Renewable and Sustainable Energy Reviews 16 (2012) 898–903

Contents lists available at SciVerse ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

Performance assessment of Alternative Energy Resources in Brazilian power

sector using Data Envelopment Analysis
Marcos Estellita Lins a , Luciano Basto Oliveira b , Angela Cristina Moreira da Silva a , Luiz Pinguelli Rosa c ,
Amaro Olimpio Pereira Jr. d,∗
a
Programa de Engenharia de Produção (PEP/COPPE/UFRJ), Brazil
b
Instituto Virtual Internacional de Mudanças Globais (IVIG/COPPE/UFRJ), Brazil
c
Programa de Planejamento Energético (PPE/COPPE/UFRJ), Brazil
d
Centro de Estudos Integrados sobre Meio Ambiente e Mudanças Climáticas (CentroClima/COPPE/UFRJ), Caixa Postal 68565, CEP 21941-972, Ilha do Fundão, Rio de Janeiro, Brazil

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

Article history: The Brazilian power sector is known for the strong participation of renewable sources. This character-
Received 23 December 2010 istic is maintained thanks to government incentives for alternative sources with particular emphasis on
Received in revised form 15 August 2011 wind generation, small hydroelectric and sugar cane bagasse fired power plants, because these are more
Accepted 7 September 2011
costly than conventional plants; they are, nevertheless, less costly than other alternative sources, such
Available online 5 October 2011
as residue-based generation. The government’s policy is based, however, on a purely economic analysis.
If socio-environmental variables were to be taken into account the government’s orientation in favor
Keywords:
of renewable sources might be different. The present article uses the Data Envelopment Analysis (DEA)
Data Envelopment Analysis
Performance indicators
method to incorporate such variables in the government’s energy policies. The results demonstrate the
Alternative Energy Resources advantage of promoting residue based generation above that from other sources.
© 2011 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 898
2. Data Envelopment Analysis (DEA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 899
2.1. Basic concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 899
3. Preparation of the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 899
3.1. Preliminary graph analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 901
3.2. Constraints on weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 901
4. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 902
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 903

by conventional thermal, nuclear and, more recently, wind power

1. Introduction
plants.
According to the 2008 National Energy Balance – BEN [1], Brazil
The Brazilian power sector is characterized by the strong
had an installed power generation system with a capacity of about
presence of hydroelectric plants, with large reservoirs of pluri-
104 GW, of which over 78 GW of hydropower, 23 GW of conven-
annual regularization, situated in different hydrographic basins,
tional thermal power plants, 2 GW of nuclear and 414 MW of wind
and distant from the consumer centers. For this reason, extensive
farms.
transmission lines are required, not only to transport the electric-
This profile, however, can change greatly, depending on the
ity generated to the consumers but also to integrate the systems to
growth of electricity demand and the availability of resources for
each geographic region. The hydraulic capacity is complemented
the generation as well as the cost of exploitation of these resources.
While, on one hand, the country has a wide variety of natu-
ral resources, their exploitation, on the other, can result in major
∗ Corresponding author. Tel.: +55 21 2562 8774; fax: +55 21 2562 8777.
investments and significant environmental impacts. In any case, the
E-mail addresses: lins@pep.ufrj.br (M.E. Lins), luciano@ivig.coppe.ufrj.br
(L.B. Oliveira), angela@pep.ufrj.br (A.C.M. da Silva), lpr@adc.coppe.ufrj.br (L.P. Rosa),
country has demonstrated the intention of maintaining a big par-
amaro@ppe.ufrj.br (A.O. Pereira Jr.). ticipation by renewable sources, going as far as to create incentive

1364-0321/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.rser.2011.09.010
 M.E. Lins et al. / Renewable and Sustainable Energy Reviews 16 (2012) 898–903 899

mechanisms for the promotion of such alternatives with empha- From the time it was developed through until today, Data
sis on wind generation, small hydroelectric and sugar cane bagasse Envelopment Analysis (DEA) has been used to study the relative
fired power plants. These technologies are not competitive with efficiency of units in many areas, such as education, hospital admin-
conventional ones, although they are cheaper than other alterna- istration, maintenance units in the US Air Force etc. In some cases,
tive sources, such as residue-based generation. the solutions obtained through DEA result in efficiency ratings and
The question of competitiveness between sources, however, is marginal substitution rates that are hard to construe, meaning that
based on a purely economic analysis. If socio-environmental vari- they are often not accepted by managers. It should be noted that
ables were to be taken into account government incentives might it is of the utmost importance to be familiar with the trade-offs
be directed at other sources. (marginal substitution rates) between the inputs and outputs in a
Urban waste, for instance, has additional advantages normally production process. For example, managers often need to know the
not taking into account, such as: (i) the equipment and input mate- additional amount of a certain input that is required to step up a
rials required for its production are sourced in Brazil and thus priced specific output, while the remaining factors continue unchanged.
in local currency; (ii) it is a labor-intensive option that requires Through the DEA technique, trade-offs between inputs and out-
screening and sorting, in order to separate out recyclable items puts may be calculated under optimum operation conditions on
from waste biomass; (iii) it is normally available close to con- the efficient frontier, through the ratio between the multiplicatory
sumers, resulting in lower electricity transmission costs and losses; factors associated with each input/output pair, two inputs or two
and (iv) it helps reducing pollution through replacing fossil fuels outputs. Thus, if u1 and u2 are the multiplicatory factors for output
by Alternative Energy Resources and avoiding the decomposition 1 and output 2 respectively obtained for a given DMU j, the u1 /u2
of solid wastes and GHG emissions. coefficient will represent the amount by which output 2 of the DMU
These advantages may be assessed through an integrated and j must be increased in order to offset a reduction unit in the out-
qualitative technical, social, economic and environmental analysis put 1 of the DMU j. Thus, the slants of the corresponding segments
of the various effects of this use, through adequate methodology. of the efficient frontier defined by the DEA technique represent the
Thus, the present study addresses the application of the Data Envel- trade-offs between the inputs and outputs, although the calculation
opment Analysis (DEA) method to a study of the performance of of the trade-offs obtained through the DEA multiplicatory factors
Alternative Energy Resources for the Brazilian power sector, in presents some difficulties. In mathematical terms, the trade-offs
order to show the importance of this tool in the energy policy represent the partial derivates on the frontier, meaning the angle of
formulation. the production frontier at a specific point. The DEA technique builds
up a linear frontier by parts, in order to approach the production
2. Data Envelopment Analysis (DEA) frontier. The DEA frontier has the following constraints:

When the electricity generated by the Alternative Energy 1. It is continuous, but at the points corresponding to the extremely
Resources is more expensive than that produced by traditional efficient DMUs, the derivates are not continuous [4]. This fact
resources, meaning when the assessment criteria are only eco- means that the problem of the DEA multiplicatory agents, pre-
nomic, the Alternative Energy Resources become feasible only pared to measure the efficiency of extremely efficient DMUs, has
where there is no supply based on traditional resources. How- multiple optimum solutions.
ever, there are social, environmental and technological aspects that 2. The DEA frontier has efficient non-Pareto-efficient/Koopmans
are well able to justify a different approach to assessment, based regions where there are no clearly defined marginal substitution
on the concept of Sustainable Development.1 To do so, method- rates because at least one of the multiplicatory agents deter-
ologies able to handle multiple criteria must be used, while also mining the hyperplane equation characterizing this region has a
considering input/output type relations [2]. In this paper, a quan- value equal to zero. In DEA models with variable returns of scale,
titative method has been selected, based on linear programming: there will be an efficient DMU for each input and output (with
Data Envelopment Analysis (DEA), which is described below. the minimum value of an input or maximum value of an out-
put). It is easy to show that each of these DMUs corresponds to
2.1. Basic concepts a non-Pareto-efficiency region. In models with constant returns
of scale, for each input/output pair, the DMU with the highest
Calculating the efficiency of organizational units has been an output/input relation will be efficient, determining the origin of
important topic for management, although hard to handle, par- a non-Pareto-efficiency region.
ticularly when multiple inputs (resources) and multiple outputs
(services, products and others) are associated with these units. Pro- The main problem of the DEA models is not the existence of non-
posed ways of addressing this problem are found in the paper by Pareto-efficiency regions, but rather the projection of inefficient
Farrell [3], from which an empirical relative efficiency frontier was DMUs in these regions. This occurs mainly because the classic DEA
derived, instead of a theoretical production function, used as basis models assume radial spoke displacement projections, for the sake
for measuring the relative efficiency of the units. of simplicity. Alternative ways of dealing with this issue include
Charnes et al. [4] created a technique based on linear program- weight restrictions [5], multi-objective models [6] or the use of
ming for calculating the relative efficiency of the units, grounded on artificial DMUs as proposed by Thanassoulis and Allen [7].
the proposal presented by Farrell [3], and established a point of ref- In this paper, we opted for the first alternative.
erence on the frontier for each inefficient unit. They called their new
approach to efficiency measurement: Data Envelopment Analysis
(DEA) and the units under assessment were called Decision Making 3. Preparation of the model
Units (DMUs).
The units whose performance will be assessed are the Alterna-
tive Energy Resources for generating electricity, listed in the first
1
column of Table 1. The criteria through which these decision units
Integrated analysis of social, environmental, technical, operating and economic
feasibilities, in order to ensure the development of today’s populations without
will be assessed are listed on the first line of this table.
adversely affecting the living conditions of future generations, which is an attitude It was decided to use a variable to deal with each of the dimen-
grounded on inter-generational accountability. sions of sustainability. Thus for the environmental dimension, gas
900 M.E. Lins et al. / Renewable and Sustainable Energy Reviews 16 (2012) 898–903

Table 1
Initial input and output data for the Alternative Energy Resources.

Alternative Energy Resources Greenhouse Gases Potential Job Creation Potential Distributed Power O&M + CC Cost Investment Cost
Emission (tCO2 /GW h) (job/TW h) Generation (GW h/year) (US$/MW h) (US$/MW h)

1. NGCC 452 7.21 83,220 52.00a 18

2. NGT 600 7.36 81,468 43.00a 27
3. Wind – 14.27 17,520 7.00 43
4. Solar PV – 6.12 49,056 4.00 76
5 SHPP 1 12.84 21,024 8.51 21.49
6. Rice Husks −1223 43.90 6833 −3.28 24.98
7. GDL −955 35,347.69 28,330 10.86 46.67
8 DRANCO −263 14,430.86 69,587 −26.52 33.48
9. Incineration −524 20,233.78 49,620 −4.71 62.63
10. CCO −278 15,818.92 63,620 8.20 51.52
11. BIG/STIG −53.57 1.88 133,296 62.53 14.96

Source: Oliveira [8].

Labels: NGCC (Natural Gas Combined Cycle); NGT (Natural Gas Turbine); Solar PV (Photovoltaic); SHPP (Small Hydro Power Plant); GDL (Garbage Gas + Conservation);
DRANCO (Garbage Gas from Digesters + Conservation); Incineration (steam from garbage + conservation); CCO (Optimized Combined Cycle with Incineration); (BIG/STIG)
Cane Bagasse + P&P.
a
Calculated based on the July 2007 electrical sector auction, in which the price of R$ 140/MW h was of no interest to the natural gas powered Thermoelectric Power
Stations. The O&M + CC cost was calculated as the difference between this value and the investment.

emissions responsible for greenhouse gases were used; for the Rows 8–10, the waste gas route is compatibilized with the oth-
social dimension, the number of jobs created; for the economic, ers, as landfill gases may be tapped regardless of the technological
the investment cost; for the operational, the operating and main- route used for the new garbage.
tenance costs, plus fuel cost; for the strategic, the energy supply As Data Envelopment Analysis (DEA) allows the variables to be
potential. classified as inputs or outputs, those selected in compliance with
In order to use the above amounts in the Data Envelopment the Sustainable Development concept were:
Analysis (DEA) it was necessary to alter the variables so as to avoid
negative or nil values. To do so, all the cells in each column with • Inputs: number of jobs generated, potential electricity provided,
negative or nil values were added to the lowest value in this column, and Greenhouse Gases Emissions;
with the addition of one unit. Thus, all the cells in the Greenhouse • Outputs: investment cost and operations and maintenance costs.
Gases Emissions column total were 1244.57, while the values in the
O&M + CC Costs column were increased by 27.52. These changes do It is noted that Greenhouse Gases Emissions constitute an unde-
not affect the performance ranking, but only the indicator value. sirable output. The undesirable outputs may be included in DEA
Rows 7–10 represent the technological routes for turning waste models through four main approaches. The first uses an inverse
into energy, enhancing power generation and conservation through transformation, according to Golany and Roll [9]; Scheel [10], and
recycling. At this point it is important to highlight the factors on Lovell et al. [11]. The second considers the output as an input,
which estimates were based: in the case of GDL recuperation the according to Omo, in Rheinhard et al. [12], Scheel [10]. The third
organic portion refers to 60% of MSW availability (55 Mt/y) and its proposes to reverse the output signal and add a positive scalar, valid
conversion into electricity is 0.15 MW h/t, while the use of dispos- only for the DEA BCC models [13] and Additive [4], as the CCR is not
able recyclable materials (35%), whose energy saving factor is of invariant for translation. The fourth option, proposed by Färe and
3 MW h/t, was of the order of 40%; in the case of Anaerobic Diges- Grosskopf [14,15], consists of considering the congestion hypothe-
tion the conversion factor is 0.5 MW h/t and the use of recyclables sis on the undesirable output, with no need to maintain the output
reaches 90%. In incineration and CCO recycling is reduced, due to level when some input increases. However, Dyckhoff and Allen [16]
the need for part of this material to be used to add calorific value stress that this proposal should only be used if the decision-taker is
to the organic fraction. In incineration, where the electricity gen- certain of the technical links between the undesirable output and
eration factor reaches 0.7 MW h/t, 85% of the MSW was used and known inputs and outputs.
recyclable represent 10% of the MSW, while in CCO, generation fac- In this paper, we decided to consider the undesirable output as
tor reaches 0.9 MW h/t, 80% of MSW was used for generation and a proxy for a depletable environmental resource, representing this
15% for recycling. as an input.

Table 2
Input and output entry data.

Alternative Energy Resources Inputs Outputs

Greenhouse Gases Potential Job Creation Potential Distributed O&M + CC Cost Investment Cost
Emission (tCO2 /GW h) (job/TW h) Generation (GW h/year) (US$/MW h) (US$/MW h)

1. NGCC 1676.57 600 83,220 79.52 18.00

2. NGT 1824.57 600 81,468 70.52 27.00
3. Wind 1224.57 250 17,520 34.52 43.00
4. Solar PV 1224.57 300 49,056 31.52 76.00
5 SHPP 1225.57 270 21,024 36.03 21.49
6. Rice Husks 1.00 300 6833 24.24 24.98
7. GDL 269.29 1,001,400 28,330 38.38 46.67
8 DRANCO 961.43 1,004,200 69,587 1.00 33.48
9. Incineration 699.84 1,004,000 49,620 22.81 62.63
10. CCO 946.35 1,006,400 63,620 35.72 51.52
11. BIG/STIG 1171.00 250 133,296 90.05 14.96

Source: Prepared by the authors.

 M.E. Lins et al. / Renewable and Sustainable Energy Reviews 16 (2012) 898–903 901

Fig. 1. O&M + CC Cost versus Potential Job Creation and Potential Distributed Gen-
eration. Labels: UTE GN CC (NGCC); UTE GN Merchant (NGT); Solar (Solar PV).
Fig. 3. Investment Cost versus Potential Job Creation and Energy Generation. Labels:
UTE GN CC (NGCC); UTE GN Merchant (NGT); Solar (Solar PV).
The two last columns of Table 2 represent the outputs (O), while
the other three represent the inputs (I). The model logic is to assign model with constraints on the weights in order to avoid the Pareto-
priority to the lowest input and the highest output. inefficient regions and adapt the weight to the specialist’s opinion.
The graph analysis presented below:
3.1. Preliminary graph analysis The first graph in Fig. 1 shows a Pareto-efficiency face and a peak
when considering the “O&M Cost Variables” (input), “Distributed
Three-dimensional graph analysis may be important for help- Generation” (output) and “Job Creation” (output). Three alterna-
ing construe the findings produced by the complete model. This is tive efficient resources are shown (CCO, Dranco and bagasse + P&P,
possible through the development of the Interactive Data Envelop- BIG/STG) which constitute only two Pareto efficient frames. It must
ment Analysis (IDEAL) software at COPPE. It is important to note be pointed out that DRANCO presents the most stable situation in
that the DMUs shown as efficient in the partial graph analyses will the DMU assembly.
also be so in the full classic model. However, as noted, we will use The second set of variables analyzed corresponds to the “Energy
Generation”, “Job Creation” and “Greenhouse Gases Emissions” out-
puts, shown in Fig. 2. It is noted that the efficient frontier consists of
three faces: one defined by the “Garbage Gas”, “DRANCO” and “cane
bagasse + P&R (BIG/STIG)” DMUs, other by “Garbage Gas”, “CCO”
and “DRANCO” and the last by “Garbage Gas”, “Cane Bagasse + P&P
(BIG/STIG)” and “Rice Husks”.
The third set of variables, consisting of “Investment”, “Job Cre-
ation” and “Energy Generation”, is shown in Fig. 3. In this case, we
have only two frames, one constituted by the sugar cane bagasse
P&P (BIG/STIG) technology and the other by “CCO”.

3.2. Constraints on weights

Wong and Beasley [17] explore the use of constraints on virtual

inputs/outputs, defined as the product of the input value by the
weight assigned to it in the DEA multiplicatory agent model. The
proportion of the total virtual of the DMU j may be curtailed, used
by the output r, i.e., the “importance” of output r by DMU j at the

Table 3
Limits assigned to constraints on input and output weights.

Inputs and outputs Lower Upper

Potential Distributed Generation (GW h/year) 0.4 0.6

Potential Job Creation (job/TW h) 0.4 0.6
Greenhouse Gases Emission (tCO2 /GW h) 0.3 0.5
Investment Costs (US$/MW h) 0.3 0.5
O&M + Fuel Costs 0.3 0.5
Fig. 2. Greenhouse Gases Emission versus Potential Distributed Generation and Job
Creation. Labels: UTE GN CC (NGCC); UTE GN Merchant (NGT); Solar (Solar PV). Source: Authors based on Wong and Beasley’s [17] methodology.
902 M.E. Lins et al. / Renewable and Sustainable Energy Reviews 16 (2012) 898–903

Table 4
Findings for the model with constraints on the virtual weights.

Alternative Energy Resources Efficiency Virtual weights

Inputs Outputs

Greenhouse Gases O&M + CC Cost Investment Cost Potential Job Creation Potential Distributed
Emission (tCO2 /GW h) (US$/MW h) (US$/MW h) (job/TW h) Generation (GW h/year)

1. NGCC 0.65 0.30 0.40 0.30 0.60 0.40

2. NGOC 0.47 0.30 0.40 0.30 0.60 0.40
3. Wind 0.46 0.30 0.30 0.40 0.60 0.40
4. Solar PV 0.39 0.32 0.30 0.38 0.50 0.50
5 SHPP 0.67 0.30 0.40 0.30 0.60 0.40
6. Rice Husks 1.00 0.30 0.30 0.40 0.60 0.40
7. GDL 1.00 0.40 0.30 0.30 0.59 0.41
8 DRANCO 1.00 0.30 0.40 0.30 0.50 0.50
9. Incineration 0.68 0.40 0.30 0.30 0.40 0.60
10. CCO 0.60 0.40 0.30 0.30 0.40 0.60
11. BIG/STIG 0.75 0.30 0.40 0.30 0.60 0.40

interval between [r , r ], with r and r being determined by the Table 5

Alternative Energy Resources ranking by hierarchical options.
decision-taker or user. Thus, the constraint on virtual output r is
presented below: Alternative Energy Ranking
Resources
ur yrj Without With With conservation
r ≤ s ≤ r (1) conservation conservation and DEA model
u y
r=1 r rj
s 1. NGCC 7 8 7
where u y represents the total virtual output of the DMU j.
r=1 r rj
2. NGOC 8 9 9
3. Wind 3 4 10
A similar constraint may be imposed on virtual inputs.
4. Solar PV 10 11 11
There are some alternatives for implementing this type of con- 5 SHPP 2 3 6
straint, stressing the following, suggested by Wong and Beasley 6. Rice Husks 1 2 3
[17]: 7. GDL 5 5 2
8 DRANCO 4 1 1
9. Incineration 11 6 5
• Impose the constraint only on the DMU under assessment, thus,
10. CCO 6 7 8
each DMU is analyzed with two additional constraints. The incon- 11. BIG/STIG 9 10 4
venient aspect of this approach is that a benchmark DMU may not
always need to be efficient, in turn.
• Add this constraint to all the DMUs. Thus, each DMU is assessed In comparing the result of hierarchisation based exclusively
with 2N additional constraints, with N being the total number on Cost Benefit Indexation and without considering the conser-
of DMUs. This procedure results in frequent blocks in the linear vation of energy provided by recycling, it can be verified that the
programming problem, as stressed by Lins et al. [18]. alternatives for energy recovery from urban waste, when energy
conservation resulting from recycling is not taken into account, do
Variables are used with associated weightings in DEA method- not figure among the best three – occupying the 4th, 5th, 6th and
ology. The ideal for restrictive application is that these should be 11th positions. This analysis is consistent with investment factors
the result of opinion research of specialists, following the establish- typical of this sector and with the levels of generation efficiency
ing of a sampling plan. When this is not applicable, the narrowest found in each technology currently in vogue.
range of each variable can be estimated by means of simulation, the Once the conservation for each alternative is taken into account
alternative applied in the present study. Thus, in order to apply the the results change; that which was in 4th position becomes the best
DEA model and ensure its feasibility, we decided to impose con- alternative while the rest remain 5th (same as in the first hierar-
straints on the virtual inputs and outputs only for the DMU under chy), 6th (previously 11th) and 7th (previously 6th). This leads to
analysis. The acceptable range is presented in Table 3. the conclusion that the effect of conservation is very significant in
The findings obtained with virtual weights are presented in technologies for energy recovery from urban waste, although this is
Table 4. not the case with the Optimized Combined Cycle – a technological
The findings show that three technologies tie for first place: alternative where increase in investment and conservation are not
“Rice Husks” and two garbage-based options: “Garbage Gas” and significant when compared with that of incineration.
“DRANCO”. Of these only the “DRANCO” presents a performance Once energy conservation resulting from recycling is taken into
indicator with maximum value in the three partial analysis graphs account in the multicriterion analysis represented by the DEA
of item 3.3 which used classical models. It is noted that “GDL” and model, the hierarchy remains in first place, the 5th placed tech-
“Rice Husk” technologies showed themselves efficient in partial nology becomes 2nd, the 6th placed becomes 6th, but that which
analyses only when greenhouse gases were considered as input. was in 7th place goes down to 8th.
The “Bagasse + P&P (BIG/STIG)” and “CCO” technologies, apart
from being found efficient in all the partial analyses made without
restriction to averages, only reached the fourth and eighth ranking 4. Conclusion
respectively in the analysis using restriction to averages and under
the imposed limits, as presented in Table 3. Renewable Energy Resources are starting to play an impor-
The application of the benchmark concept allows the tied tant role in electricity supplies again, after having lost out to
alternatives to be ranked. Table 5 presents this ranking by Data mineral resources, except in Brazil. The advantages offered by
Envelopment Analysis (DEA) for the Alternative Energy Resources. large-scale power plants have proved insufficient, when assessed
 M.E. Lins et al. / Renewable and Sustainable Energy Reviews 16 (2012) 898–903 903

on the basis of broad-ranging assumptions, taking into consider- This justifies a government policy stipulating exclusive service
ation factors that are not measurable solely in economic terms, through technologies making good use of solid urban wastes, for a
such as Job Creation, pollutant emission control and system minimum period to be defined, of a portion of the price to be stip-
safety. ulated by the Ministry of Mines and Energy (MME) for renewable
However, traditional decision-making support tools are not yet energy resources, in order to encourage the fine-tuning of equip-
adapted to display the advantages of renewable energy resources, ment production that will result in gains of scale and optimized
with only a few successful cases being known, such as Brazil’s processes.
hydro-power plants that account for over 73% of electricity gen-
erated nationwide, but with an installed capacity similar to that of References
the USA, which reaches only 30%.
In order to take into consideration the variables most rel- [1] EPE – Empresa de Pesquisa Energética. Balanço Energético Nacional – Ano Base
2008. Brasil/Ministério de Minas e Energia – MME; 2009.
evant to renewable energy resources, without neglecting those [2] La Rovere EL, Soares JB, Oliveira LB, Lauria T. Sustainable expansion of electricity
usually assessed for traditional energy resources, the Sustain- sector: sustainability indicators as an instrument to support decision making.
able Development Concept was used, which consists of complying Renewable and Sustainable Energy Reviews 2010;14:422–9.
[3] Farrell MJ. The measurement of productive efficiency. Journal of the Royal
with environmental, social, economic, technological and operating Statistical Society, Serialize THE, General 1957;3:253–81.
requirements. [4] Charnes A, Cooper WW, Rhodes E. Measuring the efficiency of decision-making
The application of this concept involves multiple criteria and units. European Journal of Operational Research 1978;2(6):429–44.
[5] Pedraja-Chapparo R, Salinas-Jimenez J, Smith P. On the role of weight
requires a methodology that can assess the efficiency of pro-
restrictions in Data Envelopment Analysis. Journal of Productivity Analysis
duction units based on multiple inputs and outputs. To do so, 1997;8:215–30.
the Data Envelopment Analysis (DEA) technique was selected, [6] Lins MPE, Meza LA, Silva ACM. A multi-objective approach to determine
based on linear programming, in order to calculate the rel- alternative targets in Data Envelopment Analysis. Journal of the Operational
Research Society 2004;55:1090–101.
ative efficiency of the units, represented by drawing up an [7] Thanassoulis E, Allen R. Simulating weight restrictions in Data Envel-
efficiency frontier. This methodology can determine a refer- opment Analysis by means of unobserved DMUs. Management Science
ence point on the frontier for each inefficient unit, taking 1998;44:586–94.
[8] Oliveira LB. Potencial de aproveitamento energético de lixo e de biodiesel
into account each of the aspects under analysis, based on a de insumos residuais no Brasil. D.Sc. thesis. Energy Planning Program.
range of variations assigned by a specialist or the decision- COPPE/UFRJ; 2004. p. 237.
taker. [9] Golany B, Roll Y. An application procedure for DEA. Omega – The International
Journal of Management Science 1989;17:237–50.
As Data Envelopment Analysis (DEA) requires the definition of [10] Scheel H. Undesirable outputs in efficiency evaluations. European Journal of
groups of input and output variables, those selected in compli- Operational Research 2001;32:400–10.
ance with the Sustainable Development Concept were divided up [11] Lovell CAK, Pastor JT, Turner JA. Measuring macroeconomic performance in
the OECD: a comparison of European and non-European countries. European
as follows: Journal of Operational Research 1995;87(3):507–18.
[12] Rheinhard S, Lovell CAK, Thijssen G. Environmental efficiency with multiple
• Inputs: investment cost and operations and maintenance costs; environmentally detrimental variables estimated with SFA and DEA. European
• Outputs: number of jobs generated, potential electricity pro- Journal of Operational Research 2000;121:287–303.
[13] Banker RD, Charnes A, Cooper WW. Some models to estimate technical
vided, and Greenhouse Gases Emissions. and scale inefficiencies in Data Envelopment Analysis. Management Science
1984;30(9):1078–92.
The findings show that technologies using solid wastes to gener- [14] Färe R, Grosskopf S. Environmental decision models with joint outputs. Eco-
nomics working paper archive, Economic Department, Washington University,
ate energy should be assigned higher priority than the other options US; 1995.
analyzed, including those fueled by natural gas and other renew- [15] Färe R, Grosskopf S. Optimization models applied to productivity and the envi-
able energy resources. ronment. In: XXXIII Simpósio Brasileiro de Pesquisa Operacional. 2001.
[16] Dyckhoff H, Allen K. Measuring ecological efficiency with data envelopment
These findings are important to assist decision taking at various analysis (DEA). European Journal of Operational Research 2001;132(2):312–25.
levels of government on this matter, as the option for distributed [17] Wong YHB, Beasley JE. Restricting weight flexibility in data envelopment anal-
power generation using renewable resources based on solid wastes ysis. Journal of the Operational Research Society 1990;41(9):829–35.
[18] Lins MPE, Silva ACM, Lovell CAK. Avoiding infeasibility in DEA mod-
may well cover 30% of consumption nationwide at competitive els with weight restrictions. European Journal of Operational Research
prices, lessening pollution and offering jobs to poorly skilled 2007;181(2):956–66.
workers.