Reliability Engineering and System Safety 96 (2011) 663–670

Contents lists available at ScienceDirect

Reliability Engineering and System Safety

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

Integrating risk analysis and multi-criteria decision support under

uncertainty in electricity distribution system asset management
M.D. Catrinu n, D.E. Nordgård
SINTEF Energy Research, N-7465 Trondheim, Norway

a r t i c l e i n f o abstract

Available online 12 January 2011 Asset managers in electricity distribution companies generally recognize the need and the challenge of
Keywords: adding structure and a higher degree of formal analysis into the increasingly complex asset manage-
Risk analysis ment decisions. This implies improving the present asset management practice by making the best use
Multi-criteria decision analysis under of the available data and expert knowledge and by adopting new methods for risk analysis and decision
uncertainty support and nevertheless better ways to document the decisions made.
Electricity distribution system asset This paper discusses methods for integrating risk analysis and multi-criteria decision support under
management uncertainty in electricity distribution system asset management. The focus is on how to include the
different company objectives and risk analyses into a structured decision framework when deciding
how to handle the physical assets of the electricity distribution network.
This paper presents an illustrative example of decision support for maintenance and reinvestment
strategies based, using expert knowledge, simplified risk analyses and multi-criteria decision analysis
under uncertainty.
& 2011 Elsevier Ltd. All rights reserved.

1. Introduction Asset management, promoted by several international man-

agement quality standards [1–3], is a relatively new concept for
Electricity distribution networks are the last in the chain of electricity distribution companies in Norway and elsewhere.
electricity supply, providing electricity from the main transmission The concept is gaining momentum as a way to cope with
system to the end-users. Typically these networks include a large the challenges distribution companies are facing in managing
number of assets: medium and low voltage installations like over- relatively aging infrastructures under increasing requirements
head lines, underground cables, substations, transformers and from stakeholders—owners, authorities, regulators, customers.
auxiliary equipment. However, the implementation of an asset management system
Traditionally, the main concern of distribution companies has imposes significant changes in how distribution companies are
been to offer access to the network to all potential end-users managed and, in particular, the way to address the risks their
and maintain an adequate reliability of power supply. Today networks and businesses are exposed to. Current asset manage-
additional challenges must be considered, which complicate the ment practices – and related research activities – focus primarily
system management process. on quantifying risk in monetary terms, on distribution system
In many countries, following the deregulation of the electricity reliability, reliability centred maintenance (or asset management)
sector, distribution companies became independent companies combined with estimations of components’ condition (lifetime
(state and privately owned). These companies maintain and estimations, etc.) [4–11]. The analyses of other risk aspects like
operate distribution infrastructures under specific concession for example the risk of personnel safety, the risk of environmental
rights and, as natural monopolies, are under state regulatory damage or the risk of a negative public acceptance are usually
control. Although the regulatory practice is different in each ‘decoupled’ from the more ‘quantifiable’ risk analyses. Often these
country, the goal is generally the same: to keep distribution aspects are taken care of by following for example safety regula-
companies costs and profits under control, while imposing high tions or investing in publicity campaigns focusing on the good
service quality (reliability and security of supply) and efficiency sides of the electricity distribution business (especially in coun-
requirements. tries where the distribution business is done by multi-utilities).
Generally, asset managers (AMs) in electricity distribution
companies recognize the need and the challenge of adding structure
n
Corresponding author. and a higher degree of formal analysis into the increasingly complex
E-mail address: [email protected] (M.D. Catrinu). asset management decisions [12]. This implies improving the

0951-8320/$ - see front matter & 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ress.2010.12.028
664 M.D. Catrinu, D.E. Nordgård / Reliability Engineering and System Safety 96 (2011) 663–670

present asset management practice by making the best use of classifications exist for uncertainty in decision making. For
available data and expert knowledge about the system and the example, in [14] two uncertainty aspects are discussed: ‘external’
assets. This in turn requires adopting new methods for risk analysis uncertainty and ‘internal’ uncertainty.
and decision support and nevertheless better ways to document the The ‘external’ uncertainty refers to the lack of knowledge about
decisions made. the consequences of a particular choice (decision). In this paper, we
Based on the work presented at the ESREL conference in 2009 consider that external uncertainty resides in the estimation of the
[13], this paper focuses on the issue of offering decision support in problem ‘data’, for example: probabilities and consequences. In this
distribution system asset management by discussing methods category we would like to include two sub-types:
and illustrating an approach for integrating risk analysis and
multi-criteria decision support under uncertainty. (a) Uncertainty that arises because of natural, unpredictable
In this paper, the use of such methods is motivated by a dis- variations associated with the system or the environment—
cussion on the criteria, risks and uncertainty aspects in distribution aleatory uncertainty. This type of uncertainty is outside
system asset management decisions. Section 3 briefly presents the control of the decision maker, e.g. the 100 years big
some of the available theoretical approaches for risk and multi- storm, variations in the material fatigue in specific system
criteria decision support, the focus being on how different methods components, etc.
allow the representation of uncertainty in modelling the decision (b) Uncertainties that stems from lack of knowledge about
problem and decision maker’s preferences. This paper brings different phenomena—epistemic uncertainty. This uncer-
forward a little explored application domain for these methods. tainty resides from the lack of data to characterize the system
In Section 4 a case study is presented to illustrate the use of an or component failure, the lack of understanding and proper
integrated approach for risk and multiple criteria assessment in modelling of asset deterioration processes, the poor under-
designing maintenance and reinvestment strategies for 12 kV MV standing of failure interdependencies in the system (physical
air insulated switch-disconnectors. The case illustrates common or other phenomena) or the poor understanding of initiating
decision situations encountered by asset managers in distribution events.
companies. Their challenge is to make decisions based on rather
limited information and modelling possibilities and mostly rely-
The ‘internal’ uncertainty can be described as ambiguity or
ing on expert opinion and limited statistical and economic data.
imprecision in decision making and most of it is due to the
The focus is on emphasising various aspects relevant for practical
uncertainty in problem ‘data’. It reflects the imprecision in human
decision support using different methods, rather than discussing
judgements: preferences, values and risk attitudes. This uncer-
theoretical background conditions for value (choice) modelling.
tainty can stem from insufficient problem understanding, insuffi-
Section 5 offers a discussion of the limitations, advantages and
cient data, insufficient modelling, little acceptance of modelling
challenges of the approach proposed, and some conclusions.
assumptions, etc.
Under many circumstances it is difficult (if not impossible) to
2. Decision criteria and uncertainty in distribution system draw a boundary between external and internal uncertainties, but
asset management this differentiation is necessary because each uncertainty aspect
has different implications for the decision support process, and
2.1. Decision criteria in distribution system asset management the design of decision support tools as it will be discussed further.

Distribution system asset management (DSAM) is a complex

process comprising the lifecycle management of a large number
3. Theoretical approaches for multi-criteria decision support
of geographically distributed assets. DSAM activities include for
under uncertainty
example: asset operation and maintenance, capital planning and
project investment analysis [1].
Most decisions involve several criteria based on which alter-
DSAM decisions are, in general, of a multi-criteria nature. In most
native actions are evaluated. The most common representation of
decision situations, asset managers must be able to balance several
a multi-criteria decision problem is in a matrix form as shown in
criteria, for example, economy (costs and profits) system reliability,
Fig. 1, where a set of alternatives (A1,y,Am) is mapped against a
quality of supply and personnel safety. These criteria are in line with
set of criteria (C1,y,Cn).
the overall objectives of a distribution company that are established
Making a decision in this setting means choosing an alter-
by its owners (and under the influence of company’s stakeholders
native based on an evaluation of consequences or outcomes aik
(the state/regulator, system’s users, and society)).
(i¼ 1,y,m; k¼1,y,n).
However, asset managers seldom make decisions taking into
account, in a structured way, all relevant criteria in the same time.
Economic analyses of course prevail in all distribution system
asset management decisions. When other criteria – safety for Criteria
instance – are also relevant in specific asset management deci- C1 C2 Cn
sions, usually the analyses tend to be ‘decoupled’. In general
adequate safety measures are adopted and their respective costs A1 a11 a12 a1n
are accepted, without an in-depth integrated safety-economic A2 a21 a22 a2n
Alternatives

analysis of alternative decisions. Moreover, the documentation

about how such decisions are made tends to be spread over
various internal, custom-made company reports.

2.2. Aspects of uncertainty in distribution system asset management Am am1 am2 amn

As in any other field, many decision elements in distribution

system asset management are inevitable uncertain. Several Fig. 1. The decision matrix.
 M.D. Catrinu, D.E. Nordgård / Reliability Engineering and System Safety 96 (2011) 663–670 665

Essential in multi-criteria decision analysis (MCDA) is the assump- For example, the evidential reasoning (ER) approach [17]
tion that the decision maker (DM) has a set of values or preferences relaxes the demands of probability theory by invoking concepts
for the decision at hand, and that these values can be modelled. from Dempster–Shafer theory of evidence. The approach provides
When there is no uncertainty about the consequences, aik are a modelling framework for both complete and incomplete assess-
deterministic and there is a direct correspondence between ments. Rule and utility based techniques are used to transform
alternatives and consequences in terms of the criteria. various types of information into a unified format, so that both
One of the most used theories for decision support in this quantitative and qualitative information can be handled in a
setting is the multi-attribute value function theory (MAVT) consistent manner.
described for example in [15]. MAVT provides the background The purpose for this short overview of methods was to
for modelling decision maker’s preferences through a value emphasise that dealing with uncertainty in multi-criteria analysis
function V(Ai) that can be constructed based on a comparison of in practise requires to
consequences in each criterion (scores/partial value functions)
and a comparison of criteria (weights). In its simplest form, this  understand, represent, and model uncertainty in problem
value function is additive and can be written as in the following: data, and
 model preferences and risk attitudes.
X
n
VðAi Þ ¼ wk vk ðaik Þ
k¼1 The main advantage of such methods is that all relevant
elements of a decision (criteria, uncertainties) problem can be
where vk(aik) are the scores and wk are the weights. The value systematically taken into account in a transparent and justifiable
function is then used to calculate overall values for each alter- decision process.
native and the alternative having the ‘highest’ value is recom- Multi-criteria decision support generally relies either on effec-
mended to the decision maker. tive facilitation by a decision analyst or on the ability and
Under uncertainty, there may exist many possible values for willingness of the decision maker to make use of such an
the outcomes aik at the time of decision (external uncertainty) approach—often without being experts in the field of multi-
and often the values (scores and weights) can be difficult to criteria analysis. In both cases the main challenge is to be able
express (internal uncertainty). Under uncertainty aik can be choose a method that can provide relevant decision support in
described quantitatively (through probabilistic, fuzzy quantities), each decision situation, i.e. a method that is suitable for the
or qualitatively (through verbal descriptions—when outcomes are information, modelling tools and expert knowledge available.
not fully known or understood). Very often scenarios are con-
structed in order to simulate the consequences (quantitative or
qualitative) the decisions alternatives might have in terms of the 4. Case study
different criteria. In the construction of scenarios, a number of
methods can be used to model chains of events and phenomena This section presents an example on how the evaluation of
that lead to different outcomes with associated probabilities. maintenance and reinvestments strategies for 12 kV MV air-
Widely used are the probabilistic risk assessment (PRA) methods insulated switch-disconnectors can be carried out, using risk
such as fault and events trees, Bayesian networks, Monte Carlo analysis and multi-criteria decision support techniques.
simulations, etc. The case is built upon research previously reported in [18,13]
There are two main approaches to resolve uncertainty in and focuses on the risk of personnel injury during the operation of
multi-criteria decision support [14]. One approach is to resolve the switch-disconnector.
first the uncertainty in outcomes by somehow reducing the set of The case illustrates real decision situations in which both
possible aik to single values and then solve the MCDA problem in a internal and external uncertainty exist. The main input to the
‘deterministic’ setting. This can be done by using a decision decision process described consists of expert knowledge and
paradigm such as: expected values, utilities, MaxMin, MinMax, limited statistical and economic data. Risk and uncertainty in
MinRegret, etc., or to define risk as a separate decision criterion. both decision problem formulation and in decision maker’s
The other approach is to define scenarios with associated prob- judgements are addressed. The example focuses on how asset
abilities of occurrence and evaluate alternatives in each managers’ judgements with respect to the decision criteria can be
scenario—however, the theoretical background for integrating modelled and how these judgements contribute to the final
MCDA and scenario planning is not yet fully developed [14]. decision. The tools for decision support used are simplified risk
The most used method for modelling preferences under uncer- assessment supported by risk matrices [19] and a multi-criteria
tainty is the multi-attribute utility theory (MAUT). In its simples decision support software being able to deal with both quantita-
(additive) form, a multi-attribute utility function resembles a tive and qualitative description of criteria (consequences).
multi-attribute value function. The way to find parameters of a
utility function is however different. While in the case of MAVT 4.1. Description of the case
the scores and weights can be determined based on direct
comparison of consequences, in the case of MAUT these compo- Distribution networks contain large numbers of 12 kV MV air-
nents are found through lottery types of questions [16]. insulated switch-disconnectors. These assets are located in med-
The practical application of MAUT is often complicated because ium/low voltage (MV/LV) substations and their function is to
the axioms underlying the existence of utility functions and the break the load current when sectioning the MV grid. In the
methods for their construction can be easily violated during the transient period after the opening of the switch – when there is
real decision support process [14]. Moreover, MAUT measures no longer physical contact between the switches’ poles – the
‘complete’ preferences under uncertainty, when outcomes are current will continue to flow through an electric arc until
precisely described (usually quantitatively). However, preferences the natural zero-crossing of the alternating current. Normally the
may not always be completely specified (internal uncertainty) and electric arc will then extinguish in a controlled manner, and the
therefore alternative methods have been developed to deal with breaking of the current is successful. In some cases, with slow
preference value intervals, qualitative estimations and incomplete- movement of the switch during operation, the arc will re-ignite
ness in judgements. and the current will continue to flow through, generating energy
666 M.D. Catrinu, D.E. Nordgård / Reliability Engineering and System Safety 96 (2011) 663–670

Table 1
Safety risk assessment of air insulated switch-disconnectors regarding the undesired event slow operation of switch-disconnector.

dispersion through heat (with accompanying pressure rise). In such (N; E; W) is a new switch in an exposed environment, with wire
cases the switch-disconnectors may pose a risk for personnel injury. fence encapsulation.
Further in this paper we will refer to this risk aspect as ‘safety risk’. Each asset type was then characterized in terms of safety risk,
The risk of a slow operation of the switch-disconnector is as illustrated in the risk matrix in Table 1. The risk matrix is
relevant when designing maintenance and reinvestments strate- divided into three zones, marked with different shadings: the
gies for these assets. The analysis in the next section focuses on ‘‘Acceptable’’ zone (lower left), the ‘‘On-the-limit’’ zone (middle)
revealing whether all switches can be treated in the same way or and the ‘‘Unacceptable’’ zone (upper right).
if some kind of differentiation in the maintenance and reinvest- The experts asked to characterize the assets in terms of risk
ment activities is needed for specific types of switches. evaluated that usually the fully encapsulated assets should not
pose high safety risk. Then, among the semi-encapsulated assets,
the ones placed in exposed environment ((N;E;S), (O;E;S)) pose
4.2. Safety risk analysis medium safety risk. Generally all wire fence switches were
characterized to pose a middle to high safety risk.
Distribution companies’ experts estimated that a slow opera- In this example, risk assessment is based on a qualitative risk
tion of the switch and possible accidents in operation may happen evaluation based on expert knowledge. This exercise is easy to
while working with specific assets [18].They found that the most perform in practise since risk matrices are common tools for risk
relevant factors for differentiating these assets are: analysis in distribution companies. However, in the construction
of risk matrices additional quantitative analyses can also be used,
1. The age of the switch-disconnector: in addition to expert opinion. For example, detailed quantitative
 New assets, age r25 years, further denoted as (N). risk analysis, modelling for example the potential loss of life (PLL),
 Old assets, age4 25 years, further denoted as (O). as done in [18,13] can be used in the definition of consequences
2. The operating environment (and probability) scales.
 Assets placed in an exposed environment, further denoted
as (E). 4.3. Risk based asset categorization
 Assets placed in a clean environment, further denoted
as (C). The study focuses on a ‘test’ distribution network having in its
3. The encapsulation of the switch-disconnector1 : structure approximately 5000 switch-disconnectors:
 Fully encapsulated—steel plate covered cubicles, with
pressure relieving outlets in safe directions, further  2500 units of fully encapsulated switch-disconnectors.
denoted as (F).  1850 units of semi-encapsulated switch-disconnectors.
 Semi-encapsulated—steel plate cubicle fronts, top and  650 units of wire fence encapsulated switch-disconnectors.
bottom open, further denoted as (S).
 Wire fence encapsulated, further denoted as (W). These assets are of different age and operate in a clean or
exposed environment. The number of assets of different types is
The age and operating environment are factors influencing the presented in Table 2.
likelihood of occurrence of the undesired event, while the The assets evaluated to pose medium and unacceptable risk, as
encapsulation type influences mostly the consequence of the illustrated in the risk matrix in Table 1, are highlighted with the
undesired event for the operator. respective colors in Table 2.
Further, several asset types have been defined considering
a combination of the above risk factors, e.g. a switch of type 4.4. Strategies for maintenance and reinvestment

1
The reason for having different encapsulations is that the substations have
The analysis of the asset population as represented in Table 2
been built over quite a long period of time, during which the technical solutions show that approx. 6% of the switches (280 units) are of high risk
have improved from the wire fence solution to the full encapsulations. and approx. 25% are in the medium risk zone (1240 units).
 M.D. Catrinu, D.E. Nordgård / Reliability Engineering and System Safety 96 (2011) 663–670 667

Table 2 The example was constructed using expert knowledge, expressed

Test population of 12 kV MV, air-insulated switch-disconnectors. qualitatively and simplified economic calculations of maintenance
and investment costs. The uncertainty in decisions modelling is of
both external and internal type, as discussed in Section 2.2 and
concerns mainly the safety criterion.
The safety risk was assessed qualitatively for each switch-
disconnector type as illustrated by the risk matrix in Section 4.2.
The other two criteria considered—the maintenance and investment
costs are assessed quantitatively and depend on the number of
assets being considered for either maintenance or reinvestment.
Table 3 shows estimates of the relative maintenance costs and
investment costs associated with Strategies 2 and 3, comparing with
the base case—Strategy 1.
To provide decision support in this decision setting, the soft-
ware IDS multi-criteria assessor has been used. IDS is a general-
purpose multi-criteria decision analysis tool developed based on
the evidential reasoning (ER) theory [20]. The software deals with
multi-criteria problems having both quantitative and qualita-
Experts indicated that this situation may require reinvestments or tive information with uncertainties and subjectivity Qualitative
changes in the maintenance strategy of specific switch-discon- evaluations or ‘degrees of belief’ can be included in a formal
nectors in order to reduce the safety risk. The following technical analysis, together with quantitative criteria. IDS can thus be used
solutions are available: to resolve both the external uncertainty (in data) and internal
uncertainty (imprecision in judgments) as discussed in Section 2
 For the wire fence switch-disconnectors (W), accounting for 13% of this paper. Applications of the ER approach in decision support
of the total population: reconstruct the encapsulation or for power system management are reported for example [21,22].
replace them with new SF6 switches.2 The first step in using the software is the definition of the
 For the semi-encapsulated switch-disconnectors (S) accounting decision problem, i.e. the definition of alternative strategies and
for 37% of the population: improve the maintenance (cleaning, their impact on the three decision criteria: safety, maintenance
lubrication, etc.). cost, investment cost. This is equivalent with the risk matrix in
Section 3 only that IDS allows the definition of a belief decision
Considering the above technical solutions, the following main- matrix of which the conventional decision matrix (discussed in
tenance and reinvestment strategies have been considered for Section 3) is a special case.
further analysis: In this example, the criterion safety risk is a qualitative measure
Strategy 1: Maintain as usual. while the maintenance and investment costs are quantitative mea-
Strategy 2 sures, depending on the number of assets being subject to a strategy.
IDS allows preferences in terms of safety risk to be defined as
(a) Replace all wire fence encapsulated switch-disconnectors the asset manager’s beliefs regarding the probability (Fig. 2) and
with SF6 insulated switches—650 units. possible safety consequences (Fig. 3) given a strategy.
(b) Improve the maintenance of semi-encapsulated switch-dis- For example, the probability of injury, considering Strategy 1
connectors, in exposed environment (O;E;S) older than 25 (status quo) can be modelled through the ‘belief’ distribution:
years—400 units. {[Improbable, 60%], [Less probable, 20%], [Probable, 20%]}; the
impact of safety given Strategy 1 can be evaluated as: {[Small,
Strategy 3 70%], [Medium, 20%], [Very serious, 10%]}. Note that this is a
‘cumulative’ risk evaluation of a strategy, over all asset categories
(a) Replace all wire fence encapsulated switch-disconnectors in concerned, while the risk matrix (in Table 1) was developed for
an exposed environment ((N;E;W) and (O;E;W)) with SF6 generic assets in each category. The two ways of characterizing
insulated switches—280 units. safety risk are purely based on expert opinion, in this example.
(b) Redesign the encapsulation of all wire fence encapsulated The initial risk evaluation for different assets types can be said to
switch disconnectors in a clean environment ((N;C;W) and have contributed to the overall understanding of risk, in the
(O;C;W))—370 units. evaluation of strategies.
(c) Improve the maintenance of all semi-encapsulated switches Further, IDS allows the combination of these beliefs into a total
in an exposed environment ((N;E;S) and (O;E;S))—870 units. safety risk evaluation for each strategy, as shown in Fig. 4. This
figure can be considered an equivalent of the risk matrix in
4.5. Decision support Table 1, but showing the ‘cumulated’ risk perception for all asset
groups, given Strategy 1.
This section illustrates an approach for analysis and decision
support in choosing a maintenance and reinvestment strategy for Table 3
the studied switch-disconnectors. A decision support software is Cost estimates for different maintenance strategies.
used to illustrate and model decision maker’s preferences based
Increase in maintenance Investment
on the theoretical background briefly described in Section 3.
cost (kNOK)a cost (kNOK)
The criteria considered in the analysis of alternatives are the
potential for reducing the safety risk and the associated mainte- Strategy 1 0 0
nance and investment costs necessary to achieve the risk reduction. Strategy 2 800 52,000
Strategy 3 1745 29,800

2 a
Switches filled with the non-flammable sulphur hexafluoride gas. NOK—Norwegian krone.
668 M.D. Catrinu, D.E. Nordgård / Reliability Engineering and System Safety 96 (2011) 663–670

and investment cost of 50,000 kNOK (Strategy 3) is likely to

contribute to a lower risk exposure. More difficult is however to
differentiate in this way advantages with Strategies 2 and 3.
Higher investment cost (Strategy 2) means that more wire fence
encapsulated switch-disconnectors (approx. 400 units) are
replaced with safer switchers. However, this may only contribute
marginally to the perceived reduction of safety risk and thus the
difference in preferences for the two alternatives can be small.
This may not be the case when evaluating maintenance cost.
Strategy 2 implies intensified maintenance and redesign of con-
siderable fewer units (400 compared to over 1200 units in
Strategy 3) which may not lead to the same perception for risk
reduction as for Strategy 3.
When all strategies are defined in IDS following the described
procedure, the next step is to define criteria weights.
Fig. 5 shows the normalised weights used in this case study.
Naturally, safety is considered the most important criterion,
Fig. 2. Degrees of belief for ‘Probability of injury’ given in Strategy 1.
followed by maintenance and investment costs which depend
on the number of assets being subject to each strategy.
The results from IDS consist in a ranking of the perceived
potential for risk reduction. The ranking is based on average
degrees of beliefs calculated based on preference and belief
information about criteria and weights.
The results in this case study are shown in Fig. 6 and indicate
Strategy 3 as the highest rank based on preference information

Fig. 3. Degrees of beliefs for ‘Consequences’ in terms of safety risk given in

Strategy 1.

Fig. 5. Criteria weights.

Fig. 4. Degrees of beliefs for ‘Safety risk’ given in Strategy 1.

In the same way, the total safety risk picture for Strategies
2 and 3 can be described.
In addition to the above evaluation grades, rules have to be
defined in IDS, to show how each criterion grade may contribute
to the overall objective – the potential for risk reduction – based
on which the alternatives will be ranked. For example no invest-
ments (Strategy 1) is likely to induce higher risk exposure while Fig. 6. The ranking of strategies in terms of risk exposure.
 M.D. Catrinu, D.E. Nordgård / Reliability Engineering and System Safety 96 (2011) 663–670 669

used, with the following degrees of belief {[Same risk expo- The focus in this paper was on the decision support process
sure, 19.8%], [Higher risk exposure, 7.2%], [Lower risk expo- and to show that such a process can be conducted by integrating
sure, 73%]}. different analyses and decision simulation tools, which in princi-
The result is rather intuitive since this strategy is proposing ple are available to the electricity distribution system manager.
improvements for the largest number of assets and has the lowest The link between different tools for risk analysis and the final
investment cost. The second ranked alternative is however not decision is often missing in real life decision making in distribu-
too far behind in terms of preferences although it proposes the tion companies and this paper illustrates possibilities for integra-
replacement of all wire fence encapsulated risk-disconnectors tion of different types of analyses. One of the main advantages of
believed to pose unacceptable risk at some point. using decision support techniques such as discussed in this paper
The purpose with this example was mainly to show possibi- is that it allows decision maker’s judgments at a given moment in
lities for illustrating and modelling preferences under uncer- time to be described and modelled. This can contribute to
tainty. The results exclusively reflect a given set of preference improving the way asset management decisions are recorded
information under the decision setting described. These results do and documented, and thus improving communication within the
not have any informative value otherwise. company’s internal decision levels or between distribution sys-
tem asset managers and external stakeholders.

5. Discussion and conclusions

Acknowledgements
This paper addresses the challenges in providing decision
support into an increasingly complex distribution system asset The work reported in this paper has been performed as part of
management. the research project ‘Risk Based Distribution System Asset Man-
It discusses available theoretical approaches for multi-criteria agement’ at SINTEF Energy Research. More information can be
decision analysis under uncertainty and illustrates its use in found on www.energy.sintef.no/prosjekt/RISKDSAM.
decision support, building upon available information in distribu- The authors thank the partners in the project consortium for
tions companies. A case study addressing the process of main- funding the project activities.
tenance and reinvestment strategy making for air-insulated
switch-disconnectors is presented. The study illustrates how References
simplified risk assessment together with quantitative economic data
can be used in a unified, software-based multi-criteria decision
[1] BSI, PAS 55-1: asset management. Part 1: specification for the optimized
support process. management of physical assets, British Standards Institution, 2008.
The software IDS multi-criteria assessor was used for decision [2] BSI, PAS 55-2: asset management. Part 2: Guidelines for the application of
analysis and support. Although the case study presented was PAS 55-1, British Standards Institution, 2008.
[3] ISO/IEC. Guide 73. Risk management – vocabulary – guidelines for use in
rather small (in terms of the number of assets and alternatives standards, 2002.
being analysed), the approach can be used for larger cases. [4] Yeddanapudi SRK, Yuan L, McCalley JD, Chowdhury AA, Jewell WT. Risk-
However, the effort in categorizing assets based on risk will based allocation of distribution system maintenance resources. IEEE Transac-
tions on Power Systems 2008;23:287–95.
increase together with the number, type and complexity of the [5] Hughes D, Pschierer-Barnfather P, Pears T. Condition based risk management
physical distribution system components analyzed. Moreover, (CBRM): bringing excellence in asset management into the boardroom. In:
the qualitative, overall evaluations of risk for different strateg- CIRED 2009—20th international conference on electricity distribution, Pra-
gue, 2009.
ies will become more difficult as the number of asset groups [6] Brown RE, Spare JH. Asset management, risk, and distribution system
will increase. planning. In: IEEE PES power system conference and exposition 2004, New
The amount of information supposed to be available to support York, 2004.
[7] Bertling L, Allan R, Eriksson R. A reliability-centered asset maintenance
the decision in this case was rather limited. The case focuses on method for assessing the impact of maintenance in power distribution
distribution system components that are rather overlooked from systems. IEEE Transactions on Power Systems 2005;20:75–82.
a system perspective. They are rarely analysed in details and [8] Brown RE, Humphrey BG. Asset management for transmission and distribution.
IEEE Power and Energy Magazine 2005;3:39–45.
limited failure statistics and condition monitoring information is
[9] Eurelectric, The operating environment for distribution companies, Eur-
available. They only count because of their large number in the electric, Brussels, Report no. 2004-233-0005 February 2005.
distribution systems. When parts of the asset fleet is approaching [10] Williams L, Jami I, Argent S. Good practice asset risk management: a
end of life, or when the technical solution used is no longer regulator’s tale. Presented at 19th international conference and exhibition
on electricity distribution, Vienna, Austria, 2007.
acceptable, then large numbers of assets have to be replaced in [11] Janjic AD, Popovic DS. Selective maintenance schedule of distribution net-
the same time—a situation which can be challenging, at times, for works based on risk management approach. IEEE Transactions on Power
distribution system managers. Systems 2007;22:597–604.
[12] Nordgård DE, Gjerde O, Sand K, Catrinu MD, Lassila J, Partanen J, et al.. A risk
The case studied mainly relied upon expert opinion for based approach to distribution system asset management and a survey of
analysing safety risk and categorizing, and also on qualitative perceived risk exposure among distribution companies. In: CIRED—19th
multi-criteria evaluations (as degrees of belief). The proposed international conference on electricity distribution, Vienna, 2007.
[13] Catrinu MD, Nordgård DE. Incorporating risk analysis and multi-criteria decision
decision support approach can also be used when more informa- making in electricity distribution system asset management. In: Bris G, Soares,
tion is available: better statistical data, more in-depth, quantita- Martorell, editors. Reliability, risk and safety: theory and applications. Proceed-
tive risk assessments and simulations that better reflect the risk ings of the European safety and reliability conference, ESREL 2009, vol. 1, Prague,
Czech Republic, 7–10 September 2009. Leiden: CRC; 2010. p. 393–400.
and uncertainties involved. [14] Stewart T. Dealing with uncertainties in MCDA. In: Figueira J, Greco S,
The case presented is a constructed case—it simulates com- Ehrogott M, editors. Multiple criteria decision analysis: state of the art
mon decision situations for a distribution system manager. More- surveys, vol. 78. International series in operations research & management
science. New York: Springer; 2005. p. 445–66.
over, we used ‘average’ expert knowledge and we did not involve
[15] Belton V, Stewart TJ. Multiple criteria decision analysis: an integrated
a decision maker with responsibility for the decision. Further work approach. Boston, MA: Kluwer Academic Publishers; 2002.
is necessary in order to test and implement such approaches in [16] Keeney R, Raiffa H. Decisions with multiple objectives: preferences and value
practice. Real life implementations should also include sensitivity tradeoffs. Cambridge: Cambrigde University Press; 1999.
[17] Yang J-B. Rule and utility based evidential reasoning approach for muliat-
analysis with respect to the preferences/choices of the decision tribute decision analysis under uncertainties. European Journal of Opera-
maker(s) involved. tional Research 2001;131:31–61.
670 M.D. Catrinu, D.E. Nordgård / Reliability Engineering and System Safety 96 (2011) 663–670

[18] Nordgård DE, Sand K. Application of Bayesian networks for risk analysis of [21] Tang WH, Spurgeon K, Wu QH, Richardson ZJ. An evidential reasoning
MV air insulated switch operation. Reliability Engineering & System Safety approach to transformer condition assessments. IEEE Transactions on Power
2010;95:1358–66. Delivery 2004;19:1696–1703.
[19] Aven T. Risk analysis. Assessing uncertainties beyond expected values and [22] Yang J, Ai X, Zhao T. An evidential reasoning approach to risk assessments in
probabilities. Chichester: Wiley; 2008. composite generation and transmission system. Presented at the electric
[20] Yang J-B, Xu D-L. On the evidential reasoning algorithm for multiple attribute utility deregulation and restructuring and power technologies (DRPT) 2008.
decision analysis under uncertainty. IEEE Transactions on Systems, Man and Third international conference.
Cybernetics, Part A: Systems and Humans 2002;32:289–304.