Download as pdf or txt
Download as pdf or txt
You are on page 1of 23

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/263152108

A multi-criteria evaluation framework for alternative light-duty vehicles


technologies

Article  in  International Journal of Multicriteria Decision Making · January 2011


DOI: 10.1504/IJMCDM.2011.039588

CITATIONS READS

10 251

3 authors:

Reza Fazeli Vítor Leal


Australian National University University of Porto
33 PUBLICATIONS   196 CITATIONS    70 PUBLICATIONS   1,371 CITATIONS   

SEE PROFILE SEE PROFILE

Jorge Pinho de Sousa


University of Porto
133 PUBLICATIONS   1,475 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

THINK Smartcities View project

Modelling, Simulation and Optimization of Lelystad Airport View project

All content following this page was uploaded by Reza Fazeli on 13 January 2016.

The user has requested enhancement of the downloaded file.


230 Int. J. Multicriteria Decision Making, Vol. 1, No. 2, 2011

A multi-criteria evaluation framework for alternative


light-duty vehicles technologies

Reza Fazeli* and Vitor Leal


Department of Mechanical Engineering,
Faculty of Engineering, University of Porto,
Rua Dr. Roberto Frias, 4200 – 465 Porto, Portugal
E-mail: reza.fazeli@fe.up.pt
E-mail: vleal@fe.up.pt
*Corresponding author

Jorge P. Sousa
Faculty of Engineering, University of Porto,
Rua Dr. Roberto Frias, 4200 – 465 Porto, Portugal
and
INESC Porto,
Campus of FEUP – Rua Dr. Roberto Frias,
378 4200-465, Porto, Portugal
E-mail: jsousa@inescporto.pt

Abstract: This paper presents a multi-criteria evaluation framework for


the choice of alternative fuel/technology options of light-duty vehicle fleets in
a mid-term horizon. These alternatives include not only fleets with one
single technology but also fleets combining several fuel/technologies. A
comprehensive literature review has identified the key metrics for comparing
different pathways: user’s acceptance, emissions of pollutants to atmosphere,
risk of the technology development, transition costs, and availability of fuel
supply. To identify the preferred alternatives, a sequential screening process
was applied, starting with a Pareto optimal (PO) approach, followed by a data
envelopment analysis (DEA) and a trade-off weights (TW) procedure. To
illustrate the approach, the methodology was applied to the case of Portugal.

Keywords: multicriteria decision making; utility function; sequential screening


process; alternative fuel vehicle; electric vehicle; Portugal.

Reference to this paper should be made as follows: Fazeli, R., Leal, L. and
Sousa, J.P. (2011) ‘A multi-criteria evaluation framework for alternative
light-duty vehicles technologies’, Int. J. Multicriteria Decision Making, Vol. 1,
No. 2, pp.230–251.

Biographical notes: Reza Fazeli is a PhD student in Sustainable Energy


Systems at the MIT Portugal Program, based in the Faculty of Engineering of
the University of Porto. He is developing a thesis on the ‘Development of a
multi criteria decision procedure to assist the design of pathways for
sustainable road transportation’. His areas of research are multi-criteria
decision analysis, energy planning in the transportation sector and assessment
of alternative fuel vehicle technologies. He received his BSc in Mechanical
Engineering and his MSc in Energy System Engineering from the Sharif
University of Technology, Iran.

Copyright © 2011 Inderscience Enterprises Ltd.


A multi-criteria evaluation framework 231

Vítor Leal is an Assistant Professor in Sustainable Energy Systems at the


Faculty of Engineering of the University of Porto. He has a Diploma in Physics
Engineering from the Technical University of Lisbon and a PhD in Engineering
Sciences from the University of Porto. His research activity started in the
energy efficiency of buildings. In recent years, and integrated in the MIT
Portugal Program, his interests have evolved into the broader context of
sustainable energy systems, comprehending integrated modelling of the energy
supply, transformation and demand chains at national, regional and municipal
geographic scales, for application in the development of forecast and backcast
evolution pathways.

Jorge Pinho de Sousa is an Associate Professor in the Faculty of Engineering of


the University of Porto and he coordinates the Manufacturing Systems
Engineering Unit of INESC Porto. He has a PhD in Operations Research by the
Université Catholique de Louvain, in Belgium. His research interests are
on decision support systems, combinatorial optimisation, metaheuristics,
operations management, transportation systems and collaborative networks.
He has been involved in several large European R&D projects. He currently
participates in the MIT Portugal Program, as a member of the steering
committee of the transportation systems area. He was the President of the
Portuguese Operations Research Society.

1 Introduction and background

This research aims at developing a multi-criteria framework for the evaluation of


alternative fuel/technology options for light-duty vehicles in a mid-term horizon. Such a
framework is intended to assist policy makers and governments to make decisions over
the next twenty years concerning the development of infrastructures or the establishment
of incentives to promote alternative fuel vehicles with a mid-term vision. In fact, this
research is based on a backcasting approach, which will start by identifying the preferred
option revealed by decision process, and then focus on how to develop mechanisms to
support the adoption of that alternative.
The first section of this paper describes the problem and presents a literature review
covering several previous studies. The model structure and key parameters are presented
in the second section. The proposed decision methodology is presented in the third
division. The fourth part illustrates the results of applying the methodology to Portugal. A
sensitivity analysis was performed to assess the impact of several critical factors on the
results. The final section presents and discusses the main findings and recommendations
achieved with this analysis.
In the literature, several different approaches have been proposed to assess alternative
fuel/technology vehicles, including life cycle analysis (LCA), societal life cycle cost
analysis and multi-criteria decision analysis (MCDA).
LCA is a ‘cradle-to-grave’ approach for assessing industrial systems. ‘cradle-to-
grave’ begins by gathering raw materials from the earth to create the product and ends at
the point when all materials are returned to the earth. LCA enables the estimation of the
cumulative environmental impacts resulting from all stages in the product life cycle,
often including impacts not considered in most traditional analyses (SAIC, 2006). Ideally,
the purpose of LCA is to analyse and assess the changes in a system considering its entire
life cycle. In practice, however, most LCAs do not specify or analyse a policy, but just
232 R. Fazeli et al.

assume (implicitly) that one simple and narrowly defined set of activities replaces another
(Delucchi, 2004).
More comprehensively, the so-called Societal Life Cycle Cost approach is proposed
to compare alternative automotive engine/fuel options based on different aspects
including vehicle first cost (assuming large-scale mass production), fuel costs (assuming
a fully developed fuel infrastructure), externality costs for oil supply security, and
damage costs for emissions of air pollutants and GHGs calculated over the full fuel cycle
(Ogden et al., 2004).
There are however, some critical concerns about these two approaches. First, these
methods are dependent on the accuracy of the collected data which may result in
considerable uncertainty of the conclusions. Secondly, decision maker’s preferences in
attribute weighting cannot be incorporated in the above methodologies.
MCDA approaches can rank alternative fuel-technologies according to different
concerns including technical, environmental, economical and social aspects. There are
several benefits of using a multi-criteria assessment approach:

• impacts do not need to be monetised as they can be expressed through a variety of


measurement units

• the approach can combine values and knowledge from experts and non-experts

• trade-offs are explicitly considered

• the results can be used to aid decision-making when several competing social and
political criteria are considered.
In terms of disadvantages, the main problems with MCDA are the fact that it often
requires much information, which sometimes is available with poor quality, and the fact
model structuring strongly depends on the modeller’s skills, which in turn may affect the
final results.
From the literature, and specifically from a review study by Delucchi (2004), it is
clear that there are many factors that can potentially affect the comparison of different
pathways. Therefore to characterise the previous studies, the following factors were
employed:

• region: the country or region covered by the analysis can affect the results (mainly
due to the feedstock characterisation and already existing infrastructure)

• time frame: the target year of the analysis

• vehicle drivetrain type: including internal combustion-engine vehicles (ICEVs),


hybrid-electric vehicles (HEVs), battery-powered electric vehicles (BPEVs),
fuel-cell powered electric vehicles (FCEVs)

• fuels: fuels carried and used by motor vehicles – gasoline, diesel, biodiesel,
bioethanol, Fischer-Tropsch diesel (FTD), compressed natural gas (CNG), liquefied
natural gas (LNG), compressed hydrogen (CH2), liquefied hydrogen (LH2),
dimethyl ether (DME)

• feedstock: the feedstock from which the fuels are made


A multi-criteria evaluation framework 233

• vehicle life cycle: the life cycle of materials and vehicles including raw material
production and transport, manufacturing of finished materials, assembly of parts and
vehicles, maintenance and repair, and disposal

• GreenHouse Gases (GHGs): the pollutants which are included in the analysis of
CO2-equivalent emissions

• infrastructure: the life cycle of energy and materials used to make and maintain
infrastructures, such as fuel stations, roads, equipment, rail lines, etc.
Another conclusion from the survey was that there are two key different perspectives to
tackle the evaluation of alternative fuel-technology vehicles:
• a policy maker’s perspective (representing the social interest)
• the user’s preferences (representing the individual interest).
From the policy maker’s perspective, the focus is on universal issues such as GHG
emissions and security of fuel supply. This approach is adequate when the target is to
incorporate the stakeholder or national-wide DM’s concerns in the decision making
process. In contrast, looking from the users’ perspective requires a different set of
metrics, in which the focus is more on vehicle characteristics such as the initial cost,
safety and performance. A categorisation of some papers with a policy maker’s approach
is presented in Table 1.
As it was mentioned before, several articles follow the user’s perspective approach.
multinominal logit models were applied in most of these studies in order to investigate
the impact of alternative vehicles’ characteristics on the vehicle choice process.
Some of those studies were classified based on the key factors presented in Table 2.
In this research, our aim is to combine these two approaches, by considering the
decision maker as an agent that takes into account both perspectives. Therefore, some
very different aspects such as GHG emissions and fuel supply security, as well as initial
cost and fuel cost of alternative fuel/technology vehicles will be included. In fact, a
significant added value of our approach is that it has the potential to avoid the conflict
between the social and the user perspective by systemically including the individual
perspective in the social perspective decision system.
The first and probably the most important step for tackling a problem in a multiple
criteria perspective is the identification of a ‘problematique’, in line with the concept
introduced into the field by Roy (1985). He proposed four different kinds of problems as
problematiques in MCDA including choice, sorting, ranking and description problems.
Another problem categorisation was added by (Belton and Stewart, 2002) including:
• the design problematique: to search for, identify or create new decision alternatives
to meet goals and aspirations; and
• the portfolio problematique: to choose a subset of alternatives from a larger set of
possibilities.
Besides the mentioned problematiques, ‘screening techniques’ in MCDA were developed
to address problems related to both choice and sorting. In practical applications of
MCDA, it is common for a DM facing complex choice problems to first identify those
alternatives that do not appear to deserve further attention (Hobbs and Meierm, 2000).
234

Table 1

GM-ANL U.S. GM-LBST Europe MIT 2035 EcoTraffic ADL CMU Japan
LEM
R. Fazeli et al.

Study (General Motors, (General Motors (Bandivadekar (Ahlvik and (Hackney and (Maclean et al., (Tahara
(Delucchi ,2003)
et al., 2001) et al., 2002) et al., 2008) Brandberg, 2001) de Neufville, 2001) 2000) et al., 2001)
Vehicle ICEVs ICEVs ICEVs ICEVs ICEVs ICEVs ICEVs ICEVs
drivetrain HEVs HEVs HEVs HEVs HEVs HEVs HEVs HEVs
FCEVs FCEVs FCEVs FCEVs FCEVs FCEVs FCEVs FCEVs
BEVs BEVs BEVs BEVs
Fuels Gasoline Gasoline Gasoline Gasoline Gasoline Gasoline Gasoline Gasoline
Diesel Diesel Diesel Diesel Diesel Diesel Diesel Diesel
Naptha Naptha Naptha Naptha
FTD FTD FTD FTD FTD FTD FTD FTD
CNG CNG CNG CNG CNG CNG CNG CNG
Methanol Methanol Methanol Methanol Methanol Methanol Methanol Methanol
Summary characterisation of previous studies

Ethanol Ethanol Ethanol Ethanol Ethanol Ethanol


CH2 CH2 CH2 CH2 CH2 CH2 CH2 CH2
LH2 LH2 LH2 LH2 LH2 LH2
Electricity Electricity Electricity Electricity
LNG LNG LNG LNG
DME DME
Source: Updated based on Delucchi (2004)
A multi-criteria evaluation framework 235

Table 2 Summary characterisation of reports with a bottom-up approach

Dagsvik Brownstone Ewing and Golob et al. Bunch et al.


et al. (2002) et al. (2000) Sarigöllü (1998) (1997) (1993)
Research TTW TTW TTW No correlation No correlation
Energy use Not Local Local pollution Local pollution Local
included pollution pollution
Emissions Top speed, Top speed, Range, Range, Range
range acceleration acceleration, refuelling time
refuelling time
Technical Purchase Purchase and Purchase and Purchase and Purchase and
cost operation cost operation cost operation cost operation and
fuel cost
Cost --------- --------- --------- --------- ---------
Availability --------- --------- --------- --------- ---------
of fuel
supply
Transitional --------- Station --------- Station ---------
issue availability availability

Screening techniques can be regarded as useful MCDA methods, supporting the choice of
an alternative. The main applicability of these techniques is when not enough information
is available to reach a final choice directly, or too many alternatives must be considered.

2 Model design and parameter specification

The initial step of problem definition consists in identifying the issues under
consideration, ensuring that all participants in the process share a broad understanding of
the problem.

2.1 Alternative construction


Based on a comprehensive literature review, a list of alternatives and attributes has been
designed to be used as a foundation for this study’s model. The focus was then on model
building, as a framework for the evaluation of alternatives. Model building was regarded
as a very dynamic process, informed by and informing the problem structuring process,
and interacting with the evaluation process. It involved several iterations, extensive
discussions and the definition of alternatives and criteria, discarding, reinstating and
redefining of previously defined ones.
Regarding the alternatives, numerous combinations of alternative fuels and
technologies exist that can be used in light-duty vehicles. Several fuels such as ethanol,
methanol, CNG, LPG, biodiesel, hydrogen, electricity as well as gasoline and diesel, have
been investigated. On technologies, beside the current common technologies [the port
injection spark ignition engine (PISI), and the direct injection compression ignition
engine (DICI)], several other technologies for mid-size light duty vehicles exist, such as
the direct injection spark ignition engine (DISI), hybrid-PISI, hybrid-DICI, fuel cell with
reformer (FC-WR) and with no-reformer (FC-NR), plug-in hybrid (PHEV), and battery
236 R. Fazeli et al.

electric vehicle (BEV). In Table 3, the possible combinations of fuels/technologies for the
future engines are presented.
Table 3 Combination of alternative fuel-technology options for light duty vehicles

PISI DISI DICI FC-NR FC-WR Hybrid-PISI Hybrid-DICI PHEV BEV


Gasoline
Diesel
LPG
CNG
Methanol
Hydrogen
Ethanol
Bio-diesel
Electricity

After having identified the alternative fuel-technology options, the decision variables
were naturally defined as the shares of each fuel-technology combination in a LDV fleet.
The decision variables are: [x1, x2,…, xn], where xi (i = 1, 2,…, n) is the share of each
alternative fuel-technology in a light duty vehicle fleet. Therefore, we have:

∑ x = 1.0
i
i

So the problem would become a continuous one with an infinite number of alternatives.
However to overcome this difficulty, a discretisation approach was applied, in order to
decrease the calculation time and to make results more useful. Therefore, it was decided
to evaluate different mixes of up to four different technologies of vehicle fleet in the case
of Portugal.

2.2 Identification of attributes


As a very critical phase of the research, a lot of attention has been paid to the selection of
the evaluation criteria. According to Baker et al. (2002), a consistent family of criteria in
order to assist the proper evaluation of potential alternatives must be:
• able to discriminate among the alternatives and to support their comparison
• complete to include all goals
• operational and meaningful.
The next phase to organise the attributes was to identify the key areas of concern, to
systematise ideas for clarifying goals and actions, and to highlight any remaining gaps.
To define clusters, similar attributes and alternatives should be moved to a position close
to others to which they relate. For this purpose it is useful to position the most general
concepts at the top of the cluster, cascading down to more specific detail at the bottom
(Wai-Ho et al., 2005). The result of the attributes classification adopted for this study is
presented in Figure 1. A comprehensive literature review led to the identification of
attributes for comparing different pathways. These attributes were grouped in five main
A multi-criteria evaluation framework 237

categories: Acceptance, emissions to atmosphere, Transition cost, risk of technology


development, and Availability of fuel supply.

Figure 1 Classification of attributes for alternative fuel-technology vehicles (see online version
for colours)

Attribute Classification

Acceptance Emissions to  Transition  Risk of  Availability of  


Atmosphere Cost Technology  Fuel Supply
Economy Maturity
Air Pollution  Energy 
Initial Cost ($)
(g/km) Security
Fuel Cost ($)
Technological
M&R Cost ($) Global GHG  Diversity
Emissions (g/km)
Technical
Share of 
Acceleration WTT Renewable
TTW
Range
M&D
Safety

Reliability
1

2.2.1 Acceptance
Regarding user’s ‘acceptance’, two factors are considered: the vehicle lifetime expenses
and its performance. In order to compare the alternative technologies in terms of lifetime
expenses, all the expenses including the Initial Cost, Maintenance and Repair (M&R)
Cost and Fuel Cost should be taken into account. The fuel cost is based on the
assumption that a fully developed fuel infrastructure is given for all the fuels. In this
research, alternatives will be compared based on their Present Value Expenses defined as
follows:
⎛ FCost + ( M & R )Cost ⎞
PVExpenses = ICost + ∑ ⎜
⎜ (1 + i ) t −1


(1)
t ⎝ ⎠
Here ICost is the Initial Cost of the vehicle, FCost is the annual fuel cost, and the (M&R)
cost is the annual maintenance and repair cost. Besides, i is the annual discount rate and
the summation is for the whole lifetime of each vehicle. In the base scenario, it is
assumed that i = 8%, the average lifetime of each vehicle is 15 years with an annual
average of 15000 km.
The data used for calculating the present expenses of each technology was collected
from several sources (Grahn and Williander, 2009; Bandivadekar et al., 2008; Yacobucci,
2005; Cao and Mokhtarian, 2003; Lipman and Delucchi, 2003; Goedecke, 2005; Waegel
et al., 2006; Bandivadekar, 2008; Gunnarsson, 2009; Kromer and Heywood, 2007). As
238 R. Fazeli et al.

there is a significant dispersion of values in the literature, figures based on the number of
articles that referred to the specific initial cost vs. the predicted value were drawn. Then a
triangular distribution function was fitted to that distribution. Then, it was agreed to use
50% as a confidence level (meaning that the specific value can be assumed as a
representative for 50% of reviewed references). Then, the resulted value will represent
the higher limit of present value expenses for each alternative.
Beside the vehicle expenses, users are willing to compare different technologies
based on their performance characteristics, such as acceleration, range, safety and
reliability. For this sub-attribute, the assessment is more a relative one with respect to
base technology (current gasoline-ICE) rather than an absolute one (Edwards et al.,
1999).

2.2.2 Emissions to atmosphere

Regarding ‘emissions’, they could be categorised into two main groups: GHG and air
pollutants (non-GHG emissions). The approach taken consisted in comparing the
alternatives with their damage cost resulting from GHG and non-GHG emissions.
Coefficient factors were based on the results of ExternE (2005) to find the overall
damage cost of different pollutants. Information on GHG Emissions for different
technologies was obtained from a significant literature review (Edwards et al., 1999;
MacLean et al., 2000; MacLean and Lave, 2003; Brinkman et al., 2005; TIAX, 2007;
Bandivadekar et al., 2008; OECD/ITF, 2009; Gunnarsson, 2009). Regarding the air
pollution data, several articles have been reviewed (Edwards et al., 1999; MacLean et al.,
2000; TIAX, 2007; Brinkman et al., 2005; Goedecke, 2005).

2.2.3 Transition cost


As for most of the alternatives there will be a need for renovating the fuel supply
infrastructures, this attribute allows the comparison of technologies according to the
required investment. Relevant data was gathered from several sources (MacLean and
Lave, 2003; Linwei et al., 2009; Kromer and Heywood, 2007).
In a later phase of this research, focus will be put on the issue of the transition period
and the aim is to incorporate other transitional costs such as required incentives for
motivating people to choose specific fuels/technologies.

2.2.4 Risk of technology development


There exists a considerable risk related to the maturity of the proposed technologies.
These technologies are under development and will indeed take a considerable amount of
time to become mature for commercial production and application. The technology
readiness level (TRL) is a measure used by some US Government agencies and many of
the world’s major companies to assess the maturity of an evolving technology prior to
incorporating that technology into a system or subsystem. Generally, when a new
technology is first invented or conceptualised, it is not suitable for immediate application.
Instead, new technologies are usually subjected to experimentation, refinement, and
increasingly realistic testing. Once the technology is sufficiently proven, it can be
incorporated into a system/subsystem.
A multi-criteria evaluation framework 239

Reviewing the literature, the TLR value of different technologies was found for a
mid-term horizon (Edwards et al., 1999; Kromer and Heywood, 2007; Steenberghen and
Lopez, 2008).

2.2.5 Availability of fuel supply


Besides the above issues, recently, availability of fuel supply also became a key issue for
decision makers. It will be tackled in this work by considering three parameters: energy
security, technological diversity and share of renewable. In order to quantify the ‘energy
security’, the methodology suggested by IEA (2007) will be applied to evaluate the
security of energy supply in different countries:

⎡ Cf ⎤
ESI price = ∑ ⎢⎣ ESMC
f
pole − f * TPES ⎥⎦
(2)

where Cf / TPES is the share of the fuel mix and ESMCpole – f is the energy security
market concentration of the targeted market for fuel f. In order to calculate ESMCpole – f,
the following expression is used:

ESI pole = ∑ ⎡⎣r * S


i
i
2
i, f


(3)

where ri is the political risk rating of country i, and Si, f is the share of each supplier i in
the market of fuel f (Si, f varies from 0 to 100%). As ri ranges from 1 to 3, the worst
possible level of political stability leads to tripling the country’s contribution to ESMC
and the best does not affect the country’s contribution. A more detailed discussion on the
calculation of the energy security market concentration for each fuel is presented here, for
the case of Portugal.

2.2.5.1 Oil-based fuels


Portugal imports oil and related products mainly from 4 countries, namely Algeria,
Nigeria, Libya and Saudi Arabia (Portugal, 2007). The information regarding the political
risk rating for these sources was obtained from a report by the PRS Group Statistics,
(2009). So, based on expression.III, the ESMC will be calculated for oil-based fuels.

2.2.5.2 Natural gas-based fuels


According to the latest report on natural gas statistics in Portugal (2007), Nigeria and
Algeria are the main suppliers of natural gas. The same reference was used for
information on risk rating factors. Applying expression.III, the ESMC for natural
gas-based fuels was calculated.

2.2.5.3 Biofuels
There seems to be a very high potential for domestic biofuel production in Portugal.
According to a study on the environmentally-compatible bioenergy potential of EU
members (EEA, 2006), Portugal has in fact a potential of producing 4.1 MtOE of
bioenergy in 2030, from domestic sources including agriculture, forestry and wastes. In
240 R. Fazeli et al.

the ‘base scenario’, it was assumed that the production capacity will reach 50% of the
potential in 2030. In order to identify the related demand of fossil fuels for the production
of biofuels, the energy return on investment (EROI) factor was applied (Hammerschlag,
2006; Murphy, 2009). The other simplifying assumption was that the production capacity
for all of these biofuels are equal. The excess demand should be supplied by import.
Following the current trend, it was assumed that all the additional required biofuels will
be provided from Brazil.

2.2.5.4 Electricity
If the domestic power generation capacity were not sufficient, the additional demand
should be imported from Spain. Accordingly, the ESMC for Electricity was calculated
using expression III.
Regarding the technological diversity in vehicle fleet, the ‘HHI’ (Herfindahl-
Hirschman Index) was applied as it is a commonly accepted measure of market
concentration (Hellmer and Warell, 2009). In order to compare the diversity of alternative
technology/fuel options in an LDV fleet, the HHI is calculated by the sum of the squares
of the market share of each alternative fuel/technology in the market.
Another critical issue is the share of renewable sources in a fuel supply mix.
Therefore, this factor is aggregated with the other two parameters of energy security and
technological diversity that have been previously presented, to assess the alternatives
from the point of availability of fuel supply. Renewable accounts for biofuels and the
share of electricity that is produced from renewable sources.

3 Decision methodology

In order to provide support to decision makers in their search for satisfactory solutions in
a multi-criteria perspective, it is necessary to construct a model to represent decision
maker preferences and value judgments. Such preference models contain two primary
components:
• preferences in terms of each individual criterion, i.e. models describing the relative
importance or desirability of achieving different levels of performance for each
identified attribute
• an aggregation model, i.e. a model allowing inter-criteria comparisons (such as
trade-offs), in order to combine preferences across criteria.
Three different classes of preference models have been identified that can be adopted in
considering multiple criteria decision problems. These approaches are: value measurement,
goal programming and outranking methods.
‘Value measurement’ is based on a numerical value or score, say V(a), that is derived
from the m attribute values for each alternative. In ‘outranking methods’, for each pair of
alternatives, an assessment is made of the strength of evidence for and against the
assertion that ‘a is at least as good as b’, from which an ordering may be inferred. In a
‘goal programming’ approach, alternatives are assessed on the basis of how well they
satisfy goals or aspiration levels of achievement expressed directly in terms of the
attribute values (Stewart and Joubert, 2006). In this research, a very popular approach in
A multi-criteria evaluation framework 241

the context of multi-criteria decision making, multi-attribute utility theory (MAUT) was
applied, mainly because it easily incorporates the decision maker’s preferences. The basic
goal of MAUT is to replace available information by ‘utility values’ allowing the
comparison of alternatives (Daellenbach, 1994). In fact, utility theory can be viewed as
an extension of a value measurement approach.
Regarding the methodology to aggregate the several attributes, there are two main
approaches: additive and multiplicative aggregation. The main advantages of the additive
approach are simplicity and the independence to outliers. But, an undesirable feature of
an additive aggregation is the full compensability it implies: poor performance in some
indicators can be compensated by sufficiently high values of other indicators. On the
other hand, multiplicative aggregation is a less compensatory approach.
In Stewart (1995), a number of simulation studies are reported in which the effects on
preference ordering of using additive rather than multiplicative aggregation under
conditions of mutual utility independence are assessed. It appears that errors introduced
by using the additive model were in general extremely small for realistic ranges of
problem settings, and were in any case considerably smaller than those introduced by
incorrect modelling of partial utility functions.
After structuring a basic MCDA model and understanding the DM preferences, a
global model to aggregate those preferences and solve a specific problem (choose, rank
or sort) may be constructed. A typical example is the linear additive value function,
which can be expressed as:

( ) ∑W
V Ai = j ( )
⋅υ j Ai (4)
j∈Q

with V(Ai) being the aggregated evaluation of alternative Ai, vj the value vector for each
sub-attribute j, and the wj the weight vector of sub-attribute j for the aggregation process.
In this model, in order to aggregate the values of sub-attributes, e.g., to determine the
value of the attribute ‘acceptance’, the suggested expression was applied aggregating the
value of vehicle expenses and vehicle performance.
For the basic MCDA problem of choosing the best alternative, it is useful for a DM to
start by eliminating those alternatives that do not seem to be interesting (Hobbs and
Meierm, 2000). This procedure is often called ‘screening’. Screening helps by allowing
the DM to concentrate on a smaller set that (very likely) contains the best alternative. In
this work, a multi-stage screening process was applied, starting with a Pareto optimal
(PO) approach, followed by a data envelopment analysis (DEA)-based screening and a
trade-off weights (TW) procedure. This seems to be a rather powerful sequential
screening technique (Chen et al., 2008).
Pareto optimality is widely used in economics and in many other fields. It provides a
very useful definition of optimality in MCDA because it can be interpreted as taking into
account multiple aspects (criteria) for attempting overall optimality. The concept of
Pareto optimality in MCDA as expressed by Chen (2006) and the procedure he suggests
were applied in this research.
DEA is a technique used to measure the relative efficiency of a number of similar
units performing essentially the same task. DEA was first put forward by Charnes et al.
(1978) who proposed the basic DEA model, called CCR (after Charnes et al., 1978). CCR
assumes that each unit operates with ‘constant-returns-to-scale’. The basic function
of DEA is to ascertain which units are efficient and which are not; in MCDA, these
242 R. Fazeli et al.

can be regarded as non-dominated and dominated alternatives, respectively. Thus, the


DEA-based screening approach is structured around the following steps:

• identify the preference direction (positive or negative) for each criterion

• apply a DEA model (usually CCR) to identify dominated alternatives

• remove dominated alternatives.


Finally, TW-based screening techniques are related to research areas such as sensitivity
analysis (Insua, 1990; Insua and French, 1991) and dominance and potential optimality
(Athanassopoulos and Podinovski, 1997; Hazen, 1986). Chen (2006), reorganised these
methods and put forward a systematic screening method, TW-based screening, that
underlies the existence of TW. The method works as follows:

• check the validity of the linear additive preference function assumption

• obtain the preference functions and values for each alternative

• for each alternative apply mathematical programming to identify a potential


optimum (the mathematical program proposed by Geoffrion (1968), can be used to
determine whether that alternative is potentially optimal)

• retain only the potential optimal alternatives.

The following conditions have been added to the mathematical model:

• the lower limit of weights for each attribute is 5%, which will tend to remove the
alternatives that are extremely weak in one attribute

• the higher limit of weights for each attribute is 50%, which will avoid excessive
biased weighting to a specific attribute.

4 Results and discussion

4.1 Base scenario


In this research, a methodology combing MAUT and this multi-stage screening approach
is used to identify a set of interesting alternative fuel/technologies.
To illustrate the approach, Portugal was chosen as a case study. MAUT was applied
to identify the utility values of each alternative for each group of attributes. In fact, some
initial weights need to be assigned in order to aggregate the utility factor of sub-attributes
(Figure 1). As it was mentioned before, the alternative fuel vehicle fleets will be a
combination of one, two, three or four technologies. Figure 2(a) to Figure 2(d) show the
evaluation results for pure alternative fuel technology fleets.
The results illustrate the value of pure technology alternatives for each of the
five attributes. As it was expected, alternatives using conventional fuels obtain a
good score not only in transition cost and risk, but also in acceptance. This is mainly
because of the low necessary investment in fuel stations, the low risk in technology
maturity, the good technical performance and the low present value expenses. Besides,
A multi-criteria evaluation framework 243

biofuel-based alternatives achieve high value in availability of fuel supply mainly


because of a high capacity of domestic production in Portugal. Hybrid-based alternatives
got a superior grade in emissions, acceptance and transition cost, essentially because of
low present value expenses and limited necessary investment. Besides, as the technology
is already mature, hybrid-based alternatives achieve a superior grade in risk evaluation.
And finally, the electric-based alternatives show a good score in emissions, when
compared to other options, primarily because of low emission intensity of the Portuguese
power sector. In contrast, the low value of electric-based alternatives in terms of
acceptance and transition costs seems to be the major challenge for the adoption of these
alternatives.

Figure 2 Multi-Attribute evaluation of pure technology alternatives, (a) alternatives using


conventional fuels; (b) alternatives using biofuels; (c) pure hybrid alternatives;
(d) electric-based alternatives

(a) (b)

(c) (d)

After analysing the pure technology alternatives, the aim became to assess the attribute
value of alternative fleets with mixes of up to four different technologies. Then the
proposed sequential screening procedure, starting with a PO approach, followed by a
244 R. Fazeli et al.

DEA-based screening and a TW procedure were applied. The final set of alternatives is
presented in Table 4.

Table 4 Screening set of alternatives at base scenario

DISI-ethanol DICI-biodiesel FCWR-methanol HEV-PISI-gasoline PHEV BEV


Alt01 100%
Alt02 50% 50%
Alt03 50% 50%
Alt04 33% 33% 33%
Alt05 33% 33% 33%
Alt06 33% 33% 33%
Alt07 25% 25% 25% 25%
Alt08 25% 25% 25% 25%
Alt09 25% 25% 25% 25%
Alt10 25% 25% 25% 25%

To choose between these alternatives, more information about the decision maker’s
attitude would be necessary. In order to be able to compare the results for different
scenarios, a weighted summation index was applied, that allows the assessment of the
likely importance of each alternative technology in the final set:
∑ i Si , j
index j = a × (5)
( Number of alternatives in final set )

where a is assumed to be 12 and Si, j is the share of each technology (j) in alternative i.
From Figure 3, it is easily understandable that hybrid-gasoline appears to be the
technology that is more often present in the most promising alternatives, followed by
biofuel-based (mainly ethanol and biodiesel) technologies.

Figure 3 Weighted summation index of screening set of alternatives for the base scenario
A multi-criteria evaluation framework 245

Based on these results, some typical weighting profiles have been investigated. Besides
an ‘equal weighting’ option, five biased profiles have been defined, towards acceptance,
emissions, risk, transition cost, and availability of fuel supply.
In these biased profiles, the weight for the important attribute is 50% and the rest is
divided equally among the other attributes. The results presented in Figure 4, in terms of
the suggested index (expression.5) clearly show that these weighting profiles would also
recommend hybrid-gasoline, ethanol and biodiesel-based technologies.

Figure 4 Weighted summation index of more interesting alternatives for predefined weighting
profiles

Comparing the results in Figures 3 and 4, it becomes clear that the trends are quite
similar. Besides, this shows that the decision methodology is comparable with the
situation of a pre-defined weighting set. Moreover, the results obviously show the
potential of the developed approach in setting a framework for supporting better and
sounder decision-making.
It is worth to mention that the only obvious difference is in the weight of PHEV and
BEV in the final set of decisions. As it was expected, according to the low utility value of
these options, in transition cost, risk of technology and also emissions to atmosphere
(based on the current power generation mix), most of the proposed weighting profiles are
not interested in these options.

4.2 Parametric sensitivity analysis

After applying the developed methodology to the multi-criteria evaluation problem of


alternative fuel technology options, some sensitivity analysis was performed, as a way to
assess the impact of several key factors on the final results.
246 R. Fazeli et al.

4.2.1 Share of renewable electricity


As one of the key factors that could play an important role to enhance the advantage of
using electric-based technology, the impact of a high share of renewable electricity
production on the final sets should be clarified.
Figure 5 clearly shows that at a higher percentage of renewable power generation
(70% compared with 35% in the base scenario), the interest on using ethanol and
methanol will decrease, while as expected the attractiveness of PHEV and BEV
technologies will increase siginificantly. This is mainly due to the effect on the two
attributes of emissions and availability of fuel supply.

Figure 5 Comparison of base results with a higher renewable penetration situation

Figure 6 Comparison of base results with a higher GHG damage cost scenario
A multi-criteria evaluation framework 247

4.2.2 GHG damage cost


The other key factor that potentially will impact the results is GHG damage cost.
Therefore, the impact of increasing the GHG damage cost from 20 Euro/tonne-CO2 Eq in
the base scenario to 50 Euro/tonne-CO2 Eq was investigated.
Figure 6 shows a significant reduction in favourability of hybrid and FCWR-
Methanol. But on the other hand, it seems that a higher GHG damage cost will result in
higher interest towards using biofuels (ethanol and biodiesel).

4.2.3 Oil price shock


The other key factor is the oil price and its impact on the price of other fossil-based-fuels.
In the base scenario the gasoline and diesel price are 21 and 20 ($/GJ), respectively. In
the oil shock scenario, they may increase to 43 and 40 ($/GJ).

Figure 7 Comparison of base results and an oil price shock situation

Figure 7 shows that, as expected, higher oil price will result in slightly lowering the merit
of using hybrid technologies in the final set, while on the other hand, PHEVs will become
more interesting.

5 Conclusions

This research aims at developing a multi-criteria evaluation framework for alternative


fuel/technology options for light-duty vehicles in a mid-term horizon. Such a framework
is intended to assist policy makers and governments to make decisions concerning the
development of infrastructures or the establishment of incentives to promote alternative
fuel vehicles with a mid-term vision.
In this research, numerous combinations of alternative fuels and technologies
that can be used in light-duty vehicles have been investigated. Following that step, a
248 R. Fazeli et al.

comprehensive literature review enabled the identification of attributes for comparing


different pathways. These attributes were grouped in five main categories: acceptance,
emissions to atmosphere, transition cost, risk of technology development, and availability
of fuel supply.
Regarding the decision methodology, a very popular approach in the context of
multi-criteria decision making, MAUT was applied, mainly because it easily incorporates
the decision maker’s preferences. Regarding the selection process, a multi-stage
screening procedure was applied, starting with a PO approach, followed by a DEA-based
screening and a TW procedure. This analysis clearly showed that the developed
methodology, combining MAUT with sequential screening is a rather straightforward
process to help decision-making in this context. Comparing the results obtained by this
approach with the results from typical attribute weighting, it was shown the potential of
applying the methodology to support decision making and to assess and choose policies.
Considering the results of decision process for Portugal, as presented in Figure 3,
Hybrid-Gasoline appears to be the technology that is more often present in the preferable
alternatives, followed by biofuel-based (mainly ethanol and biodiesel) technologies. The
reasons behind this result mainly come from the different characteristics of alternative
fuel technology vehicle in chosen attributes. Figure 2(a) to Figure 2(d) show that
Hybrid-based alternatives got a good grade in emissions, acceptance and transition
cost, essentially because of low present value expenses and limited necessary investment,
while for example the electric-based alternatives show a low score in acceptance and
transition costs, which seems to be the major challenges for adoption of these
alternatives.
Moreover, the performed sensitivity analysis helped to identify the impact of some of
the key factors – including share of renewable electricity, GHG damage cost and oil price
shock – on the final set of alternatives. The results were always coherent with qualitative
reasoning, seeming to further reinforce the validity of the model.
The next phase of this research is to adopt a backcasting approach, focusing on how
to develop mechanisms to support the adoption of the alternative that this decision
process revealed as the preferred option.

Acknowledgements

The authors are particularly grateful to Stephen Connors, Director of the Analysis Group
for Regional Energy Alternatives at the MIT Energy Initiative, for helpful discussions
and comments. This work has been developed under the financial support of the MIT
Portugal Program, Fundação para a Ciência e Tecnologia (SFRH/BD/51080/2010).

References
Ahlvik, P. and Brandberg, A. (2001) Well to Wheels Efficiency for Alternative Fuels from Natural
Gas or Biomass, p.85, Publication, Swedish National Road Administration (EcoTraffic).
Athanassopoulos, A.D. and Podinovski, V.V. (1997) ‘Dominance and potential optimality in
multiple criteria decision analysis with imprecise information’, Journal of the Operational
Research Society, Vol. 48, pp.142–150.
A multi-criteria evaluation framework 249

Baker, D., Bridges, D., Hunter, R., Johnson, G., Krupa, J., Murphy, J. and Sorenson, K. (2002)
Guidebook to Decision-Making Methods, Department of Energy, USA, WSRC-IM-2002-
00002.
Bandivadekar, A. (2008) Evaluating the impact of Advanced Vehicle and Fuel Technologies in US
Light-Duty Vehicle Fleet, PhD Thesis, Massachusetts Institute of Technology.
Bandivadekar, A., Bodek, K., Cheah, L., Evans, C., Groode, T., Heywood, J., Kasseris, E.,
Kromer, M. and Weiss, M. (2008) On the Road in 2035: Reducing Transportation’s
Petroleum Consumption and GHG Emissions, Laboratory for Energy and the Environment,
Report No. LFEE 2008-05 RP.
Belton, V. and Stewart, T.J. (2002) Multiple Criteria Decision Analysis: An integrated Approach,
Kluwer, Dordrecht.
Brinkman, N., Wang, M., Weber, T. and Darlington, T. (2005) Well-to-Wheels Analysis of
Advanced Fuel/Vehicle Systems – A North American Study of Energy Use, Greenhouse Gas
Emissions, and Criteria Pollutant Emissions.
Brownstone, D., Bunch, D.S. and Train, K. (2000) ‘Joint mixed logit models of stated and revealed
preferences for alternative-fuel vehicles’, Transportation Research Part B, Vol. 34,
pp.315–338.
Bunch, D.S., Bradley, M., Golob, T.F., Kitamura, R. and Occhiuzzo, G.P. (1993) ‘Demand for
clean-fuel vehicles in California: a discrete-choice stated preference pilot project’,
Transportation Research Part A, Vol. 27, pp.237–253.
Cao, X. and Mokhtarian, P.L. (2003) The Future Demand for Alternative Fuel Passenger Vehicles:
A Preliminary Literature Review, UC Davis – Caltrans Air Quality Project.
Charnes, A., Cooper, W.W. and Rhodes E. (1978) ‘Measuring efficiency of decision-making units’,
European Journal of Operational Research, Vol. 2, pp.428–449.
Chen, Y. (2006) Multiple Criteria Decision Analysis: Classification Problems and Solutions,
Waterloo, Ontario, Canada.
Chen, Y., Kilgour, D.M. and Hipel, K.W. (2008) ‘Screening in multiple criteria decision analysis’,
Decision Support Systems, Vol. 45, pp.278–290.
Daellenbach, H.G. (1994) Systems and Decision Making, Wiley, New York, NY.
Dagsvik, J.K., Wennemo, T., Wetterwald, D.G. and Aaberge, R. (2002) ‘Potential demand for
alternative fuel vehicles’, Transportation Research Part B, Vol. 36, pp.361–384.
Delucchi, M.A. (2003) A Life cycle Emissions Model (LEM): Life cycle Emissions from
Transportation Fuels, Motor Vehicles, Transportation Modes, Electricity Use, Heating and
Cooking Fuels, and Materials, UCD-ITS-RR-03-17, Institute of Transportation Studies,
University of California, Davis (LEM).
Delucchi, M.A. (2004) Institute of Transportation Studies, University of California, Presentation
TTP-289-002.
Edwards, W., Dunlop, R. and Duo, W. (1999) Alternative and Future Fuels and Energy Resources
for Road Vehicles, Levelton Engineering Ltd.
EEA (European Environment Agency) (2006) How much Bioenergy can Europe Produce without
Harming the Environment?, No 7/2006, Copenhagen.
Ewing, G.O. and Sarigöllü, E. (1998) ‘Car fuel-type choice under travel demand management and
economic incentives’, Transportation Research Part D, Vol. 3, pp.429–444.
ExternE (2005) ‘Externalities of energy’, in Bickel, P. and Friedrich, R. (Eds.): Methodology 2005
Update, Institut für Energiewirtschaft und Rationelle Energieanwendung – IER, Universität
Stuttgart, Germany.
General Motors et al. (2002) GM Well-to-Wheel Analysis of Energy use and Greenhouse Gas
Emissions of Advanced Fuel/Vehicle Systems – A European Study, L-B-Systemtechnik GmbH,
Ottobrunn, Germany, (GM-LBST Europe).
250 R. Fazeli et al.

General Motors, Argonne National Lab, et al. (2001) Well-toWheel Energy Use and Greenhouse
Gas Emisions of Advanced Fuel/Vehicle Systems, in three volumes, published by Argonne
National Laboratory, (GM-ANL U.S.).
Geoffrion, A. (1968) ‘Proper efficiency and theory of vector maximization’, Journal of
Mathematical Analysis and Applications, Vol. 22, pp.618–630.
Goedecke, M. (2005) Life Cycle Costs Analysis of Alternative Vehicles and Fuels in Thailand,
Master Thesis, The Joint Graduate School of Energy and Environment at King Mongkut’s
University of Technology Thonburi.
Golob, T.F., Torous, J., Bradley, M., Brownstone, D., Crane, S.S. and Bunch, D.S. (1997)
‘Commercial fleet demand for alternative-fuel vehicles in California’, Transportation
Research A, Vol. 31, No. 3, pp.219–233.
Grahn, M. and Williander, M. (2009) The Role of ICEVs, HEVs, PHEVs, BEVs and FCVs in
Achieving Stringent CO2 Targets: Results from Global Energy Systems Modeling, Stavanger,
Norway.
Gunnarsson, A. (2009) Analysis of Alternative Fuels in Automotive Powertrains, Linköping.
Hackney, J. and de Neufville, R. (2001) ‘Life cycle model of alternative fuel vehicles: emissions,
energy, and cost tradeoffs’, Transportation Research Part A, Vol. 35, pp.243–266, (ADL).
Hammerschlag, R. (2006) ‘Ethanol’s energy return on investment: a survey of the literature
1990-present’, Environ. Sci. Technol., Vol. 40, pp.1744–1750.
Hazen, G.B. (1986) ‘Partial information, dominance, and potential optimality in multi-attribute
utility theory’, Operations Research, Vol. 34, No. 2, pp.296–310.
Hellmer, S. and Warell, L. (2009) ‘On the evaluation of market power and market dominance – the
Nordic electricity market’, Energy Policy, Vol. 37, pp.3235–3241.
Hobbs, B.F. and Meierm, P. (2000) Energy Decision and the Environment: A Guide to the Use of
Multicriteria Methods, Kluwer, Massachusetts.
Hwang, C.L. and Yoon, K. (1981) Multiple Attribute Decision Making, Spring Verlag, Berlin.
IEA (2007) Energy Security and Climate Policy – Assessing Interactions.
Insua, D.R. (1990) Sensitivity Analysis in Multiple Objective Decision Making, Springer-Verlag,
Berlin.
Insua, D.R. and French, S. (1991) ‘A framework for sensitivity analysis in discrete multi-objective
decision making’, European Journal of Operational Research, Vol. 54, pp.176–190.
Jank, M.J., Kutas, G., Fernando do Amaral, L. and Nassar, A.M. (2007) EU and U.S. Policies on
Biofuels: Potential Impacts on Developing Countries, The German Marshall Fund of the
United States.
Janssen, R. (1992) Multiobjective Decision Support for Environmental Management, Kluwer,
Dordrecht.
Kromer, M.A. and Heywood, J.B. (2007) Electric Powertrains: Opportunities and Challenges in
the U.S. Light-Duty Vehicle Fleet, LFEE 2007-03 RP.
Linwei, M.A., Zheng, L.I., Feng, F.U., Xiliang, Z. and Weidou, N.I. (2009) ‘Alternative energy
development strategies for China towards 2030’, Front. Energy Power Eng. China, Vol. 3,
No. 1, pp.2–10.
Lipman, T.E. and Delucchi, M.A. (2003) Hybrid-Electric Vehicle Design Retail and Life Cycle
Cost Analysis, UCD-ITS-RR-03-01.
Lotfi, V., Stewart, T.J. and Zionts, S. (1992) ‘An aspiration-level interactive model for multiple
criteria decision making’, Computers and Operations Research, Vol. 19, pp.671–681.
MacCrimmon, K.R. (1973) ‘An overview of multiple objective decision making’, in
Cochrance, J.L. and Zeleny, M. (Eds.): Multiple Criteria Decision Making, pp.18–44,
University of South Carolina Press, Columbia.
MacLean, H.L. and Lave, L.B. (2003) ‘Evaluating automobile fuel/propulsion system
technologies’, Progress in Energy and Combustion Science, Vol. 29, pp.1–69.
A multi-criteria evaluation framework 251

Maclean, H.L., Lave, L.B., Lankey, R. and Joshi, S. (2000) ‘A life-cycle comparison of alternative
automobile fuels’, Journal of the Air and Waste Management Association, Vol. 50,
pp.1769–1779, (CMU).
Murphy, D.J. (2009) Recent Applications of Energy Return On Investment, ASPO – 2009, Denver,
Colorado.
OECD/ITF (2009) Reducing Transport GHG Emissions Opportunities and Costs: Preliminary
Findings.
Ogden, J.M., Williams, R.H. and Larson, E.D. (2004) ‘Societal life-cycle costs of cars with
alternative fuels/engines’, Energy Policy, Vol. 32, pp.7–27.
Portugal – Energy Mix Fact Sheet (2007) European Commission Statistics.
Roy, B. (1968) ‘Classement et choix en présence de points de vue multiples (la méthode Eletre)’,
Revue Francaise d’ Informatique et de Recherche Opérationnelle, Vol. 8, pp.57–75.
Roy, B. (1985) ‘Méthodologie multicritére dáide la décision’, Economica, Paris.
SAIC (Scientific Applications International Corporation) (2006) Life Cycle Assessment: Principles
and Practice, EPA/600/R-06/060.
Steenberghen, T. and Lopez, E. (2008) ‘Overcoming barriers to the implementation of alternative
fuels for road transport in Europe’, Journal of Cleaner Production, Vol. 16, pp.577–590.
Stewart, T.J. (1995) ‘Simplified approaches for multi-criteria decision making under uncertainty’,
Journal of Multi-Criteria Decision Analysis, Vol. 4, pp.246–258.
Stewart, T.J. and Joubert, A.R. (2006) Multicriteria Support for Decision-Making in the
Ecosystems Approach to Fisheries Management, Appendix 8, Department of Statistical
Sciences, University of Cape Town.
Tahara, K. et al. (2001) ‘Comparison of CO2 emissions from alternative and conventional
vehicles’, World Resource Review, Vol. 13, No. 1, pp.52–60, Japan.
The PRS Group (2009) Political Risk Rating Forecast Report.
TIAX (2007) Full Fuel Cycle Assessment: Well-To-Wheels Energy Inputs, Emissions, And Water
Impacts, Prepared for California Energy Commission, CEC-600-2007-004-REV.
Waegel, A., Tobin, D.C. Haney, B., Karki, J. and Suarez, J. (2006) Pathways to a U.S. Hydrogen
Economy: Vehicle Fleet Diffusion Scenarios, A Report to the BP Foundation.
Wai-Ho, A., Chan, K.C.C., Wong, A.K.C. and Yang, W. (2005) ‘Attribute clustering for grouping,
selection, and classification of gene expression data’, IEEE/ACM Transactions on
Computational Biology and Bioinformatics, Vol. 2, No. 2.
Yacobucci, B.D. (2005) Alternative Transportation Fuels and Vehicles: Energy, Environment, and
Development Issues, CRS Report for Congress, Code RL30758.

View publication stats

You might also like