sustainable development in practice (case studies for engineers and scientists) || multi-criteria...

30
12 Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa Jim Petrie, Lauren Basson, Philippa Notten and Mary Stewart Summary Chapters 9 and 10 considered future energy scenarios and technologies respectively. In this chapter we explore an old and well-established energy technology based on coal. However, we look into the future of this industry and use the South African example to analyse how this well-established sector can become more sustainable. Coal is likely to continue to be a dominant source of energy in South Africa (and in some other developing countries, including China). We seek to understand the broader consequences of coal-fired power generation in terms of regional sustainable development. We argue that there is a need for more transparent, legitimate and defensible decision-making processes to ensure that stakeholder concerns are cap- tured in the process, and that the implications of any trade-offs which ensue from decision outcomes are made explicit. We in particular address the issue of management of uncertainty in the decision-making process. We make use of a technology design case study to demonstrate some decision-support tools which meet these objectives. 12.1 Introduction It is not an unreasonable comment to suggest that South Africa’s economic develop- ment, like development of any other developing country, is tied inextricably to the cost Sustainable Development in Practice: Case Studies for Engineers and Scientists Edited by Adisa Azapagic, Slobodan Perdan and Roland Clift Ó 2004 John Wiley & Sons, Ltd ISBNs: 0-470-85608-4 (HB); 0-470-85609-2 (PB)

Upload: roland

Post on 06-Jun-2016

213 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

12

Multi-Criteria Decision Analysis:The Case of Power Generation

in South Africa

Jim Petrie, Lauren Basson, Philippa Notten and Mary Stewart

Summary

Chapters 9 and 10 considered future energy scenarios and technologies respectively.In this chapter we explore an old and well-established energy technology based oncoal. However, we look into the future of this industry and use the South Africanexample to analyse how this well-established sector can become more sustainable.Coal is likely to continue to be a dominant source of energy in South Africa (and insome other developing countries, including China). We seek to understand thebroader consequences of coal-fired power generation in terms of regional sustainabledevelopment. We argue that there is a need for more transparent, legitimate anddefensible decision-making processes to ensure that stakeholder concerns are cap-tured in the process, and that the implications of any trade-offs which ensue fromdecision outcomes aremade explicit.We in particular address the issue ofmanagementof uncertainty in the decision-making process. We make use of a technology designcase study to demonstrate some decision-support tools which meet these objectives.

12.1 Introduction

It is not an unreasonable comment to suggest that South Africa’s economic develop-ment, like development of any other developing country, is tied inextricably to the cost

Sustainable Development in Practice: Case Studies for Engineers and Scientists

Edited by Adisa Azapagic, Slobodan Perdan and Roland Clift

� 2004 John Wiley & Sons, Ltd ISBNs: 0-470-85608-4 (HB); 0-470-85609-2 (PB)

Page 2: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

of energy. Politicians, economic planners and the parastatal electrical utility alike allmake much of the current low cost of electricity and the critical role this has played todate in attracting economic investment to the country. In recent years, much of thisinvestment has been in the primary resources sector, including aluminium, zinc andsteel. This has been enabled, in part, by innovative electricity-pricing contracts, oftentied to commodity prices on the London Metals Exchange. This trend is set tocontinue for the foreseeable future, with several other large minerals refining projectsin the planning stages. At the same time, the national electrical utility has embarked ona highly ambitious programme to improve domestic (including rural) access to elec-tricity, resulting in an annual increase in new connections of 200 000–250 000 over thelast 5 years, and, through the African Energy Fund, promoting partnerships withother regional utilities to ‘‘light up Sub-Saharan Africa’’. This is occurring at a timewhen the industry itself is earmarked for deregulation and privatisation, and it is likelythat independent power producers will enter the market in the medium term.

Together, these trends have placed significant strain on the electricity network,and considerable increase in generation capacity is being contemplated over the next5–15 years. The challenge is how best to meet this increased demand – what supply-side generation strategies should be employed and how should these be balancedwith a comprehensive set of demand-side management strategies. Electrical powergeneration is predominantly coal-based, and the vast reserves of coal will ensure thatcoal-fired generation will continue to dominate into the medium term. This is not tosuggest that other forms of electricity generation are not in an advanced state ofdevelopment (including renewables), merely that the size of the demand, the invest-ment in existing infrastructure and current energy minerals pricing policies willensure that coal retains its dominance.

In this chapter, we seek to understand the broader consequences of this commit-ment to coal-fired power generation in terms of regional sustainable development.We argue that there is a need for more transparent, legitimate and defensibledecision-making processes to ensure that stakeholder concerns are captured in theprocess, and that the implications of any trade-offs which ensue from decisionoutcomes are made explicit. A particular challenge is to understand the implicationsof uncertainty in the various sub-models of the system under investigation (the issuealso raised in Chapter 2). We make use of a technology design case study todemonstrate some decision-support tools which meet these objectives.

Questions

1. Why does the cost of energy influence development of developing countries?2. How are developed countries affected by the cost and availability of energy?3. Where are the largest mineral reserves found: in developed or developing coun-

tries? How do you think the mineral reserves have helped the economic develop-ment of these countries?

4. Do your own research on the energy situation in the Sub-Saharan Africa. Whatdo you conclude: what are the priorities in this region with respect to energy?

5. What are the main sustainability issues for South Africa? What is the role ofenergy in the development of South Africa?

368 Sustainable Development in Practice

Page 3: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

12.2 Coal Mining and Environmental Challenges in South Africa

The official estimate indicates in situ coal resources of 121 218� 106 t in SouthAfrica, of which 55 333� 106 t are classified as economically recoverable (DME,2003a). Low-grade bituminous coal constitutes about 80% of the reserves (Bredell,1987). Opinions as to how long these reserves will last differ widely because of thedifficulty in predicting future export and usage figures. Assuming an annual increasein coal production of 1.8%, Surridge et al. (1995) predicted that coal productionwould peak by 2050, and then tail off over the next two centuries.

Coal mining is an important contributor to South Africa’s GDP, with total localand export sales amounting to R9564 million1 (152� 106 t) and R16 956 million(69� 106 t) respectively in 2001 (Prevost, 2002).

Over the years, coal mining has been a cause of major environmental degradationin South Africa (Van Horen, 1996). Coal mining, particularly opencast mining, isassociated with massive surface disruption. This results in changed land use andwater catchment patterns, in addition to the noise, visual intrusion, dust and watercontamination typically accompanying surface mining operations. Opencast miningaccounts for nearly 50% of South Africa’s coal production (Prevost, 2002). Under-ground mining is generally less disruptive. Water contamination, surface subsistenceand underground fires are some of the major impacts of underground mines (Wellset al., 1992). Only the environmental impacts of coal mining are considered here,although considerable social impacts also accompany coal mining, including theproblems of migrant labour and the high number of injuries and fatalities occurringin coal mines (Van Horen, 1996). In the first three quarters of 2002 (the mostup-to-date information available), 11 workers had died on South African coal minesand 115 had been injured (DME, 2003b). Social and ethical aspects of mining areconsidered further in Chapter 13.

Major environmental impacts stem from the high ash content of Southern Hemi-sphere coals, which require beneficiation to produce coals of acceptable quality forthe world markets. The resultant discard dumps are responsible for some of the mostserious environmental effects of coal mining, including land sterilisation and ground-water contamination. On exposure to air and water, the pyrites oxidise to formsulphuric acid, and iron oxides and hydroxides, which cause the pH to drop. Theacid produced reacts with basic minerals in the rock to form salts, in the process ofmobilising any heavy metals present. The resultant acid mine drainage (AMD) con-tains elevated levels of salts (mainly calcium and magnesium sulphates) and metals(predominantly iron, manganese and aluminium). The pyrite-rich discard is alsosusceptible to low-temperature oxidation (the so-called spontaneous combustion)and subsequent release of toxic air pollutants. AMD and spontaneous combustioncan be minimised by preventing water and air getting to the pyrite and other sulphidicminerals. The power stations mostly burn Run-Of-Mine (ROM)2 coal, so are notresponsible for discard production, although some power stations are supplied bydual-product mines. These mines maximise their coal production by producing

1 1R ¼ 0:1 US$ in 2001.2 Coal produced at the mine before any cleaning or preparation.

The Case of Power Generation in South Africa 369

Page 4: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

a high-quality coal for export (which must be cleaned) and a medium-quality powerstation coal which contains a portion of the washing discard blendedwith theROMcoal.

The location of South Africa’s coalfields is significant. The Mpumalanga/EasternGauteng/Northern Free State region, where 65% of the reserves are to be found, hasbeen extensively farmed, with little natural environment remaining. Coal miningtherefore has little residual impact on natural ecosystems and land rehabilitation isusually able to restore the land to an acceptable state (Wells et al., 1992). However,from a water quality perspective, the coalfields occur in the worst possible location,since most mines are situated in the vulnerable upper reaches of South Africa’s majorriver systems (Wells et al., 1992). In addition, regional air quality is adverselycompounded by the large number of minerals refining plants also present.

Questions

1. Why are underground mines generally less disruptive environmentally than thesurface mines? Discuss the differences in environmental impacts from these twotypes of mine.

2. Find out what are the main environmental and social issues associated withmining (you may wish to consult Chapter 13 for more detail on this topic).How do these issues affect developing countries such as South Africa?

3. Discuss the importance of coal for energy generation in South Africa. How easywould it be for this country to switch to a cleaner source of energy such as naturalgas or renewable sources? Discuss the socio-economic implications of thesechanges.

4. How could fuel cells be used in a country such as South Africa to provide energyto remote rural areas? Would this be an ‘‘appropriate’’ technology for SouthAfrica? You may wish to consult Chapter 10 for discussion on fuel cells.

12.2.1 Discard Coal as a Resource

In 2001, ROM coal production was 290�106 t of which 223:5�106 t was saleable(Prevost, 2002), the balance being described as ‘‘discard’’. Cumulative stocks ofdiscard coal exceed 800�106 t (DME, 1998). The percentage of these discard stockswhich are burning spontaneously at any one time is variable, but estimates in excessof 30% have been tabled (DMEA, 1985). The economic value and environmentalimpact of coal discards have been well documented.

Coal discards form a valuable reserve, even though they are of low quality. Since they are

located above ground they can be reclaimed by beneficiation at a competitive cost as they do

not incur a further mining cost.

There are at least two potential means of making use of discard coal, firstly the beneficiation

of discards to yield conventional coal products and, secondly, the combustion of raw

discards, for example in a fluidised bed combustor. Research has indicated that approxi-

mately 60% of the accumulated discards can be used for energy application purposes.

370 Sustainable Development in Practice

Page 5: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

Government is in support of current techno-economic investigations for power production

from discard coal. Government will continue to investigate and encourage options for the

utilisation of coal discard streams and stockpiles and will promote appropriate options for

the resultant energy and environmental benefits.

DME, 1998

We will return to the potential of fluidised bed combustion (FBC) of discards in thesubsequent case study.

Questions

1. Explain how waste from coal production can be used as a resource. Suggest whatthis might mean in the South African context. What are the barriers to this? Howcan the use of coal waste be encouraged by governments?

2. Find out how the beneficiation of coal discards could be carried out to recovermore coal. How can combustion in fluidised bed help reduce the problem of coaldiscards?

12.3 Coal-based Power Generation

South Africa’s electrical power industry is dominated by coal, a situation which willremain for the foreseeable future, given the cheap and plentiful supply available andthe strong political drive to keep electricity prices low (Notten, 2002). Significantenvironmental, social and economic effects stem from this large-scale mining andcombustion of coal.

Eskom, South Africa’s electrical utility company, is the fifth largest in the world(rated according to both sales and capacity), with a nominal installed capacity of40GW (Eskom, 2002). In 1999, it supplied 95% of South Africa’s total availableelectricity, with the balance supplied by municipalities and industries that generatepart of their electricity requirements. This translated to a total of nearly 188TWh(net) produced in Eskom stations, with a small percentage (1.4%) of the electricitysold imported from neighbouring countries (Eskom, 1999). Eskom also dominatesthe Sub-Saharan electricity supply, and is responsible for 76% of installed capacityand 83% of production and trade of electricity in this region (Lennon, 1997).Eskom’s capacity is heavily reliant on coal, with 89% of the total nominal capacitybeing provided by coal-fired stations, which consumed 96:5�106 t of coal in 2001(Eskom, 2002). The ability of the modern power stations to burn low-grade coalsmeans that primary energy costs can be kept very low, making it extremely difficultfor other fuels to penetrate the market. In addition, these coals are relatively low insulphur compared to world averages, so the stations are run without flue gasdesulphurisation (FGD) units.

Eskom’s integrated strategic electricity planning process (ISEP) has predicteda 50% increase in energy demand between the years 2000 and 2015, assuming a long-term economic growth rate for South Africa of between 1.5 and 3.5% (Lennon, 1997).

The Case of Power Generation in South Africa 371

Page 6: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

Eskom is considering a number of established and new technologies to meet thisdemand, guided by the following factors (Lennon, 1997):

– capital and operating cost;– plant reliability and availability;– access to indigenous, low-cost fuel;– lead times;– operations flexibility (base load vs peaking);– water availability;– environmental considerations (likely to move up in importance with pending

legislation);– security of fuel supply;– local capacity to sustain technology;– funding availability; and– political considerations.

The continued dominance of coal is evident from the above list, that is an emphasison keeping costs low and using an indigenous fuel supply. Any anticipated technol-ogy intervention (whether of a ‘‘supply-side’’ or ‘‘demand-side’’ management nature)needs to be assessed in terms of these potentially conflicting criteria, both at a designstage and throughout its operation. This suggests the need for a rigorous frameworkwithin which the requisite information can be analysed to explore trade-offs betweencompeting objectives. In this case study, we identify such a Decision Support Frame-work (DSF), populate it with the required arguments and demonstrate the use ofseveral tools to support decision making around technology selection. Underpinningall these tools is a commitment to the philosophy of life cycle thinking and thegeneral management science of Multi-Criteria Decision Analysis (MCDA). The DSFshould elicit all the relevant information about the problem at hand, and, through aset of systematic (and defensible) constructs and arguments, facilitate the transform-ation of this information into a decision with (hopefully) sustainable outcomes.

Questions

1. Taking a life cycle approach, identify economic, environmental and social issuesassociated with power generation from coal, from its extraction to delivery ofelectricity to end user. Discuss in turn how these issues affect developing anddeveloped countries.

2. What is the role of large energy companies such as Eskom in sustainable develop-ment of developing countries? Visit Eskom’s website to see what the companyis doing for sustainable development in South Africa.

3. What decision criteria should be considered when deciding on new energy technol-ogies in a developing country such as South Africa? Which of these criteria do youthink are likely to be more important for an energy company such as Eskom?Why?

4. Which decision criteria might be important to a local community who would bebeneficiaries of this electricity generation, but whose residences are adjacent to theproposed new power station complex?

372 Sustainable Development in Practice

Page 7: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

12.4 A Decision Support Framework

Decision making in the face of multiple objectives, uncertainty and different stake-holder values (Chapter 13) is a challenging activity, generally undertaken withoutadequate consideration of these defining characteristics. As industry, including theelectrical utility sector, seeks to align its practices with the goals of sustainabledevelopment (Chapter 1), there is increasing pressure to ensure that decision makingembraces the ‘‘triple bottom line’’ of techno-economic efficiency, environmentalstewardship and social acceptability (as discussed in Chapter 11). The challenge fordecision makers is to choose the appropriate system boundary for planning anddecision making – not only in the temporal and spatial domains, but also in a socialsense, where the latter requires a sensitivity to the needs and perspectives of all thosewho may be affected by the decision outcomes. Of significance here is the issue oflegitimacy, where the emphasis is often not so much on the outcome of the decisionor who makes it, but rather on the process by which decisions are made. Particularlyfor decisions that may affect many people with different perspectives and which mayhave ramifications over large areas and for long periods, discursive and participatoryapproaches to decision making are essential (as clearly demonstrated in Chapter 13).At the same time, an overwhelming number of guiding concepts (e.g. pollutionprevention, cleaner production, waste minimisation, industrial ecology, etc.) anddecision-support tools (e.g. Life Cycle Assessment, Environmental Risk Assessment,Social Impact Assessment, etc.) have been developed. Decision makers may find itvery difficult to know when it is appropriate to use a particular tool and how tointegrate information from several tools to ultimately make a decision with somedegree of confidence. Before considering how this approach might be developed inpractice, it is first necessary to outline some of the constructs of decision analysis.

Questions

1. Describe the following approaches: pollution prevention, cleaner production,waste minimisation and industrial ecology. What can be achieved by theseapproaches and what are the main differences between them?

2. Describe the following decision-support tools: Life Cycle Assessment, Environ-mental Risk Assessment, Environmental Impact Assessment and Social ImpactAssessment. Explain in which decision-making context each of these tools can beused and summarise the main differences between them.

3. Why do we need a more formal approach to decision making in the context ofsustainable development?

12.5 Structured Approaches to Decision Making

Decision analysis provides a structured approach to decision making (von Winter-feldt and Edwards, 1986; Keeney, 1992). The first stage of this systematic approachto decision making is called problem structuring – the aim of which is to identify the

The Case of Power Generation in South Africa 373

Page 8: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

stakeholders and obtain agreement about the exact decision at hand, the objectivesthat need to be satisfied by the decision outcome, the alternatives available, how toassess these (i.e. to what extent they meet decision objectives) and to elicit thepreferences of stakeholders for particular decision outcomes. The next step in thedecision-making process – problem analysis – involves the evaluation of the alter-natives under consideration to determine to what extent these satisfy the decisionobjectives. This is followed by the selection of a preferred alternative and sensitivityanalyses to ensure that the conclusion is robust. The decision-making processdescribed above is not linear. Iterations occur both between steps in the cycle andin the cycle as a whole, as more information becomes available or further clarity isobtained about the information that is required in each step. This structuredapproach to decision making can be applied in all decision contexts and is the basisfor the DSF used here.

Problem structuring is essentially a discursive and deliberative process, and is bestdone through direct interaction between stakeholders. Rosenhead (1989) providesa review of tools to assist in problem structuring. An important assumption of thedecision analysis methodology is that those participating in the decision-making pro-cess are willing to engage in discussion and to reach consensus position on the problemstructuring elements, to facilitate subsequent analysis of the problem – in otherwords, these participants are interested in rational decision outcomes. In general,strategic and tactical problems tend to require more discussion and deliberationthan operational ones, since the former often involve a variety of stakeholderswith divergent interests, while, in the latter, the focus is more on the decision for asingle entity, where there is a clearer basis for consensus. The outcome of aproblem structuring exercise is often summarised by way of an objectives hierarchy,which shows the criteria that will be used to evaluate the alternatives under consider-ation and the attributes (expressed quantitatively or qualitatively) that will be enum-erated (where appropriate) to determine the relative performance of the alternatives.The issue of stakeholders’ interests and the decision attributes has also been dis-cussed in Chapter 2.

During problem analysis it is necessary to obtain data on the performance of thealternatives in all the criteria. This can be done by simply rating options relative toeach other based on experience, or could involve more extensive data gatheringand modelling to obtain more accurate performance information. A variety ofmethods and tools are available to provide environmental performance informa-tion. Important aspects include the choice of system boundary for modelling andthe quality of the performance indicators used. To avoid merely transferringenvironmental impacts to other stages of a production chain or to other media,the importance of more systemic models of industrial production systems isrecognised. As discussed in many other chapters in this book, the philosophy oflife cycle thinking can be applied to guide the scope of this analysis and providethe spatial and temporal boundaries for modelling. Life Cycle Assessment (LCA)provides a range of environmental indicators based on aggregated mass andenergy balance data (see the Appendix at the end of the book). These can beinterpreted as highly aggregated impact indicators at a global and regional level.Site-specific Environmental Assessment and Ecological Risk Assessment providemore accurate environmental impact information. The type of indicators used will

374 Sustainable Development in Practice

Page 9: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

be largely determined by the type of decision sought. In general, more aggregatedindicators are used for making strategic decisions and at the early stages of tacticalproject development decisions, and indicators based on more detailed (often sitespecific) modelling are used for later project stages and operational decisionmaking.

The challenge for decision makers is to evaluate the alternatives based on a largenumber of attributes expressed in incommensurate units. In the DSF, the recom-mended approach to resolving this problem is to use a range of MCDA methodol-ogies (Stewart, 1992; Belton and Stewart, 2002). In MCDA, each decision criterion isgiven due consideration, that is the performances in different criteria are not con-verted to a common scale (e.g. monetary units), nor is a single criterion selected asmost important, thereby allowing the performance in all other criteria to be definedin terms of constraints. Instead, MCDA methods enable simultaneous comparisonof performances in all criteria. Where the focus in the project is on the evaluation offinite sets of alternatives, Multiple Attribute Decision Analysis (MADA) methodshave been applied. The choice of MADA method is based, amongst others, on thetype of decision, the stakeholders involved in the decision-making process and theinformation available to support decision making. The approach taken here is basedon Multi-Attribute Value Function Theory (MAVT), which requires explicit state-ments of acceptable trade-offs between performance in different criteria and aggre-gates the performance information across all criteria to create a single index for eachalternative. This index reflects the extent to which the alternative meets the decisionobjective. The intuitive nature of this single index approach makes it more accessibleto stakeholders involved in technology selection projects, as is the case here. We willnow briefly introduce the value function approach before proceeding to discuss thecase study.

12.5.1 Value Function Approach

Value function methods aggregate the information about preferences for differentlevels of performance in each criterion (represented as marginal or partial valuefunctions) with information about the relative importance of the criteria, in order toprovide an overall evaluation of each alternative in terms of the preferences of thoseinvolved in the decision-making process. Value function methods have been dis-cussed in detail by many authors (e.g. Keeney and Raiffa, 1976; von Winterfeldt andEdwards, 1986; Keeney, 1992; Beinat, 1997; Belton and Stewart, 2002). The mostcommonly used aggregation model is the additive aggregation in which the valuefunction V(ai) is constructed from the partial value functions vj(ai) defined over theset of criteria j:

VðaiÞ ¼Xn

j¼1

wjvjðaiÞ ð12:1Þ

where wj is known as the ‘‘weight’’ of criterion j and ai is a particular alternativewithin the set of alternatives under consideration. In turn, the partial value functions,

The Case of Power Generation in South Africa 375

Page 10: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

vj(ai), are constructed from a sense of the inherent ‘‘value’’ of the performance ofalternatives in specific criteria. It is this mapping exercise which enables performancescores to be translated from incommensurate units into a common ‘‘value’’ score,thereby allowing for aggregation of scores for a given alternative across all criteria.This is the single index, V(ai). Partial value functions are usually standardised sothat the ‘‘worst’’ and ‘‘best’’ outcomes in each criterion are assigned a value of 0 and1 (or 100) respectively. When these standardising conventions are applied, theweight of a criterion is in fact a scaling constant with a very specific meaning, thatis it indicates the relative gain associated with an improvement from the worst to thebest outcome in the criterion.

Two points are noteworthy from the above. First, the shape of the value functionrelation for each criterion is a modelling choice for the decision maker, and reflectsa strength of preference which is informed by the range of performance scores in aparticular criterion. This suggests a relation between the construction of suchpreference relations and the choice of any data normalisation rules. Secondly,whilst weights are scaling constants in the value function method, they still involvevalue judgements, and are thus sources of inherent uncertainty. Both these featuresof the value function method suggest that there is a need for any DSF employingthis technique to look critically at management of uncertainty within the decision-making process (e.g. Notten, 2002; Seppala et al., 2002; Basson, 2003). We willreturn to this consideration of valuation uncertainty in the case study whichfollows.

Questions

1. Develop a flow diagram which identifies all the key steps in a structured approachto decision making. Distinguish between those which pertain to problem structur-ing and those relevant to problem analysis.

2. Problem structuring is regarded by many as the most valuable and significant partof decision support, and particularly for decision contexts involving multiplestakeholders and potentially conflicting points of view, interaction and deliber-ation are essential. List three methods which could be used to identify the keyelements that inform the decision process and organise these to enable furtherdiscussion and analysis.

3. A distinction is often made between ‘‘means’’ objectives and ‘‘fundamental’’objectives. Returning to the objectives list of the previous section, characterisethese according to this distinction.

4. What performance measures could be used to quantify the degree to which theseobjectives might be met by a particular course of action (or decision)?

5. Value functions are a method by which performance scores (in arbitrary units)can be converted to a commensurate ‘‘value’’ scale which reflects the decisionmaker(s)’ preferences. Explain the relationship between the decision maker’ssense of ‘‘risk’’ and ‘‘value’’ for three common value functions – linear, sigmoidal,convex. For environmental criteria, can you suggest whether particular valuefunction shapes are more meaningful than others?

376 Sustainable Development in Practice

Page 11: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

12.6 Case Study: Using Decision Support Framework for Technology Choice

12.6.1 Problem Structuring: The Decision Context

The DSF is really a structure within which various models – of problem structure, ofthe alternatives under consideration, of preferences and aggregation models – arebrought together into a systematic assessment, within which various forms ofuncertainty can be identified and their residual influence on decision outcomes identi-fied. The above discussion has highlighted the importance of problem structuring,and identified an approach for consideration of stakeholders’ preferences (assumingthese can be articulated as part of the process). It is time now to turn our attention tothe technology case study, and develop appropriate models of the alternatives to beconsidered.

The problem structuring exercise generated the following statement of the pro-blem to be investigated. The potential of FBC as a technology for reducing SO2

emissions during coal combustion is well known. Of particular importance in thiscontext is its potential for burning colliery discards, thereby avoiding the signifi-cant environmental impacts of stockpiling the discard. However, burning discardin a fluidised bed results in a trade-off between impacts, primarily between thewater and air pollution from the stockpiled discard, and the emissions to air andash waste produced when the discard is burned. This case study investigates thecombustion of discard coal in a reconditioned power station. The study again takesplace at an early stage in the design process, and the operating conditions of thesystem are not yet fixed, although the environmental profile of the system isexpected to vary considerably with differing operating conditions. The study hastherefore been reformulated to some degree, and aims to determine the conditionsunder which the system need operate to achieve a particular level of certainty in netenvironmental benefit, rather than characterising the environmental performanceof a ‘‘typical’’ operation. The key question to be answered has therefore beenformulated as:

Under what operating conditions is it environmentally beneficial to re-power an old

pulverised fuel (PF) station with an FBC boiler burning discard coal?

This particular formulation of the question focuses the study on a parameter ana-lysis, rather than trying to pre-define a number of tightly defined scenarios (you mayrecall the scenario approach that has been discussed in Chapter 9). In this way,significant decision variables (e.g. the quality of the discard sourced, the type ofsorbent used, etc.) are investigated as part of the model parameter uncertaintyanalysis, rather than specified in discrete scenarios ‘‘up front’’.

Question

1. Describe in your own words the decision-making context in this case study. Whoare the decision makers?

The Case of Power Generation in South Africa 377

Page 12: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

12.6.2 Problem Structuring: System Definition and the Alternatives

The primary system to be modelled is an old station with reconditioned boiler unitsburning discard coal. Also requiring definition are the alternatives for comparison.To determine whether the FBC system is ‘‘environmentally beneficial’’ requires compar-ison against the power generating options it is displacing.Different scenarios are possible,as the comparative basis changes depending on the driver for the project. Most likelyis that the station is being re-commissioned because the additional capacity is requiredfor the grid, that is if the station is not re-powered usingFBC technology, the capacitywillbe supplied from some other source. Other possible sources include re-commissioningthe station as a conventional PF station, building a new PF station or operating theexisting stations at higher loads. The first alternative is complicated by the fact that theextent to which the station is refurbished will significantly influence its environmentalperformance, and the comparison will have to take this into account. Alternatively, thedriver behind the project may not be the requirement of new capacity, for example, theprimary aim of the project could be to remove the discard dumps, or a political drive todemonstrate a ‘‘clean coal’’ technology. In this case, re-powering the station woulddisplace existing power off the grid, and a relevant comparison would be between there-powered FBC station and the average grid mix. In this case study, we consider onlythe former possibility. The systems to be modelled are thus:

– an old plant re-commissioned with FBC boiler units; and– the plant re-commissioned with the original PF boiler units.

Although the case study is conducted without a specific power station in mind, somesite specificity is introduced, as the locations of the power stations currently instorage are known. The older stations that are no longer operating, but have thepotential for being reconditioned and brought back into service, all fall into a fairlylocalised region of the country, which mirrors the localised coal-producing region.Site-specific considerations therefore play a more significant role than usually foundin tactical studies in the selection of the relevant impacts to be considered, as well asin the importance attached to these impacts.

12.6.3 Problem Analysis

The LCA Approach

Suitable models must be constructed to determine the environmental profile of thealternatives. We have elected to use LCA as the support tool for generating thisenvironmental information (see the Appendix at the end of the book). Since the studyinvolves the comparison of alternative systems, it is also necessary to define criteriaagainst which they can be rated (e.g. impact categories or selected environmentalinterventions). This section presents the major considerations in the definition of thesestructures. The system models consist of two main components: a life cycle inventory(a complete list of all material and energy entering or leaving the system under study)and the extrapolation of the inventory data to potential environmental impacts.Coal mining and all waste disposal steps are explicitly considered. The life cyclecovered by the model is that of coal; its extraction, processing, combustion and

378 Sustainable Development in Practice

Page 13: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

disposal of its residues. Transmission, distribution and use of the power are notcovered. This model thus provides an inventory of undelivered electricity. In addition,only process-related emissions are assessed. Burdens associated with the running andmaintenance of offices, workshops and so on at the power station are not incorporatedin the assessment. The life cycle model is for an operating power station and mine.Building and commissioning the plants are not included, neither are the burdensassociated with the materials for construction. Maintenance materials are also notincluded, and only consumable materials are included in the inventory.

The flowsheet for the coal-electricity life cycle is given in Figure 12.1. This showsthe flow of coal through its life cycle and all the associated processes. The groupingsof colours indicate the top level of breakdown in the model and coincide with themajor technology combinations available at the mines and power stations. Theflowsheet is broken down into these sub-steps primarily for clarity and ease ofassessment, and the sub-processes do not necessarily stand-alone. The breakdownallows a single flowsheet to encompass all the technology combinations possible, andthe sub-processes are mixed and matched to represent the various technology com-binations of the plants. These combinations are:

– underground or opencast mining;– stockpile at the mine or at the power station;– boiler and particulate removal equipment type;– wet or dry cooling; and– wet or dry ashing.

The case study only considers discard sourced directly from the coal washing plant,and not that reclaimed from a dump. The study is limited to a consideration ofdiscard produced within a fairly close radius of the power station (approximately12 km). The coal supply is assumed to be from dual-product mines, since the studyconsiders a reconditioned station (i.e. its dedicated colliery is assumed to havestopped operating and it is supplied by an existing nearby mine adjusting its coalproduct to also produce a power station feed, or the dedicated mine subsequentlystarted producing a high-quality coal product to sustain itself when the power stationceased operating). This introduces the problem of allocating the mining burdensbetween the power station and high-quality coal product.

Factor-based models are used to generate inventory data for the foregroundsystem (the processes shown in Figure 12.1). The simple input/output models arebased on mass balance principles. Factors derived predominantly from process dataessentially act as splitter functions that steer an input to its respective output streams(e.g. the percentage of sulphur in ash determines the partitioning of sulphur in coalbetween stack emissions and retention in ash). An important feature of the processmodels is therefore that they are not based on any fundamental chemical or physicalmodelling, but on characterisation factors derived from process data. The models arethus very data intensive, and the accuracy of the calculated inventory data is verymuch dependent on the applicability of the data used. Where process data were notavailable, recourse has been made to data from process simulation models, literature,experimental data, or in the worst case, rough assumptions made.

The Eco-indicator 99 (EI99) method (Goedkoop and Spriensma, 1999) was chosento illustrate the transformation from inventory data to impact indicators, primarily

The Case of Power Generation in South Africa 379

Page 14: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

Eva

pora

tion

Eva

pora

tion

and

drift

loss

es

B

low

dow

ns, c

larif

ier

slud

ge a

nd T

RO

brin

e

Pot

able

wat

er

Raw

wat

er

Dus

t, M

etha

neF

uel c

ombu

stio

n pr

oduc

ts

Cla

rifie

r sl

udge

Sta

ck g

ases

Coa

l res

erve

sR

OM

coal

Ele

ctric

ity/F

uel

Bot

tom

ash

Fly

ash

Leac

hate

Dis

card

Leac

hate

Dus

tE

fflue

nt fo

r as

hco

nditi

onin

g an

ddu

st s

uppr

essi

on

Was

tew

ater

Roc

k/ov

erbu

rden

Leac

hate

Com

bust

ion

prod

ucts

and

dus

t

Leac

hate

Ret

urn

wat

er

CO

AL

EX

TR

AC

TIO

NC

OA

LP

RE

PA

RA

TIO

NC

OA

LS

TO

CK

PIL

E

DIS

CA

RD

and

OV

ER

BU

RD

EN

MA

NA

GE

ME

NT

BO

ILE

RP

RE

CIP

S o

rB

AG

HO

US

E

AS

HD

ISP

OS

AL

DE

MIN

WA

TE

RP

RO

DU

CT

ION

CO

OLI

NG

WA

TE

RS

YS

TE

ME

FF

LUE

NT

SY

ST

EM

CLA

RIF

IER

San

dF

ILT

ER

S

Ele

ctric

ity/F

uel

ST

OR

M W

AT

ER

/E

FF

LUE

NT

MA

NA

GE

ME

NT

Raw

wat

er

Sur

face

run

off

Effl

uent

ST

OR

MW

AT

ER

MA

NA

GE

ME

NT

Sur

face

run

off

See

page

See

page

Effl

uent

dis

char

ge

Che

mic

als

Tre

atm

ent c

hem

ical

s H2S

O4

NaO

HN

aCl

Boi

ler

blow

dow

ns

ST

OR

M W

AT

ER

/E

FF

LUE

NT

MA

NA

GE

ME

NT

See

page

Sur

face

run

off

BO

ILE

R D

RU

M,

TU

RB

INE

S a

ndC

ON

DE

NS

ER

S

Ste

am le

aks

and

soot

blow

ing

RA

WW

AT

ER

SY

ST

EM

Eva

pora

tion

Rai

n w

ater

Eva

pora

tion

Rai

nw

ater

Rai

nw

ater

Eva

pora

tion

Rai

n w

ater

Low

con

duct

ivity

effl

uent

s

Exc

ess

efflu

ent

For

egro

und

Bou

ndar

y

Bac

kgro

und

Bou

ndar

y

Prim

ary

reso

urce

sE

nerg

y

Ele

ctric

ity

Ash

use

din

cem

ent

Em

issi

ons

to a

ir

Em

issi

ons

to w

ater

Affe

cted

land

are

a

Die

sel

Hea

vy fu

el o

il

Gas

Sul

phur

ic a

cid

Filt

er b

ags

Lim

eC

aust

ic s

oda

Ret

urn

wat

er

Ion

exch

ange

and

CP

Pre

gene

ratio

n ef

fluen

ts

Figu

re12.1

Life

cycleofelec

tricityfrom

coal

Page 15: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

because data on the uncertainty of the equivalency factors are available for most ofthe impact categories. This method allows for a number of valuation scenarios to beexplored based on Cultural Theory arguments. In this case study, we consider onlythe ‘‘Hierarchist’s’’ perspective.3 Two additional criteria reflecting burdens consid-ered important in the context of the study, but not captured by the EI99 damagecategories, are included. Water use is included as a criterion because of its knownimportance in the regional context of the study. This includes raw water purchasedby the mine and power station, as well as the water ‘‘consumed’’ by the process byvirtue of its impact on the catchment area. Also of importance in the regional contextof the study is the effect of the systems on regional water quality. A key aspect of thestudy is the trade-off between the emissions from the discard dump, and those fromthe ash/gypsum dump. The level of detail at which the mine and power stationinventories are constructed is not sufficient to capture the water quality impacts asreflected in the EI99 toxicity categories, so the impacted land footprint indicator(Hansen, 2003) is used to give an indication of the potential of solid waste deposits tocontaminate water bodies through leachate generation. It incorporates the landoccupied by the dump and its leachate potential into a single indicator.

In addition to the removal of the discard dump, a key aspect of the FBC system isthe energy savings resulting from producing useful energy from an otherwise wastedenergy source. The functional unit needs to reflect this, as well as compare the systemson an equivalent basis. A dual time and product basis is therefore taken, in which theburdens calculated for an average year’s operation are normalised to the total powerproduced in the year. The normalisation is necessary because without it a moreefficient systemmerely reflects increased burdens and not an increased energy product.

Questions

1. Use the information presented in the previous section and in Figure 12.1 andconsult the Appendix at the end of the book to answer the following questions:

(a) What is the goal and scope of the LCA carried out in this case study?(b) What is the functional unit?(c) Sketch the system boundary for the system under study. Which activities in

the life cycle of coal electricity are included in and which are excluded fromthe consideration in this case study? Do you think the assumptions made arejustified given the goal of the study?

(d) Find out how the environmental burdens are translated into environmentalimpacts within the EI99 approach? How is that different from the problem-oriented approach described in the Appendix on LCA?

(e) What impacts do you think are likely to be important in the life cycle of coalelectricity?

3 The EI99 approach to Impact Assessment in LCA considers three general types of impact: damage to

human health; damage to ecosystem quality; and resource use. There are three ways to aggregate these

impact categories by using ‘‘weights’’ of importance. The ‘‘Hierarchist’s’’ approach is one of the three in

which the weights are assigned in the following way: human health, 40%; ecosystem quality, 30%; and

resource use, 30%.

The Case of Power Generation in South Africa 381

Page 16: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

2. What are the advantages and limitations of using LCA as the basis for environ-mental assessment in this case study?

3. Is the problem definition, as defined, capable of answering the question of whetherfluidised bed technology with discard coal is genuinely a ‘‘clean coal technology’’?

Reducing the Number of Possible Alternatives

The various combinations of technologies and operating conditions found withinthis case study, compounded by uncertainties in empirical parameters, model para-meters and model choices, together suggest an infinite number of alternatives forconsideration. The effort required to fully characterise and explore such a decisionspace to identify preferred alternatives using MCDA tools or similar is a dauntingtask. In this case study, we look to reduce the number of alternatives to a manage-able number, and explore their performance using uncertainty propagation tools(including scenario analysis).

The certainty required in the results depends on how conservative or ‘‘risk averse’’the decision makers are. However, the level of confidence acceptable to the decisionmakers is also influenced by the variance exhibited by the system (that due to bothempirical and model parameters), that is ‘‘high’’ confidence limits required may beunachievable in the context of the study. The uncertainty is primarily a function ofthe complexity, depth and data accuracy of the models used, which, in turn, arefunctions of the goals and scope of the study. The level of uncertainty with which thedecision makers are comfortable is therefore inextricably linked to the decisionstaken during goal and scope definition. The level of variance able to be toleratedin the results is also a function of the extent of the differences between the options.

Decision variables comprise the majority of model parameters investigated in thisstudy. The choice of these variables is under the direct control of the decision makers,and is predominantly related to the degree of refurbishment of the plant (i.e. thespecification of the boiler and water plant) and to the choice of mine supplying the coalor discard. A model parameter sensitivity analysis enabled the important parameters tobe grouped into ‘‘best’’, ‘‘worst’’ and ‘‘most likely’’ scenarios. This information was usedto define scenario states for each of the two main technologies (i.e. PF or FBC). Thisresults in a discrete set of alternative ‘‘states’’ for consideration as part of the decisionanalysis, according to Table 12.1. The uncertainty in empirical parameters for eachof these alternative ‘‘states’’ can be addressed through probabilistic techniques.A framework for such an integrated approach has been developed by Notten (2002).

Question

1. Why is it important to address uncertainty in decision making? Where do theuncertainties arise? What can be done to reduce the level of uncertainty?

Decision Criteria

Table 12.2 lists the full set of environmental indicators considered in this casestudy. The indicators have been calculated based on EI99 mid-point choices.

382 Sustainable Development in Practice

Page 17: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

The use of the EI99 impact assessment method is attractive as it allowsconsideration of different cultural perspectives as part of an exercise to unpackthe significance of valuation choices in decision outcomes. However, this aspectof valuation is not considered here, and only the EI99 default ‘‘Hierarchist’s’’perspective is used. The uncertainty in the characterisation factors for impact

Table 12.1 Decision variable values for three extreme scenario ‘‘states’’

Worst(Littlerefurbishment)

Most likely Best(Significantinvestment)

Maximum unit capacity (MW) 125 200 200Number of generator sets 4 2 2Load factor (%) 64 80 80Sorbent type dolomite limestone limestoneSO2 removed (%) 30 60 90Particulate control ESP ESP FFMine type opencast underground undergroundMine power source adjacent station grid gridCoal bypassing washing plant (%) 15 20 25Station, stockpile and mine life (years) 35 20 15Stormwater and effluent management poor average goodDistance coal transported (km) 25 10 3Method of coal transport rail conveyor conveyorTransport distances (km) 850 500 150Transport mode road road railAshing method wet wet dryWater plant configuration un-optimised optimised optimised, and

adjusted for dryashing

Stockpile size (reserve time) 3 years 3 months 3 weeks

Table 12.2 Decision criteria

Criterion Units

Generating cost aR(1996)/GWhCarcinogenic effects on humans bDALYs/GWhRespiratory effects on humans caused by organic substances(summer smog)

DALYs/GWh

Respiratory effects on humans caused by inorganic substances(winter smog)

DALYs/GWh

Climate change DALYs/GWhEcotoxic emissions PDF�m2�yr/GWhCombined effect of acidification and eutrophication cPDF�m2�yr/GWhExtraction of fossil fuels MJ surplus energy/GWhWater use Ml/GWhAffected land footprint km2/GWhNumber of jobs created total number

a R: South African rands.bDALY: Disability Adjusted Life Years. This indicator measures ill health, disability and premature deathattributable to environmental pollution.c PDF: Potentially Disappeared Fraction of vascular plant species.

The Case of Power Generation in South Africa 383

Page 18: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

categories is explored using probability distributions rather than single pointestimates.

In addition to the environmental decision criteria, the economic criterion ‘‘gen-erating cost’’ (cost of electricity generation) and the social criterion ‘‘number of jobscreated’’ have also been included as decision criteria.

Questions

1. Analyse the decision criteria in Table 12.2 and explain the meaning of each. Youmay wish to consult Goedkoop and Spriensma (1999) for more detail on the Eco-indicator approach.

2. What other criteria would you consider and why?

Normalisation

When considering familiar decision criteria such as ‘‘generating cost’’ and ‘‘numberof jobs created’’, those involved in the decision-making process are able to expresstheir preferences for different levels of performance directly, hence for these, theabsolute numbers are of interest. However, when considering performance in theenvironmental criteria, it is necessary to provide a reference base in order for thoseinvolved in the decision-making process to express their preferences. The perform-ances in the environmental criteria have thus been normalised relative to theaverage performance considering all the power stations owned by the company.

Performance Ranges

Table 12.3 shows the attribute ranges for each criterion considering the uncertaintyin the performance information (i.e. lowest and highest likely performance scores ineach criterion).

Table 12.3 Performance ranges for six alternatives under consideration

Criterion Units Minimumvalue

Maximumvalue

Generating cost R(1996)/MWh 30a 100a

Carcinogenic effects (humans) relative to company average 0.0001 13Summer smog relative to company average 0.05 177Winter smog relative to company average 0.04 20Climate change relative to company average 0.32 4.1Ecotoxic emissions relative to company average 0.0008 6.3Combined effect of acidification& eutrophication

relative to company average 0.05 4.4

Extraction of fossil fuels relative to company average 0.03 3Water use relative to company average 1 4.5Affected land footprint relative to company average �6 11Number of jobs created total number 55 165

aData scaled to protect confidential information.

384 Sustainable Development in Practice

Page 19: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

Questions

1. The performance of options in the environmental criteria are expressed ‘‘relativeto company average’’. This reduces numerical sensitivity. Explain the challengesin normalisation to ensure that stakeholder preference information can be mean-ingfully derived from performance scores.

2. Examine the minimum and maximum values for the criteria in Table 12.3. What doyou conclude: which criteria are associated with the largest level of uncertainty?

Value Functions

Two sets of value functions were considered:

1. a set in which all the value functions were assumed to be linear (set A); and2. a set in which sigmoidal value functions were selected for those criteria (other

than generating cost) in which the performance scores across the set of alterna-tives could be differentiated (set B).

The second set of value functions would thus ensure that greater value scores wereassigned to those alternatives that did well in those criteria in which the alternativescould be differentiated. The linear value function for ‘‘generating cost’’ was retainedsince this was expected to be more consistent with the attitude to improvements ingenerating cost. A summary of the value function shapes is provided in Table 12.4.

Questions

1. Why were two different sets of value functions considered?2. Examine the value functions in Table 12.4. What is the significance of the linear

and sigmoidal shapes of the value functions respectively ?3. Why is shape of the value functions for sets A and B similar for most of the

decision criteria?

Table 12.4 Summary of value function shapes

Criterion Value functionset A

Value functionset B

Generating cost Linear LinearCarcinogenic effects on humans Linear LinearSummer smog Linear LinearWinter smog Linear LinearClimate change Linear LinearEcotoxic emissions Linear LinearCombined effect of acidification & eutrophication Linear LinearExtraction of fossil fuels Linear SigmoidalWater use Linear LinearAffected land footprint Linear SigmoidalNumber of jobs created Linear Sigmoidal

The Case of Power Generation in South Africa 385

Page 20: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

Weight Elicitation

The attribute ranges for the environmental criteria spanned orders of magnitude dueto the uncertainty in the environmental performance information (Table 12.3).Experience in similar situations suggested that those involved in the decision-makingprocess are not able to express preferences with confidence when faced with trade-offs which span such large ranges. Under these circumstances, it was regarded asmore meaningful to do a sensitivity analysis for the weights rather than attempt thedirect elicitation of weights from stakeholders. Two approaches were used to con-sider the sensitivity of the rank of the alternatives to the relative weighting of thecriteria:

1. comparison of two nominal weight sets; and2. varying the weights of each criterion in turn, keeping the ratio of the other weights

constant.

The nominal weighting sets consisted of a set in which the criteria were weightedequally (i.e. trade-off across the entire attribute range of all the criteria was regardedas equally acceptable) and one in which emphasis was placed on those criteria inwhich the set of alternatives could be differentiated easily, despite inherent uncer-tainty in the information set. A summary of the weights is provided in Table 12.5.

Questions

1. How can preferences be elicited from decision makers? Explain how differentdecision contexts affect the choice of the preference elicitation method.

2. Why do decision makers (and people in general) have difficulties in expressingtheir preferences when the values of decision criteria span large ranges (e.g.several orders of magnitude)?

3. What is the implication, if any, of having multiple environmental criteria, but onlyone economic and one social criterion? Does this suggest any form of bias towardsenvironmental concerns? Explain this in terms of the objectives hierarchy.

Table 12.5 Summary of weighting sets

Criterion Equal weights Alternative weighting set

Generating cost 0.09 0.215Carcinogenic effects on humans 0.09 0.020Summer smog 0.09 0.020Winter smog 0.09 0.020Climate change 0.09 0.020Ecotoxic emissions 0.09 0.020Combined effect of acidification & eutrophication 0.09 0.020Extraction of fossil fuels 0.09 0.215Water use 0.09 0.020Affected land footprint 0.09 0.215Number of jobs created 0.09 0.215

386 Sustainable Development in Practice

Page 21: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

4. Instead of a parametric sensitivity study of ‘‘weighting’’, direct elicitation ofweights could be attempted by all stakeholders. It has been suggested here thatthis is particularly difficult given the range of performance scores in eachcriterion. However, it appears that the reported ranges are indicative more ofinherent uncertainty than they are of real performance differences for theoptions. Explain how one could target critical uncertainties to explore theirsystematic reduction.

Case Study Results

The strongest differences between the PF and FBC systems are in their fossil fuel useand their land footprint. The considerable savings in fossil fuel resources by usingFBC reflect the use of a ‘‘waste’’ energy source to generate power. The discard isdefined as a waste from the mining system and, as such, is not allocated any miningburdens other than the ‘‘avoided’’ burdens resulting from the removal of the dumps(see Chapter 10 for the ‘‘avoided burdens’’ approach). It therefore does not reflectany fossil fuel resource consumption, as all fossil fuels consumed and extractedduring mining are allocated to the coal product. Also caused by the avoidance ofmining burdens are the lower contributions to climate change, water use and summersmog of the FBC system relative to the PF system. This is less marked for climatechange, because if compared on the basis of the power station alone (i.e. without theeffects of mining), the FBC system has a slightly higher climate change burden thanthe PF system (caused predominantly by the use and transport of limestone in theFBC system).

The lower contribution to the combined effect of acidification and eutrophicationby the FBC system stems from the far lower NOx emissions from the FBC boiler.The effect is much more marked for a consideration of NOx alone, as SO2 emissionsare not always lower in the FBC system (they depend on the degree of desulphurisa-tion). The SO2 emissions of the ‘‘worst’’ FBC scenario (with only 30% SO2 removal)are higher than those of the PF scenarios, whilst those of ‘‘best’’ FBC system aresignificantly lower.

Winter smog is shown to be ambivalent between the PF and FBC systems (i.e.neither system plots strongly with or against the winter smog vector). Although theFBC system always causes an increase in particulate emissions, the volume of SO2

emitted can ‘‘swing’’ winter smog to being significantly worse or slightly better thanthe PF system. The toxicity categories also do not show a strong tendency to bebetter or worse in either of the systems.

A full consideration of empirical parameter uncertainty is beyond the scope ofthis case study. Interested readers are referred to Notten (2002), where this isexplored in some depth. What is more relevant here is to explore the significanceof valuation choices in the decision outcomes. This approach is taken not only toemphasise the importance of valuation arguments in decision making of thisnature, but also to demonstrate that such perspectives can be accommodatedwithin a decision analysis framework as part of a rational approach to complexdecision situations. This is detailed below, including consideration of model para-meter uncertainty.

The Case of Power Generation in South Africa 387

Page 22: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

Question

1. What do you conclude: which system is preferable environmentally: PF or FBC?Why?

Overall Value Scores

The overall value scores for the two-criteria nominal weighting sets are presented inFigure 12.2. These figures show the cumulative probability distribution functions foreach of the alternative ‘‘states’’ as a function of the overall desired value. In otherwords, these figures provide evidence of the confidence the decision maker can placein the separation of alternatives, despite the inherent valuation uncertainty. Theoptions which are further to the right in the figures, that is those with the highestvalue scores, are preferred overall. Detailed consideration of this approach is givenby Basson (2003).

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1(a)

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Overall value

Cum

ulat

ive

prob

abili

ty

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

(b)1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Overall value

Cum

ulat

ive

prob

abili

ty

Figure 12.2 Continued

388 Sustainable Development in Practice

Page 23: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

As can be seen from the cumulative distributions for the overall value score pre-sented in Figure 12.2, the greatest degree of separation is achieved for the value functionand weight sets which place emphasis on those criteria in which the alternativesshow the greatest differences in performance scores (Figure 12.2[d]). However, in allcases, the FBC alternatives obtain higher overall value scores than the PF alternatives,with the best scores being obtained by the FBC system for the greatest extent ofrefurbishment. The evaluation suggests that, despite the extensive uncertainty present

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

(d)

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Overall value

Cum

ulat

ive

prob

abili

ty

PF Expected Level of Refurbishment PF Significant Refurbishment

PF Little Refurbishment

FBC Expected Level of Refurbishment FBC Significant Refurbishment

FBC Little Refurbishment

Figure 12.2 Cumulative probability distribution for overall value score: (a) value functionset A, equal weights; (b) value function set B, equal weights; (c) value function set A, alternativeweighting set; and (d) value function set B, alternative weighting set

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

(c)

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Overall value

Cum

ulat

ive

prob

abili

ty

The Case of Power Generation in South Africa 389

Page 24: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

in both the performance and the preference information, it can be concluded thatre-powering the power station with a fluidised bed boiler system would be the morepreferable approach.

The sensitivity analysis conducted by varying the weights of each criterion in turn,keeping the ratio of the other weights constant, showed that the FBC alternativeswould be preferred to the PF alternatives in all cases except where there was markedvariation in the generating cost weight. This was the case regardless of value functionset or weighting set. Figure 12.3 shows that when the weight placed on the generatingcost exceeds a value of about 0.56, the FBC options with the expected level ofrefurbishment, and that with significant refurbishment, would no longer be preferredover some of the PF options.

Questions

1. Examine the results shown in Figures 12.2 and 12.3 and identify the preferredoption for different value functions and weights. What do you conclude: how dothey affect the choice of the preferred technology in this case?

2. Do you think this type of analysis is useful for decision makers? Why? What canthey gain from it?

Results of Uncertainty Analysis

Rank correlation analyses were carried out to determine which of the uncertainperformance scores made the most significant contribution to the uncertainty in the

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Value of weight

Ove

rall

valu

e

PF Expected Level of RefurbishmentFBC Expected Level of Refurbishment

PF Significant RefurbishmentFBC Significant Refurbishment

FBC Little RefurbishmentPF Little Refurbishment

Figure 12.3 Effect on median value of overall value of varying weight on generating cost(value function set B, alternative weighting set)

390 Sustainable Development in Practice

Page 25: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

overall value scores of the alternatives. The results for the different value functionsand weighting sets are presented in Figure 12.4. In these graphs the larger the‘‘Overall Value Importance’’, the more significant is the contribution of the uncer-tainty in the input parameter to the uncertainty in the overall value score.

The results of the uncertainty analysis presented above indicate that when greaterweights are placed on those criteria in which the performance scores of the alter-natives differ distinctly (Figure 12.4 [c] and [d]), the uncertainty in the overall valuescores is dominated by a few criteria (i.e. uncertainty relating to ‘‘fossil fuel con-sumption’’ in the case of the PF alternatives and uncertainty relating to the ‘‘numberof jobs created’’ in the case of the FBC alternatives). In contrast, when equal weightsare placed on the criteria (Figure 12.4 [a] and [b]), uncertainty in the performanceinformation in a larger number of criteria make similar contributions to the uncertaintyin the overall value scores. Furthermore, when relative preference for performance in

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

(a)

1

of RefurbishmentPF SignificantPF Expected LevelRefurbishment

PF LittleRefurbishment

FBC Expected Levelof Refurbishment

FBC SignificantRefurbishment

FBC LittleRefurbishment

Alternatives

Ove

rall

Val

ue Im

port

ance

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

PF Expected Levelof Refurbishment

PF SignificantRefurbishment

PF LittleRefurbishment

FBC Expected Levelof Refurbishment

FBC SignificantRefurbishment

FBC LittleRefurbishment

Alternatives

Ove

rall

Val

ue Im

port

ance

(b)

Figure 12.4 Continued

The Case of Power Generation in South Africa 391

Page 26: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

the individual criteria are modelled using sigmoidal value functions in those criteria(other than ‘‘generating cost’’) in which the performance scores of the alternativesdiffer distinctly, the uncertainty in ‘‘fossil fuel consumption’’ (PF alternatives) anduncertainty relating to the ‘‘number of jobs created’’ (FBC alternatives), which areboth allocated sigmoidal value functions, make a more significant contribution tothe uncertainty in the overall value scores. This effect is more distinct for thealternative weighting set. These results are perhaps not unexpected. Performance

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

PF Expected Levelof Refurbishment

PF SignificantRefurbishment

PF LittleRefurbishment

FBC Expected Levelof Refurbishment

FBC SignificantRefurbishment

FBC LittleRefurbishment

Ove

rall

Val

ue Im

port

ance

(c)

Alternatives

PF Expected Levelof Refurbishment

PF SignificantRefurbishment

PF LittleRefurbishment

FBC Expected Levelof Refurbishment

FBC SignificantRefurbishment

FBC LittleRefurbishment

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

1

Ove

rall

Val

ue Im

port

ance

Alternatives

Generating CostWinter SmogFossil Fuels

Number of Jobs

Climate ChangeWater Use

CarcinogensEcotoxicityAffected Land Footprint

Summer SmogAcidification and Eutrophication

(d)

Figure 12.4 Relative significance of uncertainties in performance scores for the uncertaintyin the overall value scores: (a) value function set A, equal weights; (b) value function set B,equal weights; (c) value function set A, alternative weighting set; and (d) value function set B,alternative weighting set

392 Sustainable Development in Practice

Page 27: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

scores in criteria which have large weights are expected to make a larger contributionto the overall value score. Furthermore, sigmoidal value functions allow alternativeswhich span the attribute range to have a much larger range of potential value scores.This is the case for ‘‘fossil fuel consumption’’ (PF alternatives) and ‘‘number of jobscreated’’ (FBC alternatives).

Question

1. Examine the results presented in Figure 12.4. What are the findings of theuncertainty analysis? Can the decision makers be confident in their technologychoice? What strategies do you suggest the company to adopt to improve itsconfidence in the decision outcome?

Observation of Uncertainty and Sensitivity Analyses

The results of the sensitivity analysis for weights indicate that the FBC alternativesare expected to be preferred in all circumstances considered, other than whena large weight (>0:56) is placed on the ‘‘generating cost’’. The uncertainty analysisindicates that if a distinction is to be made between the performances of theindividual PF and FBC alternatives, information leading to the estimates of ‘‘fossilfuel consumption’’ (PF alternatives) and ‘‘number of jobs created’’ (FBC alterna-tives) should be the first areas of focus of efforts to reduce the uncertainty in therelative ranking of the alternatives. What is perhaps striking about this conclusionis that it highlights the fact that the ability to make an informed decision about thistechnology choice requires more detailed and accurate information about twoaspects of the problem which, in a narrow economic assessment, run the risk ofbeing overlooked completely.

Questions

1. Summarise step by step the decision-making process carried out in this case study.Comment on the practicality of such a process: how long would it take, whowould be involved and how much would it cost?

2. Comment on the necessity of the formal approach to decision making: what arethe main advantages of the formal DSF as used in this case study? Comment inparticular on the stakeholder engagement process. What are the disadvantages ofsuch a formal approach? Do you think that the decision makers could havereached this decision without using a formal decision-making approach?

3. How important is it in your opinion to reduce uncertainty in decision making inthe context of sustainable development? How much effort should be put inreducing uncertainty and carrying out sensitivity analyses? Why?

4. Do you think that the decision makers who participated in the decision-makingprocess presented in this case study can be assured that they have identified a truly

The Case of Power Generation in South Africa 393

Page 28: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

more sustainable technological option given the alternatives that they considered?Why?

5. Given the iterative and discursive nature of decision making, how would yousuggest this analysis be communicated to the broad range of stakeholders toensure their continued participation in the decision process?

6. If this had been your decision, what would you do next, and why?

Conclusions

This case study has considered the various challenges faced by decision makerswithin the resources sector (in this case, coal-based power generation) in choosingtechnologies best suited to promoting sustainability. Based on the philosophy of lifecycle thinking and structured around the tools of decision analysis, the approachdeveloped here affords decision makers the opportunity to engage all stakeholders ina decision, thereby improving the transparency of the process and leading, hopefully,to better decisions with more robust outcomes.

The DSF places emphasis on information gathering and management of uncer-tainty. It should be emphasised that decision analysis allows for the incorporation ofinformation from a variety of sources (i.e. based on many of the decision-supporttools already in existence and used routinely). Additionally, it suggests that it is thedecision-making process (and its underlying constructs) which is most important –not the methods of data collection themselves. That said, it is important to under-stand the underlying uncertainties in the information used to characterise a decisionsituation, without which, our experience suggests, it is often impossible to arrive atmeaningful decision outcomes. In all decision situations, it is a case of ensuring thatthe requisite set of performance information is collected. This point is not to beunderstated. Too many researchers and practitioners in the field of environment andsustainability are bent on measurement and monitoring, when perhaps a focus ondecision objectives and desired outcomes might be more helpful in advancing thesustainability agenda.

The case study itself generates some interesting observations. Fluidised combus-tion of discard coals does offer advantages over refurbished PF technology whenlooking, from the context of sustainability, at re-powering mothballed stations. Thisis not an immediately obvious conclusion, given the complexity of the situation andthe inherent uncertainties in the empirical quantities and decision variables whichdefine the problem. Focusing on valuation-type uncertainties, the case study high-lighted the importance of job creation and resource depletion in determining apreferred technology alternative.

It is not a coincidence that these two criteria reflect some of the more pressingdevelopment issues within emerging economies – that of employment creation andresource management. Our opening statement that South Africa’s economy con-tinues to be driven, in large measure, by the availability of low-cost energy raises anadditional challenge – how do we engage with the interconnectedness of energy–minerals networks to deliver an analysis which is meaningful in a macro-sense? It isour belief that the approach reflected in this case study has more general applic-ability and can engage this larger perspective also.

394 Sustainable Development in Practice

Page 29: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

References and Further Reading

Basson, L. (2003) Context, Compensation and Uncertainty in Multi-Criteria Decision Mak-ing. PhD Dissertation, University of Sydney, Australia.

Baxter, B. (1993) Cleaning up the Act. SA Mining, Coal, Gold and Base Minerals, September1993, pp. 14–15.

Belton, V. and T.J. Stewart (2002) Multi Criteria Decision Analysis. Kluwer Academic Pub-lishers, Boston.

Beinat, E. (1997) Value Functions for Environmental Management. Dordrecht, Kluwer Aca-demic Publishers.

Bredell, J.H. (1987) South African Coal Resources Explained and Analysed; GeologicalSurvey of South Africa.

Department of Minerals and Energy (DME) (1998) White Paper on Energy Policy of theRepublic of South Africa. Government Printer.

Department of Minerals and Energy (DME) (2001) Report Number D2/2001: Operatingand Developing Coal Mines in the Republic of South Africa, 2001. Department of Miner-als and Energy, South Africa.

Department of Minerals and Energy (DME) (2003a) Coal Reserves; http://www.dme.gov.za/energy/coal_resources.htm (April 2003).

Department of Minerals and Energy (DME) (2003b) Quarterly Statistics: Mine Accidents 1January to 30 September 2002; http://www.dme.gov.za/mhs/mine_quarterly_stats.htm(April 2003).

DMEA (1985) South African Discard and Duff Coal – National Inventory 1985. Departmentof Mineral and Energy Affairs, South Africa.

Energy Information Administration (2002) Country Analysis Briefs: South Africa. September2002; http://www.eia.doe.gov/emeu/cabs/safrica.html (April 2003).

Energy Research Institute (2001) Preliminary Energy Outlook for South Africa. ERI,The University of Cape Town, South Africa.

Eskom (1999) Annual Report 1998. ESKOM Corporate Communication Department.Eskom (2001) Environmental Report 2000. ESKOMCorporate Communication Department;

http://www.eskom.co.za/enviroreport01/resource.htm (April 2003).Eskom (2002) Annual Report 2001. ESKOM Corporate Communication Department; http://

www.eskom.co.za/about/Annual%20Report%202002/index.html; accessed April 2003.Goedkoop, M. and R. Spriensma (1999) The Eco-Indicator 99: A Damage Oriented Method for

Life Cycle Impact Assessment. Pre Consultants, the Netherlands.Hansen, Y. (2003) Leachate Generation and Mobility in Coal-based Solid Wastes – Impact

Prediction and Environmental Assessment. PhDDissertation, University of Sydney, Australia.Keeney, R.L. (1992) Value-focused Thinking. Harvard University Press, Cambridge,

Massachusetts.Keeney, R.L. and Raiffa, H. (1976) Decisions with Multiple Objectives. Preference and Value

Tradeoffs. New York, Wiley.Lennon, S.J. (1997) Clean Coal Technology Choices Relating to the Future Supply and

Demand of Electricity in Southern Africa. J. Energy S. Afr. (May 1997), 45–51.MEI Online (2002) US grants $500 000 for South African Plant. 1 March 2002; http://

www.min-eng.com/enviro/37.html (April 2003).Notten, P. (2002) Life Cycle Inventory in Resource Based Industries – A Focus on Coal-based

Power Generation. PhD Dissertation, University of Cape Town, South Africa.Prevost, X.M. (2002) Coal. In: South Africa’s Mineral Industry 2000/01. Department of

Minerals and Energy, Mineral Economics Directorate and Mineral Bureau, South Africa.Rosenhead, J. (ed.) (1989) Rational Analysis for a Problematic World. John Wiley & Sons,

New York.Seppala, J., L. Basson and G. Norris (2002) Decision Analysis Frameworks for Life Cycle

Impact Assessment. J. Industrial Ecology, 5(4), 45–68.Surridge, A.D., C.J. Grobbelaar and J.K. Asamoah (1995) On South African Coal Reserves/

Resources and their Utilisation. Colloquium: Coal Processing, Utilisation and Control ofEmissions. South African Institute of Mining and Metallurgy, Randburg, South Africa.

The Case of Power Generation in South Africa 395

Page 30: Sustainable Development in Practice (Case Studies for Engineers and Scientists) || Multi-Criteria Decision Analysis: The Case of Power Generation in South Africa

Stewart, T.J. (1992) A Critical Survey of the Status of Multiple Criteria Decision MakingTheory and Practice. Omega, Int. J. Mgmt. Sci., 20(5–6), 569–586.

Van Horen, C. (1996) Counting the Social Costs. Electricity and Externalities in South Africa.UCT and Elan Press.

von Winterfeldt, D. and W. Edwards (1986) Decision Analysis and Behavioural Research.Cambridge Univ. Press, Cambridge, England.

Wells, J.D., L.H. Van Meurs and M.A. Rabie (1992) Terrestrial Minerals. In: EnvironmentalManagement in South Africa (eds R.F. Fuggle and M.A. Rabie), Juta.

396 Sustainable Development in Practice