Renewable Energy 66 (2014) 478e494

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

GIS-based photovoltaic solar farms site selection using ELECTRE-TRI:

Evaluating the case for Torre Pacheco, Murcia, Southeast of Spain
Juan M. Sánchez-Lozano a, *, Carlos Henggeler Antunes b, e, M. Socorro García-Cascales c,
Luis C. Dias d, e
a
Centro Universitario de la Defensa de San Javier (University Centre of Defence at the Spanish Air Force Academy), MDE-UPCT, C/ Coronel López Peña s/n,
30720 Santiago de la Ribera, Murcia, Spain
b
Department of Electrical Engineering and Computers, University of Coimbra, Polo II, Pinhal de Marrocos, 3030-290 Coimbra, Portugal
c
Department of Electronics, Computers Architecture and Projects Engineering, Universidad Politécnica de Cartagena, C/Dr Fleming s/n, 30202 Cartagena,
Murcia, Spain
d
Faculty of Economics, University of Coimbra, Av. Dias da Silva 165, 3004-512 Coimbra, Portugal
e
INESC Coimbra, Rua Antero de Quental 199, 3000-033 Coimbra, Portugal

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

Article history: The Region of Murcia has one of the highest percentages of potential solar radiation in Spain, which puts
Received 11 June 2013 it in an excellent position to host electricity generation plants through photovoltaic solar systems,
Accepted 27 December 2013 commonly known as solar farms. This paper proposes the use of a Geographic Information System (GIS)
Available online
in order to identify the best plots suitable for installing photovoltaic solar farms in the Municipality of
Torre Pacheco, in the southeast of Spain. The plots are classified according to multiple evaluation aspects,
Keywords:
by developing a multicriteria model and applying the ELECTRE-TRI method using the Decision Support
Solar farms
System IRIS. The combination of GIS and IRIS offers the user the possibility of using the information
Decision support systems (DSS)
Geographic information systems (GIS)
provided by the GIS mapping leading to an assignment of the feasible courses of action (the plots) to
Multicriteria decision analysis (MCDA) categories of merit according to multiple, conflicting and incommensurate evaluation criteria.
ELECTRE-TRI method The GIS provides a cartographic and alphanumeric database, including two factors of distinct nature:
Interactive robustness analysis and restrictions and criteria. The restrictions are entered into the GIS using layers defined from the current
parameters’ inference for multicriteria legislation (urban land, undeveloped land, special protection areas for birds, community sites, in-
sorting problems (IRIS) frastructures, etc.), which reduce the study area by eliminating those areas in which photovoltaic solar
farms cannot be implemented. The criteria are organized into a tree to be used for assessing the greater
or lesser capacity to install photovoltaic solar farms. These criteria are introduced into the GIS, taking into
account weather, environmental, location, and terrain evaluation aspects.

2014 Elsevier Ltd. All rights reserved.

1. Introduction study area, the Municipality of Torre Pacheco, the average annual
global radiation in most of its territory exceeds 5 kWh/m2 per
The European Photovoltaic Industry Association (EPIA), in its day [2].
report on the global market outlook for photovoltaic energy until As a result of the excellent climatic characteristics offered by this
2016 [1], indicates that although there was a major setback in Spain territory, it has become an attractive area to implant photovoltaic
in the year 2012 it is expected that it will continue to expand solar farms. In order to achieve higher returns on their premises,
moderately as economic conditions improve and energy policy developers and investors need to use decision support models and
stabilizes (Fig. 1). methods that enable them to maximize the efficiency of solar
The southeast of Spain, and specifically the Region of Murcia, farms. Since these problems involve the appraisal of possible
has become one of the main areas in which more solar photovoltaic courses of action/alternatives according to multiple, generally
power plants have been implemented. Many factors are responsible conflicting and incommensurate, evaluation aspects, multicriteria
for this trend, not least the fact that Murcia has one of the highest decision analysis (MCDA) approaches are the most adequate means
levels of potential solar radiation in the country; specifically in the for providing decision support.
The ELECTRE (Elimination and Choice Translating Reality)
* Corresponding author. Tel.: þ34 968 189914; fax: þ34 968 189914.
method family is a well-known MCDA approach of the so-called
E-mail address: [email protected] (J.M. Sánchez-Lozano). European MCDA school [3e5]. Among the ELECTRE methods,

0960-1481/$ e see front matter
2014 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.renene.2013.12.038
 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 479

Fig. 1. Market evolution in Spain [1].

ELECTRE-TRI is devoted to the sorting problem, which consists of development of a management and planning system on a GIS
assigning each alternative to predefined ordered categories of platform for administrators, planners or consultants in renewable
merit. In this paper the ELECTRE-TRI method is used to classify the energies. In Italy, Gemelli et al. [26] used a GIS-based approach to
different alternatives into categories using the Decision Support obtain a regional model of the low temperature geothermal po-
System IRIS (Interactive Robustness analysis and parameters’ tential and its economic exploitability.
Inference for multicriteria Sorting problems), which implements In the present article, a combined approach using GIS and the
the most common variant of the ELECTRE-TRI method (pessimistic ELECTRE-TRI method is described to classify the possible locations
variant) [6,7]. for solar farms in the Municipality of Torre Pacheco, in the south-
MCDA approaches have been used to provide decision support east of Spain, into ordered categories of merit according to multiple
in several problems in the area of renewable energies, recognizing evaluation criteria. There are two main distinguishing features of
the multiple and conflicting aspects at stake for the appraisal of this approach as regards the existing literature. First, it does not
different courses of action [8]. Beccali et al. [9] used the ELECTRE III seek to find a best location in the context of a relative evaluation
method to evaluate a plan of action for the dissemination of (among competing alternatives), but to perform a classification of
renewable energy technologies at regional level. Haralambopoulos each location based on its absolute merits and drawbacks. The
and Polatidis [10] applied PROMETHEE II to investigate and assess model can therefore be applied to assess other locations besides the
the exploitation of geothermal energy sources in the island of Chios ones considered in this work and it does not assume that only one
(Greece). San Cristóbal [11] evaluated the efficiency of 13 renewable (the best) location would be adequate for a solar farm. Second, it
energy technologies through a Multiple Criteria Data Envelopment does not require setting a precise numerical value to express the
Analysis model. Lee et al. [12] applied a new MCDA method based importance of each criterion, which can be a difficult task for a
on the Analytic Hierarchy Process (AHP) for the selection of stra- decision maker (DM).
tegies for wind farm installation. Jo and Otanicar [13] developed a The rest of the paper is organized as follows. In Section 2, GIS are
hierarchical methodology for the meso-scale assessment of build- briefly reviewed, bearing in mind the problem at hand. The
ing integrated roof solar energy systems. More recently, studies methodology proposed and the Decision Support System IRIS are
have conducted a multicriteria evaluation of photovoltaic tech- described in detail in Sections 3 and 4. GIS-based photovoltaic solar
nologies using the TOPSIS and the AHP methods [14]. farms site selection using ELECTRE-TRI is applied to the case study
Geographic information systems (GIS) are a valuable tool for in southeast of Spain in Sections 5 and 6. Finally, the results and
assisting decision making in problems with environmental impli- conclusions of this work are drawn in Section 7.
cations on a territorial base. As GIS have developed, their applica-
tion has been extending to various fields, including that of 2. Geographic information systems (GIS)
renewable energy [15e20].
GIS provide an ideal complement between technical systems According to Star and Estes [27], a GIS can be defined as an in-
and decision support by offering a collection of procedures, tech- formation system that is designed to work with data referenced by
niques and algorithms to structure data to instantiate decision spatial or geographic coordinates. In other words, a GIS is both a
problems dealing with the design, evaluation and prioritization of database system with specific capabilities for spatially referenced
decision alternatives [21]. data, as well as a set of operations for working with data. GIS are
In recent years, GIS coupled with MCDA have been applied to used for the storage, management, analysis and display of
problems in renewable energy facilities. In the state of Colorado geographically referenced data, being valuable tools for assisting
(USA), Janke [22] studied multicriteria decision models of wind and planning and decision making in multiple contexts in which geo-
solar farms using GIS. In Oman, Charabi and Gastli [23] studied the referenced information plays a relevant role.
location of solar power plants using GIS and a multicriteria fuzzy In the present work a Spanish free software application called
methodology. In Andalusia, in the south of Spain, Arán-Carrión gvSIG (Generalitat Valenciana Geographical Information System)
et al. [24] carried out research into the choice of optimal site se- has been used, which was developed by the Ministry of Infra-
lection for grid-connected photovoltaic power plants combining structure and Transport of the Generalitat Valenciana [28]. The
GIS with AHP. In the northeast of Brazil, Tiba et al. [25] analyzed the gvSIG package allows the information to be processed both as
480 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494

raster (image) and vector (shapefile) formats. This application of- alternative and the boundary actions (reference actions) defining
fers the ability to access interactive maps based on Spatial Data the categories. Let g1, g2,.gm denote the set of criteria. Each
Infrastructure of Spain (IDEE), which aims at integrating data, boundary action bh is the upper limit of the category Ch and the
metadata, services and geographic information produced in Spain, lower limit of category Chþ1 (Fig. 3). The boundaries b0 and bpþ1
using the Internet. gvSIG also provides access to map servers based may correspond to the anti-ideal and ideal solutions, respectively.
on INSPIRE (Infrastructure for Spatial Information in Europe),
which is a European Commission initiative aimed at making STEP 2 Determination of concordance indices by criteria
available relevant, coordinated and quality geographic information,
enabling the formulation, implementation, monitoring and evalu- The criterion concordance indices indicate how much each cri-
ation of policy or territorial impacts in the European Community. terion agrees with the assertion ai$S$bh, taking into account indif-
ference (qj) and preference (pj) thresholds, which characterize the
3. The ELECTRE-TRI method acceptance of imprecision in the judgment. If gj(ai), the performance
of ai on criterion gj, is equal to or better than gj(bh), the performance
The ELECTRE methods have been widely used by researchers of bh on the same criterion, then this criterion fully agrees that
and professionals in several areas of application, including in en- ai$S$bh. The criterion may also fully agree even if gj(ai) is slightly
ergy problems [8,29,30]. ELECTRE methods are based on the con- worse than gj(bh), since ai and bh are considered indifferent for cri-
struction and exploitation of an outranking relation S (a$S$b terion gj if the difference between their individual performances is
denotes that alternative a outranks alternative b, meaning that a is less than qj. If gj(ai) is worse than gj(bh) by a difference that is greater
at least as good as b). The construction of outranking relations is than qj but less than pj, then gj agrees that ai$S$bh only partially. The
based on two principles: (1) the concordance principle requires transition between indifference and preference is linear.
that a sufficient majority of criteria agree that a is at least as good as The concordance index for a given criterion cj(ai,bh) is defined as
b; (2) the non-discordance principle requires that, when the follows:
concordance condition holds, none of the criteria in the minority is
cj ðai ;bh Þ ¼ 05pj  gj ðbh Þ  gj ðai Þ
strongly opposed to the assertion a$S$b.
Several members of the ELECTRE method family have been gj ðai Þ þ pj  gj ðbh Þ
0 < cj ðai ;bh Þ ¼ < 15qj < gj ðbh Þ  gj ðai Þ < pj
proposed according to the problem to be addressed, i.e. choice, pj  qj
ranking and sorting problems: I [31], II [32,33], III [34], IV [35,36], IS
[37], and TRI [38e40]. ELECTRE-TRI [38e40] is devoted to the cj ðai ;bh Þ ¼ 15gj ðbh Þ  gj ðai Þ  qj
sorting problem, which consists of assigning each alternative to (1)
previously defined categories of merit bounded by lower and upper
STEP 3 Calculation of the overall concordance
profiles (reference actions). The main steps of ELECTRE-TRI are
depicted in Fig. 2.
The global concordance indices quantify the relative importance
of the coalitions of criteria that are in favor of the assertion ai$S$bh.
STEP 1 Definition of reference actions
The weight assigned to each criterion (denoted kj) may be inter-
preted as a true importance coefficient, in the sense of its voting
The assignment of an alternative to a given category is deter-
power for those coalitions. This means that in ELECTRE-TRI weights
mined by establishing an outranking relation between the
are scale-independent and they do not play the role of coefficients
to convert the performances of the criteria into a common value
scale. The global concordance indices are determined from the
concordance indices for each of the criteria:
Pm
j¼1 kj $cj ðai ;bh Þ
Cðai ;bh Þ ¼ Pm (2)
j¼1 kj

STEP 4 Determination of the discordance indices by criteria

The concordance relation is complemented with a discordance

relation that uses the veto (vj) and the preference (pj) thresholds to
define how much each criterion disagrees with the assertion ai S bh.
The discordance indices are calculated as:

dj ðai ;bh Þ¼05gj ðai Þgj ðbh Þpj

gj ðbh Þgj ðai Þpj
0<dj ðai ;bh Þ¼ <15gj ðbh Þvj <gj ðai Þ<gj ðbh Þpj
vj pj
dj ðai ;bh Þ¼15gj ðbh Þvj ðbh Þgj ðai Þ
(3)
STEP 5 Obtaining the degree of credibility

The degree of credibility is an indicator that considers jointly the

overall concordance indices C(ai,bh) and the discordance indices
dj(ai,bh) so it expresses in what degree “a outranks bh”. The following
Fig. 2. Main steps of the ELECTRE-TRI method. expression can be used to compute the degree of credibility [7]:
 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 481

Fig. 3. Definition of categories using limit profiles [41].

 
level, IRIS allows the DM to enter constraints on these values,
ss ðai ; bh Þ ¼ Cðai ; bh Þ$ 1  max dj ðai ; bh Þ (4) including assignment examples that it tries to restore.
j˛f1;.;mg
Using IRIS the user can state precise values (e.g. k1, the weight of the
first criterion, is equal to 0.2), intervals (e.g. k1 ˛ [0.1, 0.3]), linear con-
STEP 6 Determination of the outranking relation straints (e.g. k1  k2), or indirect constraints (e.g. a1 should be sorted in
category C3,). Sources of inconsistency among these constraints are
The outranking relation between a potential action a and a identified whenever it is not possible to satisfy them simultaneously. In
reference action bh is based on the degree of credibility and a constant such cases the combination of parameter values that minimizes the
cutting level l which corresponds to the lowest value of the degree of constraint violations is computed. IRIS also infers robust conclusions by
credibility from which the assertion “a outranks bh” is valid, i.e., such indicating the range of assignments for each alternative that does not
assertion is corroborated (ai$S$bh) only when ss(ai,bh)  l. contradict any user-defined constraint. Whenever those constraints
are consistent, IRIS infers a “central” combination of parameter values
STEP 7 Assignment of alternatives to different categories. stating the category each alternative is sorted into corresponding to
Once an alternative ai has been compared to the action or that combination, and the range of categories each alternative might be
reference profiles bh, in ELECTRE-TRI there are two ways to assigned to while satisfying all constraints.
assign an alternative ai to one of the predefined categories. IRIS offers DMs an interactive environment in which they may
The comparison of alternative ai with the reference profile gather information in an interactive and progressive manner,
may be done according either to a pessimistic or optimistic exerting a critical analysis on the results displayed to guide the
procedure: elicitation of parameter values. This enables a step-by-step approach
(a) Pessimistic procedure (or conjunctive): it consists of in which further knowledge is gathered and preferences are refined
assessing ai$S$bh successively for the different profiles through a learning process, thus reducing the scope of the search
starting with the best profile to find the bh profile for until the DM is confident about the results to make a final decision.
which ai$S$bh is verified, and once found ai is assigned In this paper, a sample of the information provided by the GIS is
to the category Chþ1. used as input into the DSS IRIS to produce an evaluation of potential
(b) Optimistic procedure (or disjunctive): it consists of locations for solar farms in the region under study, which is framed
comparing the alternative ai successively with the as an MCDA sorting problem.
different profiles starting with the worst profile to find
the bh profile for which w(ai S bh) ^ bh S ai, and once 5. A GIS-based methodology for obtaining suitable surfaces to
found ai is assigned to the category Ch. implant photovoltaic solar farms

The IRIS software, described in the next section, implements the The gvSIG program is used in two stages. In the first one, re-
pessimistic procedure, which is more intuitive and more used in strictions that prevent a solar plant from being implanted in a
practice: if ai is sorted in category Ch, then this means it is good particular area are represented. These areas unsuitable for im-
enough to outrank this category lower bound bh1 but it is not good plantation are removed, leaving only the areas that are feasible for
enough to outrank its upper bound bh. this purpose. In the second stage, all the information pertaining to
all retained alternative locations for all evaluation criteria is
4. The decision support system IRIS selected to be supplied to the DSS IRIS.
The Municipality of Torre Pacheco covers an area of 189.60 km2,
IRIS [7] is a DSS based on ELECTRE-TRI (pessimistic procedure) where the territory is divided into different types of land according
that implements the interactive methodology proposed in to its Urban Municipal General Plan. This classification is added to
Ref. [6]. IRIS enables to exploit an ELECTRE-TRI model in case the user the gvSIG program as a vector layer, as represented in Fig. 4.
is a DM with no expertise in the method or an analyst mediating the Then the layers relating to the restrictions are added using the
communication between the software and the DM. Rather than gvSIG application commands to obtain the suitable surface to
demanding precise values for the ELECTRE-TRI weights and cutting implant photovoltaic solar farms in the municipality.
482 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494

Fig. 4. Addition of layer of the Municipality of Torre Pacheco in gvSIG.

The restrictions to be inserted into gvSIG are obtained through found to be suitable to implant solar farms; this percentage cor-
the Administrations and Public Bodies of the Region. These responds to 65.36 km2. Its distribution in the municipality can be
constitute the technical and environmental restrictions of the area observed using an orthophoto or raster layer (Fig. 6).
under study and are entered into the software in the form of a In the second stage the information about the criteria is inserted
vector layer (Table 1). into the gvSIG program (in vector or raster format). The criteria to
The layers in Table 1 are added to gvSIG, and their representa- be considered are derived from a review of existing literature [22e
tions are shown in Appendix A. 24,42], which has been agreed with experts in the field of renew-
Once introduced in vector format layers, a gvSIG command able energy sources. The criteria are used to assess the merit of each
enables to remove the area affected by the restrictions and then the plot to implant a photovoltaic plant therein. This information is
resulting surface should be assessed by the decision support obtained through public and private institutions (Fig. 7, Table 2).
methodology. The above layers are added to gvSIG and their representations are
To classify and enumerate the resulting surface the classification shown in Appendix B.
made by the General Direction of Cadastre of the Region of Murcia
is used, which divides the surface of any land into zones, plots and
subplots. The cadastral layer of the Municipality of Torre Pacheco is 5.1. Vector thematic layer processing and attribute table
thus drawn in vector format (Fig. 5).
Besides indicating the zones, plots and subplots, the cadastral Once all the criteria influencing the decision process have been
layer provides information on the area of each parcel and whether defined, the layers of criteria are linked spatially with the vector
it is a building. Since to establish a photovoltaic solar farm large layer obtained in stage 1 (Fig. 6). This process is carried out in two
surfaces are required, the minimum area of each suitable plot is different ways depending on whether the criterion layer is vector or
1000 m2. It also seems logical that those parcels containing a raster.
building will be the least suitable to host a solar plant. Therefore, In the gvSIG program, vector layers are structured by rows
taking into account these considerations, filtering is carried out to (registers) and columns (fields) and each register corresponds to a
remove those parcels that are smaller than 1000 m2 or contain spatial object (e.g., a plot) referenced by an identifier which is
buildings, using a gvSIG command. A new layer of cadastral infor- usually defined in the first column, while the remaining columns
mation is obtained (classified in zones, plots and subplots) for the represent the attributes associated with each spatial object. These
Municipality of Torre Pacheco to implant photovoltaic solar farms. rows and columns are shown in a table called an attribute table.
About 35% of the total area of the municipality (189.60 km2) is then If the criterion is of vector type (agrological capacity, plot area,
distance to power lines, electricity transformer substations, towns
Table 1 or villages and main roads), the transfer of attributes is made and
Layers of restrictions.
the distances from each plot to power lines, electricity transformer
No. Denomination of the layers of restrictions substations, towns or villages and main roads are calculated using
1 Urban lands the appropriate processes in the gvSIG program. If the criterion is
2 Protected and undeveloped lands expressed in raster information, an extension of gvSIG called SEX-
3 Areas of high landscape value TANTE is used. It processes the raster layers of criterion information
4 Water infrastructure, military zones and cattle trails
(orientation, solar radiation, average temperature and slopes),
5 Watercourses and streams
6 Archaeological sites which links with the layer obtained by applying spatial links to the
7 Paleontological sites previously obtained vector layers.
8 Cultural heritage The final table obtained (Table 2) contains the alternatives
9 Roads and railroad network defined by zone, plot and subplot and the criteria indicated for
10 Community interest sites (LICs)
assessing each alternative.
 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 483

Fig. 5. Addition of layer of the Cadastral information in gvSIG.

6. Using the ELECTRE-TRI method for assigning plots to to perform the multicriteria decision support process using the DSS
categories of merit IRIS. The decision problem is structured using the criterion tree
displayed in Fig. 7 and the plots selected by the DM are shown in
To carry out the evaluation of alternatives an expert in solar Table 3.
photovoltaic facilities was involved in the model building process. In addition to selecting the alternatives (ai) and criteria (gj), the
This expert, who will be the DM, is a promoter of renewable energy application of the ELECTRE-TRI method requires the definition of
facilities with more than 10 years of experience in the industry. He the different categories with their upper and lower reference pro-
will be able to apply his judgment about a small number of alter- files (b0,b1,.,b4), as well as indifference (qj(bh)), preference (pj(bh))
natives according to his knowledge and experience. Thus, from the and veto (vj(bh)) thresholds. The values for the bounds of the cat-
table of criteria obtained using gvSIG, only a few alternatives (20 egories and the indifference, preference and veto thresholds have
plots in this case) are extracted in order to provide the information been provided by the DM according to his personal knowledge and

Fig. 6. Suitable areas after cadastral restrictions.

484 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494

Fig. 7. Criterion tree resulting from the problem structuring phase.

experience. The table of reference actions delimiting the four cat- In the next step the reference profiles bounding each category
egories is displayed in Table 4. for each criterion (also indicating whether the criterion is to be
The alternatives to be evaluated and categories are displayed in maximized or minimized), as well as the preference, indifference,
Fig. 8. and veto thresholds are entered (Fig. 10).
This information is entered into the DSS IRIS. Fig. 9 displays the The upper and lower bounds of the cutting level l ðl˛½0:5; 1Þ,
evaluation matrix, i.e. the performance of each alternative in each which states the exigency of the classification into categories of
criterion. merit, and the weights (kj refers to the weight of criterion gj), which
state the “voting power” of each criterion for establishing the
Table 2 outranking relation, are introduced. IRIS does not require the user
Attribute table of the shape with the information of criteria and attributes. to indicate precise values for the criterion weights (k1,.,k10) and
Attribute table the cutting level l; rather it allows to obtain such values through an
inference procedure [43]. The initial bounds for the criterion
Alternatives Plot
Cadastral information Zone and subplot
weights have been obtained through a system of surveys to
Criteria Agrological capacity (g1) photovoltaic solar energy experts [44], so that the maximum and
Slope (g2) minimum values are consistent with those they provided. In order
Field orientation (g3) to obtain a higher degree of credibility and consistency in the re-
Plot area (g4)
sults the interval for l˛½0:67; 0:80 has been considered, thus
Distance to main roads (g5)
Distance to power lines (g6) requiring a “qualified majority” of criteria, and the extreme values
Distance to town or villages (g7) (l ¼0.67 and l ¼0.80) are analyzed (Fig. 11).
Distance to electricity transformer substations (g8) Finally, further constraints on the range of the parameters can
Solar radiation (g9) be added; the DM may edit at any time the constraints that the
Average temperature (g10)
weights and the cutting level should respect.
 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 485

Table 3
Examples selected for evaluation by an expert.

Alternatives Criteria

Zone Plot Subplot g1 g2 g3 g4 g5 g6 g7 g8 g9 g10

2 2 
(1e8) 0e30% 1e10 m m m m m kJ/m day C

Alternatives a1 1 16 c 1.00 17.70 7.26 13559.32 68.82 126.21 1846.93 4265.78 2047.04 17.60
a2 1 19 a 4.00 10.50 7.10 37855.87 562.86 1.25 1668.60 4869.39 2049.17 17.69
a3 1 22 c 4.00 16.02 7.45 8691.09 1154.72 106.17 1738.34 5782.16 2048.65 17.60
a4 1 22 f 5.00 11.76 7.89 49659.87 1473.29 59.99 1977.82 6025.01 2050.18 17.60
a5 1 22 g 4.00 7.60 7.00 3037.95 1529.19 15.64 1851.09 6444.00 2051.64 17.60
a6 1 26 a 5.00 8.26 8.99 7891.86 982.63 197.90 1043.22 6662.57 2050.48 17.60
a7 1 27 a 2.00 5.11 8.30 4484.85 391.99 481.15 519.97 6158.91 2051.18 17.60
a8 1 32 c 1.00 7.69 4.37 9623.20 55.06 540.65 491.80 5798.14 2048.73 17.60
a9 1 33 a 4.00 20.88 5.96 193639.32 24.38 82.27 568.40 5619.33 2050.53 17.60
a10 1 33 b 5.00 32.50 6.30 57522.34 493.70 51.28 1014.88 5802.09 2050.81 17.60
a11 1 35 a 1.00 15.52 3.92 79125.52 22.69 252.74 828.56 4956.58 2049.89 17.69
a12 1 47 a 5.00 8.21 7.79 4102.16 323.52 459.22 587.23 5997.43 2051.03 17.60
a13 1 56 b 8.00 21.78 6.86 1979.96 730.25 373.08 811.91 6602.30 2051.78 17.60
a14 1 58 a 5.00 30.26 8.23 266207.77 577.19 202.73 1255.21 5251.43 2047.96 17.61
a15 1 58 i 5.00 18.76 6.08 5988.98 264.70 539.39 992.95 5429.40 2050.00 17.66
a16 1 61 a 2.00 16.78 3.25 5203.00 531.85 442.27 599.25 6328.74 2050.20 17.60
a17 1 76 a 1.00 12.88 6.59 7994.66 687.46 384.65 751.04 6529.83 2051.42 17.60
a18 1 89 a 2.00 13.99 6.11 6577.69 978.06 6.46 1179.66 6348.33 2052.02 17.60
a19 1 90 b 5.00 12.81 5.25 12170.21 1157.92 132.57 1355.16 6449.19 2052.21 17.60
a20 1 91 e 2.00 19.35 6.09 11984.62 741.60 236.24 813.25 6414.66 2050.60 17.60

6.1. First iteration category for this alternative, so that at most it is situated in category
C3 (Fig. 13).
Once all the parameters have been introduced, the IRIS program With this second iteration, besides assigning alternative a9 to
is run providing a graph as shown in Fig. 12; we refer to these input category C3, IRIS states that alternatives a7 and a16 would no longer
parameters and output results as of the first iteration. be positioned in the worst category (C1) to be classified as good
Fig. 12 shows the results (with l˛½0:67; 0:80) that indicate the alternatives (category C2). However, the DM considers that these
range of possible assignments for each alternative, i.e. the cate- alternatives deserve to be in the worst category because, besides
gories to which alternatives may be assigned without violating not presenting large areas (criterion g4), the distance of these plots
any constraints and assignments examples; these ranges appear to electricity transformers substations (criterion g8) is excessive.
in green. In each range, one of the cells has a darker shade of Therefore, a third iteration with IRIS is performed to place alter-
green, meaning it is the category assignment recommended by natives a7 and a16 in the worst category according to the DM’s
IRIS. The corresponding l and kj values are presented in the last experience and preferences (Fig. 14).
row of Fig. 12. If the DM selects any cell in a range, the penultimate Once the third iteration is made it is observed that not only it has
line in Fig. 12 shows a combination of parameter values that assign been possible to assign alternatives a7 and a16 to the worst category
the action in the cell’s row to the category in the cell’s column but also the categories of alternatives a2, a9, a10, a11 and a14 have
(e.g., Fig. 12 shows a combination of values that assign a14 to C3). been reduced from very good (category C3) to good (category C2).
The DM considers that such alternatives should not be situated in
6.2. Second and successive iterations category 2, and he even considered that alternative a14 should be
situated in the best category because not only it has the greatest area
Analyzing the results provided by IRIS, the DM observes that (criterion g4), but it is also quite far from towns and villages (crite-
alternative a9 is classified in the best category. This alternative rion g7). Therefore, a fourth iteration is performed to reflect these
(plot) is very attractive because it occupies a large area of territory preferences indicated by the DM in an interactive manner (Fig. 15).
(criterion g4); however, according to the DM it should not be in the With this fourth iteration in IRIS, the DM is satisfied with the
best category as it is very close to towns and village (criterion g7). So results obtained since in this last classification all plots are placed in
the DM considers that alternative a9 should be at most very good a single category and match his expert judgment. Indeed, each
(category C3) but not excellent (category C4). Using IRIS it is possible request of the DM consisted of excluding the vectors of parameters
to modify the input parameters in order to establish the best that were incompatible with the corresponding preference

Table 4
Reference actions.

g1 g2 g3 g4 g5 g6 g7 g8 g9 g10

b1 2 30 5 25,000 1000 10,000 100 6250 1200 16.00

b2 4 20 8 50,000 500 1000 500 2500 1700 18.00
b3 7 10 10 100,000 25 100 750 500 2000 20.00
q1(b) 1 5 4 3 100 100 100 150 0 17.50
p1(b) 4 15 7 1000 200 300 300 3000 1500 17.60
v1(b) 6 40 9 25,000 650 500 800 10,000 2050 17.70
q2(b) 1 5 4 15 75 10 100 50 0 17.50
p2(b) 4 15 7 1000 200 300 300 3000 1500 17.60
v2(b) 6 40 9 25,000 650 500 800 10,000 2050 17.70
q3(b) 1 5 4 100 20 1 100 5 0 17.50
p3(b) 4 15 7 1000 200 300 300 3000 1500 17.60
v3(b) 6 40 9 25,000 650 500 800 10,000 2050 17.70
486 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494

Fig. 8. Alternatives to be evaluated and definition of categories using reference profiles.

expression. The reduction of the set of parameter vectors accepted inferred from the DM’s judgment, through an indirect process that
in each iteration led also to reducing the range of categories in revealed more meaningful for the expert than directly providing
which other alternatives can be placed. such numerical values. It is possible to observe that the criterion
The parameter values displayed in the last line of Fig. 15 (on the “distance to town or villages” (g7) is considered to be the most
right) correspond to the cutting level l and the criterion weights important for this classification.

Fig. 9. Introduction of the evaluation matrix in IRIS.

 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 487

Fig. 10. Introduction of reference profiles for each criterion and thresholds.

Fig. 11. Upper and lower bounds of the l-cut and kj.

Fig. 12. Results output in IRIS (first iteration).

488 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494

Fig. 13. Results output in IRIS (second iteration).

7. Results The analysis of the map displaying the categorization obtained

by the DSS IRIS with selected plots enables to draw useful conclu-
The classification associated with the fourth iteration (Fig. 15) is sions to reach a final recommendation. In the 20 alternatives (plots)
shown in the gvSIG program (Fig. 16) in order to obtain a carto- chosen, it is observed that one of them (the blue colored plot) has
graphic visualization; this figure shows the sample plots with the excellent capability to host the implementation of a photovoltaic
categories in different colors. solar farm; four plots (colored in yellow) have very good capability;

Fig. 14. Results output in IRIS (third iteration).

 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 489

Fig. 15. Results output in IRIS (fourth iteration).

10 plots (green colored plots) have good capability; and the these plots. (For interpretation of the references to color, the reader
remaining plots (red plots) have poor capability. The remainder is referred to the web version of this article.)
plots, shown in light blue, have not been evaluated by the DM, Although more alternatives could have been added, the DM has
although the inferred model parameters can now be used to classify considered preferable to use a limited number of them in order to

Fig. 16. Classification and representation of the sample plots.

490 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494

run the decision process in greater depth and also to view their a final recommendation. This allows the DM to obtain robust con-
results more clearly and concisely. clusions, i.e. conclusions that hold true for all the acceptable com-
binations of parameter values, which do not need to be precisely
8. Conclusions specified (thus not imposing an excessive burden on the DM).
Future research lines include improving the combination be-
In the early stages of the study it was verified through a tween GIS and MCDA tools at the methodological level, and also
Geographic Information System (gvSIG) that the municipality of Torre analyzing larger extensions of territory and studying other
Pacheco is an optimal zone to implement photovoltaic solar farms. renewable energy technologies (wind farms, solar thermoelectric,
The main contribution of this paper is the combination of this biomass, etc.).
geographic information system with a multicriteria decision anal-
ysis Method (ELECTRE-TRI, which is based on the exploitation of an Acknowledgments
outranking relation devoted to the sorting problem) by developing
a multicriteria model to be tackled by the DSS IRIS to provide de- This work is partially supported by FEDER funds, the DGICYT and
cision support. Junta de Andalucía under projects TIN2011-27696-C02-01 and P11-
The main advantage offered by this integrated approach is using TIC-8001, respectively.
a GIS to gather and organize the information to be supplied to the C. H. Antunes and L. Dias acknowledge the support of EMSURE e
DSS, which in turn provides results that can be meaningfully dis- Energy and Mobility for Sustainable Regions Project (CENTRO-07-
played using the GIS. IRIS indicates for each alternative the category 0224-FEDER-002004) and Portuguese Foundation for Science and
of merit where it is classified according to a set of criteria, as well as Technology (FCT) under project grants MIT/SET/0018/2009 and
it provides interactive features that enable a progressive shaping of PEst-C/EEI/UI0308/2011.

Appendix A. Maps used for each restriction

 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 491

Appendix B. Maps used for each criterion

492 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494
 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494 493

References [13] Jo JH, Otanicar TP. A hierarchical methodology for the mesoscale assessment
of building integrated roof solar energy systems. Renew Energy 2011;36:
2992e3000.
[1] European Photovoltaic Industry Association (EPIA). Global market outlook for
[14] García-Cascales MS, Lamata MT, Sánchez-Lozano JM. Evaluation of photo-
photovoltaic until 2016; 2012. pp. 5e33.
voltaic cells in a multi-criteria decision making process. Ann Oper Res
[2] Gómez López MD, García Cascales MS, Ruiz Delgado E. Situations and prob-
2012;199:373e91.
lems of renewable energy in the Region of Murcia, Spain. Renew Sustain
[15] Voivontas D, Assimacopoulos D, Mourelatos A. Evaluation of renewable en-
Energy Rev 2010;14:1253e62.
ergy potential using a GIS decision support system. Renew Energy
[3] Roy B. The outranking approach and the foundations of ELECTRE methods.
1998;13(3):333e44.
Theory Decis 1991;31:49e73.
[16] Baban MJS, Parry T. Developing and applying a GIS-assisted approach to
[4] Roy B, Bouyssou D. Aide multicritère à la décision: méthodes et cas. Paris:
locating wind farms in the UK. Renew Energy 2001;24:59e71.
Economica; 1993.
[17] Amador J, Domínguez J. Application of geographical information systems to
[5] Roy B, Vanderpooten D. The European school of MCDA: Emergence, basic
rural electrification with renewable energy sources. Renew Energy 2005;30:
features and current works. J Multicrit Decis Anal 1996;5(1):22e38.
1897e912.
[6] Dias L, Mousseau V, Figueira J, Clímaco J. An aggregation/disaggregation
[18] Hoesen JV, Letendre S. Evaluating potential renewable energy resources in
approach to obtain robust conclusions with ELECTRE TRI. Eur J Operat Res
Poultney, Vermont: a GIS-based approach to supporting rural community
2002;138:332e48.
energy planning. Renew Energy 2010;35:2114e22.
[7] Dias LC, Mousseau V. IRIS: a DSS for multiple criteria sorting problems.
[19] Arnette AN, Zobel CW. Spatial analysis of renewable energy potential in
J Multicrit Decis Anal 2003;12:285e98.
the greater southern Appalachian mountains. Renew Energy 2011;36:
[8] Diakoulaki D, Antunes CH, Martins AG. MCDA and energy planning. In:
2785e98.
Figueira J, Greco S, Erghott M, editors. State of the art of multiple criteria
[20] Hossain J, Sinha V, Kishore VVN. A GIS based assessment of potential for wind
decision analysis. International series in operations research and management
farms in India. Renew Energy 2011;36:3257e67.
science, vol. 78. Springer; 2005. pp. 859e97.
[21] Malczewski J. 392pp. GIS and multicriteria decision analysis. New York: J.
[9] Beccali M, Cellura M, Mistretta M. Decision-making in energy planning.
Wiley & Sons; 1992.
Application of the Electre method at regional level for the diffusion of
[22] Janke JR. Multicriteria GIS modeling of wind and solar farms in Colorado.
renewable energy technology. Renew Energy 2003;28:2063e87.
Renew Energy 2010;35:2228e34.
[10] Haralambopoulos DA, Polatidis H. Renewable energy projects: structuring a
[23] Charabi Y, Gastli A. PV site suitability analysis using GIS-based spatial fuzzy
multi-criteria group decision-making framework. Renew energy 2003;28:
multi-criteria evaluation. Renew Energy 2011;36:2554e61.
961e73.
[24] Arán-Carrión J, Espín-Estrella A, Aznar-Dols F, Zamorano-Toro M,
[11] San Cristóbal JR. A multi criteria data envelopment analysis model to evaluate
Rodríguez M, Ramos-Ridao A. Environmental decision-support systems for
the efficiency of the renewable energy technologies. Renew Energy 2011;36:
evaluating the carrying capacity of land areas: optimal site selection for grid-
2742e6.
connected photovoltaic power plants. Renew Sustain Energy Rev 2008;12:
[12] Lee AHI, Chen HH, Kang HY. Multi-criteria decision making on strategic se-
2358e80.
lection of wind farms. Renew Energy 2009;34:120e6.
494 J.M. Sánchez-Lozano et al. / Renewable Energy 66 (2014) 478e494

[25] Tiba C, Candeias ALB, Fraidenraich N, de S. Barbosa EM, de Carvalho Neto PB, [35] Roy B, Hugonnard JC. Classement des prolongements de lignes de métro en
de Melo Filho JB. A GIS-based decision support tool for renewable energy banlieue parisien (présentation d’une méthode multicritère originale). Cah
management and planning in semi-arid rural environments of northeast of CERO 1982;24(2e4):153e71.
Brazil. Renew Energy 2010;35:2921e32. [36] Roy B, Hugonnard JC. Le plan d’extension du metro en banlieue parisien, un
[26] Gemelli A, Mancini A, Longhi S. GIS-based energy-economic model of low cas type de l’analyse multicritère. Les Cah Sci Rev Transp 1982;6:77e108.
temperature geothermal resources: a case study in the Italian Marche region. [37] Roy B, Skalka JM. ELECTRE IS: aspects méthodologiques et guide d’utilisation.
Renew Energy 2011;36:2474e83. Université Paris-Dauphine; 1985. p. 125. Document du LAMSADE No. 30.
[27] Star J, Estes J. Geographic information systems: an introduction. Englewood [38] Roy B, Bouyssou D. Aide à la décision fondée sur une PAMC de type ELECTRE.
Cliffs, NJ: Prentice-Hall; 1990. pp. 2e3. Université Paris-Dauphine; 1991. p. 118. Document du LAMSADE No. 69.
[28] Regional Ministry of Infrastructure and Transport of Valencia, gvSIG Associ- [39] Yu W. Aide multicritère à la décision dans le cadre de la problématique du tri.
ation. https://gvsig.org/web/catalog [accessed 14.03.13]. Concepts, méthodes et applications. UER Sciences de l’organisation, Université
[29] Neves LP, Martins AG, Antunes CH, Dias LC. A multi-criteria decision approach Paris-Dauphine; 1992. p. 201. Thèse de doctorat.
to sorting actions for promoting energy efficiency. Energy Policy 2008;36(7): [40] Yu W. ELECTRE TRI. Aspects méthodologiques et manuel d’utilisation. Uni-
2351e63. versité Paris-Dauphine; 1992. p. 80. Document du LAMSADE No. 74.
[30] Madlener R, Antunes CH, Dias LC. Assessing the performance of biogas plants [41] Mousseau V, Slowinski R, Zielniewicz P. A user-oriented implementation of
with multi-criteria and data envelopment analysis. Eur J Oper Res the ELECTRE-TRI method integrating preference elicitation support. Comput
2009;197(3):1084e94. Oper Res 2000;27:757e77.
[31] Roy B. Classement et choix en présence de points de vue multiples (la [42] Sánchez-Lozano JM, Teruel-Solano J, Soto-Elvira PL, García-Cascales MS.
méthode ELECTRE). Rev Fr Automat Infor 1968;8:57e75. Geographical information systems (GIS) and multi-criteria decision making
[32] Roy B, Bertier P. La méthode ELECTRE II: une méthode de classement en (MCDM) methods for the evaluation of solar farms locations: case study in
présense de critères multiples. SEMA (Metra International) Paris south-eastern Spain. Renew Sustain Energy Rev 2013;24:544e56.
1971;142:25. [43] Mousseau V, Slowinski R. Inferring an ELECTRE-TRI model from assignment
[33] Roy B, Bertier P. La méthode ELECTRE II: une application au media-planning. examples. J Global Optim 1998;12:157e74.
In: Ross M, editor. Operational research 1972. North-Holland Publishing [44] Sánchez-Lozano JM, García-Cascales MS, Lamata MT. Decision criteria for
Company; 1973. pp. 291e302. optimal location of solar plants: photovoltaic and thermoelectric. Assessment
[34] Roy B. ELECTRE III: un algorithme de classement fondé sur une représentation and simulation tools for sustainable energy systems. Green Energy Technol
floue des préférences en présence de critères multiples. Cah CERO 1978;20(1): 2013;129:79e91. http://dx.doi.org/10.1007/978-1-4471-5143-2-4. Springer-
3e24. Verlag, London.