge pressure
(psig) of the boiler.
NT: a factor accounting for the superheat temperature
and is given by:

123

334

N. Al-Azri et al.

NT = 1.5 9 10-6 Tsh2 ? 1.13 9 10-3 Tsh ? 1; Tsh is

the superheat temperature (F).
Tsh = T1 - Tsat1

The turbine is assumed a non-condensing turbine with

its cost equation derived the same way:
CTurbine $ 475 P0:45
t

Pt: the turbine shaft power output (Btu/h); Pt = m

(h1 - h2a)

To find the optimum design of a grass root utility, the

objective function will be expressed in terms of the annual
cost. In this case, the annualized capital cost of the
equipment will be used.
The annual cost 1:3 Fp Cf t ACboiler ACturbine
Pe t
10

t: the annual operation time (hours/year)

ACboiler = the annualized capital cost of the boiler
($/year)
ACboiler = (Cboiler - salvage price)/years of service
ACturbine = the annualized capital cost of the turbine
($/year)
ACturbine = (Cturbine - salvage price)/years of service

To determine the optimum operation parameters for some

existing equipments and to confine the tradeoff between the
electricity produced and the fuel consumed, the objective
function will reduce to:
The annual cost 1:3 Fp Cf t
Pe t

11

Correlations of the thermodynamic properties

To find the optimal values of the Rankine cycle parameters,
the whole problem needs to be formulated and stated as an
optimization problem with an objective function and constrains. Consequently, the steam tables usually used for
thermodynamic cycle calculations will be replaced by
mathematical equations. Therefore, it is important to have
thermodynamic property correlations that are appropriate
for use in optimization formulations. Regrettably, published correlations for steam tables are highly nonlinear
nonconvex function and often involve multiple roots. This
makes incorporating them into an optimization formulation
a tedious task that challenges the global solution (or even
convergence) of the mathematical formulation using optimization software.
To overcome the above limitations, we have developed a
new set of thermodynamic correlations for estimating
properties of steam. These correlations are attractive for

123

usage in optimization formulation. Each intensive thermodynamic property (T, P, h, s, v) can be determined with the
knowledge of any other two properties. So, each correlation
of a thermodynamic property has to be in terms of two other
properties. In order to use simple yet accurate correlations
between the pressure and other properties, the corresponding
saturation temperature will be used. The saturation temperature correlates with other properties in a simpler
function the pressure does. For that reason, the pressure will
be expressed in terms of its corresponding saturation temperature and the latter will be used to express other
properties. The saturated temperature as a function of the
pressure is calculated by using the following correlation:
Tsat 112:72 P0:2289
1

12

where P1 (psi), Tsat1 (F). The relative error is 1.87% and

average relative error is 0.64%.
For a given pressure and temperature inlet, the entropy
at the inlet of the turbine can be calculated by using the
following equation:
s
0:5549 lnTsat 3:7876T 0:1001 exp0:0017Tsat

13

where s is the entropy of the superheat in Btu/(lb.R), T is

the inlet temperature in F and Tsat is the saturation temperature, corresponding with the inlet pressure, in F. This
correlation is developed for a range of pressure values up to
2,500 psi with superheat temperature up to 1,500F. The
relative error of this correlation is 3.5%.
The enthalpy is then obtained as a function of the saturation temperature of the header and the entropy.
h 0:2029Tsat s3:647 817:35

14

where h in Btu/lb, s in Btu/(lb.R) and Tsat in F. The relative error is 0.6% and is valid in the range between
atmospheric pressure and 2,000 psi.
The turbine efficiency will be expressed using one of the
above correlations or a better expression according to the
data provided by the manufacturer. The isentropic efficiency at the outlet of the turbine will be expressed using
the same equation of enthalpy but in terms of the exit
saturation temperature and the inlet entropy. In terms of the
turbine efficiency and the inlet enthalpy and the outlet
isentropic enthalpy, the actual enthalpy will be:
h2;a h1
gturbine h1
h2;is

15

The outlet temperature will be expressed in terms of the

entropy and the enthalpy. Given the saturation temperature
and actual enthalpy, Eq. 11 is used to find the entropy.


h
817:35 0:2742
s
16
0:2029Tsat
In terms of the entropy and the enthalpy, the outlet
temperature is found using the following correlation:

Optimization of process cogeneration

335

T h
Bs =A
AS
0:7918S3 3:4575S2
4:5513S 2:1267
3

17

BS 710:22S
3910:6S 7117:3S
3253:5

This correlation is valid for a superheat of up to 500F
with a relative error less than 5%.
The boiler and condenser loads are assumed to be the
latent heat added to the superheat. The saturated fluid
enthalpy is calculated using the following correlation:
1:2127
hf 0:2674Tsat

18

where hf in Btu/lb and Tsat in F. This correlation is

developed for a range of pressure up to 2,500 psi. The
maximum relative error is less than 3%.
The mass flow rate will be expressed in terms of the heat
required by the process, the actual enthalpy at the exit of
the turbine and saturated fluid enthalpy of the exit pressure.
m

Qprocess
h2a
hf 2

19

The heat output of the boiler will be expressed in terms

of the mass flow rate, the enthalpy at the inlet of the turbine
and the saturated fluid enthalpy of the exit pressure.
Qboiler mh1
hf 1

20

The model solution

The optimization problem incorporates a mathematical
model of the turbine cycle by bringing together the thermodynamic correlations, equipment performance and the
relationships between the performance and the thermodynamic properties of steam on one side and the relationships
between the different steps within the cycle on the other
side. The main assumption made in this model is that the
steam is discharged from the process after heating as a
saturated liquid and so does it enter the boiler. In other
words, the work of the pump is assumed negligible.
The purpose of the optimization problem is to find the
economically and practically best operation conditions.
The optimization problem is fed with the operation range
of the equipment (the gray area in Fig. 5). Using a model of
the equipments performance or data provided by the
manufacturer, the algorithm finds the optimal path of the
cycle using the thermodynamic correlations and the relationships between the different blocks within the cycle.
The economic judgment is based on a tradeoff between
the fuel consumed at the boiler and the power produced at
the turbine for a given amount of heat to be delivered to
the process. Operating the boiler at the highest pressure
and load will result in the maximum power production.
That benefit, however, might not offset the high cost of

Fig. 5 The optimal solution of a given problem

equipment and fuel and so an optimal pressure and load

needs to be investigated. The same argument is valid if it is
needed to operate the turbine at the highest load in order to
achieve the maximum efficiency. In that case, an additional
cost of fuel may render the process suboptimal.

Case study
The above formulation gives a universally solvable optimization problem for the CHP in a steam turbine. An
optimization software such as LingoTM 10.0 that is used in
this case study can obtain the global solution of the problem. An incremental loop of pressure with a nested loop of
superheat temperature can also be used in any programming language. The increment can be set reasonably such
that the runtime is reasonable.
Case study I
A chemical process requires 20 MMBtu/h heat at 100 psi.
The turbine used in the facility can operate at a maximum
inlet temperature of 450F and maximum flow rate of
25,000 lb/h (Fig. 6). The boiler efficiency is 40% and it
operates at a maximum pressure of 280 psi. The minimum
superheat temperature at the exit of the turbine is 5F. The
power price is 0.05 $/kWh and the fuel price is 2.80
$/MMlb, the generator efficiency is 40%, the pressure
factor in the generation cost is unity.
In order to determine the optimum working conditions,
the isentropic efficiency of the turbine is estimated using
(Mavromatis and Kokossis 1998) model. Solving the
problem can be either performed using a nonlinear programming technique to solve the optimization problem
using a global solver like Lingo 10.0 or it can also be

123

336

N. Al-Azri et al.

choices by establishing an MINLP that includes selecting

the proper equipment along with the optimal operation
parameters as shown in the next case study.
Case study II

Fig. 6 The initial range of operation in case study I

solved by using an iterative algorithm that tests the optimization variables at reasonable increments.
The optimum solution of this problem is located at an
inlet pressure of 242 psi and inlet superheat temperature of
52F (Fig. 7) with the superheat at the turbine exit 5.8F.
The flow rate is 22,669 lb/h. The turbine isentropic efficiency is 50% and the maximum efficiency is estimated
around 61%.
Although the pressure factor is assumed one, there is a
clear tradeoff between power production and fuel
consumption.
This problem was solved using Lingo 10.0 on Windows
XP system with Intel processor 1.79 GHz and 1.00 GB of
RAM. The runtime was less than a minute. An exhaustive
search using Matlab on the same machine retrieved the
same results in 3 s.
The developed optimization problem can be used for a
wide range of applications. The applications are not only
limited to operation parameters but also to equipment

A chemical process that requires an amount of heat that is

20 MMBtu/h at 100 psi pressure; available selections of
turbines and boilers are given in Tables 1, 2.
All turbines tolerate the maximum pressure of the
boilers. The minimum superheat at the exit of the temperature should at least be 5F. The generator efficiency is
50%. The pressure factor is taken as unity.
This price equation is to account for the expense
resulting from producing the same amount of heat at higher
pressure values. The power price is 0.05 $/kWh.
The optimum of the nine possible combinations was
found to be Boiler 1 and Turbine 3. The inlet pressure is
chosen to be 250 psi and a superheat of 52F at inlet and
5F at exit and flow rate of 22,679 lb/h.
This problem was formulated as a MINLP on Lingo 10.0
on Windows XP system with Intel processor 1.79 GHz and
1.00 GB of RAM. The runtime was 78 s.

Conclusions
A mathematical-programming approach has been introduced for the optimal design of a process cogeneration
system. It addresses the inter-dependent nature of the
Rankine cycle variables with the process heating requirements. The formulation includes various constraints
pertaining to the power cycle and its interaction with the
process. In order to properly model the steam system in a
way which is appropriate for optimization formulations, a
new set of steam-property correlations has been developed.
Table 1 Boilers specifications in case study II
Fuel cost
($/MMBtu)

Efficiency

Maximum
operating
pressure (psi)

Boiler 1

5.50

0.8

250

Boiler 2

3.40

0.7

230

Boiler 3

2.80

0.4

280

Table 2 Turbines specifications in case study II

Fig. 7 The optimal solution of case study I

123

Maximum
inlet temp. (F)

Maximum steam
load (lb/h)

Turbine 1

500

25,000

Turbine 2

450

25,000

Turbine 3

600

30,000

Optimization of process cogeneration

337

Table 3 The regression coefficients used in the isentropic efficiency equation

Back pressure turbines
max

W
a0 (Btu/s)

Condensing turbines

\ 4.10 MMBtu/s

-0.1518

max

[ 4.10 MMBtu/s

Wmax \ 5.12 MMBtu/s

Wmax [ 5.12 MMBtu/s

-0.062488889

-1.038755556

-0.115877778

a1 (Btu/s. F-1)

0.00065

0.003461111

0.000555556

0.000777778

a2

0.961977778

1.111644444

1.195233333

1.166466667

a3 (F-1)

0.000844444

0.000261111

0.000333333

0.000166667

The coefficients in the table are taken from Varbanov et al. (2004) and Mavromatis and Kokossis (1998). They exist in the SI units in those
references but converted here into the American customary system for consistency with the other examples and quantities in this work

Table 4 The regression

coefficients used in the
isentropic efficiency equation

Back pressure turbines

Condensing turbines

Wmax \ 2,000 kW

Wmax [ 2,000 kW

Wmax \ 2,000 kW

Wmax [ 2,000 kW

a0 (kW)

-463

a1 (kW/C)

1.08

4.23

0.662

3.53

a2

1.097

1.155

1.191

1.220

a3 (C-1)

0.00172

0.000538

0.000759

0.000148

Each intensive thermodynamic property (T, P, h, s, v) can

be determined with the knowledge of any other two
properties. The accuracy of the developed correlations is
acceptable for conceptual design purposes and facilitates
the solution of the optimization formulation. A case study
with two scenarios has been solved to demonstrate the
applicability of the developed procedure.

gis

For a given turbine, the isentropic efficiency is a function

of the flow load and hardware parameters provided by the
manufacturer. This expression is based on Willans line
that correlates the load to the power output. The isentropic
efficiency is expressed by Eq. 21 (Mavromatis and
Kokossis 1998).



6
3:41443  106 A
M max
gis
1

1

21
5B
Dhis M max
6M
A and B are constants dependent on the turbine and are
functions of the inlet saturation temperature and the flow
rate is measured in lb/h. A good approximation, within 2%,
is given by the following straight line segments
(Mavromatis and Kokossis 1998; Varbanov et al. 2004):
A a0 a1 T sat

22

sat

23

B a2 a3 T

Another expression of the isentropic

developed by (Varbanov et al. 2004) is:

efficiency

24

Whereas n and Wint are the slope and the intercept of the
linear Willans line respectively and given by this equation:
L
a
Wint DHis

b
mmax
n

Appendix: Expressions for isentropic efficiency

of the turbine

n:m
WINT
m:gmech :DHis

L1
DHis mmax
a
b

25
26

With L is to be estimated by correlating the performance

of each specific turbine. The parameters a and b depends
on the type pf pressure, maximum power load and the
saturation temperature differences as in the following table
(Table 4; Smith 2005).
a a0 a1 DT sat
b a2 a3 DT

sat

27
28

References
Al-Thubaiti M, Al-Azri N, El-Halwagi M (2008) Optimize heat
transfer networks: an innovative method applies heat integration
to cost effectively retrofit bottlenecks in utility systems. Hydrocarbon Process 87:109115
Branan C (2002) Rules of thumb for chemical engineers. Gulf
Professional Publishing, Houston
Dhole V, Linnhoff B (1993) Total site targets for fuel, co-generation,
emissions and cooling. Comput Chem Eng 17:s101s109
El-Halwagi MM (1997) Pollution prevention through process integration. Academic, San Diego
El-Halwagi MM (2006) Process integration. Academic/Elsevier,
Amsterdam

123

338
El-Halwagi MM, Harell D, Spriggs HD (2008) Targeting cogeneration and waste utilization through process integration. Appl
Energy. doi:10.1016/j.apenergy.2008.08.011
Evans FL Jr (1971) Equipment design handbook for refineries and
chemical plants. Gulf Publishing Co, Houston
Fronseca JG Jr (2003) Analise Energetica e Exergetica de um Ciclo
Rankine com Aquecimento Distrital: Estudo de uma Planta
Termoeletrica. Universidade Federal do Rio Grande do Sul,
Porto Alegre
Grossmann IE (1985) Mixed-integer programming approach for the
synthesis of integrated process flowsheets. Comput Chem Eng
9:463482
Hul C-W, Natori Y (1996) An industrial application using mixedinteger programming technique: a multi-period utility system
model. Comput Chem Eng 20:S1577S1582
Kumana and Associates (2003) How to calculate the true cost of
steam, Rep. No. DOE/GO-102003-1736. US Department of
Energy, Washington, DC
Maia LOA, Qassim RY (1997) Synthesis of utility systems with
variable demands using simulated annealing. Comput Chem Eng
21:947950

123

N. Al-Azri et al.
Mavromatis SP, Kokossis AC (1998) Conceptual optmisation of
utility networks for operational variations I. Targets and level
optimisation. Chem Eng Sci 53:15851608
Papandreou V, Shang Z (2008) A multi-criteria optimisation approach
for the design of sustainable utility systems. Comput Chem Eng
32:15891602
Papoulias SA, Grossmann IE (1983) A structural optimization
approach in process synthesis-I utility system. Comput Chem
Eng 7:695706
Rodrguez-Toral MA, Morton W, Mitchell DR (2001) The use of new
SQP methods for the optimization of utility systems. Comput
Chem Eng 25:287300
Seider WD, Seader JD, Lewin DR (2004) Product & process design
principles. Wiley, New York
Smith R (2005) Chemical process design and integration, 1st edn edn.
Wiley, Chichester
Varbanov PS, Doyle S, Smith R (2004) Modelling and optimization
of utility systems. Trans ICheE 82:561578
Wilkendorf F, Espuna A, Puigjaner L (1998) Minimization of the
annual cost for complete utility systems. Chem Eng Res Des
76:239245