Energy 193 (2020) 116679

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Co-optimization of thermal power plant flowchart, thermodynamic

cycle parameters, and design parameters of components
Aleksandr M. Kler, Yulia M. Potanina*, Andrey Y. Marinchenko
Melentiev Energy Systems Institute of Siberian Branch of the Russian Academy of Science (ESI SB RAS), 130 Lermontov Str., Irkutsk, Russia

a r t i c l e i n f o a b s t r a c t

Article history: The improvement in the efficiency of an energy plant depends on a rational development of its flowchart
Received 6 September 2019 and choice of parameters along with the load schedule, equipment reliability, operating mode, etc. It is
Received in revised form advisable to study such complex technical systems with the methods of mathematical modeling and
26 November 2019
optimization. The paper presents an approach to the development of optimal flowcharts and selection of
Accepted 1 December 2019
parameters of energy plants. The approach is based on the combination of a method for optimization of
Available online 10 December 2019
the most complex flowchart and a method for solving discrete-continuous problems of nonlinear
mathematical programming. A case study of the co-optimization of design parameters, operating pa-
Keywords:
Mathematical modeling
rameters and equipment mix for the integrated gasification combined-cycle plant is demonstrated. The
Thermal power plants optimization calculations were carried out by the criterion of the minimum price of electricity for a given
Methods of optimization internal rate of return on investment and the maximum energy efficiency of the plant. Several optimal
Mixed-integer programming solutions meeting the different criteria are obtained. The proposed approach can be used for optimi-
Reversible processes zation of flowcharts and parameters of other complicated energy plants (high-efficiency combined-cycle
Branch-and-bound method plants, ultra-supercritical steam cycles, integrated power plants for electricity and synthetic liquid fuel
co-production from coal, etc.).
© 2019 Elsevier Ltd. All rights reserved.

1. Introduction emphasize that the calculations of the thermal power plants can be
conducted even if the values of these parameters do not meet the
The efficiency of thermal power plants is determined by their discreteness (integrality) requirements, i.e. the systems of equality
flowchart, values of thermodynamic cycle parameters and design constraints can be solved, inequality constraints can be solved and
parameters of components. The flowchart is specified by a set of the criteria of energy and economic efficiency can be determined.
components and a structure of material and energy flows that Many researchers solve such problems by different methods of
connect these components. Optimization of the plant parameters direct enumeration. These methods are robust and have rather
under the specified flowchart is reduced to the problem of simple algorithms. They, however, optimize only a small number of
nonlinear mathematical programming. In this case, most of the parameters. The examples of their application to the optimization
parameters can take any real value lying within some set limits. of energy plant parameters can be demonstrated with the following
Such parameters are called continuously varying. They include the publications.
flow rates of working media and heat carriers, their pressure and Roberto Carapelucci and Lorena Giordano [1] compare two
temperature; and geometrical sizes of components. optimization methodologies based on the energy-economic and
The other parameters can take only discrete values that lie exergy-economic analyses. The considered case study is based on a
within specified limits. These are mainly design characteristics of combined cycle gas turbine plant with different configurations of
components, such as the number of parallel heat exchanger tubes, heat recovery boiler and various gas turbines. A genetic algorithm
the number of nozzle and working blades in the turbine stage, etc. was used in the calculations.
Such parameters are called discretely varying. It is worthwhile to Mahmoud Nadir et al. [2] present the results of thermo-
economic optimization of a heat recovery boiler for different tem-
peratures of turbine exhaust gases, and a study of the electricity
price effect on the optimal calculated parameters of the boiler. The
* Corresponding author. calculations involved the particle swarm optimization method.
E-mail address: [email protected] (Y.M. Potanina).

https://doi.org/10.1016/j.energy.2019.116679
0360-5442/© 2019 Elsevier Ltd. All rights reserved.
2 A.M. Kler et al. / Energy 193 (2020) 116679

Nomenclature b exc
Q vector of intensive parameters of external flows of i-
i
th component
b inc
Q vector of intensive parameters of internal flows of i-
Abbreviations i
NLP nonlinear programming th component
IGCC integrated gasification combined cycle Wfl system of equations that define the constant part of
GAS gasifier the flowchart
FC furnace chamber Vfl system of inequality constraints of the constant part
AC1, AC2 air compressors of the flowchart
GT gas turbine L the number of discrete optimized parameters
SEP1, SEP 2 drums separators I the number of components included in the most
CC combustion chamber complex flowchart
DAR dry ash removal wi subsystem of equations of the i-th includable
WS wet scrubber component
RH regenerative gas/gas heat exchanger vi subsystem of inequality constraints of the i-th
WGG1-5 heat exchangers on gasification products includable component
WGS1-8 heat exchangers on combustion products tce tonne of coal equivalent
P1eP7 pumps
C condenser Greek symbols
TD1-3 steam turbine departments ε coefficient
g specified accuracy of the approach of parameters zi to
Symbols discrete values
P pressure
T temperature Superscripts
H enthalpy ct parameters of component
G mass flow rate in parameters of input flows
zi parameter that determines the presence of the out parameters of output flows
component in the flowchart fl parameters of flows between the flowchart
F optimization criterion (objective function) components
xH vectors of continuously varying optimized exc external parameters of the component
parameters of a constant part of the energy plant inc internal parameters of the component
diagram compr compression
x H ; xH vectors with the components setting the lower and ex expansion
upper constraints for the components of vector xH
y vector of variables calculated from the governing Subscripts
system of equations max maximum
U vector of initial data that sets external conditions for min minimum
the constant part of the diagram i number of the component
~ exc j number of the input flow of component
Q vector of extensive parameters of external flows of i-
i k number of output flow of component
th component
l number of discrete optimized parameter
~ inc
Q vector of extensive parameters of internal flows of i-
i
th component

CGAM problem, an example to compare the thermo-economic in the analysis of the power plant flowchart and the decomposition
optimization methodologies, was proposed in 1990 by a group of of the general problem of optimization of the thermal power plants
researchers (Christos Frangopoulos, George Tsatsaronis, Antonio into a number of iteratively solved subproblems for optimizing its
Valero, Michael von Spakovsky) and named after their first initials subsystems.
[3]. A natural gas-fired cogeneration plant with five optimized The research in Ref. [7] by Ligang Wang et al. is focused on multi-
parameters was considered as a case study. In the following pub- objective optimization of coal-fired power plants based on differ-
lications, Frangopoulos [4], von Spakovsky [5], Valero et al. [6] ential evolution. The case study represents a coal-fired plant for
considered the optimization of this plant using nonlinear mathe- ultra-supercritical steam parameters. The efficiency criteria are the
matical programming methods, the classical method of undeter- maximum efficiency and minimum electricity price.
mined Lagrange multipliers, and decomposition methods. As evidenced by an analysis of the above-mentioned studies and
Optimization was carried out according to the criterion of economic many others dedicated to this topic (for example, E. Godoy et al. [8],
efficiency (reduced costs). The model included both the de- A. Lazzaretto and A. Toffolo [9], Emanuele Martelli et al. [10], X.J.Liu
pendences between the thermodynamic parameters of the plant et al. [11], Rajesh Arora et al. [12], Ming Liu et al. [13], Chaoyang
elements and the expressions for determining their costs. The Wang et al. [14], etc.), this problem is of great interest for the
proximity of the optimal solutions obtained by various methods is researcher. It is sensible to study the methods for the improvement
shown. In addition, the exergy prices of the flows transmitted be- of the plant efficiency with more detailed mathematical models,
tween the plant components were determined at the optimum and much larger number of optimized parameters.
parameters of the plant. It is indicated that these prices can be used Some studies employ gradient methods enabling a great
 A.M. Kler et al. / Energy 193 (2020) 116679 3

number of parameters (one hundred and more) to be optimized 2.1.2. Evolutionary optimization methods
(A.M.Kler et al. [15], Dekanova N.P. et al. [16], etc.) It is worth noting These methods suggest setting an initial flowchart and formu-
that these methods have a downside (however, like the others lating optimization problem of its parameters. At the point of so-
considered). If the optimization problem has several local extrema, lution to this problem, we make an analysis that allows finding such
the optimization process can “get stuck” in one of them and fail to changes in the initial flowchart that can lead to an improvement in
reach the global extremum. This downside can be overcome by the efficiency criterion. In Ref. [23], such an analysis is based on the
starting the optimization process from different initial points. The duality theory in nonlinear programming and makes it possible to
solution with the highest efficiency is chosen from those obtained assess how the organization of additional small thermal or material
in such a way. flows between different points of the chart will affect the optimal
In the event that it is impossible to disregard the discrete nature value of the efficiency criterion. A.M.Kler et al. [27] do not consider
of some parameters of the energy plant to be optimized, we have to the “small” flows; however, they deal only with those ensuring the
solve a mixed nonlinear programming problem with continuously maximum improvement in the efficiency criterion. Such an analysis
varying and discretely varying optimized parameters. The main is reduced to solving special linear programming problems. The
method of solving such a problem is the branch-and-bound evolutionary methods include an expert stage of transition from
method (A.H.Land and A.G.Doig [17]). The examples of its applica- recommendations of the initial flowchart analysis to a new flow-
tion are presented in Ref. [18] (Juan I. Manassaldi et al.) [19], (Si- chart which can result in inconsistency of the optimization results.
Doek Oh et al.), and in some other studies.
A.M.Kler and Y.M.Potanina [20] apply a simplified method to 2.1.3. Methods for optimization of the most complex flowchart
solve the mixed nonlinear programming problem. The method These methods are intended to design the most complex flow-
suggests constructing a tree of potentially optimal options. In this chart from which we can obtain any flowchart of a plant interesting
case, the continuously varying parameters of the coal-fired steam for research by excluding “extra” components. These methods are
turbine unit and discretely varying parameters that specify a set of divided into two groups.
steels to manufacture steam boiler heating surfaces are optimized.
The problems of flowchart optimization are much more difficult 2.1.3.1. Methods for constructing a specific flowchart by setting
to reduce to some known mathematical problems than the prob- logical variables each responsible for including one component.
lems of parameter optimization. The only problem that was In the case that the value of such a variable is “truth”, the compo-
completely formalized was the optimization problem of a system of nent is added to the flowchart. Its cost is taken into account, its
heat exchangers. In Ref. [21] (Kafarov V.V. et al.) [22], (Kesler M.G., operating and design parameters are included into a set of pa-
Parker R.O.), this problem is reduced to the problem of allocation rameters to be optimized, the constraints on the component pa-
and in Ref. [23] (Kler A.M. et al.) e to a transportation problem of rameters are included in a set of inequality constraints, and the
linear programming. inclusion of the component in the flowchart is “activated”. In the
R. Anantharamama et al. [24] consider a sequential framework case that the value of a logical variable is “false”, the parameters
for synthesis of a network of heat exchangers. The objective of this and constraints related to the component are not considered [28].
methodology is to solve the problems of a commercial scale As a result, based on the specified set of logical variables, a specific
(dimension). Leandro Vitor Pavao ~ et al. [25] focus on automated flowchart is constructed and corresponding optimization problem
synthesis of a network of heat exchangers by using the genetic of the plant parameters is formulated. An advantage of this group of
algorithms, the method of particle swarm and parallel computing. methods is that there is no need to develop an individual mathe-
Zorka Novak Pintari and Zdravko Kravanja [26] propose a meth- matical model for each option of the flowchart. Their downside is
odology for synthesis of a network of heat exchangers with a great the necessity to solve an optimization problem of continuously
number of uncertain parameters. This methodology combines varying parameters for each option of the flowchart.
several heuristic methods.
However, we have not managed to formalize the flowchart 2.1.3.2. Methods for optimization of a most complex flowchart based
optimization problems of complex steam turbine, combined-cycle on discretely varying parameters. Discretely varying parameter z is
and other thermal power plants completely. Next, the methodo- placed in correspondence to each component that can either be
logical approaches developed earlier in the ESI SB RAS will be present (z ¼ 1) or absent (z ¼ 0) in the plant flowchart. The values of
considered, as well as a new approach to optimizing the flowcharts objective function, equality constraints and inequality constraints
of thermal power plants will be presented. can be calculated for any real values of z from 0 to 1 but only the
values that belong to the set {0, 1} are feasible.
The advantage of the methods of the second group is that the
2. Methods of research optimal flowchart is determined by solving one mixed nonlinear
programming problem. The considered approach is employed in
2.1. Existing approaches to a flowcharts optimization Ref. [28] to select the equipment mix for a combined heat and
power plant. Groups of parallel included, one-type and equally
The following methods were used to solve these problems. loaded components (groups of cogeneration power units) were
considered. The problem was to determine the number of units of
each group included into the work in a specific mode. At the same
time, the relationship between the extensive parameters (steam,
2.1.1. Assignment of a set of flowcharts to be considered in advance water, and electric power consumption) of the group and the
For each of such flowchart, we formulate a problem of param- separate unit had the form
eter optimization. The values of an assumed efficiency criterion at
the point of solution to these problems for different flowcharts are Ygr
Yun ¼ ; (1)
compared and the one for which the criterion has the best value is zgr
selected. The main advantage of the method is its simplicity, while
its disadvantage is the impossibility of considering a large number where Yun , Ygr - extensive parameters of the group and the unit, zgr
of options. - the number of components included into the work. In this case,
4 A.M. Kler et al. / Energy 193 (2020) 116679

Ygr - parameter characterizing the material and energy flows con-

necting the group of elements with the rest of the flowchart, Yun - X in:exc
ijl ¼ X in:inc
ijl ; (4)
parameter describing the energy flows connecting the separate
unit with the flowchart. All extensive flow parameters of one block
X out:exc
ikl ¼ X out:inc
ikl ; (5)
must take such values at which a valid mode of operation of the
block is possible. The intensive flow parameters (pressure, enthalpy
where X in:exc , X in:inc - l-th intensive parameter of the j-th external
temperature, etc.) of the group and similar flow parameters cor- ijl ijl

responding to one components should be equal. and internal input flows of the i-th component, X out:exc
ikl
out:inc
, X ikl - l-th
In Ref. [29] (Lidia S. Guerras, Mariano Martín), for an integrated intensive parameter of the k-th external and internal output flows
optimization of the exhaust gas cleaning system of a coal-fired of the i-th component.
power unit, an approach is used that is close to the optimization It should be noted that for z ¼ 0 in expressions (2), (3) the di-
method for the most complex scheme. However, the integrity vision by zero will occur. Therefore, instead of 0 it is proposed to
requirement of optimized parameters was not taken into account. equate the smaller discrete value of the parameter z to some pos-
The presented analysis of the previously conducted studies itive magnitude ε. This magnitude should be selected so that at z ¼
related to the optimization of discrete parameters of thermal power ε the material and energy “external” flows are small and do not
plants indicates that most of them use the methods of direct affect the results of the calculation of the flowchart. In this case, the
enumeration that can optimize a small number of parameters. The “internal” material and energy flows must be within feasible limits.
co-optimization of cycle parameters of a thermal power plant and In the considered problems of flowchart optimization, the
design parameters of its components can be performed only by includable component is included in the flowchart through the
gradient methods of nonlinear mathematical programming. This is flow separation or mixing nodes (points). At flow separation points,
due to the need to optimize the significant number of parameters. some initial flow going from one flowchart component to another
The effectiveness of directional search methods with increasing one is divided into two flows. One of them still goes to a corre-
dimension of the optimization problem decreases sharply. sponding component of the flowchart, while the other e to the
The problem of flowchart optimization can only be formalized includable component. At mixing points, the flow from the
for certain subsystems, in particular a subsystem of heat ex- includable component joins the flow going from one flowchart
changers. The approach based on the method of the most complex component to another one.
flowchart optimization is rather promising to optimize the flow- At the flow separation point, from which the outgoing j e th flow
charts of thermal power plants. This is especially relevant to the of the i-th component goes, the mass balance equation and
power plants to be newly designed. equalities of intensive flow parameters hold true

out:fl in:fl
Gij ¼ Gij  Gin:exc
ij ; (6)
2.2. Description of the proposed approach
out:fl in:fl
In this study, the approach to optimizing the number of working P ij ¼ P ij ; (7)
elements of a group described in Ref. [28] is applied to optimizing a
most complex scheme containing components that can either be in:fl
present or absent in the power plant flowchart. Let us call such a
P in:exc
ij ¼ P ij ; (8)
component “includable”. As already mentioned, the parameter z
 
can take two possible discrete values: 1 - the component is present T out:fl ¼ T in:fl or Н out:fl ¼ Hin:fl ; (9)
ij ij ij ij
in the flowchart, 0 - the component is absent. However, the oper-
ation of the includable component with extensive flows parameters  
in:fl in:fl
close to zero connecting this component with other components of T in:exc
ij ¼ T ij or Hin:exc
ij ¼ Hij ; (10)
the flowchart is impossible. The limitations on the technically
minimal load and the others will not be fulfilled. Therefore, in the
where G, P, T e mass flow rate, pressure and temperature (other
proposed approach, the “external” flows connecting the compo-
notations are presented in section “Nomenclature”).
nent with the flowchart and the “internal” flows at which the
Moreover, at the separation point we should take into account
component is calculated are separated (by analogy with the pre-
the constraint on non-negativity of the flow rate of the flow going
sented separation of the flows parameters of the component and of
from the separation point to a flowchart component
the group).
The relationship between the extensive parameters of these
Gout:fl 0 (11)
flows has the form ij

At the flow mixing point, to which the k-th outgoing flow of the
Y in:exc
ijl i-th component comes, the equations of mass and energy balances
Y in:inc
ijl ¼ ; (2)
zi and the identical equation for pressure of flows between the
flowchart components hold true
Y out:exc
Y out:inc
ikl ¼ ikl
; (3) out:fl
Gik
in:fl
¼ Gik þ Gout:exc ; (12)
zi ik

where Y in:inc
ijl , Y in:exc
ijl - l-th extensive parameter of the j-th external Gout:fl H out:fl ¼ Gout:fl Hin:fl þ Gout:exc
ik Hout:exc ; (13)
ik ik ik ik ik
and internal input flows of the i-th component, Y out:inc
ikl , Y out:exc
ikl - l-th
extensive parameter of the k-th external and internal output flows P out:fl ¼ P in:fl : (14)
ik ik
of the i-th component.
Similar intensive parameters of the corresponding external and Furthermore, at this point, the inequality constraint that re-
internal input and output flows should be equal quires the pressure of the flow from the includable component to
 A.M. Kler et al. / Energy 193 (2020) 116679 5

in:fl
be no less than the pressure P ik , is taken into account incorporated at the interruption point. It is worth noting that in a
general case one component can be inserted at the interruption
in:fl point of not one but of several flows. The components that are
P out:exc
ik  P ik : (15)
included in such a manner are additional steam coolers installed in
Only if this condition is met, the flow between the group of the steam flows going from steam turbine to regenerative heaters.
components and the mixing point of the flowchart can be orga- Another example of such components is additional heating surfaces
nized without additional pumps or compressors. placed in the flow of combustion products of a steam boiler. In
It is worth noting that includable component is defined by a order to transfer gradually from the state when a new component is
system of equations relating the above parameters of incoming and present in the flowchart, to the state when it is absent, we
outgoing flows of the component and its internal parameters. Part conventionally introduce a bypass line to the flowchart. This bypass
of these internal parameters is specified, and part is calculated from line goes from the point of flow separation and enters the point of
the indicated system of equations. In turn, part of the specified flow mixing (Fig. 2). It is assumed, that this line does not have
parameters is determined by external conditions, while the other hydraulic resistance.
part can vary within some limits and is to be optimized. Let us call It is worth noting that, if there is even small internal hydraulic
the latter group of the parameters the internal optimized param- resistance of the component, conditions (7), (14) and (15) cannot be
eters of a component. The internal component parameters to be met simultaneously, since the hydraulic resistance of the bypass is
calculated should meet the system of inequality constraints. We assumed to be zero, and the hydraulic resistance of the element at a
will call these constraints internal inequality constraints. positive flow rate through it must also be greater than zero.
To calculate the criteria of the plant cost-effectiveness, the in- For such components, in order to eliminate the indicated
ternal capital investments of a component are determined from of contradiction, we propose using a model of an ideal mixer at the
its calculation, and the external capital investments used in calcu- point of mixing of the flows in which a new component is placed.
lating the total capital investments into a plant are determined This model, as the one above, takes into account the equations of
from the expression mass balance (12) and energy balance (13). Instead of conditions
(14) and (15), however, the following conditions are considered.
K exc inc We assume that the flow coming to the mixer with a higher
i ¼ K i ,zi (16)
pressure (bypass line flow) expands in the course of the adiabatic
It is worth noting that for part of the components that were process to some pressure Pmix , while the flow coming to the mixer
considered in optimization, conditions (14) and (15) are easily met with a lower pressure (the flow going through a component),
and do not limit the optimal solutions. This refers to the compo- shrinks in the course of the adiabatic process to the same pressure
nents for which the flow rates of all flows, connecting the group Pmix . In this case, pressure Pmix is selected so that the work of bypass
with the flowchart become small at zi ¼ ε (Fig. 1). flow expansion is equal to the work of compression of the flow
Using inequality (15) for a mixing point is not always possible. going through the component. The mixer outlet pressure P out:fl is
This applies to the case when the separation point and the mixing taken equal to pressure Pmix . Obviously, the pressure Pmix lies in the
point are located on the same flow of the working media or heat range determined by the pressures of flows entering the mixer.
carrier. Such a situation arises if there is material flow in the In the general case, it can be assumed that the pressure is
flowchart that goes from component l to component k in the determined from the equality
absence of the i-th includable component. When the i-th compo-
 
nent is included (zi ¼ 1), the flow is completely directed from compr
Nctij P out:exc:ct:
ij ; Gout:exc:ct:
ij ; Hout:exc:ct:
ij ; P mix
ij ¼
component l to component i, and from component i to component   (17)
by by by mix
k. N ex
byij P ij ; Gij ; H ij ; P ij ;
Imagine that this flow is interrupted and a new component is

Vapor
Vapor
Feedwater Feedwater Bypass

Ideal mixer
Condensate
Fig. 1. An “includable” flowchart component connected in parallel to the existing
flows. Fig. 2. An “includable” flowchart component inserted in the vapor flow.
6 A.M. Kler et al. / Energy 193 (2020) 116679

compr
where N ctij - compression capacity of the j-th external output
flow of the i-th component, Nex
byij - expansion capacity of the bypass

in.exc
G
flow.
It is worthwhile to note that the use of the indicated condition
makes it possible to accurately determine the pressure after the
mixer in the case where zi ¼ 1 (the entire flow goes through the
component, Gin:fl ¼ 0), and with a small error, when zi ¼ ε, and the
flow through the component is small, and virtually the entire flow

in.inc
G
goes through the bypass line. With the intermediate values zi , the
proposed approach ensures a gradual change in the pressure be-
tween the extreme values, which provides a good convergence of
the computational process. It should be noted that the equality of
inlet and outlet pressures is only at the points of separation of
flows, and in such elements as heating surfaces, etc., hydraulic
resistance is taken into account.
At the flow separation point, to which the external flow going to
component is connected, the equation of mass balance is written as
follows

Gin:exc
i ¼ Gin:fl ,zi ; (18)

Gout:fl ¼ Gin:exc ,ð1  zi Þ:

out.inc
G
i i (19)

by
where Gi - flow rate through the bypass line.
At the flow mixing point, the equation of mass balance will have
the form

out.exc
G
by
Gin:exc
i ¼ Gin:exc
i þ Gi : (20)

Here, the flow rate of the internal flow of the flowchart

component is determined from the expression

Gin:exc
Gin:inc
i ¼ ¼ Gin:fl : (21)
zi
Fig. 3. Relation between internal and external flows of a component.
In the flowchart, these flows can be denoted as follows (Fig. 3).
Given the above descriptions of flow separation and mixing
points, as well as the relations between the parameters of internal
and external flows, the mixed optimization problem of the  inc inc 
maximum complex flowchart will have the following formFind ~ ;Q
vi Q b ; y ; u ; xH  0; (29)
i i i i i
 
~ exc ; …; Q
minF xH ; y; U; Q b exc ; …; Q
~ exc ; Q b exc ; (22)
1 I 1 I
ε  zi  1; (30)
subject to
 
~ exc ;…; Q
Wfl xH ; y; U; Q b exc ;…; Q
~ exc ; Q b exc ¼ 0; (23) zi 2 ½ε; 1; i ¼ 1; …; I: (31)
1 I 1 I
The vectors that refer to i-th includable component have the
  same notation as the corresponding vectors that refer to the con-
~ exc ;…; Q
Vfl xH ; y; U; Q b exc ;…;
~ exc ; Q b exc
Q  0; (24)
1 I 1 I stant part of the flowchart, but with the subscript index i. The
optimality criterion F can be represented by a criterion of energy
efficiency (specific fuel consumption per unit of electricity output)
x H  xH  xH ; (25) and a criterion of cost-effectiveness (discounted costs, electricity
price at a specified value of internal rate of return, etc.).
b exc ¼ z Qb inc ; In the above description, we assume that all optimized flow-
Q i i i (26)
chart parameters x and internal optimized parameters of the
includable components xi are continuously varying. The equations
b exc ¼ Q
b inc ; of separation and mixing points of external flows of includable
Q i i (27)
components refer to the system of equations Wfl . Inequality con-
 inc inc  straints for the mixing and separation points belong to the system
~ ;Q
wi Q b ; y ; u ; xH ¼ 0; (28) Vfl . Let us call problem (24e33) with problem (I). The branch-and-
i i i i i
bound method is used in this paper to solve the mixed problem of
non-linear mathematical programming [17].
 A.M. Kler et al. / Energy 193 (2020) 116679 7

2.3. Application of the branch-and-bound method within the branch-and-bound method, the set of potentially optimal options is
proposed approach divided into two subsets by excluding one discretely varying
parameter from those to be optimized. In this case, this parameter
The general scheme for applying this method is as follows. A set can take a maximum discrete value for all options of one subset,
of all possible options of the values of optimized discretely varying and the minimum discrete value for all options of the other subset.
parameters is divided into disjoint subsets. Moreover, in all the We call this parameter a branching parameter.
options of one subset, part of the optimized discretely varying At each step, this parameter is suggested to be selected as fol-
parameters takes fixed discrete values, which are the same for all lows. Among the discretely varying parameters, which were opti-
options of the subset. The remaining discretely varying parameters mized when determining the lower bound of the objective function
are optimized without discreteness constraints. For each subset, (J qi ¼ 0), we search for a parameter whose optimal value is the most
the lower estimate (bound) of the optimal value of the objective distant from the nearest feasible discrete value. At the current step,
function is determined. This estimate must have the following this parameter is taken as a branching parameter.
properties. If for a set of potentially optimal options all values of the
discretely varying parameters are with certain accuracy close to the
1. The optimal value of the objective function of any option from feasible discrete values, they can be rounded up to these values.
the subset (for fixed values of discretely varying parameters and These values are used to solve an additional optimization problem
optimal values of continuously varying optimized parameters) q
II. It should be noted the set J i corresponding to this problem does
should be no lower than the bound.
not have components that equal to zero. The resulting discrete
2. If the bound of the objective function of a subset coincides with
solution is considered optimal for this set of options. Rounding up
the value of the objective function for a discrete solution
the values of discretely varying parameters allows reducing the
belonging to this subset, then the bound is exactly equal to the
number of branches and the amount of calculations necessary to
minimum value of the objective function of the specified subset
solve the mixed problem.
of options.
A tree of sets (subsets) of options of discretely varying param-
eters can be used to illustrate how the branch-and-bound method
A subset with the lower bound is chosen. Let us call this subset a
works. The tree of options is constructed according to the following
“subset of potentially optimal options”, since it seems advisable to
rules. Each vertex of the tree with number q corresponds to its own
search, primarily, for the optimal option in this subset. If the
variant of fixing part of the discretely varying parameters and its
boundary value of the objective function of this subset is achieved
own optimization problem II corresponding to this fixing. The root
for the option with discrete values of all parameters zi , then this
(zero) vertex of the tree corresponds to the option when no
option will be optimal. Otherwise, the subset of potentially optimal
discretely varying parameter is fixed. Each vertex corresponds to
options is divided (branches) into several (at least two) subsets, for
the set J q , the value of the objective function F q and the vector of
each of which the bounds of the objective function are determined.
discretely varying parameters Z q obtained as a result of optimiza-
This process is repeated until the above optimality condition is
tion. If the top of the tree does not have outgoing vertices, it is
satisfied. The branch-and-bound method will converge if, when
called a pendant vertex.
branching, the bound of the objective function of the subset ap-
In each t-th step of the branch-and-bound method, we search
proaches its minimum value.
for the pendant vertex of a tree, which corresponds to the lower
To apply the general scheme of the method for specific prob-
bound. We will denote the number of this vertex by mt . We search
lems, it is required to find the method of determining the bound of
for the number Vt of branching parameter, such that J m Vt ¼ 0 and
t
the objective function and the method of splitting the set of options
of discretely varying parameters into subsets. meets the condition
In the considered problem, the set J q of I components is intro-
duced to construct the bound of the objective function of the initial
h   i n h
set of options of problem I. Each component of the set J q can have min zm min
; zmax mt
Vt  zVt Vt  zVt ¼ max
n . o min
t

three values: 1, 0 and 1. If J qi ¼ 1, then zi is fixed on the value 1. If ci2 j

m
J j t ¼0
J qi J qi
¼ 1, then zi is fixed on the value ε. If ¼ 0, then zi is optimized    io
in the interval [ε, 1].  zm
i
t
 zmin
i ; zm
i
t
 zmax
i ;
The system of additional conditions is introduced at the form
(33)

q
Ji ¼ 1/zi ¼ 1; where zmin
i ; zmax
i - minimum and maximum discrete values of
parameter zm mt
Vt , zVt - optimal value of parameter zi at point of so-
t

Jiq ¼ 0/ε  zi  1;
lution to the optimization problem corresponding to vertex mt .
Jiq ¼ 1/zi ¼ ε; Two branches go from vertex mt of the tree. The first one cor-
i ¼ 1; …; I (32) responds to the minimum value of parameter zVt , and the second e
to the maximum one. The first branch enters the vertex tA, and the
To find the bound of the objective function of the options second branch enters the vertex tB. When solving the problems
determined by the composition of fixed discretely varying param-
that correspond to these vertices, the elements of sets J tA tB
i and J i are
eters defined by the set, problem (24e32), (34) is solved. Let us call
defined as follows
this problem with problem II.
If this problem does not have a feasible solution, then a suffi-
ciently large number D (larger than any possible feasible value of (
the objective function) is assumed as the bound of the objective Jm t
; if isVt
function. J tA
i ¼ i ; (34)
1; if i ¼ Vt
In the problem under consideration, any discretely varying
parameter can take only two discrete values. At each step of the
8 A.M. Kler et al. / Energy 193 (2020) 116679

( present some set in which only the part of discretely varying pa-
Jm t
; if isVt rameters take values close to discrete with the required accuracy F с.
J tB
i ¼ i (35) If the difference between the value of the objective function for the
1; if i ¼ Vt
optimal discrete solution F opt and the lower bound for the set
If, the condition mentioned above meets the condition

n h   io F opt þ f  F c ; (39)

g max min zN min
i  zi ; zmax
i  zN
i ; (36)
c i2fj=ðJ j ¼0Þ∧ðj¼1;IÞg
N

where f is the computational deviation of the optimization prob-

lem solution, then the branching of this set occurs. This process is
is met in any of the tree vertices with number N, when solving the
repeated until there are many options for which the above condi-
optimization problem that corresponds to this vertex, then all pa-
tion is true. As a result, several discrete solutions can be found for
rameters zi are rounded up to the closest discrete values and an
which the difference in the objective functions with the optimal
additional optimization problem is solved. In this problem, the
solution is less than the optimization error. We can assume that
values of all discrete parameters are fixed. A new vertex N0 and a
these solutions form many optimal options.
branch connecting the vertices N and N0 are included into the tree.
The vertex N0 corresponds to the set J N’ with all of components are
3. A mathematical model of the plant
not equal to zero. The value of the objective function at the point of
solution of the additional problem is assumed lower bound of this
The proposed approach is demonstrated by the integrated
vertex. This vertex will correspond to one option of values of
gasification combined cycle power plant (Fig. 5). The investigations
discretely varying parameters.
of such plants are discussed in many papers, for example, by Kler
It should be noted that at each optimization step, the set of
A.M. et al. [30], Ryzhkov A. et al. [31], Tola V. and Pettinau A. [32],
pendant vertices of the tree of options determines all disjoint
Emun F. et al. [33], Ryouhei Hoya and Chihiro Fushimi [34], Mona
combinations of discretely varying parameters. In this case, three
Gharaie et al. [35], to name but a few.
types of vertices can be distinguished. If problem II at the vertex q
We developed a mathematical model for the design calculation
does not have a feasible solution (F q ¼ D), then all options of the set
of the plant. The dotted lines in the Figure demonstrate the
represented by the vertex q do not have it either. Let us introduce
bypassing sites of the included components for different heat car-
the set Q tD including numbers of all pendant vertices of the indi- riers (vapor, combustion products, gasification products, feed-
cated type that are available in the tree at step t. If all optimized water). The model offers a possibility of excluding both one
discrete parameters at the optimal point of solution of problem II component and a group of different-type components that are
corresponding to the hanging vertex q are close to discrete values excluded in a coordinated manner. For example, the evaporation-
with the required accuracy (i.e. met condition (36), then the best of heating surface is excluded with a pump and a drum separator. In
all the options represented by vertex q is the option with discrete total, we consider 11 excludable components in the flowchart.
values of optimized parameters corresponding to the vertex q’. The These are heat exchangers operating on gasification products and
assumption is made that if problem II, corresponding to the vertex combustion products (high-pressure evaporators are not excluded).
q, has a feasible solution, then problem II, corresponding to the The mathematical models of the components contain the
vertex q’ also has it. equations of heat and mass balances, relationships between the
The set Q t0 includes the numbers of all hanging vertices that the input and output variables (pressure, temperature, flow rate, etc.),
vectors of discrete optimized parameters with discrete values and design characteristic of the components. The process of coal
available in the tree at step t correspond. The set Q tB includes the gasification is calculated taking into account the conditions of
numbers of all pendant vertices of the tree at step t that are not thermodynamic equilibrium. Steam and gas turbines are calculated
comprised in the sets Q tD and Q t0 . These are the numbers of those by the compartments (group of stages with constant gas or steam
vertices, for one of which a branching procedure can be performed flow rate). In the design calculation of a gas turbine, the initial
at step t. parameters are the flow rate, inlet pressure and temperature of the
gas mixture, as well as the outlet pressure. In addition, the me-
If at some step t the set Q t0 is not empty, then we find the
chanical and adiabatic efficiency of the compartment are set. The
number of the pendant vertex S belonging to the best value of the
purpose of the calculation is to determine the mechanical power of
objective function. This number meets the following condition
the turbine and the gas temperature at the outlet of the compart-
  ment. The initial data for calculating the compartment of the
F S S 2 Q t0 ¼ minFti (37) compressor are the air flow rate, inlet and outlet pressure, inlet
ci2Q 0
temperature, adiabatic and mechanical efficiency. The output air
If Q tB ¼ ∅, then the vertex with number S corresponds to the temperature and the consumed mechanical power are determined.
optimal discrete solution. If Q tB s∅, then we find the number of the The mathematical model of the steam turbine compartment is
designed to calculate the process of steam expansion. In the design
vertex т 2Q tB , which corresponds to the best value of the objective
calculation of the technical specifications for each turbine
function
compartment, the input flow rate, steam pressure, enthalpy and the
steam output pressure are initially set, and the output steam
F n ¼ minFtj (38)
cj2Q B enthalpy and the mechanical power of the compartment are
determined.
If the condition F S  F n is met, then the number S corresponds Models of convective heat exchangers on gasification or com-
to the optimal discrete solution. A flowchart of an optimization bustion products are focused on determining the temperature and
algorithm using the branch-and-bound method is presented in pressure of the gas at the outlet, the outlet pressure of the heated
Fig. 4. fluid, the heat transfer surface, the mass of the tubes, the temper-
It should be noted that the solving of optimization problems ature and mechanical stress of the tube metal and other structural
occurs with some with some computational deviation. Let us characteristics. The initial data for the models are the inlet
 A.M. Kler et al. / Energy 193 (2020) 116679 9

The number of step (t) is assumed equal to zero

Solving the problem (II) corresponding to the root

vertex of the tree, finding the F0, z0

No
There is a feasible
solution The problem (I) does not have a feasible solution

Yes

All of zi0 Yes

Solving the additional problem (II) for the
are close to the
discrete values discrete values of zi

No

Increasing the step t by 1 The problem (I) is solved

Finding the number of pendant vertex mt

with the minimum boundary value of the
objective function

Searching the branching parameter number

Vt

Solving the problems (II) corresponding to the

vertices tA and tB, determining ztA, ztB, FtA, FtB

Analysis of optimal solutions ztA , ztB for proximity

to discrete values. Solving the additional problems
(II) corresponding to the vertices ztA and / or ztB in
the case of such proximity

All of the problems (II) Yes

with no feasible solution
correspond to all of the pendant The problem (I) does not have a feasible solution
vertices at step t

No
Yes

The optimality condition is The problem (I) is solved

satisfied
No

Fig. 4. A flowchart of an optimization algorithm using the branch-and-bound method.

temperature and gas pressure, the flow rate and the thermody- software modules.
namic parameters of the heated coolant at the inlet and its output The calculation was performed with the software package
enthalpy, some design characteristics. The algorithm for calculating “System for Machine Building of Programs”, developed at the
convective heat exchangers is based on standard methods of Melentiev Energy Systems Institute SB RAS [23,28]. The mathe-
thermal, hydraulic and aerodynamic calculations of boiler units matical model of the plant contains 802 information input pa-
[36]. In this work, a parameter was introduced into the mathe- rameters, 918 information output parameters and 22 iteratively
matical models of the heat exchangers that determines the inclu- specified parameters. The optimized (independent) parameters of
sion of the component in the flowchart. process relations between the components are flow rate, pressure
Technical and economic indices are determined using individual and enthalpy of live steam, pressure and enthalpy of reheat steam,
10 A.M. Kler et al. / Energy 193 (2020) 116679

GAS
WGS1
TD1 TD2 TD3
TOP
SEP1

AC2 WGS2
P1

WGG1 P4
FC
WGS3

WGG2 AC1 GT

WGS4

P5
WGG3

WGS5 SEP3

WGG4
P1
WGS6

WGG5 SEP2
P3 SEP4
WGS7
P7
DAR
P2
WGS8

RH

WS

Fig. 5. A calculated flowchart of the plant (the notations are presented in the section “Nomenclature”)

heated medium enthalpy at the outlet of all economizer and  Capital investment depending on electric power (generators,
evaporation heating surfaces, etc. The following internal parame- etc.);
ters of the components were optimized: inner diameters of tubes in  Capital investment depending on the amount of heat extracted
convection heat exchangers, tube spacing in heat exchangers, and from steam turbine condensers;
mass velocity of medium in tubes, width and depth of gas duct in  Capital investment depending on the volume of consumed fuel;
the heat recovery boiler.  Capital investment in the construction, that are fixed within
The optimized parameters also included the parameters that some range;
determined the inclusion of a component in the flowchart, and the  Capital investment depending on the cost of the main thermal
number of parallel flows of heat exchangers on gasification prod- equipment.
ucts, which varied in the range from 1 to 2. The inequality con-
straints were the constraints on non-negativity of end temperature For the first three types of the investment, we set a specific cost
differences for all the heat exchangers, maximum temperature of of related systems (fuel preparation, electric equipment, etc.). The
external walls of their tubes, mechanical tension of the tube metal, construction cost is assumed expertly for the considered capacity.
etc. In addition, for convective heat exchangers, the minimum When determining the cost of the main equipment, the cost of
allowable number of pipes along the gas flow was taken into ac- turbine machines and pumps is assumed in proportion to their
count. The introduction of these restrictions does not allow us to capacity, while the cost of heat exchangers is taken in proportion to
reduce the optimization to the NLP problem with only continuously their mass or area of heat transfer surface. The calculation of the
changing parameters. Totally, we optimized 127 parameters main equipment cost allows for the share of assembling costs and
(including 12 parameters considered as discretely varying) and the share of the other equipment costs. The determination of
took into account 169 constraints. Brown coal was considered as a capital investment also takes into account some coefficients that
fuel. Its characteristics were: consider unforeseen costs, annualizing the equipment prices to the
prices of the calculation year, etc.
The minimum price of electricity Cel for a given value of internal
 rate of return (IRR) and the maximum efficiency of the power plant
Qir ¼ 17291 kJ kg; Wtr ¼ 22%; Ar ¼ 15; 6%; Sr ¼ 0; 9%; is used as optimization criteria. It is assumed that the plant works at
C r ¼ 46%; Hr ¼ 3; 6%; Nr ¼ 0; 9%; Or ¼ 11% a base load; therefore, one mode of operation (rated) is considered.
The deviation of the load from the rated one, as well as the power
The following components were taken into account to deter-
unit being in downtime associated with different types of
mine capital investment in the combined-cycle plant:
 A.M. Kler et al. / Energy 193 (2020) 116679 11

scheduled and emergency repairs, is taken into account through 56,009

the number of hours of use of the installed capacity. The approach 0
to calculate the Cel is presented in more detail in Ref. [15]. WGG3 z4 1
z4 0
55,995 56,006
1A 1B
4. Calculation results WGS3 z10 0 WGS6
z8 1
z8 0 z10 1
It is worth noting that the method used to solve the optimiza- 55,842 6A 6B 2A 2B 56,005
tion problem belongs to the interior point methods. These methods
have good convergence and stability [37]. Therefore, the values of 55,975 54,92 WGS3
the parameters zi , that determine the inclusion of a component in z8 0 z8 1
the flowchart, obtained at the optimal point, cannot strictly belong
to their boundary values ε, and 1 and are assumed to be 0.00001 55,993 3A 3B 55,999
z1 1
and 0.99999. With these values of the specified parameters, an WGS4 z9 1 z1 2
z9 0
additional optimization problem is solved. Optimization by the 55,985
criterion of the minimum price of electricity was carried out at two 55,994 55,971 55,997
7A 7B 4A 4B
values of the fuel price. The value of the specified accuracy g was WGS8 z12 1 WGS8 z12 0 WGS8
taken to be 0.05. z12 1
z12 0 z12 0 z12 1
With the optimization problem solved at point 0 for both op- 55,993
tions, the values of all discretely varying parameters turned out to 54,671 8A 8B 9A 9B 5A 5B
be close to a given accuracy to their nearest integer values. There- 55,985
55,998 54,642 53,17
fore, the branching procedure by the branch-and-bound method
was not used. An additional optimization problem (0*) was solved
to clarify the optimal values of continuously varying parameters 8B* 56,003 9B* 55,989 5B* 55,999
and the objective function. The optimal values of discretely varying
parameters at point 0, obtained by solving problem (22)e(30), are Fig. 6. Tree of potentially optimal options for including the component in the
flowchart.
presented in Table 1.
The values of the objective function at point 0* were 7.746 and
8.967 cent/kWh, respectively. As can be seen, components WGS3, meet condition (36)), so the tree of potentially optimal options was
WGS4 and WGS7 (the first stages of the convective and interme- constructed (Fig. 6).
diate superheaters and the low-pressure evaporator) are excluded For the optimal discrete-continuous solution, it was necessary to
in both options. perform nine branching operations. Each vertex corresponds to a
Optimal prices for electricity at points 0 * in both cases turned separate optimization calculation, and each branch corresponds to
out to be slightly less than optimal values at points 0. This is a discrete parameter fixed at the next step, and its value. Calcula-
explained as follows. At the points obtained as a result of an iter- tions performed at the minimum discrete value of the parameter
ative search for the optimum by the method of internal points, all are marked with the letter A, at the maximum - with the letter B.
inequality constraints on zi are satisfied as strict, i.e. ε < zi < 1. The The abbreviations on the branches of the tree indicate the names of
optimization process ends if zi  gi or zi  1  gi for all i (with heat exchangers in the scheme (Fig. 5), in which the discrete values
gi ¼ 0.05). Then zi is being assigned with the value that lies close to are set at the corresponding step. The values of the step number
zero or 1 (for example, 0.0001 or 0.9999, respectively). In this and objective function (energy efficiency) are given on the vertices
process, the new point (N*) may turn out to be closer to the optimal of the tree. On branches 3 B-4A and 3 Be4B, the selection was made
point (with zi is zero or 1) than the point N. Accordingly, F (N*) will for the number of parallel flows of gasification products.
be slightly less than F (N). Three options were obtained from optimization. The difference
When the optimization problem was solved by the criterion of between them is less than the specified accuracy (0.02) and they
maximum energy efficiency at point 0, the values of some param- can be considered non-distinguishable. Although their efficiency
eters zi were far from their nearest integer values (i.e. they did not

Table 1
Results of discretely varying parameters optimization (efficiency criterion - minimum price of electricity).

Name Parameter zi Values of parameters at different fuel prices

50 USD/tce 100 USD/tce

Number of parallel flows of gasification products Z1 1,000 1,004

Optimal values of parameters that determine the degree of component inclusion in the flowchart
WGG1 z2 0,999 0,999
WGG2 z3 0,999 0,998
WGG3 z4 0,999 0,998
WGG4 z5 0,998 0,967
WGS1 z6 0,997 0,999
WGS2 z7 0,999 0,999
WGS3 z8 0,00005 0,0005
WGS4 z9 0,0008 0,0001
WGS6 z10 0,999 0,999
WGS7 z11 0,00003 0,00005
WGS8 z12 0,999 0,999
Objective function
Electricity price, cent/kWh e 7,763 8,969
12 A.M. Kler et al. / Energy 193 (2020) 116679

Table 2
Optimization results for discretely varying parameters (heat exchangers on gasification products).

Number of vertex Name

Parameters that determine the degree of component inclusion in the flowchart Number of parallel flows of
gasification products z1
Z2 Z3 Z4 Z5

0 0,999 0,995 0,517 0,007 1,932

1 A 0,999 0,999 0 0,004 1,994
1 B 0,999 0,997 1 0,011 1,897
2 A 0,999 0,998 1 0,002 1,981
2 B 0,999 0,998 1 0,005 1,871
3 A 0,999 0,997 1 0,005 1,971
3 B 0,999 0,998 1 0,001 1,903
4 A 0,999 0,998 1 0,007 1
4 B 0,999 0,992 1 0,003 2
5 A 0,999 0,999 1 0,669 2
5 B 0,999 0,997 1 0,003 2
6 A 0,994 0,878 0 0,067 1,471
6 B 0,999 0,998 0 0,001 1,983
7 A 0,999 0,998 1 0,007 1,945
7 B 0,999 0,998 1 0,003 1,954
8 A 0,999 0,997 1 0,003 1,856
8 B 0,999 0,999 1 0,005 1,977
9 A 0,999 0,999 1 0,001 1,993
9 B 0,999 0,999 1 0,002 1,981
5 B* 1 1 1 0 2
8 B* 1 1 1 0 2
9 B* 1 1 1 0 2

Table 3 achieved by lowering the waste gas temperature, which leads to a

Optimization results for discretely varying parameters (heat exchangers on com- reduction in temperature difference in convective heat exchangers
bustion products).
of the recovery boiler and an increase in the area of heating sur-
Number of vertex Name faces. This, in turn, leads to a significant increase in investment in
Parameters that determine the degree of component the plant and in the price of electricity. The tables show that some
inclusion in the flowchart of the parameters zi take values close to discrete as a result of
Z6 Z7 Z8 Z9 Z10 Z11 Z12 continuous optimization, and the other part of zi takes discrete
values as a result of branching. Options with the lower price of
0 0,999 0,999 0,471 0,023 0,873 0,006 0,917
1 A 0,999 0,999 0,382 0,164 0,668 0,009 0,884
electricity have lower capital investments.
1 B 0,998 0,999 0,987 0,004 0,714 0,003 0,92
2 A 0,999 0,999 0,445 0,003 0 0,999 0,976
2 B 0,999 0,999 0,492 0,011 1 0,002 0,791 5. Conclusions
3 A 0,999 0,999 0 0,277 1 0,004 0,861
3 B 0,999 0,999 1 0,002 1 0,002 0,94
4 A 0,999 0,999 1 0,003 1 0,007 0,907 The researchers in different countries have been investigating
4 B 0,999 0,999 1 0,002 1 0,002 0,975 the optimal selection of flowсharts and parameters of energy plants
5 A 0,003 0,998 1 0,0001 1 0,999 0 for several decades. They calculated thermodynamic cycles, design
5 B 0,999 0,999 1 0,004 1 0,006 1 parameters of the plants and their economic indices.
6 A 0,991 0,993 0 0,211 0,848 0,05 0,759
The optimization problem of continuously varying parameters
6 B 0,999 0,999 1 0,007 0,709 0,001 0,965
7 A 0,999 0,999 0 0 1 0,005 0,922 of thermal power plants is solved by various methods of direct
7 B 0,999 0,999 0 1 1 0,004 0,873 enumeration, gradient methods and others; while the problem
8 A 0,999 0,999 0 0 1 0,999 0 with continuously varying and discretely varying parameters is
8 B 0,999 0,999 0 0 1 0,001 1
normally solved by the branch-and-bound method. The optimiza-
9 A 0,999 0,999 0 1 1 0,999 0
9 B 0,999 0,999 0 1 1 0,001 1
tion problems of flowcharts of complex thermal power plants are
5 B* 1 1 1 0 1 0 1 not completely formalized. There are several approaches to solve
8 B* 1 1 0 0 1 0 1 them. This can be an assignment of a set of flowcharts to be
9 B* 1 1 0 1 1 0 1 considered, a successive improvement in the flowchart, optimiza-
tion methods of a most complex flowchart (using the variables
setting the presence of a component). In this paper, we propose a
values are close, the options differ in capital investments. At methodological approach to flowchart-parameter optimization of
vertices 5 B, 8 B and 9 B, the values of all discrete-varying param- thermal power plants based on combination of a method for
eters meet condition (36), consequently additional optimization optimization of the most complex flowchart and a branch-and-
problems 5*, 8 B* and 9 B* were solved. Tables 2 and 3 show the bound method. The authors are not aware of other works in
values of the parameters at the points corresponding to all the which the optimization of the flowchart of the thermal power plant
vertices of the tree. The results of the calculations at optimal points would be reduced to the problem of nonlinear mathematical pro-
for different criteria are presented in Table 4. For options with gramming. The method has the following advantages: a) wide
maximizing energy efficiency, the fuel price for calculating the coverage of possible flowcharts compositions, b) combination of
electricity price was taken as 100 USD/tce. continuous optimization of a large number of thermodynamic cycle
As can be seen from Table 4, the increase in plant efficiency is parameters, design parameters of components and parameters that
 A.M. Kler et al. / Energy 193 (2020) 116679 13

Table 4
Technical and economic indices of the plant.

Index Optimization criteria

Minimum price of electricity at Maximum efficiency, %

different fuel prices, cent/kWh

50 USD/tce 100 USD/tce 5 B* 8 B* 9 B*

Fuel flow rate in gasifier, kg/s 48,0 48,0 48,0 47,96 47,97

Gas temperature at gasifier outlet, С 1297,0 1297,0 1297,0 1297,0 1297,0

Air temperature at gasifier inlet, С 334,0 344,0 447,0 447,0 447,0
Air pressure in gasifier, MPa 1,13 1,19 2,01 2,01 2,01
Live steam pressure, MPa 10,0 10,1 11,9 11,96 11,96

Live steam temperature, С 513,0 514,0 523,0 524,0 524,0
Live steam flow rate, kg/s 145,5 146,5 137,6 137,5 137,5
Reheat steam pressure, MPa 0,97 1,04 1,95 1,94 1,92

Reheat steam temperature, С 454,0 456,0 474,0 475,0 475,0
Pressure of low pressure loop steam, MPa 0,87 0,88 0,99 0,99 0,99

Waste gas temperature, С 144,0 129,0 100,0 100,0 100,0
Weight of carbon steel heat exchangers, t 316,5 354,3 3775,3 3152,4 3410,1
Weight of pearlitic steel heat exchangers (with cases), t 115,3 119,6 815,3 897,1 843,9
Number of parallel flows of gasification products 1 1 2 2 2
Useful capacity of the plant, MW 404,2 409,8 452,2 452,0 451,9
Annual consumption of coal equivalent per plant, thousand tce 694,6 694,6 694,6 694,1 694,2
Specific consumption of coal equivalent per plant, gce/t 245,8 242,4 219,6 219,6 219,6
Electricity production, mln kWh 2829,4 2868,6 3165,4 3164,0 3163,3
Specific capital investment, USD/kW 2070,0 2078,6 4935,6 4596,7 4589,0
Net efficiency, % 50,04 50,74 55,99 56,003 55,989
Electricity price, cent/kWh 7,746 8,967 17,73 16,67 16,64

determine the composition of the flowchart, c) use of sufficiently optimization of heat recovery steam generator for a range of gas turbine
exhaust temperatures. Appl Therm Eng 2016;106:811e26. https://doi.org/
detailed mathematical models of power plants, taking into account
10.1016/j.applthermaleng.2016.06.035.
a large number of physical and technical inequality constraints on [3] Valero A, Lozano MA, Serra L, Tsatsaronis G, Pisa J, Frangopoulos CA, et al.
their parameters, d) possibility of obtaining a set of solutions close CGAM problem: definition and conventional solution. Energy 1994;19(3):
to the optimum. The disadvantage is the difficulty of optimization 279e86.
[4] Christos A. Frangopoulos. Application of the thermo-economic functional
by gradient methods in the presence of several local optima. approach to the CGAM problem. Energy 1994;19(3):323e42.
However, it should be noted that almost the only way to search for [5] Michael R, Spakovsky von. Application of engineering functional analysis to
global optimums is to start the optimization process using the the analysis and optimization of the CGAM problem. Energy 1994;19(3):
343e64.
gradient method from various starting points. [6] Valero A, Lozano MA, Serra L, Torres C. Application of the exergetic cost theory
The research is exemplified by an integrated gasification to the CGAM problem. Energy 1994;19(3):365e81.
combined-cycle plant. A mathematical model of the plant is con- [7] Wang Ligang, Yang Yongping, Dong Changqing, Morosuk Tatiana,
George Tsatsaronis. Multi-objective optimization of coal-fired power plants
structed. It can be used for co-optimization of a set of the flowchart using differential evolution. Appl Energy 2014;115:254e64. https://doi.org/
components, operating and design parameters. 10.1016/j.apenergy.2013.11.005.
The optimization calculations were carried out by the criterion [8] Godoy E, Scenna NJ, Benz SJ. Families of optimal thermodynamic solutions for
combined cycle gas turbine (CCGT) power plants. Appl Therm Eng 2010;30:
of the minimum price of electricity for a given internal rate of re- 569e76. https://doi.org/10.1016/j.applthermaleng.2009.10.022.
turn on investment and the maximum energy efficiency of the [9] Lazzaretto A, Toffolo A. Energy, economy and environment as objectives in
plant. The options are obtained in which the values of the consid- multi-criterion optimization of thermal systems design. Energy 2004;29:
1139e57. https://doi.org/10.1016/j.energy.2004.02.022.
ered discretely changing parameters are as close as possible to their
[10] Martelli Emanuele, Kreutz Thomas G, Gatti Manuele, Chiesa Paolo,
boundaries. For the options with discrete values, additional opti- Consonni Stefano. Numerical optimization of steam cycles and steam gener-
mization calculations were performed. The paper shows that ators designs for coal to FT plants. Chem Eng Res Des 2013;91:1467e82.
solving a problem on maximum efficiency, several optimal solu- https://doi.org/10.1016/j.cherd.2013.02.026.
[11] Liu XJ, Kong XB, Hou GL, Wang JH. Modeling of a 1000 MW power plant ultra
tions are possible with close efficiency values, but differing in super-critical boiler system using fuzzy-neural network methods. Energy
equipment set (i.e. heat exchangers that included in the flowchart). Convers Manag 2013;65:518e27. https://doi.org/10.1016/
The proposed approach can be used for optimization of flowcharts j.enconman.2012.07.028.
[12] Arora Rajesh, Kaushik SC, Kumar Raj, Arora Ranjana. Soft computing based
and parameters of other energy plants. multi-objective optimization of Brayton cycle power plant with isothermal
heat addition using evolutionary algorithm and decision making. Appl Soft
Acknowledgements Comput 2016;46:267e83. https://doi.org/10.1016/j.asoc.2016.05.001.
[13] Liu Ming, Zhang Xuwei, Ma Yuegeng, Yan Junjie. Thermo-economic analyses
on a new conceptual system of waste heat recovery integrated with an S-CO2
This study was carried out at Melentiev Energy Systems Insti- cycle for coal-fired power plants. Energy Convers Manag 2018;161:243e53.
tute of Siberian Branch of the Russian Academy of Science (ESI SB [14] Wang Chaoyang, Zhao Yongliang, Liu Ming, Qiao Yongqiang, Chong Daotong,
Yan Junjie. Peak shaving operational optimization of supercritical coal-fired
RAS) (Irkutsk, Russia) and supported by the grant of the Russian power plants by revising control strategy for water-fuel ratio. Appl Energy
Science Foundation (project No. 16-19-10174). 2018;216:212e23.
[15] Aleksandr M.Kler, Pavel V. Zharkov, Nikolai O. Epishkin. Parametric optimi-
zation of supercritical power plants using gradient methods. https://doi.org/
References 10.1016/j.energy.2019.116230.
[16] Dekanova NP, Kler AM, Moskalenko LF, Shchegoleva TP. Mathematical
[1] Carapellucci Roberto, Giordano Lorena. A comparison between exergetic and modeling and study of integrated gasification and combined-cycle power
economic criteria for optimizing the heat recovery steam generators of gas- plants//. Proc Int Cent Heat and Mass Transf 1990:829e42.
steam power plants. Energy 2013;58:458e72. https://doi.org/10.1016/ [17] Land AH, Doig AG. An automatic method for solving discrete programming
j.energy.2013.05.003. problems. Econometrica 1960;28:497e520.
[2] Nadir Mahmoud, Ghenaiet Adel, Carcasci Carlo. Thermo-economic
14 A.M. Kler et al. / Energy 193 (2020) 116679

[18] Manassaldi Juan I, Scenna Nicola s J, Mussati Sergio F. Optimization mathe- [28] Kler AM, Dekanova NP, Skripkin SK, others. Mathematical modeling and
matical model for the detailed design of air cooled heat exchangers. Energy optimization in the tasks of operational control of thermal power plants.
2014;64:734e46. https://doi.org/10.1016/j.energy.2013.09.062. Novosib: Sci Sib Enterprise Russ Acad Sci 1997:120 [in English)].
[19] Oh Si-Doek, Lee Ho-Jun, Jung Jung-Yeul, Kwak Ho-Young. Optimal planning [29] Guerras Lidia S, Martín Mariano. Optimal gas treatment and coal blending for
and economic evaluation of cogeneration system. Energy 2014;64:734e46. reduced emissions in power plants: a case study in Northwest Spain. Energy
https://doi.org/10.1016/j.energy.2006.05.007. 2019;169:739e49.
[20] Kler AM, Potanina YM. An approach to optimization of the choice of boiler [30] Kler AM, Dekanova NP, Tyurina EA. Heat-power systems: optimization
steel grades as to a mixed-integer programming problem. Energy 2017;127: studies. Novosibirsk: Science; 2005. p. 236 [in English)].
128e35. https://doi.org/10.1016/j.energy.2017.03.089. [31] Ryzhkov A, Bogatova T, Gordeev S. Technological solutions for an advanced
[21] Kafarov VV, Meshalkin VP, Guryeva LV. Optimization of heat exchange pro- IGCC plant. Fuel 2018;214:63e72. https://doi.org/10.1016/j.fuel.2017.10.099.
cesses and systems. Moscow: Energoatomizdat; 1988. p. 192 [In Russian)]. [32] Tola V, Pettinau A. Power generation plants with carbon capture and storage:
[22] Kesler MG, Parker RO. Optimal networks of heat exchange//. Chem Eng Prog a techno-economic comparison between coal combustion and gasification
Symp Ser 1969;65:111e20. N 92. technologies. Appl Energy 2014;113:1461e74. https://doi.org/10.1016/
[23] Kler AM, Dekanova NP, Shchegoleva TP, others. Methods of optimization of j.apenergy.2013.09.007.
complex heat power plants - Novosibirsk: VO “Nauka”. Siberian Publishing [33] Emun F, Gadalla M, Majozi T, Boer D. Integrated gasification combined cycle
Company; 1993. p. 116 [(in Russian)]. (IGCC) process simulation and optimization. Comput Chem Eng 2010;34:
[24] Anantharamana R, Nastad I, Nygreen B, Gundersen T. The sequential frame- 331e8. https://doi.org/10.1016/j.compchemeng.2009.04.007.
work for heat exchanger network synthesis d the minimum number of units [34] Hoya Ryouhei, Fushimi Chihiro. Thermal efficiency of advanced integrated
sub-problem. Comput Chem Eng 2010;34:1822e30. https://doi.org/10.1016/ coal gasification combined cycle power generation systems with low-
j.compchemeng.2009.12.002. temperature gasifier, gas cleaning and CO2 capturing units. Fuel Process
[25] Pav~ao Leandro Vitor, Costa Caliane Bastos Borba, Ravagnani Mauro Antonio da Technol 2017;164:80e91. https://doi.org/10.1016/j.fuproc.2017.04.014.
. Automated heat exchanger network synthesis by using hybrid nat-
Silva Sa [35] Gharaie Mona, Jobson Megan, Hassan Panjeshahi M, Zhang Nan, Smith Robin.
ural algorithms and parallel processing. Comput Chem Eng 2016;94:370e86. Techno-economic optimization of IGCC integrated with utility system for CO2
https://doi.org/10.1016/j.compchemeng.2016.08.009. emissions reductiond simultaneous heat and power generation from IGCC.
[26] Novak Pintari Zorka, Kravanja Zdravko. A methodology for the synthesis of Chem Eng Res Des 2015;94:428e39.
heat exchanger networks having large numbers of uncertain parameters. [36] Grigor’yev VA, Zorin VM. Industrial power system and thermotechnics.
Energy 2015;92:373e82. https://doi.org/10.1016/j.energy.2015.02.106. Moscow: Energoatomizdat; 1991. p. 588 [in Russian)].
[27] Kler AM, Maximov AS, Epishkin NO. Methods for analysis and synthesis of [37] Fiacco Anthony V, MacCormick Garth P. Nonlinear programming: sequential
technological schemes of thermal power plants based on solving auxiliary unconstrained minimization techniques. New York: Published by John Wiley
problems of linear programming. Thermophys Aeromechanics September and Sons, Inc.; 1968.
2017;24(5):787e98. https://doi.org/10.1134/S0869864317050146.