Evaluation of natural gas supply options for Southeast and Central Europe: Part 2. Multi-criteria assessment

Abstract

Decision making methods are used as a tool for the selection of alternatives to be evaluated on the basis of several criteria. Evaluation of the potential routes for natural gas supply to the Southeast and Central European countries is studied using single and multi-criteria evaluation. The potential options included in this analysis are the: Yamal Route; Nabucco Route; West Balkan Route; LNG Neum Route; and Gas by Wire Route. In part 1, the paper was devoted to the definition of the indicators and to single indicator analysis. In part 2, the analysis is based on multi-criteria evaluation, which comprises the possibility to assess the options under predefined con-straints amongst indicators. The paper also describes different methods used for multi-criteria evaluation. This analysis focuses on cases with different priorities defined among individual weighting coefficients with the others having the same value. It was shown that all options under specific constraints can be qualified as promising in the decision making process. It is also concluded that favourite indi-cator constraints may exaggerate some options.

Full text

Page 1

Evaluation of natural gas supply options for Southeast
and Central Europe: Part 2. Multi-criteria assessment
Naim H. Afgana, Maria G. Carvalhob, Petros A. Pilavachic, Nelson Martinsd,*
aTechnical University of Lisbon, Av. Rovisco Pais 1049-001, Lisbon, Portugal
bBEPA – Bureau of European Policy Advisers, European Commission, Brussels, Belgium
cUniversity of Western Macedonia, Kozani, Greece
dUniversity of Aveiro, Engenharia Meca ˆnica, Campus Santiago, 3810-193 Averio, Portugal
Received 3 June 2007; accepted 14 January 2008
Available online 7 March 2008
Abstract
Decision making methods are used as a tool for the selection of alternatives to be evaluated on the basis of several criteria. Evaluation
of the potential routes for natural gas supply to the Southeast and Central European countries is studied using single and multi-criteria
evaluation. The potential options included in this analysis are the: Yamal Route; Nabucco Route; West Balkan Route; LNG Neum
Route; and Gas by Wire Route. In part 1, the paper was devoted to the definition of the indicators and to single indicator analysis.
In part 2, the analysis is based on multi-criteria evaluation, which comprises the possibility to assess the options under predefined con-
straints amongst indicators. The paper also describes different methods used for multi-criteria evaluation. This analysis focuses on cases
with different priorities defined among individual weighting coefficients with the others having the same value. It was shown that all
options under specific constraints can be qualified as promising in the decision making process. It is also concluded that favourite indi-
cator constraints may exaggerate some options.
? 2008 Elsevier Ltd. All rights reserved.
Keywords: Natural gas route evaluation; Decision making methods; Southeast Europe
1. Introduction
Decision making methods are used as a tool for selec-
tion and ranking of alternatives on the basis of an eval-
uation with several criteria. Extensive studies devoted to
the overview of multi-criteria decision methods, applied
to the analysis of sustainable energy systems can be
found at Refs. [1,2]. Decisions are, most times, based
on a trade off or compromise amongst a number of crite-
ria, which are in conflict with each other. These methods
provide a better understanding of the inherent features of
decision problems, promote the role of participants in
decision making processes, facilitate compromise and col-
lective decisions and provide a good platform for under-
standing the perception of models by analysts in a
realistic scenario. A single criterion approach, aimed to
identify the most energy efficient supply option at low
cost has been quite popular. However, in the 1980s, a
growing environmental awareness has slightly modified
the above decision framework [3]. The need to incorpo-
rate environmental and social considerations in energy
planning resulted in an increasing use of multi-criteria
approaches.
Multi-attribute decision making and the multi-objective
decision making are among the multi-criteria decision mak-
ing methods. Multi-objective linear programming is a plan-
ning methodology used for illustrating the trade off
between environmental and economic parameters and for
assisting in the selection of a compromise solution [4–6].
This was popular in energy planning with conventional
fuels in the 1970s.
0196-8904/$ - see front matter ? 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enconman.2008.01.024
*Corresponding author. Tel.: +351 234378171; fax: +351 234370953.
E-mail address: nmartins@mec.ua.pt (N. Martins).
www.elsevier.com/locate/enconman
Available online at www.sciencedirect.com
Energy Conversion and Management 49 (2008) 2345–2353

Page 2

Nowadays, renewable energy sources are being pro-
moted for a wide variety of applications worldwide. This
compels planners and decision makers to identify the bar-
riers for penetration and suggest interventions to overcome
them. The role of different actors in decision making, thus,
becomes important. Methods of group decisions are, there-
fore, of primary interest for implementation of the decision
sciences in real life problems.
1.1. Multi-criteria decision making (MCDM)
Multi-criteria decision making is a well known branch of
decision making. It is a branch of a general class of opera-
tional research models, which deal with decision problems
under the presence of a number of criteria. This major class
of models is very often called MCDM. This is further
divided into multi-objective decision making (MODM)
and multi-attribute decision making (MADM) methods
[7]. Multi-criteria decision making (MCDM) methods deal
with the process of making decisions in the presence of
multiple objectives. A decision maker is required to choose
among quantifiable or non-quantifiable and multiple
criteria.
There are several methods in each of the above catego-
ries. Priority based, outranking, distance based and mixed
methods are also applied to various problems. Each
method has its own characteristics, and the methods can
also be classified as deterministic, stochastic and fuzzy.
There may be combinations of the above methods.
Depending upon the number of decision makers, the meth-
ods can be classified as single or group decision making
methods. Decision making under uncertainty and decision
support systems are also prominent decision making tech-
niques [8].
1.2. The elimination and choice translating reality
(ELECTRE)
The ELECTRE method is capable of handling both
quantitative and qualitative discrete criteria and provides
complete ordering of the alternatives. The problem is to
be formulated in such a way that it chooses alternatives
that are preferred over most of the criteria and that do
not cause an unanticipated level of discontent for any
of the criteria. Concordance, discordance indices and
threshold values are used in this method. Based on these
indices, graphs for strong and weak relationships are
developed. These graphs are used in an iterative proce-
dure to obtain the ranking of the alternatives [9]. This
index is defined in the range 0–1, provides a judgment
on the degree of credibility of each outranking relation
and represents a test to verify the performance of each
alternative. Finally, the ELECTRE method yields a whole
system of binary outranking relations between the alterna-
tives. Because the system is not necessarily complete, the
ELECTRE method is sometimes unable to identify the
preferred alternative. It only produces a core of leading
possibilities by eliminating lesser alternatives in a decision
making problem [10].
1.3. Preference ranking organization method for enrichment
evaluation (PROMETHEE)
The PROMETHEE method uses the outranking prin-
ciple to rank the alternatives and combines ease of use
and decreased complexity. It performs a pair wise com-
parison of the alternatives in order to rank them with
respect to a number of criteria. Brans et al. [11] have
offered six generalized criteria functions for reference,
namely, usual criterion, quasi-criterion, criterion with
linear preference, level criterion, criterion with linear pre-
ference and indifference area and Gaussian criterion. The
method uses a preference function Pj(a,b), which is a
function of the difference dj between two alternatives for
any criterion j, i.e. dj = f(a,j) ? f(b,j), where f(a,j) and
f(b,j) are values of the two alternatives a and b for crite-
rion j. The indifference and preference thresholds q0and p0
are also defined depending upon the type of criterion
function.
1.4. Analytical hierarchy process (AHP)
The analytical hierarchy process was developed by Saaty
[12,13]. The essence of the process is the decomposition of a
complex problem into a hierarchy with goal (objective) at
the top of the hierarchy, criteria and sub-criteria at levels
and sub-levels of the hierarchy and decision alternatives
at the bottom of the hierarchy. The elements at a given
hierarchy level are compared in pairs to assess their relative
preferences with respect to each of the elements at the next
higher level. The procedure is repeated upwards for each
level until the top of the hierarchy is reached. The overall
weight coefficient with respect to the goal for each decision
alterative is then obtained. The alterative with the highest
weight coefficient value should be taken as the best alterna-
tive. One of the major advantages of the AHP is that it cal-
culates the inconsistency index as a ratio of the decision
maker’s inconsistency and randomly generated index. This
index is important for the decision maker to assure him
that his judgments were consistent and that the final deci-
sion is well made.
1.5. Multi-attribute utility theory (MAUT)
Multi-attribute utility theory takes into consideration
the decision makers’ preference in the form of the utility
function that is defined over a set of attributes. The utility
valuecanbedetermined
utility functions followed by verification of preferential
and utility dependent conditions and derivation of
multi-attribute utility functions. The utility function can
be either additively separable or multiplicatively separable
with respect to single attribute utility. The multiplicative
equation for the utility value is defined as follows:
throughsingleattribute
2346
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353

Page 3

1 þ kuðx1;x1;x1...xnÞ ¼
Y
n
j¼1
ð1 þ kkjujðxjÞÞð1Þ
where j is the index of each attribute, k is the overall scaling
constant (greater than or equal to ?1), u(...) is the overall
utility function operator, kk, is the scaling constant for
attribute j and uj(...) is the utility function operator for
each attribute [14].
2. Multi-criteria evaluation of natural gas supply options
2.1. ASPID – analysis and synthesis of parameters under
information deficiency method
‘‘Uncertainty”: ‘‘ambiguity,” ‘‘fuzziness” and other sim-
ilar concepts are quite usual in multi-criteria evaluation of
real large scale systems, time long projects, variants of cru-
cial financial decisions and other complex objects. Thus, it
seems rather natural to use the flexible language and appa-
ratus of fuzzy sets theory for a new multi-criteria decision
making technique expounding and representing the tech-
nique’s computer realization – a decision support system
shell (DSSS) ‘‘ASPID-3W” (analysis and synthesis of
parameters under information deficiency), which operates
in Windows 95 and is widely used to evaluate different
complex objects under uncertainty [15,16].
The decision support system shell (DSSS) ‘‘ASPID-3W”
is based on the following reasoning. It is supposed that
complex objects are identified with vectors x(j) = (x1(j),...,
xn(j)), x1(j) ? E1x(j) ? En, where x1(j) is a value of the ith
initial parameter xi for the jth complex object, i =
1,...,n, j = 1,...,k (where k is the number of objects under
investigation). In other words, the jth complex object is
identified with a vector x(j) = (x1(j),...,xn(j)), which is a
value of the n-variable vector x0= (x1,...,xn) of the initial
parameters. All objects under consideration compose a
finite set X = {x(j), j = 1,...,k}.
The subject of evaluation is complex objects quality (effi-
ciency, reliability, productivity and others). Taking into
account that the main goal of objects evaluation is decision
making, that is, choosing the most preferable object(s), we
can use the word ‘‘preferability” for any quality under eval-
uation. This preferability of complex objects is evaluated
by a number of criteria (e.g. qi; i = 1,...,m) with each of
them being a function of the initial parameters vector
xi= (x1,x2,...,xn),
q = q(xi) = q(x1? xn),
Here, we deal with the simple case where every particular
criterion qiis a function of only one relevant initial param-
eter xi:q = q(xi), i = 1,...,m = n.
A function qi= (xi) may be treated as a particular mem-
bership function of a fuzzy set of objects that are preferable
from the point of the ith criterion’s. Without loss in gener-
ality, it can be supposed that all particular criteria are nor-
malized, that is, the ith criterion meets the inequality
1 P q P 0. As this normalization takes place, the minimal
value q(j) = 0 of the ith criterion is correlated with object
x(j), which has minimal preference, and the maximal value
i = 1,...,m.
qi(j) = 1 is correlated with an object x(j), which has maxi-
mal preference. Thus, the membership function sets up a
correlation between an object x(j) = (xi(j), ...,xm(j)) from
the set X and its multi-criteria estimation q(x(j)) =
q(x1(j),...,xn(j)), q(j) = (q1(j),...,qm(j)) where qi(j) = qi(x),
i = 1...,m, j = 1...,k.
The fuzzy sets synthesis technique may be interpreted
in probabilistic terms. An uncertain choice of a weight
vector w = (w1,w2,...,wm) from the set W(I;m,n) may
be modelled by a random choice, which is determined
by the uniform distribution of the set. Such randomiza-
tion of uncertainty gives random weight coefficients
w1(I),..., wn(I). The fuzzy sets synthesis technique out-
lined is a choice of a synthesizing (aggregative) function.
Usually, a modification of the generalized weighted
mean:
QuðqÞ ¼ Quðq;wÞ ¼ Quðq1...qm;w1...wmÞ
¼ u?1X
is selected as an aggregative function. In this formula, u is
an arbitrary monotonically increasing continuous function,
and w = (w1,...,wm), w P 0, w1+ w2+ ??? + wm= 1, is a
vector of weight coefficients w1,w2,...,wm(weight vector).
A weight coefficient wiis a measure of the relative signif-
icance of the corresponding particular criterion qi for
aggregative estimation Q(q(j)) of the general preferability
of an object x(j).
The generalized weighted mean Q = (q,w) is a very flex-
ible tool and many practically wide spreading synthesizing
functions may be treated as its special case. That is, if u is
selected as an exponential function u(z) = zk, z > l, k > 0,
then we have an exponential weighted mean:

m
i¼1
wiuðqiÞð2Þ
Qkðq;wÞ ¼
X
m
i¼1
wiqk
i
!1
k
ð3Þ
Interpretation of the mathematical tool presented in this
section can be outlined in the form of the following
procedure.
What is a base for generating the most popular synthe-
sizing function?
That is, if k = 1, then Q(q,w) transforms into an additive
function (weighted arithmetic mean):
X
The General Index is formed through the following
procedure:
Q ¼ ðq;wÞ ¼
m
i¼1
wiqi
ð4Þ
1. Formation of vectors x = (x1,...,xm) of all input attri-
butes (characteristics) that are necessary for full evalua-
tion of the quality of the options under consideration.
Attributes are expressed by four groups of indicators:
resource indicators, environmental indicators, social
indicators and economic indicators.
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353
2347

Page 4

2. Formation of vectors of specific criteria q = (q1,...,qm)
by which input attributes (indicators) x1,...,xmare to
be evaluated.
3. Introducing weighting factors for each indicator, in the
multi-criteria assessment, the General Sustainability
Index for the options under consideration is defined.
With multiplication of individual indicators with their
respective weighting coefficients, the participation of
individual criteria is expressed. The additive function
of the product of indicators with their respective weight-
ing coefficient gives the General Index, as per Eq. (3).
4. In order to define the weight coefficient vector, random-
ization of uncertainty is introduced. Randomization
produces stochastic realizations from corresponding sets
of functions and a random weight vector. It is assumed
that the measurement of the weight coefficients is accu-
rate to within a step h = 1/n, with n a positive integer. In
this case, the infinite set of all possible vectors may be
approximated by the finite set W(m,n) of all possible
weight vectors with discrete components. In our case,
m = 5, and n = 40 will be used so that the total number
of elements of the set W(m,n) is N(m,n) = 92,251.
5. As the final result of this procedure, the list of priorities
measured by the General Index of options under consid-
eration is obtained.
2.2. Multi-criteria analysis of potential options for natural
gas supply
An individual criterion for evaluation of the potential
gas transport options is leading to a limited guidance for
the respective decision making process. In this respect, indi-
vidual indicators are leading to the priorities of specific
options, which will strongly depend on the selected
indicator.
It is of interest to investigate the effect of multi-indica-
tors in order to establish the priorities of the options under
consideration. A trivial solution can be obtained if it is
assumed that all indicators equally participate in validation
of an individual option. Another possibility is that the rat-
ing among options is established from the most probable
combination of the weighting factors in all potential com-
binations. For multi-criteria analysis, there are a large
number of combinations describing potential situations of
weighting factors. In order to overcome this arbitrariness
in the evaluation of these options, the multi-criteria evalu-
ation method is introduced [15,16,18]. This method is
based on the numerical values of the indicators used in
the multi-criteria analysis and is presented in Table 1 [17].
The multi-criteria assessment method is based on the
decision making procedure [17] reflecting the combined
effect of all criteria under consideration and is expressed
in the form of a General Index of Sustainability. A selected
number of indicators is taken as the measure of the criteria
comprising specific information of the options under con-
sideration [19]. The procedure aims to express the property
of the options by a respective set of indicators, (Fig. 1).
The first step in the preparation of the data for the
multi-criteria sustainability assessment is the numerical
allocation of the data. This step consists in the formation
of particular membership functions q1(x1),...,qm(xm).
For every Indicator xi, we have to (1) fix two values
MIN(i), MAX(i); (2) indicate if function qi(xi) is decreasing
Table 1
Indicators
Option Indicators
Designation Capacity
Bcm/yr
Length
km
Environment
103t/yr
NG Cost
Euro/103m3
Transport Cost
Euro/103m3
Investment
Euro/m3/yr
NG Demand
103m3/capita/yr
1
2
3
4
5
Yamal Gas Route
Nabucco Project
West Balkan Route
LNG Terminal Neum
Gas by Wire
30
20
15
10
15
4000
3600
3400
600 + 1600
5000
9930
6400
3500
2380
3250
90.47
37.00
37.00
47.20
37.00
40.50
79.00
96.60
38.95
73.60
0.91
0.50
0.60
0.15
0.24
1.00
0.65
0.31
0.53
0.437
General
Index
GI
Environ.
Indicator
EI
NG
Cost
Indicator
NGCI
NG
Transport
Indicator
NGTI
Investment
Indicator
II
NG
Demand
NGD
Fig. 1. Graphical presentation of indicators.
2348
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353

Page 5

or increasing with argument xiincreasing; (3) choose the
exponent value k in equation:
if xi6 MINðiÞ;
MAXðiÞxi?xi
MAXðiÞ?MINðiÞ
0;
if xi> MAXðiÞ
qiðxiÞ ¼
1;
??k;
if MINðiÞ < xi6 MAXðiÞ;
8
>
>
>
>
:
<
ð5Þ
for the decreasing function qi(xi).
The functions q1(x1),...,qm(xm) are calculated and pro-
duce a matrix ðqðjÞ
where an element qðjÞ
i
is a value of the ith particular crite-
rion for the j th option. In this analysis, it is assumed that
linear functions q1(x1),...,qm(xm) are used.
Decreasing functions are adopted for the q1, q2, q3and
q4membership functions.
In Table 2, the values of these functions q1(x1),...,
qm(xm) are presented.
In order to define the weight coefficient vectors, random-
ization of uncertainty is introduced. It is assumed that the
weight coefficients are accurate to within a step of h = 1/n
where n is a positive integer. In this case, the infinite set of
all possible vectors may be approximated by the finite set
W(m,n) of all possible weight vectors with discrete compo-
nents. In the present case, m is set to 5 and n is set to 40 so
that the total number of elements N(m,n) of set W(m,n) is
135751.
Non-numeric, inexact and incomplete information is
used for reduction of the set W(m,n) of all possible vectors
w to obtain the discrete components set W(I,n,m). It is
defined as a number of constraints reflecting non-numeric
iÞ having i = 1,...,m and j = 1,...,k,
information about the mutual relations among the criteria
under consideration.
3. Results
3.1. Selection of cases
For evaluation of any complex system by the ASPID
method, appropriate parameters, needed for its application
have to be selected.
As a fist step we have to select a number of criteria and
respective indicators. In this analysis, m = 5 will be used.
Randomization of the weight coefficient will need the deci-
sion of scale n = 40. Finally, we have to select constraints
to impose on the analysis and evaluation of the system
under consideration.
In this exercise, two groups of constraints are defined,
namely: a group with priority given to the single indicator,
while the other indicators have equal values; and a second
group with all indicators defined by the internal preference
amongst the criteria.
3.1.1. Group 1
In evaluation of the priority list amongst the alterna-
tive options, some cases have to be selected to represent
potential constraints between the indicators. This means
that the cases that are to be representative for the deci-
sion making procedure have to be defined. The evalua-
tion procedure implies that constraints have to be
defined among the options in order to obtain the respec-
tive values of their weighting coefficients [20–22]. In this
evaluation, attention will be focused on a number of
cases to be analyzed that correspond to the individual
priority of every indicator with the other indicators hav-
ing the same value. In this analysis, the following cases
are taken.
3.1.1.1. Case 1: EI > NGC = NGT = INV = NGD. Case 1
(Fig. 2) is designed with priority given to the Environmen-
tal Indicator while the other indicators have the same value
of weighting coefficient. For this case, it is noticed that
priority is obtained for the LNG Terminal Neum and the
Gas by Wire options followed by the West Balkan and then
the Nabucco options. One characteristic of this case is that
the Yamal Gas Route option is at the lowest position in the
priority list.
Fig. 2. Weight coefficient and General Index for Case 1.
Table 2
Normalized indicators
Option EnvironmentNG
Cost
Transport
Cost
Investment
cost
NG
Demand
Yamal
Route
Nabucco
Route
West Balkan
Route
LNG
Terminal
Neum
Gas by Wire
0.000 0.000 0.3680.000 1.000
0.264 0.8060.809 0.485 0.636
0.7870.8061.0000.2790.000
0.989 0.5630.000 1.0000.381
0.8320.806 0.6980.9470.189
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353
2349

Page 6

3.1.1.2. Case 2: NGC > EI = NGT = INV = NGD. Case 2
(Fig. 3) is characterized by priority given to the Natural
Gas Cost Indicator. It is noticed that the Gas by Wire,
the Nabucco and the West Balkan options have similar pri-
orities. The LNG Terminal Neum option has a slightly
lower priority while the Yamal Route option is much
lower.
3.1.1.3. Case 3: NGT > EI = NGC = INV = NGD. In Case
3 (Fig. 4), priority is given to the Natural Gas Transport
Cost indicator. Here, the West Balkan, the Nabucco and
the Gas by Wire options have the highest priority. It can
be noticed that the dispersion for some options of this
group is smaller than for those of Case 2.
3.1.1.4. Case 4: INV > EI = NGC = NGT = NGD. Case 4
(Fig. 5) is characterized with priority given to the Invest-
ment Indicator. The results obtained focus our attention
on the Gas by Wire and on the LNG Terminal Neum
options,whichhavesubstantially
General Index. It should also be mentioned that this case
is very probable among all combinations taken into
consideration.
highervaluesof
3.1.1.5. Case 5: NGD > EI = NGC = NGT = INV. Case 5
(Fig. 6) is designed with priority given to the Natural
Gas Demand Indicator. Here, there is very high dispersion
of the General Index values for all options. This implies
that this situation takes into account very different combi-
Fig. 4. Weight coefficient and General Index for Case 3.
Fig. 5. Weight coefficient and General Index for Case 4.
Fig. 6. Weight coefficient and General Index for Case 5.
Fig. 3. Weight coefficient and General Index for Case 2.
2350
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353

Page 7

nations within the set of all combinations being considered
in this evaluation.
3.1.2. Group 2
Group 2 cases are aimed to emphasise the role of the
cases when internal priorities amongst them are defined
by ordinal information that defines the mutual relations
of the criteria and the respective indicators. It is obvious
that the number of such cases can be very much larger
than that of Group 1. The cases are formed by ordering
the criteria, always keeping another criterion at the first
position. In this group, the results are presented as Cases
6–10.
3.1.2.1. Case 6: EI > NGC > NGT > INV > NGD. Case 6
(Fig. 7) is designed with the aim to give priority to the
Environmental Indicator with the other indicators having
the priorities shown in Fig. 7. It is of interest to notice that
this constraint amongst the indicators leads to marginal
differences amongst the Gas by Wire, West Balkan and
LNG Terminal Neum. It also becomes obvious that by
introducing constraints amongst the indicators by priority
gives a smaller dispersion of the weight coefficients.
3.1.2.2. Case 7: NGC > EI > NGT > INV > NGD. Case 7
(Fig. 8) is defined with priority given to the NG Cost Indi-
cator. As for Case 6, the Gas by Wire and the West Balkan
options have priority in comparison to the Nabucco Pro-
ject and the LNG Terminal Neum options. There is also
a substantial decrease in the dispersion of the General
Index.
3.1.2.3. Case 8: NGT > EI > NGC > INV > NGD. Case 8
(Fig. 9) is characterized with priority given to the Gas
Transport Indicator. This case shows that only marginal
differences exist compared with Case 7. Again, the West
Balkan Route and the Gas by Wire options have priority
in comparison with the other options.
3.1.2.4. Case 9: INV > EI > NGC > NGT > NGD. If the
Investment Cost Indicator (Fig. 10) has priority compared
to the other indicators hierarchically ordered, then the Gas
by Wire and LNG Terminal Neum options get marked pri-
ority. The West Balkan Route and the Nabucco Project
occupy second place on the priority list.
3.1.2.5. Case 10: NGD > EI > NGC > NGT>INV. Finally,
in Case 10 (Fig. 11), priority is given to the NG Demand
Indicator. As can be noticed, in this case, there is no sub-
stantial difference in the General Index amongst the
options under consideration. It is shown that if priority is
given to the NG Demand and the Environmental
Fig. 7. Weight coefficient and General Index for Case 6.
Fig. 8. Weight coefficient and General Index for Case 7.
Fig. 9. Weight coefficient and General Index for Case 8.
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353
2351

Page 8

Indicator, the difference in the General Index is marginal,
with a slight advantage for the Nabucco Project.
4. Discussion of the results
Two groups of cases were considered in this analysis.
One group (Cases 1–5) was designed with priority given
to individual indicators, while the other indicators had
the same value, and the second group (Cases 6–10) con-
sisted of cases where the priorities were defined always
keeping one criterion in first position, while the priorities
of the other criteria were defined by ordinal information
defining the mutual relations of the criteria and the respec-
tive indicators.
In the fist group, the number of cases analyzed is rather
limited and will represent only a partial number of the
potential cases amongst the indicator constraints. By
assessing all five cases, it is concluded that depending on
the constraints imposed amongst the indicators, different
rating lists among alternative options are obtained. It can
be noticed that Case 1 (EI highly weighted) and Case 4
(INV highly rated) provide essentially the same discrimina-
tion between options. The same applies for Case 3 (NGT
highly weighted) and Case 5 (NGD highly weighted) where
discrimination does not appear to be as good as that using
Cases 1 and 4.
The second group comprises cases with hierarchical con-
straints, with changing priority in constraints in each case.
Amongst these cases, priority is obtained for Gas by Wire
in Case 6 (EI highly weighted), Case 7 (NG Cost highly
rated) and Case 9 (INV highly weighted).
In this analysis, it is assumed that the results obtained
for each case have equal importance for the decision mak-
ing evaluation. In the final rating list based on the summa-
tion of values determined by the position of each
alternative option, the final Option List Rating can be
established. Therefore, the following rating list among
options is obtained (Table 3).
Even if this type of analysis contains arbitrariness in the
evaluation of the priorities among the alternative options,
it is noticed that the Gas by Wire option and the LNG Ter-
minal Neum option are the best choices under the con-
straints used. By increasing the number of cases to be
analyzed, a better result for decision making should be
obtained. It should also be noticed that, in this type of eval-
uation, further improvement of the data might lead to
higher quality results.
5. Conclusions
This paper is devoted to priority ratings amongst
selected alternative options of gas transport systems in
Southeast and Central Europe. The evaluation is based
on the multi-criteria decision making procedure. The pri-
mary goal of this analysis is to use the method based on
non-numerical information as the constraint for the design
of the options under consideration.
Cases in the first group considered combinations where
individual indicators are given priority, while the other
indicators have the same weight coefficient. The second
group of cases are obtained by hierarchically rated indica-
tors used in the design of constraints amongst the indica-
Fig. 11. Weight coefficient and General Index for Case 10.
Table 3
Option rating
OptionRate
1
2
3
4
5
Gas by Wire
LNG Terminal Neum
Nabucco Route
West Balkan Route
Yamal Route
21
25
28
30
41
Fig. 10. Weight coefficient and General Index for Case 9.
2352
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353

Page 9

tors. Even if this analysis is based on a limited number of
cases, it is noticed that the priority on the final rating list
is the result of the respective relations amongst the criteria
under consideration.
It is of interest to notice that the Gas by Wire and the
LNG Terminal Neum options are the most attractive solu-
tions obtained through evaluation of the alternative
options for the final decision. Obviously, one of the major
deficiencies of this analysis is that most data used are
derived from the available literature and may not be suffi-
ciently accurate.
References
[1] Haung JO, Poh KL, Ang BW. Decision analysis in energy and
environmental modelling. Energy 1995;20:843–55.
[2] Poher SD, Ramachandran M. Application of multi-criteria decision
making to sustainable energy planning. Renew Sustain Energy Rev
2004;8:365–81.
[3] Samouilidis J, Mitropoulos C. Energy economy models: a survey. Eur
J Oper Res 1982;11(3):222–32.
[4] Meirer P, Mubayi V. Modeling energy–economic interactions in
developing countries – a linear programming approach. Eur J Oper
Res 1983;13:41–59.
[5] Kavrakoglu I. Multi-objective strategies in power system planning.
Eur J Oper Res 1983;12(2):159–70.
[6] Schulz V, Stehfest H. Regional energy supply optimization with
multiple objectives. Eur J Oper Res 1984;17:302–12.
[7] Climenco J, editorMulti-criteria analysis. New York: Springer Verlag;
1997.
[8] Gal T, Hanne T, editorsMulti-criteria decision-making: advances in
MCDM models,algorithms,
York: Kluwer Academic Publisher; 1999.
theory,and applications. New
[9] Roy B. Multicriteria methodology to aid the de ´cision. Collection
Gestion. Paris: Economica; 1985.
[10] Goicoecheva A, Hansen D, Duckstain L. Introduction to multi
objective analysis with engineering and business application. New
York: Wiley; 1982.
[11] Brans JP, Vincke Ph, Marechal B. How to select and how to rank
projects: the PROMETHEE method. Eur J Oper Res 1986;24:228–38.
[12] Saaty TL. The analytic hierarchy process. New York: McGraw Hill;
1980.
[13] Saaty TL. Derision making for leaders. Pittsburgh: RWS Publica-
tions; 1992.
[14] Keeny RL, Raiffa H. Decisions with multiple objectives: preferences
and value tradeoffs. New York: Wiley; 1976.
[15] Hovanov Nikolai V, Fedorov Yuri V, Zakharov Viktor V. The
making of index number under uncertainty. In: Pykh Y, Haytt
DE, Lenz RJM, editors. Advances in sustainable development
environmentalindices. Oxford,
1999.
[16] Afgan NH, Carvalho MG, Hovanov N. Indicators for sustainable
energy systems. Thermal Science 2005;9(2).
[17] Hovanov N, Kornikov V, Seregin I. Qualitative information
processing in DSS ASPID-3W for complex objects estimation
under uncertainty. In: Proceedings of the international confer-
ence informatics and control, St. Petersburg, Russia; 1997. p.
808–16.
[18] Afgan NH, Carvalho MG. Sustainability assessment method for
energy systems. New York: Kluwer Academic Publisher; 2000.
[19] Afgan NH, Carvalho MG. Quality, sustainability and indicators of
energy systems. Begell House; 2006.
[20] Afgan NH, Pilavachi PA, Carvalho MG. Multi-criteria evaluation of
natural gas resources. Energy Policy 2007;35:704–13.
[21] Pilavachi PA, Roumpeas CP, Minett S, Afgan N. Multi-criteria
evaluation forCHPsystems.
2006;47:3519–29.
[22] Zhou Zupeng, Hua Jiang, Liancheng Qin. Life cycle sustainability
assessment of fuels. Fuel 2007;86(1–2):256–63.
UK:EOLSS PublishersCo.;
Energy ConservManage
N.H. Afgan et al./Energy Conversion and Management 49 (2008) 2345–2353
2353