1
Degree Project in Energy Systems Analysis
(MJ248X)
Thesis report
on
Development of a multi-criteria analysis framework to study
low carbon transition pathways for the European Union
By:
Balasubramanian Viswanathan
TRITA-ITM-EX 2018:748
Master Programme in Sustainable Energy Engineering
Department of Energy Technology
2 MJ248X Degree Project in Energy Systems Analysis
TRITA-ITM-EX 2018:748
Development of a multi-criteria analysis framework to study low carbon transition pathways for the European Union
Approved Examiner Supervisor
Mark Howells
Georgios Avgerinopoulos
Commissioner Contact Person
ABSTRACT
A crucial element of the global action against climate change is transforming our current carbon-intensive energy system to a sustainable, low-carbon one. Researchers are trying to study this transformation by modelling the pathways. REEEM is one such project with a scope set to the European Union. In parallel, multi-criteria analysis is gaining prominence among decision makers and academics for sustainable energy planning. In this research a multi-criteria analysis framework has been developed to study the low-carbon pathways for the EU. REEEM acts as a background and has been used at various points in the framework development process. Some of the key elements of the framework include a review of academic trends, a logically-derived decision tree and consideration of issues related to data collection and processing. A case study has been codified in R and its results are indicative of the usage of the framework. Further, an analysis of extending the scope of the code, supporting files, and the framework, to other regions has been made. The fundamental goal of this research is to support science-based policy making.
SAMMANFATTNING
3 ACKNOWLEDGEMENTS
I wish to start by thanking my parents, Viswanathan and Poorna, and my sister, Meenakshi. This spectacular journey would not have been possible without their love and support. Next, I would like to thank my supervisor, Georgios Avgerinopoulos whose enthusiasm and expertise helped me navigate through this research.
Next, I would also like to thank my classmates Camilo Ramirez and Jairo Mosquera for lending a patient ear and providing me valuable feedback. Further, I would like to thank my fellow Sustainable Energy Engineering students for the shared learning experiences and instilling a resolute optimism on action against climate change.
4 TABLE OF CONTENTS Abstract ... 2 Sammanfattning ... 2 Acknowledgements ... 3 List of Abbreviations ... 8 Chapter 1: Introduction ... 9 1.1 Background ... 9
1.2 REEEM Energy Systems Modelling Project ... 9
1.3 Intentions ... 11
1.4 Terminology ... 11
Chapter 2: Literature Review ... 14
2.1 Theoretical Principles ... 14
2.2 State-Of-The-Art ... 15
Chapter 3: Creating The Framework ... 18
3.1 Objectives and Alternatives ... 18
3.2 Criteria Setting ... 21
3.3 Indicator Selection ... 26
3.4 Weighing criteria ... 26
3.5 Selecting the MCDM Method ... 29
3.6 Practical Considerations ... 31
Handling Missing Data ... 31
Tools for Data Collection and Processing... 33
3.7 Sensitivity Analysis ... 34
3.8 Summary of the Framework ... 35
3.9 Limitations of the framework ... 36
Chapter 4: Case Study ... 38
5
4.2 Applying the framework ... 38
4.3 Results and Discussion ... 41
Chapter 5: Discussion ... 45
5.1 Intentions Revisited... 45
5.2 Conclusions... 46
Chapter 6: Future Work ... 47
Appendices: ... 49
Appendix 1: Proposed Indicators for MCA framework ... 49
Appendix 2: Criteria Weights Survey ... 51
Appendix 2.1 Survey... 51
Appendix 2.2 Survey Data... 55
Appendix 2.3 Criteria Weights ... 56
Appendix 3: Case Study Data Input Table ... 59
Appendix 4: MCDM Calculation Code ... 60
Appendix 4.1: Proposed AHP Tree ... 60
Appendix 4.2: Case Study AHP Tree ... 67
Appendix 5.3: Case Study AHP Calculation Code ... 70
Appendix 5: MCDM Calculation Guidelines ... 72
Appendix 6: Case Study Results ... 73
Appendix 6.1: Case Study Code Output ... 73
Appendix 6.2: Case Study Run Results Data ... 74
Appendix 6.3: Case Study Sensitivity Analysis Data ... 76
References... 78
LIST OF TABLES AND FIGURES Figure 1: Distribution of papers using Multi-Criteria Analysis for Energy Management (Source: Mardani, et al. [5]) ... 9
Figure 2: Proposed REEEM IAM Framework[8] ... 11
6
Figure 4: General Multi Criteria Analysis Framework[11] ... 16
Figure 5: Visual representation of interlinkages between objectives ... 18
Figure 6: Ten key actions of the Integrated-SET Plan[45] ... 20
Figure 7: Assigning criteria from Criteria Themes ... 22
Figure 8: Comparison of criteria used in literature ... 23
Figure 9: Sub-Criteria Tree for Ecosystem ... 24
Figure 10: Sub-Criteria Tree for Resource ... 25
Figure 11: Criteria Tree for Low Carbon Pathways ... 26
Figure 12: Example of DM data input for criteria weighing using Simos method ... 28
Figure 13: Example of DM data input for criteria weighing using the AHP method ... 28
Figure 14: Frequency of MCDM Methods used for Energy Management (Constructed from Mardani, et al. [5] ) ... 30
Figure 15: Visual representation of proposed MCA framework ... 36
Figure 16: Case Study analysis flowchart ... 38
Figure 17: Case Study Decision Tree ... 39
Figure 18: Case Study pathway criteria scores, 2040 ... 42
Figure 19: Comparison of weighing methods for Case Study pathways, 2040 ... 42
Figure 20: Case study pathway comparison 2015-2040, Base and HighRES ... 43
Figure 21: One-dimensional sensitivity analysis, Biodiversity losses ... 44
Figure 22: Two-dimensional sensitivity analysis for Pilot 1 Biodiversity Losses, 2015 ... 44
Figure 23: Open decision-making platform concept ... 48
Figure 24: Survey Screenshot- Introduction ... 51
Figure 25: Survey Screenshot- Description of scales ... 51
Figure 26: Survey Screenshot- Example of user interface ... 52
Figure 27: Survey Screenshot- Criteria weights ... 53
Figure 28: Survey Screenshot- Ecosystem sub-criteria weights ... 53
Figure 29: Survey Screenshot- Welfare sub-criteria weights ... 53
Figure 30: Survey Screenshot- Security sub-criteria weights ... 53
7
Figure 32: Screenshot- Description of choices ... 54
Figure 33: Screenshot- Randomizing comparisons order ... 55
Figure 34: Screenshot- Case study pathways, 2015 ... 73
Figure 35: Screenshot- Case study pathways, 2040 ... 73
Figure 36: Screenshot- Case study pathways alternate calculation, 2015 ... 73
Table 1: Literature comparison of MCA used in Energy Planning ... 16
Table 2: Comparison of AHP and Simos method for weighing criteria ... 28
Table 3: MCDM methods selection criteria comparison ... 31
Table 4: Recommendations for specific types of missing data ... 33
Table 5: Proposed Sensitivity Analysis framework for the MCA framework ... 34
Table 6: Case Study Indicators ... 39
Table 7: Case Study Criteria Weights ... 40
Table 8: Case Study Missing Data Assumptions... 40
Table 9: Case study runs ... 41
8 LIST OF ABBREVIATIONS
AHP Analytical Heirarchy Process DM Decision Maker, Decision Making
ELECTRE Elimination Et Choix Traduisantla Realité
EU European Union
GHG Greenhouse Gas
HighRES High Renewables
IEM Integrated Energy Model MADM Multi Attribute Decision Making MCA Multi-Criteria Analysis
MCDM Multi-Criteria Decision Making MODM Multi-Objective Decision Making
NEWAGE National European Worldwide Applied General Equilibrium Openmod Open Energy Model
OSeMOSYS Open Source Energy Modelling System PROMETHEE
Preference Ranking Organization Method for Enrichment Evaluation
REEEM
Role of technologies in an energy efficient economy– model-based analysis of policy measures and transformation pathways to a sustainable energy system
SA Sensitivity analysis
SET Strategic Energy Technology
SI Storage Innovation
TIMES The Integrated MARKAL-EFOM System
9
CHAPTER 1: INTRODUCTION
1.1 BACKGROUNDAs the global dialogue on combating climate change gathers momentum, it becomes imperative to approach the issue with sound scientific reasoning. This implies not just predicting the harsh outcomes but also planning to avoid them. At the heart of the solution, lies a need to transform our society into a low carbon one. This requires development of low carbon transition pathways i.e., the various means that can be adopted to avoid the said harsh outcomes. Unfortunately, there is not a perfect solution and these low carbon pathways come with trade-offs. This is particularly significant in the energy system as it is constrained by the trilemma[1] of having energy security, environmental sustainability and energy equity.
Through the years, it can be stated that the European Union (EU) has been leading the global action against climate change. In a 2012 publication by da Graça Carvalho [2], a comprehensive EU-level energy and climate change strategy was laid out. This was published along with a roadmap on achieving a low-carbon economy[3]. The ultimate target of this roadmap is to reduce Greenhouse Gas (GHG) emissions by 80% below 1990 levels[4]. However, with an ever-changing socio-political landscape, it has become difficult to track progress and plan for the future.
Meanwhile, there has been a rise in the use of a technique called Multi-criteria Analysis (MCA) within the field of energy planning. As can be seen in Figure 1, nearly 200 publications have used MCA in their research. Fundamentally, MCA is a technique to help decision making when there are multiple areas of interest with varying preferences. This has been explained in detail in later chapters.
Figure 1: Distribution of papers using Multi-Criteria Analysis for Energy Management (Source: Mardani, et al. [5]) However, it was observed through literature review and expert opinion that MCA hasn’t been used within the context of achieving the low-carbon economy for the EU. This led to the formulation of the primary research question of this thesis: “Can Multi-criteria Analysis help EU reach its objective of
becoming a low-carbon economy? If so, how?”. While looking at this question, an opportunity arose
which has been described in the next section.
10 Studying the energy system at any level requires establishment of boundaries as it is an open system1.
The elements that lie outside the boundaries but influence the output are called externalities. The system analysis has no control over the externalities. In case the externalities have a strong influence on the system, the analysis is rendered weak as there is minimal control over the output. This situation must be avoided for reliable energy system planning.
An integrated energy system analysis brings externalities with a strong influence within the system boundaries. A quantitative model designed with such an approach in mind is an Integrated Energy Models (IEM). They are hard or soft linked, digital simulation of an energy system which has been designed with an intention of using an integrated energy analysis framework. One such example is the model being developed by the REEEM Energy Systems Model Project[6] which focuses on the European Union.
“Role of technologies in an energy efficient economy – model based analysis policy measures and
transformation pathways to a sustainable energy system”, abbreviated as REEEM is a Horizon 2020
Project funded by the European Commission undertaken by 11 partners. It is driven by four core objectives[7]:
1. Developing an integrated assessment framework 2. Defining and assessing pathways and case studies 3. Creating a science-policy interface
4. Establishing transparency
These objectives are observed by 8 work packages[7] led by the various participants in the project. A central task to these work packages is the development of the IEM. Various modelling approaches will be taken to study individual issues which will then be connected through TIMES Pan-EU and OSEMOSYS models. It is expected to include understanding of consumer behaviour, electricity markets, health impact, life cycle analysis, water and land use of energy systems as well as uncertainty in technology growth. An early draft of such a framework was made in Gardumi, et al. [8] as shown in Figure 2.
11 Figure 2: Proposed REEEM IAM Framework[8]
REEEM is relevant to this research work because it provides an opportunity to answer the research question. To develop a multi-criteria assessment tool, the need for data is crucial. This project puts the various EU level low carbon pathways into quantitative metrics. Further, REEEM’s objective of maintaining transparency augments the trust capital that is brought about by proposals built on MCA. 1.3 INTENTIONS
In this section, the thesis intention or research objectives have been stated. They are built on the background of growing use of MCA in energy planning, EU’s quest for a low carbon society and REEEM’s planned IEM. The four stated intentions are designed with the objective of answering the research question. They are as follows:
Intention 1: Identify gaps in current energy planning literature within the context of Multi-Criteria Analysis (MCA) and ways to cover them.
Intention 2: Develop an MCA framework to study low carbon transition pathways in the European Union Intention 3: Develop an application of the proposed framework using the case study of REEEM to demonstrate the framework’s practical feasibility
Intention 4: Discuss the proposed MCA framework within the context of re-creating the research in other regions
1.4 TERMINOLOGY
12 key terms as used in this project are clarified. This terminology is derived from the analysis done by Kurka and Blackwood [9].
The most broadly used term is Multi-Criteria Analysis (MCA). This refers to a general research approach that moves beyond conventional single factor analysis like cost-based analysis. One of the techniques used in an MCA is called Multi-Criteria Decision Making (MCDM). This is also referred to as Multi-Criteria Decision Analysis (MCDA) or Multi-Attribute Decision Analysis (MADA). The terminology used in this project is MCDM and is a technique for an MCA. Within MCDM, there are two types of studies- Multi-Objective Decision Making (MODM) and Multi-Attribute Decision Making (MADM). In this project, MADM is used as explained in section “3.5 Selecting the MCDM Method”. MADM is further divided in several types of models with two prominent classifications being Value Measurement Models and Outranking Models. Both these Models require four elements which are Alternatives, Objectives, Criteria and Indicators. This terminology tree is visualized in Figure 3.
Figure 3: Terminology Tree for Multi-Criteria Analysis
In addition to this, several other terms are used in the project. They are defined as follows- Alternatives: They refer to a set of entities which are compared by the decision maker in an MCA Criteria: They refer to a set of aspects upon which the decision maker aims to base their decision on Decision Maker: One or a group of persons who wish to evaluate a set of alternatives to arrive at their objectives
Energy Model: A quantitative assessment of the energy system which uses measurable data to produce metrics
Indicators: Count-able metrics used to describe the characteristics of a criteria
Integrated Energy Model: An energy model that combines several energy models while maintaining consistency in the data input
Multi-Criteria Analysis: A research concept that aims to include several aspects to an issue.
Multi-Criteria Decision Making: A type of MCA that quantifies the contradicting preferences of the decision maker
13 Objectives: The results that the decision maker wishes to achieve through the MCA
Pathways: A set of pre-defined conditions which are modelled in energy models
14
CHAPTER 2: LITERATURE REVIEW
In this chapter, literature has been reviewed based on two aspects of the study. The first is to review the theoretical principles behind MCA and the second is to understand current research on MCA for sustainable energy systems planning.
2.1 THEORETICAL PRINCIPLES
Fundamentally, MCA is a general description used to for studies that tries to move beyond single factor studies like costs benefit analysis. Since the methodology behind the MCA varies from case to case, the terminology used in literature also has significant variance. Kurka and Blackwood [9] reviewed and categorized the MCA terminology which are relevant to the sustainable energy planning. Using this study as a starting point, it was concluded that the methodology used in this project could be termed as Multi-Criteria Decision Making (MCDM) or Multi Attribute Decision Making (MADM). In this report MCDM has been used.
The MCA manual by Dodgson, et al. [10] gave an overview of the popular MCDM techniques. It also highlighted the characteristics that must be considered when selecting a MCDM method for a study. Mateo [11] provided a more detailed description of the MCDM methods and their mathematical formulations. It also mapped the MCDM calculation methods to appropriate criteria weighing techniques. These two studies were used in conjecture to arrive at the MCDM techniques focussed on this report.
This was followed up by detailed studies on selected methods. Saaty [12]and Handfield, et al. [13] were used to study Analytical Hierarchy Process (AHP) and Siskos and Tsotsolas [14] was used to review Simos method and its variants. To achieve robustness in results of the MCA, a combination of MCDM can be used. Marttunen, et al. [15] reviewed the combinations of MCDM used in literature and their respective advantages and disadvantages.
A practical consideration while development of MCA is the data availability. Since, missing data can prevent the completion of study or lead to a study with poor results, strategies to handle the same must be developed. Mareschal and Hadouchi [16] list the multiple ways in which data can be missing and the prominent techniques used to overcome the issue. In Jimenez, et al. [17], a simple statistical technique was demonstrated with an example while Ma, et al. [18] developed a theoretically richer method specific to the case of AHP. The report by Image [19] on MCDA for conservation assessment was used as a reference for developing guidelines on how to deal with missing data.
The last aspect of the MCA framework is the sensitivity analysis. While it is not included by the framework proposed by Mateo [11] (See: Figure 4), it has been included in the MCA manual by Dodgson, et al. [10] as the last step. Its underlying importance in all forms of quantitative analysis was studied by Saltelli [20] who defined it as, “the study of how the uncertainty in the output of a model
(numerical or otherwise) can be apportioned to different sources of uncertainty in the model input”. It
15 Within the area of MCDM, the basic sensitivity techniques were comprehensively studied by Triantaphyllou and Sánchez [21] following which more advanced methods were reviewed. The widely used Monte Carlo Simulation was studied in a decision making context by Mateos, et al. [22] while Insua and French [23] provided a mathematical forumaltion of sensitivity analysis for dominance based MCDM methods. Additionally, the research done by Butler, et al. [24] gave an insight into the role of MCDM and sensitivity analysis in handling a major issue such as disposing the stockpile of weapon-grade plutonium in the United States.
When developing the case study, a survey of experts was needed. Sato [25] was used to design the survey to obtain weights used for the AHP method. A review of the latest MCA tools by International Society on MCDM [26] was used to select the best tool for data processing and visualization.
2.2 STATE-OF-THE-ART
As was shown in Figure 1, the use of MCA within the field of energy management has been on the rise. Consequently, there have been publications which review the literature. Three such review articles were identified that were relevant to this study.
The first of these was published by Wang, et al. [27] in 2009. They break down the MCA into several stages and analyse literature at those levels. These include survey of various criteria selection methods, weighing techniques and Multi Criteria Decision Making (MCDM) algorithms. The most popular options for each of these stages were then identified. These popular options form the basis for some of the analysis done in Chapter 3.
In Kumar, et al. [28], a review has been made with a focus on the MCDM methods. The focus is primarily on the mathematical aspects of MCA, which involves the creation of an evaluation matrix. The prominent MCDM methods were then cross referenced to the objectives of the research. Additionally, prominent software and tools for MCA were also listed.
The review article by Mardani, et al. [5], published in 2017, surveys 196 journals published between 1995 to 2015 in the area of energy management. The study breaks down each of the reviewed papers into the following categories: nationality of authors, region of MCA study, technique, number of criteria, research purpose, gap and contribution, solution and modelling, and research.
16 Figure 4: General Multi Criteria Analysis Framework[11]
Table 1: Literature comparison of MCA used in Energy Planning Region Defining problem Generati ng alternati ves Establishin g criteria Assigning criteria weights Construc ting evaluati on matrix Selecting MCDA method Citati ons Switzerlan d[29] Impact of policies Scenarios (1 ref, 2 climate issues) 12 interdiscpli nary indicators Profiles (manually assigned) Swiss Markal Model Min-Max Normaliz ation 16 EU[30] Comparin g scenario desirabili ty across two models Scenarios (Planets Model) EU Energy Policy Priority, 4 criteria Egalitarian+ Monte Carlo (sensitivity) TIAM, WITCH Models WASPAS, ARAS and TOPSIS 2
Turkey[31] Best RES for Turkey Renewabl e Energy Sources 11 criteria from various sources Grey ANP w/ expert opinions Presume d multiple national studies Grey Based MCDA 11
17 onal standard s China[36] Preferred Energy Alternativ e Energy Sources 6 criteria, used literature AHP and ANP w/survey and sensitivity National studies, internati onal standard s WASPAS 59
On completion of the study, the following key patterns were identified:
• Defining the problem can be of two types - direct and indirect. Direct problem setting, like in [32, 34, 36], has been used to identify a single best alternative like a renewable energy technology while indirect problem setting aims to provide a broader analysis, like in [29, 33] to study the effect of policy changes.
• In general, there are three types of alternatives which are considered in MCA for sustainable energy planning. They are pathways or scenarios[29, 30], project or policy[33, 34], and technology solutions[31, 32, 35, 36].
• There is considerable variation in selection of criteria ranging from 4[30] to 20[35] most of which compiled through studying literature. While all of them cover aspect of sustainability, there is no standardization in criteria selection. This is one the issues addressed in this thesis project in section “3.2 Criteria Setting”
• Weighing of criteria is prominently done through three methods. They are assigning equal weights[30], Analytical Hierarchy Process (AHP)[32, 35] and Simos Method[33]. However, there are mathematical variations within these methods. These weights are usually obtained through a combination of experts[31, 32, 35], locals[34] or provided by the user of the MCA[29]. • Some studies have opted to do a sensitivity analysis on the weights. They are done using
conventional sensitivity on a range of possible weights[33, 34, 36] or a Monte Carlo Simulation[30].
• The evaluation matrix is constructed using data from local, national and international studies[35, 36] or energy systems models[29, 30].
• The technique used for ranking the alternatives vary considerably. A detailed technical analysis of the MCDM methods was done in the review paper by Wang, et al. [27] which was described earlier.
18
CHAPTER
3:
CREATING
THE
FRAMEWORK
In this chapter, the MCA framework for low carbon pathways in EU designed. The various elements are defined by a set of assumptions and logical reasoning. They are underpinned by learnings from literature review and first principles are used when conceptualizing new segments. The presence of IEMs like REEEM is also considered as the data output from them are crucial for applying the framework. The primary structure of the chapter follows the MCA framework proposed by Mateo [11] as shown in Figure 4, An important aspect of the framework is the logical constraints imposed on each of its sections by other sections. For example, the objectives drive the selection of criteria which in turn determines the use of indicators. However, the indicators themselves are constrained by data availability. The framework is conceptually summarized at the end of the chapter.
3.1 OBJECTIVES AND ALTERNATIVES
In this section, the objectives and alternatives are considered. These options are traditionally set early in the MCA. According to Mateo [11], they are decided in the first step of MCA which is, “Defining the
problem, generating alternatives and establishing criteria”. At this stage it must be made clear that the
objectives here refer to that of the MCA and not that of this thesis itself. The relationship between the objectives and alternatives of the MCA framework, the thesis intentions and the energy planning goals behind the energy models is shown in Figure 5.
Figure 5: Visual representation of interlinkages between objectives
As can be seen in Figure 5, the MCA Objectives and Alternatives are linked by the low carbon pathways. These pathways act as alternatives but are built through a set of energy planning objectives. Given the strong linkage, these two elements must be decided in parallel. The literature review gave an indication of how these linkages were considered in other MCA studies.
19 However, the studies lose relevance with change in time and increase in geography considered. On the other end of the abstraction spectrum lies studies like that of Brazilian Amazon[34] and Crete[33] which aim to pick the “best energy solution” among a series of possible solutions. These studies offer significant support to decision makers, but their relevance is set to the location.
This research focuses on EU as has been described in 1.1 Background with a focus on low carbon pathways. This implies an MCA set to a longer time scale and large energy systems. There are several directives set at the EU level to achieve a transformation of the energy system which can serve as the objectives of the MCA.
Energy Union and Climate [39] is one of the 10 priorities laid out by the 2015-19 European Commission under Jean-Claude Junker[40]. It has five agenda points of which one is “Research, Innovation & Competitiveness”. This agenda listed six targets which were then combined with the SET Plan in the Integrated SET Plan document[41].
Strategic Energy Technology (SET) Plan[42] is an initiative by the European Commission to accelerate the development to low carbon pathways. It is overseen by a steering group and annually reports progress. In 2015 an attempt was made to integrate the SET Plan with Energy Union’s targets[43] which resulted in a document by Directorates-General for Energy Research & Innovation and Joint Research Centre [41] which lists 10 key actions. This integrated target with 10 keys actions is shown in Figure 6.
20 Figure 6: Ten key actions of the Integrated-SET Plan[45]
While these EU directives serve as guiding principles for future policy making and setting preferences for research grants, they have two fundamental flaws when it comes to acting as objectives for an MCA analysis.
• Some of the targets set by these directives are often used as base assumptions while building energy models. However, MCA is done after the energy model provides results (considered in the MCA through indicators). Hence such assumptions are checked again at the end creating a fallacy. For example, India’s ambitious target of 100 GW of solar power capacity by 2022 or other Nationally Determined Contributions are often embedded into assumptions for new policy scenarios. Checking these scenarios for targets achieved can cause a fallacy.
• These directives do not clearly translate into alternatives. While energy models are capable of generating the relevant alternatives, there are no such models currently, hence rendering an MCA practically unfeasible.
A way around these constraints is by using REEEM’s objectives as the focus of the MCA and using its pathways as alternatives. In doing so consistency can be maintained between objectives and alternatives and at the same time producing meaningful results. Further, REEEM’s pathways have been defined through elaborate discussions[8] which strengthens the MCA’s results.
Hence the driving objectives of the MCA are derived from that of REEEM’s which states that, “REEEM
aims to gain a clear and comprehensive understanding of the system-wide implications of energy strategies in support of transitions to a competitive low-carbon EU energy society, given the objectives and framework outlined in the Strategic Energy Technology Plan.”[6] As an addition, this also accounts
21 point of time but is expected to be available at the end of the project[46]. This delivers the alternatives of the MCA.
3.2 CRITERIA SETTING
After objectives and alternatives, the next aspect to be decided is the criteria. This forms the last part of the first step of MCA as defined by Mateo [11]. Unlike objectives and alternatives, criteria setting can be generalized. While it is given that the criteria must be relevant to the purpose of the study, a set of common criteria can help compare across studies. Further, this generalization can help set norms which can make public debate over energy management policies on an even platform. It can prevent certain studies from being undercut by their inclusion or exclusion of certain criteria.
Keeping this value in mind, the criteria used in the literature review were compared to look for norms and key differences. As stated in section 2.2 State-Of-The-Art, there were no established standards for criteria selection. However, five major criteria themes could be observed. These were: economy, society, environment, technology, and institutions or politics. They are referred to as criteria themes as they represent the broad areas covered by the criteria used in literature. Some researchers considered criteria belong to all five criteria themes, while other chose a subset. The terminology used varied significantly.
Another aspect of criteria setting is the logic behind their selection. Only a few papers attempted to logically derive a reasoning for the selection of the criteria adopted. This reasoning is crucial as it defines the boundaries of research. Volkart, et al. [29] used one of their earlier work[47] which had an explanation on their criteria selection logic. Most other research used criteria defined through multiple other research without providing an explanation on why they were chosen as in [31-36]. Some offered no justification altogether like in [30, 38].
Hence, criteria setting was done with two requirements- it must be in line with standard criteria themes and it must be logically reasoned.
Since this MCA deals with studying pathways as alternatives, there are significant assumptions which are built into the data that feeds the analysis. A pathway is defined by setting futures based on certain economic, political and global assumptions. Then technology related assumptions are exogenously provided to build the pathways. Hence, studying a “political” or “technology” dimension is not meaningful as it is already built into the results. A simplified way of viewing this logic would be that the aim is to “study the impact of the pathways” and not “assess the likelihood of the pathways”.
The Brundtland Report in 1987[48] spoke about three pillars of sustainability which has since become used to define the term sustainability. Those three pillars were economy, society and environment. Since there is an established norm on these three themes and the exclusion of the institution and technology can be logically reasoned, economy, society and environment were chosen as the three themes of criteria setting.
22 weighed against each other. Further, an indicator can be relevant under multiple themes which can confuse the decision maker.
To arrive at the criteria, each of the three main themes were divided into 7 criteria. Of these seven, 3 refer to just the themes in isolation, 3 refer to overlap of two themes and 1 which is the overlap of all 3 themes. It is visually represented in Figure 7. The terms used to refer to the criteria are chosen to best represent the issues studied. However, they are prone to interpretations as there is no established norm for their usage in energy management studies. To see how these criteria, match established literature, a comparative study was made. Studies with four different types of alternatives- policy, pathway, technology and project, were compared in Figure 8. This figure also shows the lack of standardization mentioned at the start of this section.
23 Figure 8: Comparison of criteria used in literature
The seven criteria as derived in Figure 7 are further described as follows:
Climate: Impact of human activities on the climate resulting in global warming and increased number of natural catastrophic events.
This criterion is at the centre of sustainable planning and is the overarching motivation to transform the energy system.
Costs: Investments needed for the energy system transformation
This criterion represents the economic aspect of sustainable planning in isolation.
Ecosystem: Preserving the ecosystem comprised of living components and non-living components i.e. land, air and water.
This criterion represents the environmental aspect in isolation.
Welfare: The improvement in lives of the residents by higher incomes and better quality of life.
This criterion represents the social aspect in isolation. While there is a possibility of alternative interpretation based on de-growth concepts[49] developed by ecological economists for this criterion, a classical understanding of growth and welfare is used.
24 This criterion represents the confluence of societal and environmental aspects of sustainable planning. It deals with lives of the society (humans) in the region considered.
Security: Here it refers to cost of using energy (affordability) and security of supply (accessibility).
This criterion represents the combination of social and economic aspects. Despite being labelled as security, it represents only two of the four elements as described by Kruyt, et al. [50]. The two elements considered are affordability which is studied through economic output per unit fuel consumed and accessibility which looks at import dependence.
Resource: Availability of natural resources which can be used for consumption or trade.
This criterion represents the combination of the environment and economic aspects. It covers the aspects of security labelled under “availability” by Kruyt, et al. [50]. These resources are subject to depletion.
In addition to this categorization, a sub-criteria level analysis is required for four criteria i.e., ecosystem, welfare, security and resource. This stems from the fact that these criteria have a range of interpretations depending on the metric used. The primary logic behind the sub-criteria design is to separate the criteria based on its constituent elements. They are described as follows:
Ecosystem is branched as shown in Figure 9.
Figure 9: Sub-Criteria Tree for Ecosystem
Living beings Comprises all the lifeforms in the region excluding humans.
Non-living beings Comprises of the remaining elements of the biosphere which is water, land and air Water Comprises of all the water bodies in the region i.e., saltwater, freshwater and ice caps. Land Comprises of the lithosphere which hold the soil and minerals
25 Freshwater Includes the rivers, lakes and other water bodies used for non-human consumption Soil It refers to soil quality for non-human use
Minerals It refers to the rock formations and natural resources used by non-human entities. It must be noted that the ecosystem criterion is vastly complex and an entire MCDM analysis can be made with just this focus. This research has been abstracted to the level where data can be obtained through the underlying model.
Welfare is branched into three elements:
Income It refers to the money earned by the residents of the region considered Jobs It refers to the state of employment of the residents in the region
Recreation It refers to the time spent by the residents involved in activities that bring non-monetary
returns
Security as considered has two elements:
Affordability It refers to the amount of energy required by the community for a unit of value
produced.
Accessibility It refers to assurance in supply of energy to community given the geopolitical
constraints of energy trade.
Resources is branched as shown in Figure 10.
Figure 10: Sub-Criteria Tree for Resource
Water It refers to the available freshwater that can be used for human or industrial consumption. Land It refers to the arable land that can be used to produce edible and fuel crops.
Minerals It refers to mineable ores that can be used to produce metals or industrial minerals. Fuel It refers to fuel reserves that are economically viable to be exploited to produce energy for
26 Natural Gas It refers to the determined natural gas reserves.
Oil It refers to the crude or refined oil reserves.
Coal It refers to metallurgical coke or coal reserves that can be used by the power or industry sector. Biomass It refers to the biomass that can be processed to produce biofuel.
The completed criteria tree after inclusion of the sub-criteria is represented through Figure 11.
Figure 11: Criteria Tree for Low Carbon Pathways 3.3 INDICATOR SELECTION
An aspect of MCA closely aligned with criteria setting is the indicator selection. They are qualitative or quantitative metrics which are used to represent the criteria and sub-criteria. It is at this stage that data collection or an energy model is needed.
Often, a criterion can encompass more than one relevant indicator. In such cases, ignoring other relevant indicators limits the use of the underlying study. One option is to weigh each of these indicators within the criteria. Then, it is combined with the weights of the criteria as done in [34-36]. Alternatively, the decision makers can be allowed to choose the indicators for each of the criteria as done by McCollum, et al. [38]. A third method is to use the range of indicators as a form of uncertainty analysis. While this method may make the results more inclusive, it requires a lot more data and can be difficult to compare results at the end.
Since this MCA framework is based on the pathways considered in REEEM, it is also constrained by the REEEM integrated energy model. Only the indicators which are the outputs (primary) or derived from the outputs (secondary) can be used as opposed to indicators that are relevant in the context of criteria but can’t be obtained. A list of primary and secondary indicators from the REEEM model was provided in Avgerinopoulos, et al. [51]. Out of the 105 listed indicators, 35 were chosen to represent the criteria described in the previous section. The indicators are tabulated in Appendix 1: Proposed Indicators for MCA framework. Data is required for each of these indicators for all the alternatives. In addition to this, a temporal dimension is needed as the pathways cover a period from 2010 to 2050.
27 Once the goals, alternatives and criteria are established, the next step in the MCA is “assigning criteria weights”[11]. It is at this stage where the preferences of the decision makers are considered into the analysis. They are quantified in the form of weights for the criteria and sub-criteria. There are several mathematical and logical methods to quantify the preferences. A review of prominent weighing methods in literature with reference to the objective of the MCA is done by Yeh, et al. [52]. It must be noted that weighing of criteria and the next step, which is the select of the MCDM, are interlinked. Hence, even though the report follows a linear structure, the actual process requires iterative thinking. In Siskos and Tsotsolas [14], six prominent weighing methods were briefly described as follows -
• the method of cards as proposed by Simos [53]
• the method of centralized weights by Solymosi and Dombi [54], which requests from the DM a number of ordinal comparisons of criteria which are formulated as linear inequalities, to obtain the centroid of the vertices of a polyhedron
• the TACTIC method[55] in which the relative importance of the criteria is depicted and assessed as a system of functional representations of relations
• DIVAPIME[56], which has been adapted to the ELECTRE methods and is implemented by making pairwise comparisons of fictitious alternatives, in order to identify the importance variation intervals
• the analytic hierarchy process (AHP), proposed by Saaty [12], where the DM is asked to provide pairwise comparisons over the priority of criteria on a prespecified numerical scales
• MACBETH[57] which infers the weights as values of attractiveness from pairwise comparisons of the criteria on a qualitative scale, thereby measuring the magnitude of attractiveness. When cross-referencing these weighing methods with the sample literature review, two methods emerge as popular choices- AHP and its variant ANP [31, 32, 35, 36]and Simos method [33, 34]. This statement is further supported by Mateo [11] as well as Wang, et al. [27] who state in their review article that, “Subjective weighing methods such as pair-wise comparison, AHP and Simos, were the most
methods in sustainable energy DM.” Using this as an early filter, AHP and Simos methods were studied
in detail.
28 Figure 12: Example of DM data input for criteria weighing using Simos method
The AHP method developed by Saaty [12] is a complete framework for MCA that includes weighing of the criteria as well as a method to evaluate the MCDM matrix. Unlike Simos method, AHP relies on a pairwise comparison of the criteria. The DM assigns a value from 1 to 9 to the favoured criteria. 1 indicates “equal importance”, 3 indicates “weak importance”, 5 indicates “strong importance”, 7 indicates “dominant importance” & 9 indicates absolute importance. Other values are middle ground between the levels. In the example shown in Figure 13, Apples show “weak importance” when compared to Oranges, Mangoes show “dominant importance” over Oranges, and they show “strong importance” over Apples. Once these choices provided by the DM, the consistency of the importance is checked and weights for the individual criteria are calculated. In Table 2, the AHP method is compared to the Simos method.
Figure 13: Example of DM data input for criteria weighing using the AHP method Table 2: Comparison of AHP and Simos method for weighing criteria
Simos Method AHP Method
DM data input time is quick even with a large number of criteria
DM data input time increases dramatically with increase in number of criteria
Requires minimal mathematical analysis post DM data input
Requires mathematical processor to arrive at the weights from DM data input
Usually robust in its results[14] Occasional rank reversal may occur[58] Difficult and an unintuitive data input process
for the DM
Easy and intuitive data input process for the DM
Rigid and requires additional assumptions if some criteria are omitted at a later stage
Flexible and requires no assumptions if some criteria are omitted at a later stage
Does not check for consistency of DM choices Checks for consistency of DM choices
29 Based on the number of criteria and sub-criteria, the number of comparisons to be made by the DM is 21 for the criteria and 21 for the sub-criteria. These comparisons can be taken in by any simple survey tools. A method similar to the AHP survey method as used by Sato [25] is proposed. The DMs will have to assign a value from 1 to 9 in favour of a side for each comparison. The descriptions associated with these values were taken from Park [60]. In addition to this, a sample comparison is shown for the DMs to understand the user interface. A MS Excel based survey for the proposed criteria is shown in Appendix 2.1 Survey.
The ability to compound multiple DMs’ choices, check for consistency and capability to handle criteria omission due to missing data, supports the utilization of AHP. These aspects are analysed in sections 3.6 Practical Considerations and 3.7 Sensitivity Analysis. However, if Simos method is used for assigning criteria weights, it will still be consistent with the MCA framework.
3.5 SELECTING THE MCDM METHOD
After the criteria weights are assigned, the next step, as proposed by Mateo [11], is to select the appropriate MCDM method. This is a crucial step as the MCDM method determines the core logic and output of the analysis. Further, it also constraints the criteria weighing methods. A MCDM method is fundamentally an algorithmic framework that combines the values of alternatives under various criteria and the relative importance of these criteria. Since these algorithmic frameworks are well-documented, they have often been referred to as a MCDM by itself. However, in this report they are referred to as MCDM methods or just methods.
30 Figure 14: Frequency of MCDM Methods used for Energy Management (Constructed from Mardani, et al. [5] ) However, at a fundamental level the MCDM methods shown in Figure 14 can be classified in two categories- outranking methods and value measurement methods. A third category was included in this categorization by Belton and Stewart [64]. However, the third category, Goal and Reference Point methods, is not represented through any of the described methods2[65]. Outranking methods use a
direct comparison between alternatives across categories to identify preferences, indifference and incomparability. Examples of this method are PROMETHEE, VIKOR and ELECTRE. On the other hand, the value measurement method assigns a final score to each alternative and this score is used to rank the final preferences. Examples of this method include AHP, ANP and TOPSIS. In addition to this categorization, a “fuzzy” variant for these methods is also used. A fuzzy analysis tries to include the uncertainty in the DM’s mind by using a probabilistic model for the preferences as opposed to fixed values.
Given the range of methods, the selection of an MCDM method is context specific but needs to be done in a logical fashion. In the MCA guidance manual by Dodgson, et al. [10], seven categories are provided to help this selection process- internal consistency and logical soundness; transparency; ease of use; data requirements not inconsistent with the importance of the issue being considered; realistic time and manpower resource requirements for the analysis process; ability to provide an audit trail, and software availability, where needed.
Using this guideline, three MCDM methods have been compared. These methods have been selected to represent the simpler variants of more complex methods. They are AHP, PROMETHEE and TOPSIS whose more complex variants are ANP, ELECTRE & VIKOR and MACBETH. This comparison is shown in Table 3 and where a classification of high, medium or low preference is shown by the colour of the cell. This classification technique follows the one used by Kurka and Blackwood [9] where three MCDM methods were compared. In Baležentis and Streimikiene [30], a comparison of MCDM methods was done for two energy models. However, the other steps in the MCDM like criteria selection and weighing were simplified.
2 The method refers to MCDM which aims to meet given targets. For example, if this study was done to see the pathways’ success in meeting the Sustainable Development Goals, then it would be relevant.
0 10 20 30 40 50 60 N o . o f Pu b lic a tio n ( 995 -2015) MCDM Method Hybrid MCDM and FMCDM AHP and Fuzzy AHP
TOPSIS Fuzzy TOPSIS ANP and Fuzzy ANP PROMETHEE and Fuzzy PROMETHEE
31 Table 3: MCDM methods selection criteria comparison
AHP PROMETHEE TOPSIS
Internal consistency and logic Documented and tested algorithm Documented and tested algorithm Documented and tested algorithm Transparency
High with consistency check for criteria weights
Low with preference function requiring several assumptions Medium with requirement to calculate n-dimensional Euclidean Distance Ease of use High with 1
computation step for preference matrix
Low with several unintuitive steps for calculating pre-order Medium with 3 computation steps to evaluate the preference matrix Data requirements consistency Medium as complete data set is needed to compute final score.
Low as outranking involves comparing two data points
High requirements for min and max values in a criterion
Time and manpower resource
Medium due to long survey to get DM preferences
High with multiple computation steps Medium due to computation time Ability to provide an audit trail Medium as some inconsistency in weights may be observed Low as preference functions can be arbitrarily defined. High as no assumptions required in the evaluation step
Software availability
High with new work package available in R
Medium with slightly outdated work packages in R
Medium with slightly outdated work packages in R
Based on this analysis, the AHP method has been selected for this MCA framework. This is consistent with the AHP criteria weighing method chosen in the previous section. Aside from the consistency in logic, the selection of this method comes down to its ease of use, short computational time and tool support. In addition to this, the AHP method is also capable to handle several forms of missing data as has been discussed in section 3.6 Practical Considerations along with a review on software tools used for MCA.
It must be noted that complex variations of these methods were avoided due to keeping the MCDM simple and easy to explain for the DM. Further, the “fuzzy” variants were also avoided as the pathways by themselves do not provide a single recommendation. Such complex variants are best suited for MCA where alternatives involve policy action or technological solutions. Meanwhile, other simple MCDM methods like MACBETH or PROMETHEE can be used for studying low carbon transition pathways provided consistency with the criteria weighing method is ensured.
3.6 PRACTICAL CONSIDERATIONS HANDLING MISSING DATA
32 effective analysis as the algorithms requires data to run. Given the significant impact of the issue, counter measures must be planned in advance. An overview of the popular techniques to handle missing data in MCDM was given by Mareschal and Hadouchi [16]. These techniques were categorized as external and internal techniques.
The external techniques refer to solutions that do not depend on the MCDM used. One of them is “simple elimination” where a portion of the MCA relevant to missing data is excluded from the analysis. Another is “substitution” of the missing data. This can be done by placing an arbitrary value, choosing central values i.e., mean, median or mod, or, by applying regression among the available data. On the other hand, the internal techniques depend on the MCDM method selected with particular relevance for outranking methods such as PROMETHEE or ELECTRE. Conceptually, these techniques rely on “pairwise comparison” of the alternatives and elimination before the need for the missing data is felt[16]. In Jimenez, et al. [17], the substitution technique using central values has been explained with an example. While this is a form of internal technique, the given example is for an AHP analysis where the upper and lower limits for the missing data is known.
In the case of studying transformation pathways, missing data can occur in many forms. A fundamental assumption here is that data for these pathways are largely sourced from energy models. The key ones are as follows:
• Pathways: When comparing pathways, some of them might not have been completely defined and modelled yet, so they might have insufficient data. For example, the base pathway might be fully modelled while the new policy pathway is still at a conceptual stage.
• Criteria: Often in case of integrated energy models or MCDM involving qualitative and quantitative analysis, not all criteria might be modelled at the same time. For example, the “costs” and “climate” criteria might be modelled while the “ecosystem” and “resource” has not been integrated to the model yet.
• Indicator: Since each of the criteria is underpinned by one or a set of indicators, the available data might not match the indicators used while developing the MCDM. For example, the planned indicator for “costs” might be levelized cost of energy but the closest available indicators for the criterion is system costs.
• Time-steps: A pathway by its design covers a time span but there can be a mis-match between the pathways on the time-steps for which data is available. For example, the base pathway starts from 2010 and has 5-year time steps while the new policy pathway starts from 2011 and has 10-year time steps.
• Geographical: Excluding world energy models, all other energy models are defined by a geographical boundary. In such cases, not all data refer to same region. For example, the economic analysis may include all of Europe, but the social analysis is done only for a sample of six European countries.
• Random: Finally, data can be missing with no underlying pattern. These can occur due to model errors or random human errors.
33 is the elimination from analysis. While this doesn’t affect the MCDM structure, it can defeat the purpose of the MCA unless acted upon at a different stage. This can be done through descriptive case study of the missing elements or using the knowledge available among experts in a workshop-like setting. Table 4: Recommendations for specific types of missing data
Type of missing data Recommendation Trade-off
Pathway with all criteria missing Eliminate from analysis Smaller scope of analysis Pathway with some criteria
missing
Substitution by regression Assumed correlation Substitution by central value Central tendency bias Criteria data missing Eliminate from analysis Smaller scope of analysis Indicator data missing/not
matching
Re-assess MCDM structure Time consuming
Standardize across alternatives May weaken MCDM logic Time-step data
missing/mis-aligned
Substitution by regression Central tendency bias Geographical data misaligned Standardize across alternatives May weaken MCDM logic Random missing data Substitution by central value Central tendency bias In addition to specific recommendations, some general counter measures can be taken. A case study of missing data treatment can be seen in Image [19] where guidelines on preference between the recommendations are given. One general suggestion is to keep the statistical analysis simple. An MCA is meant to clarify the decision-making process and unite the stakeholders. A complex analysis will hamper the ability to process the results. Similarly, using professional judgement may improve the analysis but reduce the transparency.
Before the start of the MCDM, it is recommended to have a plan on how to deal with missing data but room for flexibility must be kept. In particular, the sensitivity analysis must be done depending on the type of missing data and the form in which it impacts the MCA. One positive takeaway from missing data is that it guides the researcher to look for the missing data and saves time from finding redundant or irrelevant data. Finally, it is crucial that all the assumptions while handling missing data is documented along with the analysis.
TOOLS FOR DATA COLLECTION AND PROCESSING
Another practical consideration when developing an MCA framework is the set of software tools for data collection, processing and visualization. A lack of such options can discourage the use of MCDM for the analysis or make it a tiring process. Some of the requirements in these tools are as follows:
• Easy to use: The software must be intuitive and simple to set up. Often the MCDM planners are not specialists in the software but rather specialize in their respective areas of work.
• Available at a low price: Purchasing a full license for a software can be expensive given that the MCDM is likely to be used only a few times
• Compatible with the MCDM requirements: A basic requirement is that the tool is capable of handling data input and computation of the MCDM method
34 In International Society on MCDM [26], a list of latest software for MCDM has been compiled. However, such software is constantly being developed and disposed so it is recommended that the latest software is identified during the time of the MCDM. In this research framework the proposed software tools used are Microsoft Excel and RStudio. Excel is a staple across most devices and is used for basic data collection and clean-up. It can also be used for designing the survey for getting the criteria weights. Meanwhile, RStudio is a free and open-sourced platform for developing programmes using R[66]. RStudio is specialized for data handling and processing. It has free packages for AHP calculation and data visualization making it handy for this MCA framework.
3.7 SENSITIVITY ANALYSIS
The final section of the MCA framework is the SA. It must be understood that the SA is dependent on the mathematical methods used to obtain the result, so it varies from case to case. This is explained by Pilkey and Pilkey-Jarvis [67] as, “It is important, however, to recognize that the sensitivity of the
parameter in the equation is what is being determined, not the sensitivity of the parameter in nature”.
The basic SA methods3 used in MCA as inferred from literature [21-23] and demonstrated by Butler, et
al. [24], are one dimensional SA, two dimensional SA and Monte Carlo simulation. These three SA methods considered in this report and are elaborated later within the context of their recommended usage.
As described by Saltelli [20], an SA starts by identifying the uncertainty in the input. Since the MCDM is built on an energy model, the uncertainty can arise at either of those two stages. In this section, a SA framework has been proposed in alignment with the designed MCA framework. The proposed SA frameworkstarts by identifying the uncertainty source in its various forms. The next step is to measure the uncertainty which varies depending on the quantitative method used. Finally, a SA method is proposed to study the impact of the uncertainty. In case the uncertainty is determined to be at a basic level, then a separate study is suggested where the entire modelling or MCDM process is repeated with a different methodology. The proposed SA framework is shown in Table 5.
Table 5: Proposed Sensitivity Analysis framework for the MCA framework
Uncertainty Source Uncertainty Measurement Sensitivity Method DMs individual preferences
can change
Fuzzy calculations Requires separate study Difference in preferences
among DMs
Surveying multiple DMs Monte Carlo simulation Models' algorithms Calculating theoretical
uncertainty associated with the model
Requires separate study
Data input into energy models Comparing probabilistic and deterministic models with same data input
Requires separate analysis
Missing data that has been left out
Using dummy variables One-dimensional sensitivity analysis
35 Missing data that has been
filled through quantitative reasoning
Variance associated with the mathematical method
One or two-dimensional sensitivity analysis Missing data that has been
filled through qualitative reasoning
Surveying multiple experts One-dimensional sensitivity analysis
Several indicators representing a criterion
Swapping across indicators Iterating evaluation matrix Difference between MCDM
methods
Comparison of methods Requires separate study Difference between weighing
methods
Comparison of methods Requires separate study Uncertainty levels change
across pathway years
Context specific Requires separate research method
Among the uncertainties, that could be the SA are difference in preference among the DMs and the missing data in the MCDM. Other uncertainties like the MCDM method, weighing method or the energy model input are more fundamental in nature. To study these, a secondary study must be done with the alternate method and then the results could be compared. For example, a study using PROMETHEE method can be done and the final pathways ranks can be compared with that done from the AHP study. When it comes to missing data, the one-dimensional SA varies a single parameter across a range of values and sees its impact on the results. For example, if the indicator “biodiversity losses” is assumed to be zero in a pathway due a lack of data, then its value can be changed across a range, like -25% to +25%, and the consequent change in pathway score can be observed. However, this method doesn’t account for correlation among the data. In the two-dimensional method, two variables which are correlated are varied across a range of values. For example, let us take a case where the indicator, “biodiversity losses” for two pathways are filled by doing a regression with another indicator, “external health costs” which is available for all pathways. In this case the change in the degree of correlation can result in a change in score for both the pathways which can be studied simultaneously.
When a final value is obtained by mathematically combining several values with its own variances, then a Monte Carlo simulation can be used to study the variance associated with the final value. For example, consider calculating the criteria weights. If there are five DMs, then the average of their preferences for each criterion is used to calculate the final weights. However, for each criterion there is a variance associated with the preferences. Using a Monte Carlo simulation, the variance in final weights can be calculated. This is demonstrated by Mateos, et al. [22].
3.8 SUMMARY OF THE FRAMEWORK
36 Figure 15: Visual representation of proposed MCA framework
When viewing this figure an observation can be made. The only parts of the framework dependent on REEEM are the objectives, alternatives and indicators (as seen on the left side of Figure 15). This implies that this framework could be modified to study low carbon pathways in other regions. The requirements to do so would be the presence of either- an IEM like REEEM focussed on that particular region, or, a set of indicators calculated from databases where the pathways are designed consistently. The latter is possible for smaller regions where an IEM is not required for calculating the pathways. Hence, while the scope of this research is set to the EU specifically, the findings could be expanded to a wider context. Approaches to do so has been discussed in Chapter 5 and Chapter 6.
3.9 LIMITATIONS OF THE FRAMEWORK
In this last section of the chapter, the limitations of the MCA framework are discussed. At the fundamental level, this study aims to bridge the science-policy divide by helping DM better understand energy pathways as represented via energy models. Hence, there are assumptions at all the stages which can limit the veracity of the framework. However, it is important to acknowledge that not all limitations are equally significant. When reviewing the study, focus must be given to improve upon the key limitations of the study. The categorization of key limitations is based on the author’s interpretation of the research and is open to alternative interpretations.
Keeping this in mind, the key limitations are:
1. This framework studies transition pathways indirectly. The narratives of the pathways are first translated to energy models whose results are then used. If the pathway narratives are at fault or biased at the energy modelling stage, the results from this framework would be flawed. 2. Indicators are not true representations of the criteria. Selecting any indicator to represent a
criterion comes with an assumption and this selection process can be manipulated by the DM. 3. Stakeholders were not surveyed while designing the framework. Hence, there may be
37 1. The pathways as formulated by REEEM are complex and not intuitive. A DM will be expected
to know about the pathways along with the MCDM results when formulating their actions. 2. The sub-criteria logic is not as elaborated on as the criteria logic.
3. The logic for selection of the MCDM method and criteria weighing is a case of “best practises” rather than analytically-derived
4. The AHP method has its inherent flaws such as rank reversal
38
CHAPTER 4: CASE STUDY
4.1 OVERVIEWIn this chapter, the framework proposed in Chapter 3 is demonstrated through a case study. The background of the case study is the same as the MCA framework with the REEEM project[6] as the source of the data and the low carbon transition pathways. The goals of the case study are as follows:
• Demonstrate the elements of MCDM as proposed in the framework
• Obtain results and visuals that give an indication of how a completed MCDM will be used by the DMs
• Identify early trends in the studied low carbon pathways
In the next section, the case study is reported in stages as shown in Figure 16. This corresponds to how practically the analysis is done and is developed based on the stages proposed by Mateo [11](See: Figure 4).
Figure 16: Case Study analysis flowchart 4.2 APPLYING THE FRAMEWORK
Step 1: Objective Setting
This case study aims to represent the MCA framework and the objectives is to study the low carbon pathways. These pathways are obtained from the ongoing REEEM project. In all, five pathways have been considered- Base, High Renewables (HighRES), Storage Innovation (SI), Pilot 1 and Pilot 2. These pathways have been described in Deliverable 1.2[51]. Further, only two time steps have been considered- 2015 and 2040.
Step 2: Criteria and Indicator Selection
39 Figure 17: Case Study Decision Tree
The data was obtained from energy models built on The Integrated MARKAL-EFOM System (TIMES), National European Worldwide Applied General Equilibrium (NEWAGE) and Eco-Sense or calculated as secondary indicators from these models. The indicator used for each of the criterion is shown in Table 6 along with its source.
Table 6: Case Study Indicators
Criteria Indicator Unit Source
Climate GHG Emissions Mt TIMES
Costs Total Investment Costs Million Euros Calculated from TIMES Ecosystem Biodiversity losses Billion Euros Eco-Sense
Health Disability Adjusted Life-years thousand years Eco-Sense
Security-Accessibility Net Import Dependence # Calculated from TIMES Security-Affordability Average Fuel Costs Euro/MWh Calculated from TIMES Welfare Gross Domestic Product Billion Euros NEWAGE
Step 3: Getting Criteria Weights
The next step is to get the criteria preferences from the DM. In line with the framework, the AHP method was used. In this case study, five researchers at the KTH Division of Energy Systems Analysis were consulted. Given their experience in energy modelling, their expertise is representative of high-level energy policy decision makers.
A survey as described in the framework was created on Excel and shared digitally. The data from the survey is shown in Appendix 2.2 Survey Data. This was then calculated according to the AHP method. The criteria and sub criteria weights and the inconsistency for all five DMs as well as a combined average is shown in Appendix 2.3 Criteria Weights.
