Energy 164 (2018) 1011e1020

Contents lists available at ScienceDirect

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

Recent developments and trends in optimization of energy systems

Christos A. Frangopoulos
School of Naval Architecture and Marine Engineering, National Technical University of Athens, Heroon Polytechniou 9, 157 80 Zografou, Greece

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

Article history: An overview of recent developments, trends and challenges in the synthesis, design and operation
Received 30 May 2018 optimization of energy systems is presented in this manuscript. The static and dynamic optimization
Received in revised form problems are stated mathematically, solution methods are mentioned in brief and classification of
31 August 2018
optimization problems based on the presence and treatment of time is presented. Examples of objective
Accepted 31 August 2018
functions in single-objective optimization are given and the need for multi-objective optimization is
Available online 7 September 2018
highlighted. Special reference is made to Prof. Szargut's early statement about the need for optimization
with ecological concerns.
Keywords:
Energy systems
Emphasis is given to optimal synthesis of energy systems, a subject that is still between (or combines)
Developments in optimization art and science/technology. Furthermore, it is interesting to note that optimization is applied also during
Synthesis the development of models of systems. Hints are given on further considerations and research needs in
Design subjects such as global optimization, as well as optimization with uncertainty, reliability, maintenance
Operation and social aspects. Examples of algorithms appearing in recent publications are given as an indication of
the related strong activity. The manuscript closes with general remarks and a rather comprehensive,
even though not exhaustive, list of references.
© 2018 Elsevier Ltd. All rights reserved.

1. Introduction components, which convert primary energy to useful forms of

energy such as mechanical, electric and thermal energy. Examples
Building a system with as good performance as possible has are power plants, cogeneration systems, propulsion plants, heating,
always been in an engineer's mind during the design of the system. ventilating and air conditioning (HVAC) systems, etc. Electrical
Today, the world is facing with a ‘trilemma’: how to have economic networks and district heating networks are beyond the limits of
growth without depleting the energy and other physical resources this presentation, even though they are also subject to
and without destroying the environment [1]. In order to help in optimization.
coping with this trilemma, engineers have to build the best systems The focus of the manuscript is on recent developments, trends
and operate those in the best way. Thus, application of optimization and challenges in the synthesis, design and operation (SDO) opti-
methods and procedures are necessary. In general, mization, as it is described in Section 2. The meaning of the word
‘recent’ is clarified, if we look at the years of publications of the
Optimization is the act of obtaining the best result under given references: out of 105 references, 29 are of the last two years
circumstances and, in particular, (2017e2018), 44 are of the last four years (2015e2018), 49 are of
Mathematical optimization is the process of finding the con- the last nine years (2010e2018) and 71 are of the last 19 years
ditions that give the minimum or maximum of a function, called (2000e2018).
‘objective function.’ The remaining sections are as follows.
In Section 3, the static and dynamic optimization problems are
As ‘energy system’ can be considered any system that trans- stated mathematically, solution methods are mentioned in brief
forms energy from one form to another (or to other forms) or and classification of optimization problems based on the presence
transfers energy from one place to another or from one body to and treatment of time is presented.
another. For clarity, it is necessary to set the limits of this manu- Section 4 gives examples of objective functions in single-
script: it is concerned with complex systems of interconnected objective optimization and explains the need for multi-objective
optimization. Special reference is made to Prof. Szargut's early
statement about the need for optimization with ecological
E-mail address: [email protected]. concerns.

https://doi.org/10.1016/j.energy.2018.08.218
0360-5442/© 2018 Elsevier Ltd. All rights reserved.
1012 C.A. Frangopoulos / Energy 164 (2018) 1011e1020

Section 5 is dedicated to optimal synthesis of energy systems, a

subject that is still between (or combines) art and science/tech-
nology. In Section 6, it is mentioned that on one hand optimization
is applied also during the development of models of systems, while
on the other hand the fact that a model will be used for optimi-
zation affects the development of the model itself.
Section 7 gives hints of further considerations and research
needs in subjects such as global optimization, as well as optimi-
zation with uncertainty, reliability, maintenance and social aspects.
Also, it gives examples of algorithms appearing in recent publica-
tions. The manuscript closes with general remarks and a rather
comprehensive, even though not exhaustive, list of references.

2. Levels of energy systems optimization

The optimization of an energy system can be considered at three

levels [2e4]:

A. Synthesis optimization. The term ‘synthesis’ implies the compo-

nents that appear in a system and their interconnections. After
the synthesis of a system has been composed, the flow diagram
of the system can be drawn.
B. Design optimization. The word ‘design’ here is used to imply the
technical characteristics (specifications) of the components and
the properties of the substances entering and exiting each
component at the ‘design’ point (nominal load) of the system. Fig. 1. The three inter-related levels of optimization.
One may argue that design includes synthesis too. However in
order to distinguish the various levels of optimization, the word
“design” will be used with the particular meaning given here.
gj ðxÞ  0 i ¼ 1; 2; …; p (4)
C. Operation optimization. For a given system (i.e. one in which the
where
synthesis and design are known) under specified conditions, the
optimal operating point is requested, as it is defined by the
x set of the independent variables
operating properties of components and substances in the sys- hi equality constraint functions, which constitute the simulation
tem (speed of revolution, power output, mass flow rates, pres-
model of the system and are derived by an analysis of the system
sures, temperatures, composition of fluids, etc.). (energetic, exergetic, economic, etc.),
gj inequality constraint functions expressing design and opera-
Of course if complete optimization is the goal, each level cannot tion limits, state regulations, safety requirements, etc.
be considered in complete isolation from the others (Fig. 1).
Consequently, the complete optimization problem can be stated by For energy system optimization, in particular, it is often helpful
the following question:
to arrange the independent variables into three sets [2]:
What is the synthesis of the system, the design characteristics of
the components and the operating point (or operating strategy) x≡ðv; w; zÞ (5)
that lead to an overall optimum?
where

3. The time in optimization of energy systems v set of independent variables for operation optimization (load
factors of components, mass flow rates, pressures and temper-
3.1. The static optimization problem: mathematical statement and atures of streams, etc.),
solution methods w set of independent variables for design optimization (nominal
capacities of components, mass flow rates, pressures and tem-
In the simplest case, time is not considered, i.e. a static opti- peratures of streams, etc.),
mization problem is formulated, which is mathematically stated as z set of independent variables for synthesis optimization; there
follows [5e7]: is only one variable of this type for each component, indicating
whether the component exists in the optimal configuration or
minimize f ðxÞ (1)
x not; it may be a binary (0 or 1), an integer, or a continuous
variable such as the rated power of a component, with a zero
with respect to value indicating the non-existence of a component in the final
configuration.
x ¼ ðx1 ; x2 ; …; xn Þ (2)
Then, Eq. (1) is written
subject to the constraints
minimize f ðv; w; zÞ (1a)
hi ðxÞ ¼ 0 i ¼ 1; 2; …m (3) v;w;z

For a given synthesis (structure) of the system, i.e. for given z,

 C.A. Frangopoulos / Energy 164 (2018) 1011e1020 1013

the problem becomes one of design and operation optimization: algebraic equation (DAE) formulation). The differential equations
describe the behavior of the system, such as mass and energy
minimize fd ðv; wÞ (1b) balances, and the algebraic constraints ensure thermodynamic
v;w
consistency or other physically meaningful relations or limits
Furthermore, if the system is completely specified (both z and w imposed on the problem. A general DAE optimization problem can
are given), then an operation optimization problem is formulated: be stated in implicit form as follows [17e19]:
h       i
minimize fop ðvÞ (1c)
v mininimize J z tf ; y tf ; u tf ; tf ; w (7)
zðtÞ;yðtÞ;uðtÞ;tf ;w
Maximization is also covered by Eq. (1), since:
subject to
min f ðxÞ ¼ maxff ðxÞg (6)
x x
_
HðzðtÞ; zðtÞ; yðtÞ; uðtÞ; t; wÞ ¼ 0 (8)
In energy systems, static optimization can only be performed
under the assumption of steady-state operation of the system at a
_
GðzðtÞ; zðtÞ; yðtÞ; uðtÞ; t; wÞ  0 (9)
certain point (usually the design point, also called nominal point).
A variety of methods has been developed for the solution of the
with initial conditions
optimization problems stated by Eqs. (1)e(4). Each method, even
though it is claimed as being general, it can in fact be effective in zð0Þ ¼ z0 (10)
certain types of problems. The various methods are known as
mathematical programming methods and they are usually available point conditions
in the form of mathematical programming algorithms. They can be
h i
classified in three broad categories:
Ps ðzðts Þ; yðts Þ; uðts Þ; ts ; wÞ ¼ 0; ts 2 t0 ; tf (11)
(i) Search methods: They calculate the values of the objective
function at a number of combinations of values of the inde- and bounds
pendent variables and seek for the optimal point. They do not
use derivatives. The search may be random or systematic, the
zL  zðtÞ  zU (12a)
second one usually being more efficient [5e8].
(ii) Calculus methods: They use first and (some of them) second yL  yðtÞ  yU (12b)
derivatives; this is why they are called also gradient methods.
In general, gradient methods converge faster than the search uL  uðtÞ  uU (12c)
methods, but in certain cases they may not converge at all
[5e8].
(iii) Stochastic or Evolutionary methods: Methods and algorithms wL  w  wU (12d)
such as Genetic Algorithms (GA), Simulating Annealing (SA),
Particle Swarm Optimization (PSO), Neural Networks belong tf  tf  tf (12e)
to this category. Even though some of those are in fact search where
methods, they are usually placed in a separate category
[9e14]. J scalar objective functional
H differential-algebraic equality constraints
If the objective function is continuous, by applying a search G differential-algebraic inequality constraints
method the exact optimum can only be approached, not reached, Ps additional point conditions at times ts (including tf)
by a finite number of trials, because only discrete points are z differential state profile vector
examined. However, the region, in which the optimal point is z0 initial values of z(t)
located, can be reduced to a satisfactorily small size at the end of y algebraic state profile vector
the procedure. On the other hand, there are problems for which u control (independent variables) profile vector
search methods may be superior to calculus methods, as for w time-independent variables vector
example in optimization of systems with components available tf final time.
only in finite, discrete sizes.
Two of the most successful methods for optimization of energy The objective function (functional) can have various forms. A
systems are the Generalized Reduced Gradient (GRG) and the general form for continuous time, commonly used in optimization
Sequential Quadratic Programming (SQP) method [6,7]. A combi- of energy systems, is the Bolza form:
nation of a stochastic algorithm (Genetic or Particle Swarm Algo- h       i      
rithm) with a deterministic algorithm (GRG or SQP) has been also J z tf ; y tf ; u tf ; tf ; w ¼ Q z tf ; y tf ; tf ; w
successful in energy systems optimization:the first one performs a
coarse search of the feasible space and locates a number of possible Ztf
optimum points, while the second one locates the exact optimal þ FðzðtÞ; yðtÞ; uðtÞ; t; wÞdt
point [15,16]. t0
(13)
3.2. The dynamic optimization problem: mathematical statement It consists of the function Q at the end of the time interval tf and
and solution methods the integral of the function F over the time horizon.
If the optimization problem is discrete in time, then the math-
The dynamic optimization problem is, in general, formulated as ematical statement takes a discrete form, as follows. The time
a system of differential and algebraic equations (differential- period [t0, tf] is divided in N time intervals of length Dtn, so that tf 
1014 C.A. Frangopoulos / Energy 164 (2018) 1011e1020

t0 ¼ N$Dtn and the integral in Eq. (13) is replaced by a summation

over the N time intervals, while the variables are discrete vector The optimization problem is formulated as follows: The real
sequences (e.g. u ¼ fu1 ; u2 ; …; uN g). The discrete problem then is operating profile of a system for any selected length of time (e.g.
stated as follows: day, week, year, the whole life cycle, etc.) is approximated with
h i time intervals properly selected from the point of number and
minimizeJ z; y; u; tf ; w ¼ Q ðzN ; yN ; N; wÞ length of each one, so that a steady-state operation can be
z;y;u;tf ;w
considered in each time interval, while the transients are ignored,
X
N as being very short in time compared with the rest of the operating
þ Fðzn ; yn ; un ; n; wÞ (14) period. It is also considered that the operation in a time interval
n¼1
does not affect and it is not affected by the operation in other time
intervals. Thus, the operation optimization can be performed for
subject to constraints appropriately written for each time interval.
one interval after the other and the whole period can then be
The optimization problem, as it is stated by Eq. (14), is additively
embedded in an upper level optimization, which works on the
separable across time, and it can be solved by optimal control
synthesis and design of the system. In other words, time dis-
theory or dynamic programming.
cretization is applied. Such an optimization of energy systems has
In general, the methods that have been developed for the dy-
appeared in numerous publications as, for example, in
namic optimization problem as stated by Eqs. (7)e(13) or (14) can
Refs. [2,3,15,16,34,35].
be classified in two main categories:Indirect Methods and Direct
If, however, the assumption of time intervals with independent
Methods.
operation from each other is not satisfactory, then dynamic opti-
The Indirect Methods include Calculus of Variations (COV) and
mization is necessary (it goes without saying that dynamic opti-
Dynamic Programming (DP) based on principles stated by Pon-
mization is intertemporal anyway).
tryagin and Bellman, respectively [20e22].
In COV, the problem is transformed into a two-point boundary
3.4. Optimization of transient conditions e optimal control
value problem (TPBVP) and solved accordingly. The procedure
works fairly well for unconstrained problems, but the solution of
Transient conditions appear, for example, during load increase
the TPBVP is still difficult to be achieved especially with the addi-
or decrease or during loading and unloading of energy storage
tion of the profile inequalities. These methods are, in general,
units, e.g. Refs. [36e38]. Depending on the exact formulation,
focused on using necessary conditions for optimality [4].
problems of this type belong to trajectory optimization (a term
Dynamic Programming is suitable for solving complicated and
coming from the early attempts to determine the optimal rocket
multi-stage decision problems by tracing the optimal strategy. It is
thrust profiles in the atmosphere and in vacuum [39]) or to optimal
not particularly useful in problems studied here, and therefore it
control, e.g. Refs. [40e42]. This field is beyond the limits of the
will not be further discussed.
present manuscript.
In the Direct Methods, the problem is approached by applying a
certain level of discretization that converts the original continuous
3.5. Dynamic optimization of synthesis, design and operation
time problem into a discrete one. The Direct Methods can be
including transients
divided in two sub-categories, according to the level of discretiza-
tion applied: Sequential Methods and Simultaneous Methods. Both
The most general case is the synthesis, design and operation
approaches have been effective in solving dynamic optimization
optimization of energy systems during their whole life taking the
problems of energy systems. Detailed description and application
complete operating profile into consideration, which includes
examples can be found in the related literature, e.g.
interrelated periods of practically steady-state operation, as well as
Refs. [18,19,23e29]. More close to the systems considered here are
transient conditions.
applications such as the dynamic operation optimization of a tri-
In these cases, the energy system is described by dynamic
generation system [30] and the dynamic synthesis-design-
models that include different stages or operating modes of the
operation optimization of ship or aircraft energy systems
system. The system characteristics (state equations, constraints,
[4,31e33]. In these works, the variation of the system operating
etc.) may vary from stage to stage, and this type of variation may be
conditions with time is explicitly taken into consideration and,
formulated as a mixture of continuous and discrete functions of
consequently, models of the systems at both design- and off-design
time. Thus, the operation of the whole system must be described as
points are first developed.
a sequence of different sets of DAEs (multi-stage systems) with the
objective to find the duration and operating conditions of each
stage in order to achieve an overall optimal result for the whole
3.3. Intertemporal static optimization
system. The mathematical statement consists of the objective
function in continuous or discrete form, Eq. (13) or Eq. (14),
As mentioned in the preceding, in static optimization it is
respectively, and the pertinent constraints.
considered that the system operates at a single point, which is not
Publications tackling this complex problem are very scarce in
realistic. On the other hand, the formulation and solution of the
the open literature. An example is given by a series of two papers
dynamic optimization problem, which takes the variation in oper-
[43,44], which optimize the energy system of an aircraft that in-
ating conditions with time into consideration, can be formidable or
cludes the phases of take-off, flight and landing. Each phase is
the required software may not be available. In between is the
described by a different DAE system and the optimization must be
intertemporal static optimization problem, with the word ‘inter-
performed for the whole trip.
temporal’ having the following meaning [4]:
A work towards this direction is Ref. [45] concerning the design
Intertemporal optimization is the optimization that takes into of an ORC system exploiting low-temperature waste heat of the
consideration the various operating conditions that a system en- main engines of an LNG carrier. It does not solve the complete
counters throughout its life time and determines the mode of optimization problem, but dynamic simulations are performed and
operation at each instant of time that results in the overall mini- the system is so designed that a satisfactory performance during
mum or maximum of the general objective function. transients (time needed for the system to reach stability after a
 C.A. Frangopoulos / Energy 164 (2018) 1011e1020 1015

change of the ship speed) is achieved. objective optimization is to combine the various objectives in a
single function by means of weighting factors, and then the prob-
4. Objective functions e single- and multi-objective lem is reduced to the one stated by Eq. (1) [56]. The subjectivity can
optimization be reduced or eliminated, if the objectives are studied in separate
and a characteristic point on the Pareto diagram is selected, such as,
There are several criteria that can be selected as objective for example, the point with minimum distance from the ideal one
functions of optimization. The selection of the criterion depends on or the point with maximum distance from the non-ideal one
the particular application and may have a strong effect on the final [57,58].
result (optimal system).
In the early years of optimization, technical criteria were pri- 5. On synthesis optimization of energy systems
marily selected, followed later on by economic criteria. In recent
years, the need to protect the environment introduced environ- 5.1. The possibility of finding the optimal synthesis
mental criteria, while concern for the society introduced societal
criteria too. With such a classification, the following examples of Even though not always simple, the design and/or operation
objectives are mentioned (of course, the list is not exhaustive). optimization problem can be effectively solved by existing algo-
Thermodynamic and other technical objectives: rithms. The case is completely different, however, with the syn-
thesis optimization. The following is written in Ref. [59]:
 Minimum weight of the system (particularly relevant for air- “In the usual design process of an energy system, the designer
crafts or space vehicles) uses knowledge and experience to select the type, configuration
 Minimum volume (e.g. systems on cars) and technical characteristics of a workable system (i.e. a system
 Maximum efficiency (or minimum fuel consumption) that is technically feasible and satisfies a given set of needs), which
 Maximum net power density he/she then evaluates for its technical and economic performance
 Minimum exergy destruction. and for ways of improving it. If the system synthesis (type and
configuration) is given, the decisions to be taken are of a rather
Economic objectives: quantitative nature. If, however, the synthesis is not given, in
addition to quantitative decisions there is need for many qualitative
 Minimum life cycle cost (LCC) decisions, which may be non-deterministic. In such a case, inno-
 Maximum net present value (NPV) vation and creativity play a vital role. Given the multitude of energy
 Maximum internal rate of return (IRR) system types and the variations in each type, one may question
 Minimum dynamic payback period (DPB). whether it is ever possible to replace the experienced designer's
mental process with an algorithm consisting of a set of formulae
Environmental objectives: and rules. On the other hand, in today's complex world, this same
multitude of types and variations makes it rather impossible even
 Minimum CO2 and other gaseous pollutants for an experienced designer to evaluate all possible alternatives.
 Minimum thermal pollution Consequently, an automated procedure, if properly used, can be of
 Minimum noise invaluable help to the designer.
 Minimum land deterioration. Several methods have been developed for the synthesis opti-
mization of processes and systems. Some of these are applicable
Social objectives: only to particular classes of systems (e.g., heat exchanger net-
works). Other methods are applied to more complex energy sys-
 Maximum job creation (or welfare, in more general terms) tems. However, up until now there has been no single method that
 Minimum adverse effects on health. can tackle the synthesis optimization problem in all its generality
and completeness. The field is, thus, still open to research.”
It is interesting to note here that, even though publications with All these statements are still valid today.
environmental objective functions in optimization appeared much
later [46e50], Prof. Szargut in his 1979 presentation [51] defined 5.2. Classes of methods for optimal synthesis of energy systems
the cumulative exergy consumption index and mentioned:
“… it may be called an ‘index of ecological cost. With the help of The various methods that have appeared in the literature on the
exergy it is, therefore, possible to establish an ecological economy optimal synthesis of energy systems can be classified into three
for the purpose of saving natural resources.” classes [59]:
This idea was taken further and the thermo-ecological cost was
defined, which includes not only the direct costs, but also the costs (a) Methods based on heuristics and evolutionary search.
for extracting the material used for construction of equipment, as (b) Methods attempting to reach pre-determined targets, which
well as the cost caused to the environment and the society by waste have been identified by the application of physical rules.
products of a system. The thermo-ecological cost was used as the (c) Methods starting with a superstructure, which is reduced to
objective function for energy systems optimization in publications the optimal configuration.
such as [52,53].
A single objective may lead to a system that does not satisfy A brief description of each class is given here, while further
other objectives. For example, a system of minimum cost most details as well as characteristic methods are presented in Ref. [59].
probably will not have maximum efficiency or minimum adverse In class (a), rules based on engineering experience and on
environmental effects. The need to take two or more objectives into physical concepts (e.g., exergy) are applied to generate feasible
consideration led to the development of multi-objective (or multi- configurations, which are subsequently improved by applying a set
criteria) optimization methods [54,55]. The system obtained does of evolutionary rules in a systematic way. These rules may come
not satisfy each objective in isolation, but it is the result of a from special techniques, such as exergy analysis. Artificial Intelli-
compromise, more or less subjective. A simple approach to multi- gence and Expert Systems have been proven effective in generating
1016 C.A. Frangopoulos / Energy 164 (2018) 1011e1020

appropriate configurations. For each acceptable configuration, the A different and very interesting approach is followed in
objective function is evaluated (e.g., efficiency, cost, etc.) and the Ref. [75]: Computer-aided molecular design of the working fluid is
system with the best performance is selected. The best of a certain combined with the ORC system optimization into a single frame-
set of configurations, however, does not guarantee that the optimal work. Thus, the fluid is synthesized during the optimization pro-
configuration has been revealed. In most cases, though, at least a cedure using several molecular groups (e.g. eCH3,
near-optimal configuration has been obtained [13,60,61]. eCH2e, ¼CH2,¼CHe, etc.). The aim of the optimization is to
In class (b), principles from thermodynamics and other physical determine the optimal combination of the molecular groups and
sciences are applied to obtain targets for the optimal system thermodynamic variables that maximize the power output gener-
configuration. These targets can correspond to upper or lower ated by the ORC for a specified heat source and heat sink. In this
bounds on the best possible configuration and provide vital infor- particular application the synthesis of the system is fixed. The road
mation for improvement of existing configurations. Pinch analysis towards complete SDO with molecular synthesis is open.
is a characteristic example of such a method [62,63].
In class (c), a superstructure is considered with all the possible 6. Optimization in modeling of energy systems and modeling
(or necessary) components and interconnections. An objective for optimization
function is specified and the optimization problem is formulated.
The solution of the optimization problem gives the optimal system One more interesting application of optimization is in the
configuration, which, inevitably, depends on (and is restricted by) development of the model of a system, no matter whether this
the initial superstructure [2,3,43,44,64e67]. model will be used for optimization or not. For example, in Ref. [76]
It is noted that the distinction among the three classes may not particle swarm optimization is applied in order to determine the
be so clear-cut. For example, the targets of class (b) can serve as set of model parameters that minimize the difference between
heuristics or rules in class (a) and they can be embedded in the estimated and experimental data of photovoltaic cells and modules,
optimization procedures of class (c) to the benefit of the whole while in Ref. [77] an algorithm named multi-verse optimizer is
process. applied in order to identify the optimal set of model parameters
Recently, methods with the words “superstructure-free syn- that give polarization curves of a proton exchange membrane fuel
thesis” have appeared. They may not constitute a separate class, but cell that are close to manufacturers data.
they can be classified in class (a) or (b) or a hybrid of those. A few During the numerical solution of the optimization problem, the
examples are the following. objective function may be evaluated tens of thousands of times. For
In Refs. [68,69] the HEATSEP method presented in earlier pub- this purpose, the model of the system, which may consist of hun-
lications [70,71] is further developed for the synthesis and design dreds or even thousands of algebraic and/or differential equations,
optimization of energy systems, in general, and it is applied for the is called by the optimizer. The computational burden can be beyond
optimization of an ORC system. A hybrid evolutionary/traditional the limits of the available computers. Thus, there is often need to
optimization algorithm is used, which is organized in two levels. A develop reduced models. It is written in Ref. [78]:“Whereas suitable
complex original codification of the topology and the intensive methods are well known for linear model reduction, the efficient
design parameters of the system is managed by the upper level reduction of nonlinear models, i.e., the development of reduced
evolutionary algorithm according to the criteria set by the HEATSEP models that are both accurate and computationally beneficial, is
method, which are used to automatically synthesize a “basic” sys- still challenging …
tem configuration from a set of elementary thermodynamic cycles. Other approaches use surrogate models that are tailored to
The lower SQP (sequential quadratic programming) algorithm op- state-of-the-art optimization algorithms … Consequently,
timizes the objective function(s) with respect to cycle mass flow modeling and optimization should not be considered purely
rates only, taking into account the heat transfer feasibility sequential tasks as it is commonly done but rather in an integrated
constraint within the undefined heat transfer section. fashion.”
In Ref. [72] a hybrid algorithm combining evolutionary with
deterministic optimization of thermal power plants is used. The 7. Further considerations and research needs
upper-level evolutionary optimization combined with graph theory
generates structural alternatives, i.e., unit selection and in- 7.1. Global optimization
terconnections. The generated alternatives are optimized with
respect to unit sizing and operation in the lower-level deterministic The type of objective functions and constraints in optimization
optimization. of energy systems result in a multi-modal optimization problem,
A method based on graph theory was presented also in Ref. [73] i.e. a problem with many local optima. Many optimization methods
for the synthesis optimization of energy systems, and applied for locate one local optimum only that depends on the starting point.
the optimization of a combined-cycle plant with two different Stochastic or evolutionary algorithms, such as genetic algorithms
objectives: (i) maximization of the power output for given fuel and particle swarm optimization, search the solution space starting
input, (ii) minimization of the annualized cost of owning and from many points and are more likely to locate the global optimum.
operating the system for given power output. It is interesting to Global optimization is, in general, computationally heavy.
note that the objective function had a very strong effect on the Development of methods which are both effective and efficient is
optimal synthesis of the system. still a challenge [67,79e81].

5.3. Extension of synthesis 7.2. Optimization under uncertainty

As defined in Section 2, synthesis refers to the configuration The values of various parameters in the objective function and/
(structure) of the systems, as it is determined by the components or the constraints, in particular the cost model, may be uncertain.
and their interconnections. The working fluids are selected in Uncertainties exist also in the load profile of a system under design
advance. If there are several candidate fluids, as in ORC systems, the as well as in future prices. In the simplest approach, at least a
system is optimized for each fluid in separate and the best system is parametric study or sensitivity analysis with respect to uncertain
selected [35,74]. parameters should be conducted. In a more systematic way,
 C.A. Frangopoulos / Energy 164 (2018) 1011e1020 1017

uncertainties are taken into consideration during the optimization 7.4. Optimization of/with maintenance
procedure.
Two main approaches exist for optimization under uncertainty: There are several maintenance strategies, e.g. maintenance at
stochastic programming and robust optimization [78]. A thorough fixed and pre-specified periods, condition-based maintenance,
discussion on these subjects and presentation of appropriate reliability-centered maintenance, maintenance after failure (run till
methods appear in Refs. [78,82], while examples of recent appli- failure), etc. Thus, maintenance itself can be the subject of
cations on energy systems such as power plants and cogeneration optimization.
systems are reported in Refs. [33,83e87]. In Ref. [92] for example, a network of compressors operating in
There is significant research activity on this subject. parallel in an air separation plant is considered and optimization of
operation and maintenance schedule is performed simultaneously.
Objective function is the minimization of the total startup, shut-
7.3. Optimization with consideration of system reliability down and power consumption costs of the compressors as well as
the procurement cost of products from external sources. Three
The word ‘reliability’ appears in engineering publications with different maintenance policies are considered: (i) fixed mainte-
various meanings, e.g. reliability of power supply to customers, nance plan, (ii) flexible maintenance plan through the simulta-
reliability in the performance of algorithms, etc. Therefore, it is neous optimization of operational and maintenance tasks, and (iii)
necessary to give the context in which reliability is used here: flexible maintenance plan considering maintenance workforce
limitations. Optimization is performed for policies (ii) and (iii) in
‘Reliability is the probability that an item will perform a required
separate.
function under stated conditions for a stated period of time’ [88].
In Ref. [93], an electric power system is considered and the task
is to minimize the total cost, which consists of maintenance cost,
The usual practice in optimization of energy systems is to solve operation cost, interruption cost, and environmental cost during
the optimization problem under the assumption that the equip- and after the maintenance. In fact, there is no formal optimization,
ment is always available for operation except, perhaps, of pre- but evaluation of the cost for three pre-determined maintenance
specified periods of maintenance. Any modification of the system strategies and the selection of the one with the lowest cost.
in order to obtain redundancy is performed empirically afterwards, In both these publications, the system is fixed. So, there is much
with a consequence the system thus designed to be non-optimal. work to be done in order for maintenance aspects to be included in
A different approach is followed in Ref. [15]: reliability and the synthesis, design and operation optimization of energy systems
availability are introduced in the mathematical model of the system and more so since reliability and maintenance are directly
using the state-space method (SSM) for reliability analysis com- interrelated.
bined with the Intelligent Functional Approach [2,3] for decision
making in case of partial failure of equipment. As an example, the
SDO optimization of a cogeneration system was performed without
and with reliability considerations, which revealed that (i) reli-
7.5. Social aspects in optimization
ability considerations have a strong effect on all three aspects
(synthesis, design, operation) of the system, and (ii) if reliability is
Year after year, the scope of optimization, in other words the
not considered, i.e. if the probability of failure of equipment is
objective, has been broadened: Thermodynamic and economic
ignored, then the economic performance of the system (as
objectives were combined in what is known as ‘thermoeconomics’,
expressed by the net present value) is overestimated.
while later on environmental aspects were considered either in
It is interesting to note that publications on this direction are
separate or included in the objective function together with ther-
still scarce. A few examples are given here.
modynamic and economic aspects [46e50].
In Ref. [89] uncertainties (in weather conditions, internal heat
There is one more step we need to take: inclusion of social as-
sources and indoor set-points) and reliability of chillers are com-
pects. Assessment of projects from the point of view of their effect
bined in order to determine the optimal capacity of chillers in a
on the society have been widely performed with criteria such as job
building. Uncertainties are quantified with the Monte Carlo
creation, general welfare, standard of living, etc. The titles of certain
method, while reliability is quantified with the Markov method.
publications give the impression that social aspects are included in
Minimization of the total annual cost is the objective.
the optimization, but a closer look reveals that these aspects are
The state-space and the continuous Markov methods in com-
assessed after the solution of the optimization problem, for the
bination are used in Ref. [90] in order to determine the optimal
system(s) obtained with optimization.
configuration and design point specifications of a system of boilers
An exception is Ref. [94], where a system of forest-based bio-
and steam turbines supplying a new petrochemical complex with
refineries and biofuel supply chain is studied and a multi-objective
electrical and thermal energy. Two objectives are considered in
optimization is performed with three objective functions: (i)
constructing the Pareto front: minimization of product cost rate
maximization of newly created jobs, (ii) maximization of the net
and minimization of exergy destruction.
present value, and (iii) maximization of the GHG emission savings
In Ref. [91] a steam turbine trigeneration system operating on
compared to the current supply chain.
biomass is optimized with respect to synthesis and design, taking
It seems that applications of this type on systems we are
into consideration the change of available biomass and loads in
studying here (as mentioned in the Introduction) are still missing.1
three periods of the year. Chance-constrained programming and k-
Of course, a prerequisite of such applications is the selection and
out-of-m system modeling are used. Maximization of the economic
perhaps the definition of new quantitative criteria, appropriate for
performance of the system is the objective.
formal optimization.
A study of the available publications reveals that there is still
much work to be done for further development of efficient
methods for the triple optimization (SDO) with reliability consid- 1
Perhaps my literature search was not successful in revealing related publica-
erations and more so for combinations of reliability with other tions. If this is so, I apologize and I would highly appreciate, if the reader could give
aspects such as uncertainty, dynamic conditions, etc. me information about such publications.
1018 C.A. Frangopoulos / Energy 164 (2018) 1011e1020

7.6. Optimization algorithms GA Genetic Algorithm

GEM Grenade Explosion Method
The improvement of existing and development of new algo- GRG Generalized Reduced Gradient
rithms for optimization is a never-ending process, which is sup- HVAC Heating, Ventilating and Air Conditioning
ported by a continuous increase in the performance of computers. IRR Internal Rate of Return
Here are a few examples of algorithms from recent publications LCC Life-Cycle Cost
(the list is not exhaustive). LFOPC Leap-Frog Optimization Program with Constraints
LNG Liquefied Natural Gas
 Parallel AlgoRithms for Global Optimization of coNstrAined MOGA Multi-Objective Genetic Algorithm
grey-box compUTational problems (p-ARGONAUT) [67]. NPGA Niched Pareto Genetic Algorithm
 Particle generating set-complex algorithm (PGS-COM) [95]. NPV Net Present Value
 Co-constrained multi-objective Particle Swarm Optimization ORC Organic Rankine Cycle
(CC-MOPSO) [96]. PSO Particle Swarm Optimization
 Global extremum seeking algorithm (GES) [40]. RSM Response surface method
 Guaranteed Convergence Particle Swarm Optimization [76]. SA Simulating Annealing
 Time-varying acceleration particle swarm optimization (TVAC- SDO Synthesis, Design and Operation
PSO) [87]. SQP Sequential Quadratic Programming
 Differential evolution and particle swarm optimization hybrid SSM State-Space Method
algorithm (DE-PSO) [56]. TLBO Teaching-Learning Based Optimization
 Non-dominated sorting genetic algorithm II (NSGA-II) [97e99]. TPBVP Two-Point Boundary Value Problem
 Gravitational search algorithm (GSA) [100].
 Micro-time variant multi-objective particle swarm optimization References
algorithm (micro-TVMOPSO) [101].
 Self-adaptive Jaya algorithm [102]. [1] Yoda Susumu, editor. Trilemma: three major problems threatening world
survival. Tokyo: Central Research Institute of Electric Power Industry; 1995.
 Niched Pareto genetic algorithm (NPGA) [102]. [2] Frangopoulos CA. Intelligent Functional Approach: a method for analysis and
 Response Surface Method (RSM) [102]. optimal synthesis-design-operation of complex systems. A Future for Energy.
 Leap-frog optimization program with constraints algorithm In: Stecco SS, Moran MJ, editors. Florence world energy research symposium.
Florence, Italy: Pergamon Press; 1990. p. 805e15. Oxford. Published also in
(LFOPC) [102].
International Journal of Energy-Environment-Economics 1991;1(4):
 Teaching-learning based optimization algorithm (TLBO) [102]. 267e274.
 Grenade explosion method (GEM) [102]. [3] Frangopoulos CA. Optimization of synthesis-design-operation of a cogene-
 Multi-objective genetic algorithm (MOGA) [102]. ration system by the intelligent functional approach. A future for energy [2],
pp. 597-609. Publ Int J Energy Environ Econ 1991;1(4):275e87.
 Enhanced multi-objective bee colony optimization algorithm [4] Frangopoulos CA, Sakalis GN, Tzortzis GJ. Intertemporal and dynamic opti-
[103]. mization of synthesis, design and operation of energy systems. In: 4th in-
 Front-based Yin-Yang-Pair optimization algorithm [104]. ternational conference on contemporary problems of thermal engineering.
Katowice, Poland: CPOTE 2016; 14e16 October 2016.
 Multi-objective grey wolf optimizer [104]. [5] Papalambros PY, Wilde DJ. Principles of optimal design: modeling and
 Multiobjective strength firefly algorithm (MOSFA) [105]. computation. second ed. Cambridge, UK: Cambridge University Press; 2000.
[6] Ravindran A, Ragsdell KM, Reklaitis GV. Engineering optimization: methods
and applications. second ed. New York: J. Wiley and Sons, Inc.; 2006.
A detailed description and a comparison among the algorithms [7] Rao SS. Engineering optimization: theory and practice. fourth ed. New York:
are beyond the scope of this manuscript, but it can be found in the John Wiley & Sons; 2013.
aforementioned publications. In Ref. [102], in particular, the self- [8] Stoecker WF. Design of thermal systems. third ed. New York: McGraw Hill,
Inc.; 1989.
adaptive Jaya algorithm is compared with the algorithms NPGA,
[9] Goldberg DE. Genetic algorithms in search, optimization and machine
RSM, LFOPC, TLBO, GEM and MOGA, while in Ref. [104], the front- learning. Reading, Massachusetts: AddissoneWesley; 1989.
based Yin-Yang-Pair optimization algorithm is compared with the [10] van Laarhoven P, Aarts E. Simulated annealing: theory and applications.
Boston: D. Reidel; 1987.
multi-objective grey wolf optimizer and a genetic algorithm.
[11] Freeman JA, Skapura DM. Neural networks. Addison-Wesley; 1992.
[12] Masatoshi Sakawa. Fuzzy sets and interactive multiobjective optimization.
8. Closure New York: Plenum Press; 1993.
[13] Sciubba E, Melli R. Artificial intelligence in thermal systems design: concepts
and applications. Commack, New York: Nova Science Publishers, Inc.; 1998.
Judging from the number of publications, which has being [14] Dimopoulos GG, Frangopoulos CA. Optimization of energy systems based on
increasing exponentially during the last ten or twenty years, and evolutionary and social metaphors. Energy 2008;33(2):171e9.
from the quality of many of those, there is a tremendous and [15] Frangopoulos CA, Dimopoulos GG. Effect of reliability considerations on the
optimal synthesis, design and operation of a cogeneration system. Energy
impressive development in methodology and applications of 2004;29(3):309e29.
mathematical optimization in energy systems. The preceding sec- [16] Kalikatzarakis M, Frangopoulos CA. Multi-criteria selection and thermo-
tions indicate that the field has many areas that still need ‘culti- economic optimization of organic Rankine cycle system for a marine appli-
cation. Int J Therm 2015;18(2):133e41.
vation’ and further ‘growth’ not only in isolation from each other, [17] Allgor RJ, Barton PI. Mixed integer dynamic optimization. Comput Chem Eng
but also in combination of two or more areas. This is extremely 1997;21:451e6.
interesting and challenging, in particular for young researchers. It is [18] Biegler LT, Cervantes AM, Wachter MA. Advances in simultaneous strategies
for dynamic process optimization. Chem Eng Sci 2002;57:575e93.
worth mentioning also that optimization can be a way of thinking [19] Biegler LT, Grossman IE. Retrospective on optimization. Comput Chem Eng
and acting, beyond the particular technological applications. 2004;28:1169e92.
[20] Pontryagin LS, Boltyanskii VG, Gamkrelidze RV, Mishchenko EF. The math-
ematical theory of optimal processes. New York: Interscience; 1962.
Abbreviations
[21] Gelfand IM, Fomin SV, Silverman RA. Calculus of variations. Englewood Cliffs,
New Jersey: Prentice Hall, Inc.; 1963.
COV Calculus of Variations [22] Bellman RE. Dynamic programming. Princeton, NJ: Princeton University
DAE Differential-Algebraic Equation (formulation) Press; 1957. Republished by Dover; 2003.
[23] Ray WH. Advanced process control. New York: McGraw-Hill; 1981.
DP Dynamic Programming [24] Edgar TF, Himmelblau DM. Optimization of chemical processes. New York:
DPB Dynamic Payback Period McGraw-Hill; 1988.
 C.A. Frangopoulos / Energy 164 (2018) 1011e1020 1019

[25] Vassiliadis VS, Sargent RWH, Pantelides CC. Solution of a class of multistage Energy 1980;5:709e18.
dynamic optimization problems. 1. Problems without path constraints. Ind [52] Szargut JT. Optimization of the design parameters aiming at the minimiza-
Eng Chem Res 1994;33(9):2111e22. tion of the depletion of non-renewable resources. Energy 2004;29:2161e9.
[26] Vassiliadis VS, Sargent RWH, Pantelides CC. Solution of a class of multistage [53] Szargut J, Stanek W. Thermo-ecological optimization of a solar collector.
dynamic optimization problems. 2. Problems with path constraints. Ind Eng Energy 2007;32:584e90.
Chem Res 1994;33(9):2123e33. [54] Cohon JL. Multiobjective programming and planning. New York: Academic
[27] Kameswaran S, Biegler LT. Simultaneous dynamic optimization strategies: Press; 1978.
recent advances and challenges. Comput Chem Eng 2006;30:1560e75. [55] Eschenauer H, Koski J, Osyczka A. Multicriteria design optimization: pro-
[28] Renfro JG, Morshedi AM, Asbjornsen A. Simultaneous optimization and so- cedures and applications. Berlin: Springer-Verlag; 1990.
lution of systems described by differential/algebraic equations. Comput [56] Ma Weiwu, Fang Song, Liu Gang. Hybrid optimization method and seasonal
Chem Eng 1987;11(5):503e17. operation strategy for distributed energy system integrating CCHP, photo-
[29] Bock HG, Platt KJ. A multiple shooting algorithm for direct solution of voltaic and ground source heat pump. Energy 2017;141:1439e55.
optimal control problems. In: Proceedings of 9th IFAC world congress; 1984. [57] Ahmadi MH, Ali Jokar MA, Ming T, Michel Feidt M, Pourfayaz D, Astaraei FR.
p. 242e7. Budapest. Multi-objective performance optimization of irreversible molten carbonate
[30] Vallianou VA, Frangopoulos CA. Dynamic operation optimization of a tri- fuel celleBraysson heat engine and thermodynamic analysis with ecological
generation system. Int J Therm 2012;15(4):239e47. objective approach. Energy 2018;144:707e22.
[31] Tzortzis GJ, Frangopoulos CA. Dynamic optimization of synthesis, design and [58] Sadeghi M, Mehr AS, Zar M, Santarelli M. Multi-objective optimization of a
operation of marine energy systems. J Eng Marit Environ 2018. http:// novel syngas fed SOFC power plant using a downdraft gasifier. Energy
journals.sagepub.com/doi/full/10.1177/1475090217749370. 2018;148:16e31.
[32] Rancruel DF, von Spakovsky MR. Development and application of a dynamic [59] Frangopoulos CA, von Spakovsky MR, Sciubba E. A brief review of methods
decomposition strategy for the optimal synthesis/design and operational/ for the design and synthesis optimization of energy systems. Int J Appl
control of a SOFC based APU under transient conditions. In: Proceedings of Therm 2002;5(4):151e60.
the international mechanical engineering congress and expositiondIMECE, [60] Kott AS, May JH, Hwang CC. An Autonomous Artificial designer of thermal
Orlando, Florida, USA, November 5-11. New York: ASME; 2005. paper No. energy systems: Part 1etheoretical considerations, Part 2esolution algo-
82986. rithm. J Eng Gas Turbines Power 1989;111:728e39.
[33] Kim K, von Spakovsky MR, Wang M, Nelson DJ. A hybrid multi-level opti- [61] Sciubba E. Toward automatic process simulators: Part I e modular numerical
mization approach for the dynamic synthesis/design and operation/control procedures, Part II e an expert system for process synthesis. J Eng Gas
under uncertainty of a fuel cell system. Energy 2011;36:3933e43. Turbines Power 1998;120:1e16.
[34] Dimopoulos GG, Kougioufas AV, Frangopoulos CA. Synthesis, design and [62] Linnhoff B, Townsend DW, Boland D, Hewitt GF, Thomas BEA, Guy AR,
operation optimization of a marine energy system. Energy 2008;33(2): Marsland RH. A user guide on process integration for the efficient use of
180e8. energy. Rugby, Warks, England: Institution of Chemical Engineers; 1982.
[35] Kalikatzarakis M, Frangopoulos CA. Thermo-economic optimization of syn- [63] Linnhoff B. Pinch technology for the synthesis of heat and power systems.
thesis, design and operation of a marine organic Rankine cycle system. J Eng J. Energy Resour Technol 1989;111:137e48.
Marit Environ 2017;231(1):137e52. [64] Floudas CA. Nonlinear and mixed-integer optimization: fundamentals and
[36] Bausa J, Tsatsaronis G. Dynamic optimization of startup and load-increasing applications. New York: Oxford University Press; 1995.
processes in power plants - Part I: Method. J Eng Gas Turbines Power [65] Olsommer B, von Spakovsky MR, Favrat D. An approach for time-dependent
2001;123:246e50. - Part II: Application. 123:251-254. thermoeconomic modeling and optimization of energy system synthesis,
[37] Girnik IS, Aristov YI. Dynamic optimization of adsorptive chillers: the design and operation: Part I e methodology and results, Part II e reliability
“AQSOA™-FAM-Z02Water” working pair. Energy 2016;106:13e22. and availability. Int J Appl Therm 1999;2(3):97e114. and 1999;2(4):177-186.
[38] Wang C, Zhao Y, Liu M, Qiao Y, Chong D, Yan J. Peak shaving operational [66] Zhu Q, Luo X, Zhang B, Ying Chen Y. Mathematical modelling and optimi-
optimization of supercritical coal-fired power plants by revising control zation of a large-scale combined cooling, heat, and power system that in-
strategy for water-fuel ratio. Appl Energy 2018;216:212e23. corporates unit changeover and time-of-use electricity price. Energy Convers
[39] Phillips CA, Drake JC. Trajectory optimization for a missile using a multitier Manag 2017;133:385e98.
approach. J Spacecraft Rockets 2000;37(5):653e62. [67] Beykal B, Boukouvala F, Floudas CA, Pistikopoulos EN. Optimal design of
[40] Bizon N. Energy optimization of fuel cell system by using global extremum energy systems using constrained grey-box multi-objective optimization.
seeking algorithm. Appl Energy 2017;206:458e74. Comput Chem Eng 2018. https://doi.org/10.1016/
[41] Hübel M, Meinke S, Andre n MT, Wedding C, Nocke J, Gierow C, Hassela E, j.compchemeng.2018.02.017.
Funkquist F. Modelling and simulation of a coal-fired power plant for start- [68] Toffolo A. A synthesis/design optimization algorithm for Rankine cycle based
up optimization. Appl Energy 2017;208:319e31. energy systems. Energy 2014;66:115e27.
[42] Mua B, Lib Y, House JM, Timothy I, Salsbury TI. Real-time optimization of a [69] Lazzaretto A, Manente G, Toffolo A. SYNTHSEP: a general methodology for
chilled water plant with parallel chillers based on extremum seeking control. the synthesis of energy system configurations beyond superstructures. En-
Appl Energy 2017;208:766e81. ergy 2018;147:924e49.
[43] Munoz JR, von Spakovsky MR. A decomposition approach for the large scale [70] Lazzaretto A, Toffolo A. A method to separate the problem of heat transfer
synthesis/design optimization of highly coupled, highly dynamic energy interactions in the synthesis of thermal systems. Energy 2008;33:163e70.
systems. Int J Appl Therm 2001;4(1):19e33. [71] Toffolo A, Lazzaretto A, Morandin M. The HEATSEP method for the synthesis
[44] Munoz JR, von Spakovsky MR. The application of decomposition to the large of thermal systems: an application to the S-Graz cycle. Energy 2010;35:
scale synthesis/design optimization of aircraft energy systems. Int J Appl 976e81.
Therm 2001;4(2):61e76. [72] Wang L, Voll P, Lampe M, Yang Y, Bardow A. Superstructure-free synthesis
[45] Rech S, Zandarin S, Lazzaretto A, Frangopoulos CA. Design and off-design and optimization of thermal power plants. Energy 2015;91:700e11.
models of single and two-stage ORC systems on board a LNG Carrier for [73] Frangopoulos CA, Grekas DN. Automatic synthesis of mathematical models
the search of the optimal performance and control strategy. Appl Energy using graph theory for optimization of thermal energy systems. Energy
2017;204:221e41. Convers Manag 2007;48(11):2818e26.
[46] Frangopoulos CA. An introduction to environomic analysis and optimization [74] Soffiato M, Frangopoulos CA, Manente G, Rech S, Lazzaretto A. Design opti-
of energy-intensive systems. In: Valero A, Tsatsaronis G, editors. Interna- mization of ORC systems for waste heat recovery on board a LNG Carrier.
tional symposium on efficiency, costs, optimization and simulation of energy Energy Convers Manag 2015;92:523e34.
systems, ECOS '92. New York: ASME; 1992. p. 231e9. Zaragoza, Spain, June [75] White MT, Oyewunmi OA, Haslam AJ, Markides CN. Industrial waste-heat
15-18. recovery through integrated computer-aided working-fluid and ORC sys-
[47] Frangopoulos CA, Bulmetis VA. Optimization of a combined-cycle plant with tem optimisation using SAFT-g Mie. Energy Convers Manag 2017;150:
thermodynamic, economic and environmental considerations. In: 851e69.
Pilavachi PA, editor. Energy efficiency in process technology. London: [76] Nunes HGG, Pombo JAN, Mariano SJPS, Calado MRA, Felippe de Souza JAM.
Elsevier Applied Science; 1992. p. 140e50. Vouliagmeni, Greece, October 19- A new high performance method for determining the parameters of PV cells
22. and modules based on guaranteed convergence particle swarm optimization.
[48] Frangopoulos CA, von Spakovsky MR. A global environomic approach for Appl Energy 2018;211:774e91.
energy systems analysis and optimization - Part I. In: Szargut J, et al., editors. [77] Fathy A, Rezk H. Multi-verse optimizer for identifying the optimal parame-
Energy systems and ecology ENSEC '93; 1993. p. 123e32. Cracow, Poland, ters of PEMFC model. Energy 2018;143:634e44.
July 5-9. [78] Mitsos A, Asprion N, Floudas CA, Bortz M, Baldea M, Bonvin D, Caspari A,
[49] Von Spakovsky MR, Frangopoulos CA. A global environomic approach for Sch€afer P. Challenges in process optimization for new feedstocks and energy
energy systems analysis and optimization e Part II. In: Szargut J, et al., ed- sources. Comput Chem Eng 2018;113:209e21.
itors. Energy systems and ecology ENSEC ‘93; 1993. p. 133e44. Cracow, [79] Grossmann IE, editor. Global optimization in engineering design. Dordrecht:
Poland, July 5-9. Kluwer Academic Publishers; 1996.
[50] Agazzani A, Massardo AF, Frangopoulos CA. Environmental influence on the [80] Floudas CA. Deterministic global optimization: theory, algorithms and ap-
thermoeconomic optimization of a combined plant with NOx abatement. plications. Dordrecht: Kluwer Academic Publishers; 2000.
J Eng Gas Turbines Power 1998;120:557e65. [81] Floudas CA, Gounaris CE. A review of recent advances in global optimization.
[51] Szargut J. International progress in second law analysis. Second law analysis J Global Optim 2008;45(1):3e38.
of energy devices and processes, Washington, D.C., 14-16 August 1979. [82] Sahinidis NV. Optimization under uncertainty: state-of-the-art and
1020 C.A. Frangopoulos / Energy 164 (2018) 1011e1020

opportunities. FOCAPO 2004;28(6e7):971e83. 2003 Special Issue. [95] Scaccabarozzi R, Gatti M, Martelli E. Thermodynamic analysis and numerical
[83] Sy CL, Aviso KB, Ubando AT, Tan RR. Target-oriented robust optimization of optimization of the NET Power oxy-combustion cycle. Appl Energy
polygeneration systems under uncertainty. Energy 2016;116:1334e47. 2016;178:505e26.
[84] Charitopoulos VM, Dua V. A unified framework for model-based multi- [96] Soheyli S, Mayam MHS, Mehri Mehrjoo M. Modeling a novel CCHP system
objective linear process and energy optimisation under uncertainty. Appl including solar and wind renewable energy resources and sizing by a CC-
Energy 2017;186:539e48. MOPSO algorithm. Appl Energy 2016;184:375e95.
[85] Gholaminezhad I, Jafarpur K, Paydar MH, Karimi G. Multi-scale multi- [97] Wang R, Li G, Ming M, Wu G, Wang L. An efficient multi-objective model and
objective optimization and uncertainty analysis of methane-fed solid oxide algorithm for sizing a standalone hybrid renewable energy system. Energy
fuel cells using Monte Carlo simulations. Energy Convers Manag 2017;153: 2017;141:2288e99.
175e87. [98] Lee S, Esfahani IJ, Ifaei P, Moya W, Yoo CK. Thermo-environ-economic
[86] Urbanucci L, Testi D. Optimal integrated sizing and operation of a CHP sys- modeling and optimization of an integrated wastewater treatment plant
tem with Monte Carlo risk analysis for long-term uncertainty in energy with a combined heat and power generation system. Energy Convers Manag
demands. Energy Convers Manag 2018;157:307e16. 2017;142:385e401.
[87] Shaabani YA, Seifi AR, Kouhanjani MJ. Stochastic multi-objective optimiza- [99] Ahmadi MH, Ahmadi MA, Pourfayaz F, Hosseinzade H, Acıkkalp E, Tlili I,
tion of combined heat and power economic/emission dispatch. Energy Feidt M. Designing a powered combined Otto and Stirling cycle power plant
2017;141:1892e904. through multi-objective optimization approach. Renew Sustain Energy Rev
[88] Villemeur A. Reliability, availability, maintainability and safety assessment. 2016;62:585e95.
In: Vol. 1: methods and techniques. Vol. 2: assessment, hardware, software [100] Naserbegi A, Aghaie M, Minuchehr A, Alahyarizadeh Gh. A novel exergy
and human factors. Chichester: John Wiley & Sons; 1992. optimization of Bushehr nuclear power plant by gravitational search algo-
[89] Gang W, Wang S, Xiao F. Gao D-ce. Robust optimal design of building cooling rithm (GSA). Energy 2018;148:373e85.
systems considering cooling load uncertainty and equipment reliability. Appl [101] Rey A, Zmeureanu R. Multi-objective optimization framework for the se-
Energy 2015;159:265e75. lection of configuration and equipment sizing of solar thermal combisys-
[90] Rad MP, Manesh MHK, Rosen MA, Amidpour M, Hamedi MH. New procedure tems. Energy 2018;145:182e94.
for design and exergoeconomic optimization of site utility system consid- [102] Rao RV, More KC. Design optimization and analysis of selected thermal de-
ering reliability. Appl Therm Eng 2016;94:478e90. vices using self-adaptive Jaya algorithm. Energy Convers Manag 2017;140:
[91] Andiappan V, Tan RR, Aviso KB, Ng DKS. Synthesis and optimisation of 24e35.
biomass-based tri-generation systems with reliability aspects. Energy [103] Zhou J, Lu P, Li Y, Wang C, Yuan L, Mo L. Short-term hydro-thermal-wind
2015;89:803e18. complementary scheduling considering uncertainty of wind power using
[92] Kopanos GM, Xenos DP, Cicciotti M, Pistikopoulos EN, Thornhill NF. Opti- an enhanced multi-objective bee colony optimization algorithm. Energy
mization of a network of compressors in parallel: operational and mainte- Convers Manag 2016;123:116e29.
nance planning e the air separation plant case. Appl Energy 2015;146: [104] Punnathanam V, Kotecha P. Multi-objective optimization of stirling engine
453e70. systems using front-based Yin-Yang-pair optimization. Energy Convers
[93] Shayesteh E, Hilber JYP. Maintenance optimization of power systems with Manag 2017;133:332e48.
renewable energy sources integrated. Energy 2018;149:577e86. [105] Wang H, Zhang R, Peng J, Wang G, Liu Y, Jiang H, Liu W. GPNBI-inspired
[94] Cambero C, Sowlati T. Incorporating social benefits in multi-objective opti- MOSFA for Pareto operation optimization of integrated energy system. En-
mization of forest-based bioenergy and biofuel supply chains. Appl Energy ergy Convers Manag 2017;151:524e37.
2016;178:721e35.