Sie sind auf Seite 1von 11



Bouchra Frikh, Hamza Sellak, Afaf Dadda, Brahim Ouhbi

To cite this version:

Bouchra Frikh, Hamza Sellak, Afaf Dadda, Brahim Ouhbi. A DECISION SUPPORT SYSTEM AP-
PLIED TO RENEWABLE ENERGY FIELD . MOSIM 2014, 10ème Conférence Francophone de
Modélisation, Optimisation et Simulation, Nov 2014, Nancy, France. <hal-01166632>

HAL Id: hal-01166632
Submitted on 23 Jun 2015

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
10the International Conference onModeling, Optimization and Simulation - MOSIM’14–November 5-7-2014-
Nancy – France “Toward circular Economy”




FSDM –UniversitySidi Mohammed Ben Abdellah ENSAM- University My Ismail
Fès, Morocco. Meknes, Morocco.,,

ABSTRACT:With the introduction of new concepts in operation and planning of renewable energy sources, decision
making is becoming more critical than ever before. These concepts include renewable energy planning and policy,
evaluation and assessment,project selection and finally the importance of environmental concerns. This paper aims to
present a decision support system (DSS) applied to renewable energy field, based on a new developed relational multi-
criteria clustering algorithm. The presented tool is a Renewable Energy WEB DSS application (RE-WEB-DSS)
combining different layers to achieve the main goal of a stakeholder trying to select the best renewable energy plant
projects (REP).To do so, a comparison between several existing decision support systems and multi-criteria models,
highlighting their applications and limitations in the context of renewable energy.This is followed by a discussion of the
different factors that should be considered to ensure optimal results. A subsequent section introduces and analyses the
RE-WEB-DSS approach and methodology. The paper explains also the approach and illustrates the tool’s algorithm
application through a simulation.Experimental studies of real cases are conducted. These cases studies demonstrate the
effectiveness and feasibility of the proposed DSS.

KEYWORDS:Renewable Energy Projects, Decision Support System,Multi-criteria Decision Making,Clustering Algorithm.

manage distributed and delocalized manufacturing

1. INTRODUCTION information regarding energy efficiency.
To the best of our knowledge there are no decision
The encouragement of renewable energy is a key policy systems supporting the operational environment of energy
to reduce greenhouse gas emissions and the energy bill. companies. The existing enterprise systems are generally
Over the last three decades there has been rapid growth limited to monitoring, analysis and reporting. According
in renewable energy sources (RES) development and to Zampou et al. [21], there is a gap, with respect to
energy efficiency. Nowadays, RES are considered as an energy-awareness in manufacturing, between the
engine for increasing economic competitiveness and a solutions available and the actual implementation.
way to reduce dependency on fossil fuel source.However, Therefore, an interesting issue to be addressed is: how
for stakeholders, it is clearly observed that the real information systems can be used to reduce energy
decision-making process regarding the choice of consumption? Such type of information systems should
renewable energy project plant is multidimensional, made collect and analyze energy data sets, integrate energy
up the number of aspects at different levels – economic, flows with other information flows, support optimization
technical, environmental and social. Therefore, getting of energy consumption and contribute to energy efficient
clear and explicit solutions might be very complicated. business process.
Hence, the need arises to develop tools that should enable The present paper aims at depicting the context and the
the decision maker (e.g. investor, government, association need of strengthening the use of renewable energies to
and electricity supplier) to select a series of alternatives achieve sustainability goals, acknowledging the need of
(objectives are usually conflicting) and so, the solution tools that facilitate the decision-making process in this
which highly depends on the preferences of decision- field in operational options for decision makers. After a
makers must be a compromise. review of the state of the art in the existent DSS as well as
On the other hand, transitions to decentralized and the MCDM-methods used in renewable energy field, we
renewable production pose many challenges and have discuss the main functionalities and components of a DSS
recently attracted the attention of decision makers and (section 3). ‘New Renewable Energy DSS’ subsections
academic researchers. One of the most important describe the proposed algorithm, present the main
parameters for the renewable energy sources development guidelines that can be applied to achieve the final purpose
is the enhancement of all the involved producers. The of the algorithm and finally illustrate how it works
success of these producers is based on the formulation of through simulations.Also, experimental studies of real
an intelligent decision support system, which is able to cases are conducted to demonstrate the effectiveness and
MOSIM’14–November 5-7-2014 - Nancy - France

feasibility of the proposed DSS (section 5).Finally, the methods (MCDA), is a branch of a general class of
‘Conclusions’ section draws some conclusions about Operations Research models which deal with the process
theadvantages, limitations andfuture ameliorations to of making decisions in the presence of multiple
overcome the identified barriers. objectives. These methods, which can handle both
quantitative and qualitative criteria, share the common
2. OVERVIEW OF THE DSS IN THE FIELD OF characteristics ofconflict among criteria,
RENEWABLE ENERGY incommensurable units, and difficulties in
design/selection of alternatives [10].The main purpose of
2.1 Literature review Multi-criteria Decision Analysis is to provide decision
aiding tools that help finding solutions for real-world
Decision supportsystems have, since the 1980s, become problems, most often, problems having conflicting points
an established part of mainstream management of view [25]. The multi-criteria problem is related to the
science[22], and the increased sophistication and methods and procedures by which the different criteria
computational prowess ofcomputer systems make them can be formally involved in the decision process.
an ideal tool for the rather complex model requirements The complexity of energy planning and energy projects
that seembest suited for solving decision problems of the in particular makes multi-criteria analysis a valuable tool
present age. Indeed, there are many researchedand in the decision-making process. A number of conflicting
documented instances of decision support systems being factors, technological, economic, environmental, risk,
critical to decision outcomes the renewable energy social, etc., must be taken into account each time,
field.In most cases, the applications of DSS in renewable whereas different groups of decision makers (DMs) get
energy field include the problems of: Electricity involved in the process. Each group brings along different
Transport problem, Electricity Production, Energy criteria and points of view, which must be resolved within
Efficiency, Energy Demand/Supply problems, Buildings a framework of understanding and mutual compromise.
Energy Management, Carbon Emissions, Cost of Product MCDA field contain several different methods of which
Designs and finally the impact of Pollution on the most important are: Analytic Hierarchy Process
Environment, etc. (AHP), Preference Ranking Organization METHod for
From an investment proposal to complex scientific Enrichment Evaluations (PROMETHEE), ELimination Et
analysis, decision-making is a key aspect in projects. ChoixTraduisant la REalité (ELimination and Choice
Only a few works were focusing their studies on the Expressing REality or more commonly-ELECTRE), and
renewable energy plant projects selection. Patlitzianas et Multi-attribute Utility Theory (MAUT). The comparison
al [7]. presents an information DSS towards the of MCDM methods related to renewable energy planning
formulation of a modern energy companies environment. is discussed in the literature [6, 10–14]. The use of Multi-
It consists of an expert subsystem and a multi-criteria criteria Decision Aid (MCDA) techniques has a long
decision making subsystem based on ELECTRE III and history in energy projects and outranking methods like
developed with Java. This important tool suffers from ELECTRE and PROMETHEE have proved valuable
weaknesses of the ELECTRE III method. Another where many different criteria should be taking in
interesting work was proposed by Zampou et al. consideration.In a previous analysis byPohekarand and
[21]discuss the industrial requirements for an energy M.Ramachandran[10], multi-attribute utility theory
aware information system and system functionalities that (MAUT) was the most common MCDM method used in
should be supported by an energy-aware information energy planning literature.
system. This system doesn’t use a multi-criteria model to At present, we can observe that the most developed
manage the decision maker. DSS especially in renewable energy field,in the greater
On the other hand a number of mathematical theories part, apply Multi-Criteria Decision Making (MCDM)
have been evolved specifically for solving decision methods in order to compile and analyze data input, to
questions. Many of these theories have been found useful produceresults.
in real world decision support models.Most of the
theories tackle the question of optimization [23]. Linear 3. DSS FUNCTIONALITIES ANDCOMPONENTS
programming, mixed integer programming, fuzzy sets,
bayesian nets, and other probabilistic linear and nonlinear One of the most important decision
modeling procedures consistently recur in decision makerresponsibilitiesisthe problem of selection. Decision
support system design [24]. In addition to that, recently, support systems are usually defined as a software
almost the majority of developed DSS choose to use the packageable to assist decision making in complex
so called Multi-criteria Decision Aid methods (MCDA) decision situation, where no simple solution methodology
as a mathematical models to resolve the problems of can be applied.The main functions of aDSS are:
decision-making. a. Integrating decision analysis techniques with data
The next section will undertake the MCDA methods access and management.
and their own contributions in the problems related to b. A user-friendliness “easily used by the decision
renewable energy field especially the selection of the best makers”.
plants project. c. Adopting the changes in the decision environment as
well as the possible adjustments in the decision-making
2.2 Multi-criteria Decision Aid (MCDA) policy and the preferences of the stakeholders.
d. Applying an interactive mode to enable real-time
Multi-criteria analysis, often called Multi-criteria decision-making.
Decision Making (MCDM) or Multi-criteria Decision Aid
MOSIM’14–November 5-7-2014 - Nancy - France

In other way to achieve all of these functionalities, a One of the most common techniques in data analysis is
DSS should have at least some basic elementswhich to group together objects that display similar properties,
are(Figure 1): based on certain criteria. This is a very natural process
a. A database thatincludes all the needed information and and plays a very important role in our lives. In fact, one of
the required data to execute the analysis of the problem at the most basic activities of human beings, when trying to
hand. understand a new object or phenomenon, is to compare it
b. Apowerfulmodel base of a DSS which is a collection of with other known ones by identifying common
decision analysis algorithm, used to support decision characteristics [29, 30]. From a methodological point of
making.The model base management system is view, the process of grouping objects together in data
responsible for the treatmentof the model base including analysis is split in two: classification and
the storage and retrieval of models that are developed, clustering.Classification is a supervised grouping process
their update and adjustment. that relies on a priori information regarding the groups in
c. An Interfaceto conduct the communication between the which to place the objects, which are called classes.
end users and the developed DSS. A special part of the Clustering, on the other hand, does not benefit from any
user interface, the dialog generation and management knowledge on the structure of the data. The purpose of
system are specifically designed to manage this clustering is to group objects together, based on the
communication. natural structure of the data, using measures of similarity.
Hence, objects that are similar should be placed in the
same cluster while objects that are dissimilar should be
placed in different ones.
One of the most well-known partitioning clustering
algorithms is the K-means algorithm, which was
independently discovered in different fields by Steinhaus
[31], Lloyd [32], Ball and Hall [33] and MacQueen [34].
The algorithm can be summarized in the following way: a
set of centroids is generated, usually randomly; the
objects in X are assigned to the centroid that is closest to
them, therefore constructing a partition; a new set of
centroids is then generated from the average evaluations
of the objects from each set inside the partition; all these
steps, except the first, are then repeated until a certain
amount of time, or a certain number of iterations have
passed, or if the centroids no longer change significantly
across consecutive iterations.
Since more than 50 years ago, when K-means was first
proposed, many variations have been created, in order to
deal with some of the shortcomings of this method. Some
of the more important ones pertain to the choice of the
number of clusters the algorithm should retrieve, the
In addition to all these functionalities,the proposed DSS convergence towards locally optimal solutions and the
tool in the next sectionis a Web-based application fact that the resulting clusters are hyper-spherical in
assisting decisions makers to choose among a various shape.
alternatives, the best renewable energy plant, thus making
it easier and less costly to be accessed by all relevant b. Multi-criteria clustering algorithm
stakeholders in geographically distributed locations. Some clustering problems cannot be appropriately
solved with classical clustering algorithms because they
require optimization over more than one criterion. In
general, solutions optimal according to each particular
criterion are not identical. Thus, the problem arises of
how to find the best solution satisfying as much as
In this section, we describe the proposed algorithm and possible all criteria considered. In this sense the set of
the new hybrid developed MCDM-method which is an Pareto efficient clustering was defined a clustering is
aggregation of the well-known K-means algorithm, Pareto efficient if it cannot be improved on any criterion
Relational Multi-criteria Clustering and Profiles concept without sacrificing some other criterion.
based on calculating a Preference functions between all Traditional clustering uses attributes, which means that
the participated alternatives. By combining all these there is no preference order on scale of attributes and no
concepts, we become able to use at the same time the order between clusters [35, 36].
important results of the clustering research, the widely In the field of multi-criteria decision aid, we can usually
used algorithm K-means, and the basic notion of the distinguish three main problems [25]. The first one can be
preferences. formulated as identifying a subset of alternatives that are
considered to be the best or the more interesting ones.
This is referred to as the choice problem. The second one
4.1 Background is related to the ranking of the actions from the best to the
worst one. Finally, in sorting problems, the decision
a. K-means clustering algorithm
MOSIM’14–November 5-7-2014 - Nancy - France

maker focuses on the assignment of alternatives to pre- b. Profile

defined ordered categories [37-39]. Recently some To build the profiles of alternatives, could be
researchers have addressed the question of detecting considered the application of classic multi-criteria tools -
clusters in multi-criteria contexts. In an earlier paper, De Utility Theory, Outranking methods (Electre,
Smet et al. [19] have developed an extension of the Promethee,etc.), AHP, etc. -and the conceptsof preference
traditional K-means algorithm to tackle this problem. The modeling mentioned in theprevious subsection lead to
main contribution of their work relies on the definition of define (Preference,Indifference and Incomparability)
a distance measure that is based on binary preference relations .However, it is important to note
relations between the alternatives. thatonly outranking methods lead to define the
The aforementioned work was the start for many incomparabilityrelation J. When applying othermethods,
researchers in this context. But, he had one important AHP or Utility Theory for instance, theJ relation will
issue which is the lack of information about the relations remain empty and the comparisonbetween pairs of actions
between the detected clusters. In a second work [42] De will be restricted to P andI relations.
Smet and Eppe was notified this issue in their work and Once these relations are elicited, any action
they tried not only to detect clusters in a multi-criteria canbe characterized with a profile which is defined
context but also to identify relations between these as being 4–uple〈 〉 where:
clusters. Typically we are looking for information such as
“two clusters A and B are incomparable while they are { }
both preferred to a third cluster C”. As mentioned in
Cailloux et al. [40], this is referred to as relational multi- { }
criteria clustering.
On the other hand, multi-criteria clustering uses at least { }
criteria. In a special case of multi-criteria clustering, { }
multi-criteria ordered clustering (MCOC), there is also
order on clusters [40, 41].
As aforementioned, note that majority of works don’t
4.2 The Model define the preference function which is the main factor to
calculate the profile of each alternative. One of the main
Let { }be a set of n elements called contributions of our work is a new preference function
‘alternatives’ and let { } be a set of m that we will detail in a next section.
elements called‘criteria’. The clustering model uses the
idea that inside the same cluster, all the alternatives are c. Distance
similar; this means that these alternatives have the same Due to the multi-criteria nature of the problem the
preference relation to more or less the same concept of distance, so widely used in
alternatives.In the following subsections, we will focus on classificationtechniques, does not seem to be well
three basic concepts needed to develop thisclustering suitedhere. This observation constitutes a
model. fundamentalbarrier to the extension of famous
clusteringtechniques, like the K-means algorithm. To
a. Preference modeling avoidthis bottleneck, De Smet [19] defines adistance
This method compares alternatives using preference between alternatives that will take into accounttheir
modeling that constructs so-called profile of each multi-criteria nature.Basically, the intuition behind this
alternative. concept isthat two alternatives will be as close as their
The comparison between alternatives , profiles arealike.
results on one of the following relations: Preference (P), So, let be the profile of , thedistance between two
Indifference (I) or Incomparability (J). alternatives is definedas follows:

∑ | ( )|
( )

Only one of these relations is true between each couple 4.3 The Algorithm
of given alternatives.Indeed, these relations translate
situations ofpreference, indifference and incomparability Our work is based on a particular clustering approach
and itcan be assumed that they satisfy the following from MCDA, the Multi-criteria Relational Clustering
requirements: approach from [19]. This is, to our knowledge, one of the
first contributions to clustering in MCDA, which
additionally takes into account the preferential
information available in this context. Ouralgorithm is an
extension of the k-means algorithm with a MCDA
background. The Multi-criteria clustering (MC clustering)
method utilizes criteria, and not attributes when compared
to classic clustering approaches, however, alternatives are
{ assigned to the K suited clusters. That is why we can say
that it is a Relational Multi-criteria Clustering (RMCC)
MOSIM’14–November 5-7-2014 - Nancy - France

according to taxonomy of clustering methods. In addition o are respectively the index sets of
to that, another contribution of our algorithm is that benefit criteria and cost criteria and is the
always possible to order the resulting clustering, which normalized value of .
makes it not only a RMCC but also an Ordered Multi-
criteria Clustering (OMCC).  Step 3: Now we can calculate values of two
While now let’s set the steps of our approach: functions of Select-ability and Reject-ability for
each alternative as following:
 Step 0: The starting point of this approach is a
decision matrix or eventually an alternatives matrix ∑
which had the following structure: o ∑ ∑

[ ] o ∑

[ ][ ]  Step 4: Afterward, the profile of each

alternative is determined based on one of the 4
relations of comparison between objects, based on
preference relations that express whether an
o { } is a set of the evaluated alternative is indifferent, incomparable or preferred
alternatives, to the other, as follows:
o { } is a set of criteria according ( )
{ }
to which the decision problem will be ( )
evaluated, ( )
{ }
o is the corresponding value of an alternative i ( )
to a criterion j. ( )
( )
 Step 1: Splitting the criteria into two categories: ( )
Select-ability criteria (Set of criterion to ( )
{ }
Maximize/Benefit) and Reject-ability ones (Set of
criterion to Minimize/Cost). Note that each criterion { }
{ }
had a weight that reflects the decision maker
The vector of weightings { }  Step 5: Finally those created profiles will be the
should respect the following conditions: input of our Multi-criteria Relational Clustering
Algorithm. In our case it’s a K-means clustering
o algorithm based on the definition of the
o ∑ aforementioned distance between two alternatives.
Then the method follows the general lines of the K-
 Step 2: Establish a normalized version of the initial means approach, with a slight change:
decision matrix. This step is very important in order
to unify the different unit of each criterion. The o Unlike in classical K-means, an initial set of
structure of the new matrix can be expressed as K clusters or partitions is generated first
follows: and not the set of centroids. We have used an
approach of randomly constructing this initial
o o The centroids are then generated from this
initial set of clusters, through their profiles,
in the following way [43] considering a
Where: centroid of a cluster, an alternative from the
dataset will be placed in one of the four sets
o { } of the profiles of this centroid with the
o { }
condition that it also appears in the same set
MOSIM’14–November 5-7-2014 - Nancy - France

for a majority of the alternatives of that

cluster. For each cluster a centroid is
generated through a profile as defined
the following formula:

|{ ( ) }|
 Step 8: Finally an additional step could be present, if
o A new set of clusters is built from the a small test is validated: if there is no
centroids by assigning each alternative to the incomparability’s relations between the K clusters we
cluster whose centroid it is closest to, using can prooced to aranking between the n alternatives
the distance measure defined before. based on the relations between the K clusters (K =
o This process is repeated until that the K Number of alternatives in this case). Else we can just
partitions are no longer change, or a certain use the relations between the different created
number of iterations have passed in order to clusters to choose the best and worst ones).
eliminate a potential infinite loop.
4.4 Simulation
 Step 6: A relation between any two clusters
is also proposed through a simple To illustrate how the proposed algorithm works, a
simulation of different iterations from the algorithm
majority rule. We calculate the relations between all
process applied on a decision matrix(Table 1) is presented
the K created clusters based on the preference matrix below. Let’s consider a set of six alternatives
(amatrix that contents relations between all the and a set of five criteria Note that,
existents alternatives). only the criterion is a benefit one, while all the rest are
cost criterion. First of all, a preliminary phase consists of
[ ] preparing the input for the algorithm (Normalized
decision matrix, Calculate Preference functions and
[ ][ ] finally Identify sets of profiles for each alternative).
The result of this phase can be summarized in a
preferences matrix (Table 2). The created profiles will be
the input of our algorithm. In our case it’s a K-means
Where: clustering algorithm based on the definition of the
aforementioned distance between two alternatives. Then
the method follows the general lines of the K-means
( ) approach, with a slight change.
An initial set of K=3 clusters or partitions C is
{ generated first and not the set of centroids (Figure 2,
Iteration 1). We have used an approach of randomly
constructing this initial partition. The centroids are then
generated from this initial set of clusters, through their
 Step 7: The definition of a relation between two profiles, in the following way considering a centroid of a
cluster, an alternative from the dataset will be placed in
clusters will be based on the determination of the one of the four sets of the profiles of this centroid with
dominant relation between the objects belonging to the condition that it also appears in the same set for a
each cluster. Formally, we define the frequency of majority of the alternatives of that cluster. In this case,
while choosing 3 clusters on this input data, the algorithm
occurrence of a preference relation when considering
proceeds on 3 iterations until there are no changes (Figure
a couple of clusters as follows: 2).
Now, we should identify the nature of relations between
( ) the created clusters(Figure 3). For example, a cluster
|{ }| means that all alternatives in are better than
alternatives in , same for the relation , a cluster
{ } means that all alternatives in are worse than
alternatives in . In addition, if K = 6 (number of clusters
The proposed measure is computed for each equal number of alternatives) we can order the set of
preference relation belonging to the set { }. alternatives, in a condition that there are no
Then, the dominant relation can be determined by: incomparability relations between the different clusters.
MOSIM’14–November 5-7-2014 - Nancy - France


P+ P- P+ I P+ P+

Our RE-WEB-DSS is a JEE (Java Enterprise Edition) P+ P- P+ P- I P-

dynamic web application based on many technics at
different layers. P+ P- P+ P- P+ I
First of all, we are using the last JSF (Java Server
Table2Preference matrix.
Faces) implementation “Mojarra 2.2.6 included in
Glassfish server 4” as an MVC (Model-View-Controller)
component-based framework to separate the different
layers of our application. Then we used the ORM
EclipseLink JPA implementation (Java Persistence API)
for the data access layer to communicate with our
database for basic CRUD (Create, Read, Update and
Delete) operations.
Finally the presentation layer is designed with
combining the libraryPrimefaces5.0 with the most used
web languages: CSS, JavaScript and Ajax to create end-
users interfaces and to present the results of our decision
making algorithm. AUI (user-interface) example from our
application is shown in Figure 4. Note that many other
API are used to achieve different objectives such as the
encryption/decryption of user’s personal information or
even the parsing of date type objects,etc.
All the used frameworks, API and technologies are open


In order to demonstrate the usefulness and effectiveness

of our DSS, a comparative study of more than 10
published researches had been made. However, we had
chosen to present in this section just one of the published
researchers that facesthe RE selection problematic.

11,4 119,0 5,8 105,0 22080,0

8,6 164,0 9,6 160,0 28100,0

9,0 119,0 4,5 143,0 24650,0

7,9 128,0 6,7 177,0 32700,0

9,4 151,0 7,6 136,0 22750,0

7,9 164,0 8,4 180,0 27350,0

Table1Decision matrix.

I P- P+ P- P- P-

P+ I P+ P+ P+ P+

P- P- I P- P- P-
MOSIM’14–November 5-7-2014 - Nancy - France

conventional power plant) of P >50MW is the best ranked

by both methods.


In this paper we have developed a Decision Support System

(DSS) applied to Renewable Energy field. The established
tool is developed based on an aggregation of the well-known
K-means algorithm, Relational Multi-criteria Clustering and
Profiles concept based on calculating a Preference function
between all the participated alternatives.The originality of
this DSS is due to the used frameworks, API’s and
technologies which all are open sources. DSS based on these
various technologies will provide an analytical and
decisional mechanism from higher level (Evaluation matrix)
to detailed reports ( a sample BI Bar Chart ) to answer many
important questions determining the final decision of the
participant stakeholders.The developed DSS has been tested
on more than ten international published researches. Each
time, the obtained results completely matched the published

Future developments in our RE-DSS explore the inclusion of

We consider the case study of selecting the best electrical a full RE Information System that should cover all RE field
generation technology based on the renewable energy problems such as the Energy Efficiency, Forecasting,
sources in Spain. This study had used the VIKOR method Knowledge Management and last and not least the semantics
to resolve the problem of selection. The decision making as an attempt to make the decision makers aware of aspects
matrix presented in Table 1 include7 criteria and a set of of the decision process that they might not have considered
13 evaluated project. The designed systems will be in first time, and to run consistency checks. Ontology (a
evaluated according to the considered criteria: Power (P), technique for semantic representation) will be used to
Investment Ratio (IR), Implementation Period (IP), represent the domain knowledge of different renewable
energy technologies, requirements and policies to support
Operating Hours (OH), Useful Life (UL), Operation and
the decision process.
Maintenance Costs (O&M) and tons of emissions of CO2
avoided per year (tCO2/y). These emissions are estimated
Criteria/ C1 C2 C3 C4 C5 C6 C7
by the Spanish Government according to the increase in Alternatives
RE projects [1].
Finally as shown in Figure 5, the result of our ranking A1 5000 937 1 2350 20 1.470 1,929,936
method is 100% matching the results find by the VIKOR
method. Effectively, the concerned study gives us, as a A2 10,000 937 1 2350 20 1.470 3,216,560
compromise solution the alternative A12. This
alternative, a Biomass plant (Co-combustion in a A3 25,000 937 1 2350 20 1.510 9,649,680
MOSIM’14–November 5-7-2014 - Nancy - France

A4 50000 1,500 1.5 3100 25 1.450 472,812

multicriteriagroup decision-making framework,
Renewable Energy 28, 961–973.
A5 20,000 700 2 2000 25 0.700 255,490 [4] B. Vucijak, T. Kupusovic, S. MidzcKurtagic, and A.
Ceric, 2013.Applicability of multicriteria decision aid to
A6 35,000 601 2.5 2000 25 0.600 255,490 sustainable hydropower, Applied Energy, 101, 261-267
[5] S.Theodorou, G.Florides and S.Tassou, 2010.The use
A7 50,000 5,000 2 2596 25 4.200 482,856 of multiple criteria decision making methodologies for
the promotion of RES through funding schemes in
A8 5000 1,803 1 7500 15 7.106 2,524,643 Cyprus, A review.Energy Policy 38, 7783–7792
[6] S.PavlosandGeorgilakis, 2006.State of artof decision
A9 5000 1.803 1 7500 15 5.425 2,524,643 support system for the choice of renewable energy
sources for energy supply in isolated regions.
A10 5000 1.803 1 7500 15 5.425 2,524,643
International Journal of Distributed Energy Resources, 2,
A11 5000 1.803 1 7500 15 2.813 2,524,643
[7] D.Konstantinos. Patlitzianas, A. Pappa and J.
Psarras,2008.An information decision support system
A12 56,000 856 1 7500 20 4.560 4,839,548
A13 2000 1.503 1.5 7000 20 2.512 5,905,270
the formulation of a modern energy companies’
environment. Renewable and Sustainable Energy
Table3Alternatives evaluation Matrix. Reviews 12, 790–806
[8] A.DaddaandB.Ouhbi, 2013.A new hybrid MCDM-
Method & Application to renewable energy project
management.Proceeding of IEEE International
Conference on Industrial Engineering and Systems
Management (IESM’2013), Rabat, Morocco.
[9] N.K Sidda, B.Espejo-Garcia, F. . Lope - ellicer,
M. ngelLatre, and F. ara aga-Soria, 2000.Role of
Decision Support System for Renewable Energy
[1] J.R. San Cristóbal, 2011. Multi-criteria decision-
Outreach.Computer Science and System Engineering
making in the selection of a renewable energy project in
Department,University of Zaragoza, María de Luna 3,
spain: The Vikor method. Renewable Energy, 36, 498-
50015, Zaragoza, Spain.
[10] S.D.PohekarandM.Ramachandran, 2004. Application
[2] M.Kabak and M.Dagdeviren,2014. Prioritization of
of multi-criteria decision-making to sustainable energy
renewable energy sources for Turkey by using a hybrid
planning–a review. Renewable and Sustainable Energy
MCDM methodology. Energy Conversion and
Rev 8, 365–381.
Management 79, 25–33
[11] J.Sliogeriene, Z Turskis and D.Streimikiene, 2012.
[3] D.A. Haralambopoulos and H. Polatidis,
Analysis and choice of energy generation technologies:
2003,Renewable energy projects: structuring a
MOSIM’14–November 5-7-2014 - Nancy - France

the multiple criteria assessment on the case study of decision aid clustering.Procedia Computer Science 17,P.
Lithuania. Energy Procedia,32, 11-20. 134 – 140.
[12] H. Polatidis et al, 2006.Selecting an appropriate [29] M. R. Anderberg (1973). Cluster Analysis for
multi-criteria decision analysis technique for renewable Applications Academic Press. (Cited on pages v, 4, 8 and
energy planning. Energy Sources Part B 1, 181–193 11.).
[13] P. Zhou et al, 2006. Decision analysis in energy and [30] B.S. Everitt, S. Landau, and M. Leese (2001).
environmental modeling: an update.Energy.31, P.2604– Cluster Analysis. A Hodder Arnold Publication Hodder
2622. Arnold.(Cited on pages v, 4, 11, 14 and 42.).
[14] F.Cavallaro and L.Ciraolob, 2005. A multicriteria [31] H. Steinhaus (1956). Sur la division des
approach to evaluate wind energy plants on an Italian corpmateriels en parties. Bull. Acad. Polon. Sci, 1, 801–
island. Energy Policy, 235–24. 804. (Cited on page 12.).
[15] G.Munda. 1995. Multicriteria evaluation in a fuzzy [32] S. P. Lloyd (1982). Least squares quantization in
environment Theory and applications in ecological pcm. IEEE Transactionson Information Theory, 28, 129–
economics. hysica−Verlag, Heidelberg. 255 pp. ISBN: 137. (Cited on page 12.).
3-7908-0892-X. Review in Journal of Economic. [33]G. H. Ball and D. J. Hall (1965).ISODATA. A
[16] A.H.I Lee, H.H Chen and H.Y Kang, 2009.Multi- novelmethod of data analysis and pattern classification..
criteria decision making on strategic selection of wind Technical Report., MenloParkStanford Research Institute.
farms. Renewable Energy, 120–126. (Cited on pages 12 and 13.).
[17] C.Macharis, J.Springael, D.Brucker and AVerbeke, [34]J. B. Macqueen (1967). Some methods of
2004. PROMETHEE and AHP: The design of operational classification andanalysis of multivariate
synergies in multicriteria analysis, Strengthening Proceedings of the Fifth Berkeley Symposiumon
PROMETHEE with ideas of AHP. European Journal of Mathematical Statistics and Probability, pages.281–
Operational Research; 307–317. 297.(Cited onpages viii, 12 and 13.).
[35] A. Sharma, C.W. Omlin. Performance Comparison
[18] S.Opricovic and G.Tzeng, 2007.Extended VIKOR of Particle Swarm Optimization with Traditional
method in comparison with outranking methods.European Clustering Algorithms used in Self-Organizing Map.
Journal of Operational Research.514–529. World Academy of Science, Engineering and Technology
[19] De Smet, Y., Montano Guzmán, L.Towards 51, 2009.
multicriteria clusteringAn extensionof the k-means [36] A. J. Hartugan, Clustering algorithms. New YorkJ.
algorithm. European Journal of Operational Research Wiley, 1975.
158, 390–398(2004). [37]Nemery, P., Lamboray, C.FlowSorta flow-based
[20] A.Dadda, B.Ouhbi, 2012. Multi-criteria Decision sorting method with limiting andcentral profiles. TOP
Making in the selection of the Renewable Energy (Journal of the Spanish Society of Statistics and
Project.3ème congrés de la SM2A, Marrakech-Maroc. OperationsResearch) (to appear).
[21] E. Zampou, S. Plitsos, A. Karagiannaki and I. [38] Yu, W.Aide multicritère à la décision dans le cadre
Mourtos, 2014.Towards a framework for energy-aware de la problématique du tri méthodes et applications,
information systems in manufacturing, Computers in PhDthesis, LAMSADE, Universit’e Paris Dauphine, Paris
Industry, 65, 419-433. (1992).
[22] P. Korhonen, 1991. Two Decision Support Systems [39] Zopounidis, C., Doumpos, M.Multicriteria
for Continuous and Discrete Multiple Criteria Decision classification and sorting methods. A literature
Making: VIG and VIMDA. In A. Lewandowski, P. review.European Journal of Operational Research 138,
Serafini, and M.G. Speranza, editors, Methodology, 229–246 (2002).
Implementation and Applications of Decision Support [40] Cailloux, O., Lamboray, C., Nemery, P.A taxonomy
Systems, volume 320 of CISM Courses and Lectures, of clustering procedures. InProceedings of the 66th
International Centre for Mechanical Sciences, Springer Meeting of the EWG on MCDA, Marrakech, Maroc
Verlag, Berlin. (2007).
[23] C.Weber , 2005. Uncertainty in the Electric Power [41] Richard O. Duda, Peter E. Hart, David G. Stork.
Industry: Methods and Models for Decision Support. Pattern Classification ( 2nd edition) 2000.
Springer Publishing, NY. [42] Y. de Smet and S. Eppe (2009). Relational
[24] E.Turban, J.Aronson, T.Liang and R.Sharda, 2007. MulticriteriaClusteringThe Case of Binary Outranking
Decision Support and Business Intelligence Systems, 8th Matrices. in. M. et al. Ehrgott(Editor.)Evolutionary Multi-
Edition, Prentice-Hall, NJ. Criterion Optimization.Fifth international
[25]P.Vincke, 1992. Multicriteria Decision-aid, John conference,EMO 2009.Proceedings, vol. 5467 of Series.
Wiley & Sons, Inc., Chichester. Lecture Notes in ComputerScience, pages. 380–
[26] S.Eppe, J.Roland and Y.DeSmet. On the use of 392Springer Berlin.(Cited on pages viii, 36, 37, 64, 65,
Valued Action Profiles for Relational 79and 125.).
MulticriteriaClustering.CoDE-SMG.Technical Report [43]Lee SK et al (2008) The competitiveness of Korea as
Series Technical Report No.TR/SMG/2011-003. a developer of hydrogen energytechnologythe AHP
[27] Y.DeSmet and S.Eppe, 2009.Multicriteria Relational approach. Energy Policy 361284–1291.
Clustering: The Case of Binary Outranking Matrices. M.
Ehrgott et al. (Eds.): EMO 2009, LNCS 5467, pp. 380–
392, 2009. Springer-Verlag Berlin Heidelberg.
[28] B.Rouba and S.N Bahloul, 2013. Towards the
definition of relations between clusters in multicriteria