CIRED 18th International Conference on Electricity Distribution
QUANTITATIVE RISK ASSESSMENT; A KEY TO
COST-EFFECTIVE SAIFI AND SAIDI REDUCTION
Marko KRUITHOF
Eneco Netbeheer – The Netherlands
m.g.kruithof@eneco.nl
SUMMARY
The Dutch utility ‘Eneco Utrecht’ (part of the ENECO group)
was faced with increasing financial risks due to increasing
risks of failure of their ageing infrastructure. The challenge
was, to mitigate these risks in the most cost-effective manner.
KEMA developed a practical methodology for identifying and
ranking these measures on the basis of maximum risk
mitigating for minimum costs. The methodology was based on
a quantitative risk assessment (QRA) with input from data-
analyses and staff knowledge. Results were very encouraging,
as the structured methodology not only ensured optimum use
of available knowledge but also initiated open communication
between the different participating departments. Also, the
quantified results in some cases proved the anticipated results
wrong.
PREPARATION
In January 2003, Eneco implemented an organisational
change to achieve the unbundling of the Asset Manager and
the Service Provider. The tasks of the new departments were
directly related to the Eneco business drivers and hence,
included clearly defined accountabilities [1].
To fulfil her task adequately, the Asset Management
department of Eneco Netbeheer needed to have a good
understanding of the risks Eneco was exposed to due to their
ageing Medium Voltage (MV) distribution network. Both
kwalitative and quantitative valuations were required to
manage the risks and retain optimum power quality in a cost-
effective manner.
100
Current activities
Milestone 2:
Recommended
% Complete
improvement
projects
Milestone 1:
Step 3: Develop mitigation measures
Plan of approach
(when required)
Step 2: Undertake quantitative risk assessement
Step 1: Establish method of approach - Initiate risk analyses
0
June, 2003
Time
Figure 1: Risk Management – Planning Steps
CIRED2005
Session No 5
Turin, 6-9 June 2005
John HODEMAEKERS Rob VAN DIJK sr.
Eneco Netbeheer – The Netherlands KEMA – The Netherlands
j.p.hodemaekers@eneco.nl robert.vandijk@kema.com
To achieve this, KEMA developed a step-wise planning as
depicted in figure 1. The first 2 steps have been completed
and Eneco is now working on development and
implementation of the mitigating measures (steps 3 and 4).
APPROACH
To control the operational and financial risks in a utility, it is
essential to inventorise all relevant existing risks, assess their
potential consequences and develop risk-mitigating actions.
The risk-mitigating actions are to be verified for their
economic viability. Therefore, a quantitative risk assessment
(QRA) was considered the most appropriate approach. A
short introduction to a QRA is described below.
Quantitative Risk Assessment
In principle, a QRA consists of answering the following five
questions:
- What are the hazards / what can go wrong?
- What are the consequences / costs, if it goes wrong?
- How often is it expected to happen?
- Is this acceptable?
- If not: what could we do about it?
After having answered the above questions
for any (group of) component(s) and
organisational issues, a cost/benefit
analysis is undertaken. Hereby the
Active Risk Management
in MV grids for optimum
estimated costs of the mitigating measures
cost/benefit ratio are compared to the prevented costs of a
failure. If economically viable, a measure
Step 5:
Verify effectiveness
could be implemented. It should always be
of mitigation measures considered however that more than one
and adjust, if required solution could be available to solve the
Step 4: Implement problem!
mitigation measures
If a measure is not economically viable, the
risk should be accepted or some other
means of risk mitigating should be
developed. When an accepted risk-
mitigating measure is implemented, it
becomes part of the overall risk
management system. A graphical
representation of risk management
principles is depicted in figure 2 overleaf.
CIRED 18th International Conference on Electricity Distribution Turin, 6-9 June 2005

"RISKS"

Damage
Example
Hazards Consequential limitation
preventive
damage measures
potential measures
causes:
- not knowing training, PPE, escapes,
Human error, health &
- not able procedures, fire fighting eq't,

Normal Operation
Normal Operation

deliberate action safety

- sabotage protection rescue equipm.

- design maintenance, effect restriction

Internal technical
- condition replacement, OR environment measures Recovery
failure
- loading protection (containment)

- lightning system- redundancy,

External technical,
- subsidence protection, repair org.,
technical failure direct costs
- digging civil meas. risk transfer

normal
failure
operation

1. REPORTING
Figure 2: Risk Management Principles 2. EVALUATING
"CONTROL MEASURES" 3. DIRECTING

Information requirements
All risks in relation to the MV grid are related to the
performance of the separate components in the grid and their
functionality. To quantify the risks, information is required
from all levels of the grid: from component-level up to
system-level. To obtain this information, several workshops
were organised in which Eneco staff were interviewed for
their perception of the problems.
Different sessions were held with representatives from the
management teams, planning, design and operations
departments. The representatives were selected on the basis
of their knowledge of the particular issues. The sessions
resulted in a list of most-feared risks, supplemented with
potential mitigating measures and their rough cost estimates.
BASIC DATA AND ESTIMATES
To verify cost savings from implementing risk-mitigating
(control) measures, a limited amount of numerical data is
required.
To obtain this data, each risk ‘block’ in figure 2 was treated
as a separate entity and discussed in detail with the relevant
groups.
As the study commenced in the year 2003, the only complete
outage figures and statistics that were available, were from
2002. Although some improvements had already been made,
these were the figures that were used throughout the study.
The entire QRA process, as executed, is depicted in figure 3.
Costs of failure
From statistics and verbal information from Eneco, the
average cost of an outage failure was calculated as follows:
Average cost = the total costs paid to recover from the
failures, divided by the total number of failures. In the Eneco
2002 case: (€ 2.2 Million) / (201 failures) = € 11.000 per
failure. This € 2.2 Million comprised € 1.4 Million of time-
depending costs and € 0.8 Million of fixed costs (e.g.
component costs). Obviously, if recovery times could be
reduced, a cost saving could be obtained. Saving one percent
in total time for recovery from the problems would yield a
cost saving of: € 1.4 Million / 100 = € 14.000. The 201
failures accounted for approximately 9 million customer
minutes lost.

Risk-reduction
Failure (mitigation) measures Recommended
Database Risk Assessment Convergence Cost-benefit analyses actions

Failures
Ranking on C/B ratio

SAIDI reduction
Information A (short term) 1
on all failures B 2
Costs &
and outages C ….
Benefits
Criteria & D SAIFI reduction
Risk
Evaluation E (short term)
Analyses
F
Interviews Information G
Discussions from the … SAIFI reduction To the Planning
Workshops organisation
(long term) Department; new
(staff)
construction items

Figure 3: The entire QRA process

CIRED2005

Session No 5
CIRED 18th International Conference on Electricity Distribution
Estimates of mitigating measure quality and costs
To assess the improvement potential of mitigating measures,
estimates are required of their costs, benefits and anticipated
service life duration. To facilitate providing these estimates,
four categories were developed, each with a different rating
of severity: L (Low), M (Medium), H (High) and VH (Very
High).
For outage duration reduction, the following categories were
used: L = 0-2%, M = 2-6%, H = 6-14% and VH = in excess
of 14%.
Analogues categories for costs [units: € x 1000]:
0-150, 150-750, 750-1500 and in excess of 1.5 Million.
Analogues categories for anticipated minimum service life
[years]: 0-2, 2-6, 6-14 and in excess of 14 years.
To estimate the service life of mitigating measures, the
following assumptions were made:
- New equipment lasts at least 20 years
- Modifications and upgrades last at least 10 years
- IT and organisational upgrades last 4 years
- Inspections, plans, etc. are only valid for one year
To compare investments, all costs and benefits are expressed
on a ‘per year’ basis. The approximation was made that new
investments were written off linearly over the anticipated
service lives of the particular investments. This is not as
accurate as working with Net Present Values (NPV’s), but the
difference lies well within the uncertainty of the other input
parameters and therefore has not much added value (at this
stage).
RESULTS
Due to the fact that several independent workshops were held,
many risks appeared more than once on the list. After
screening for duplication and combining similar risks, a total
of 40 essentially different risks remained, with in total 50
potential mitigating measures. These measures were sorted by
their nature of influence, i.e. reducing the frequency of the
failures (SAIFI) and reducing the recovery (repair) time of the
failures (SAIDI). Another sub-division was made to group the
timing of their influence; short term or long term. The
measures that are merely effective on the long term mainly
addressed new construction projects. These were left for later
study, as the goal of this exercise was to focus on ‘quick
results’ only.
Outage duration (SAIDI) reduction
The risks that are of influence on the outage duration are
mainly proces-related. This is illustrated by the fact that
remedial action is required when an outage occurs and the
CIRED2005
Session No 5
Turin, 6-9 June 2005
efficiency of this action mainly determines the time the
outage persists. Improving the related business processes
therefore can reduce the average outage times. Some
(random) examples are:
x Hazard: Increased traffic delays repair crews
- Potential mitigating measure: Spread out home
bases of crews so travel distance becomes shorter
x Hazard: Small stations are difficult to find in the
cities due to new buildings covering or hiding them
- Potential mitigating measure: Equip all vehicles
with GPS location finders
x Hazard: Equipment expertise and area knowledge is
disappearing through retiring of staff
- Potential mitigating measure: Hire and train new
staff prior to old staff leaving the company
The above are just a few examples of the many proces-related
risks and mitigating actions that were identified. At this stage,
i.e. prior to the cost/benefit analyses, it was by no means
certain that any of the above mitigating measures would
actually be implemented.
After the cost-benefit analyses were carried out, the following
four mitigating measures appeared to be economically viable:
- Install GPS equipment in each repair crew vehicle
- Review protection philosophy and procedures
- Readjust protection relay settings
- Earth the isolated neutral system (i.e. earthing the
‘floating neutral’ across the grid)
Further investigation with more accurate data would have to
prove whether these four projects indeed are economically as
good as they appear. Some of these studies have been, and
others are currently being carried out.
Outage frequency (SAIFI) reduction
The risks that are of influence on the outage frequency are
mainly related to the components and their operating
environment. This is illustrated by the fact that the condition
and/or the environment of a component basically determine
when it fails [2]. Improving the condition or environment of
the components therefore can reduce the average failure
frequencies. Some examples are:
x Hazard: Some types of circuit breakers are well
overdue for replacement
- Potential mitigating measure: Replace breakers
soonest
x Hazard: Cables become damaged by unsupervised
digging activities
- Potential mitigating measure: Supervise digging
activities
x Hazard: Some types of old cable joints are of poor
quality
- Potential mitigating measure: Replace cable joints
pro-actively
CIRED 18th International Conference on Electricity Distribution Turin, 6-9 June 2005

The above are just a few examples of the many component-
related risks and mitigating actions that were identified. At
this stage, i.e. prior to the cost/benefit analyses, it was by no
means certain that any of the above mitigating actions would
actually be implemented.
To obtain an estimate of the costs of each particular failing
component type, the failure database was analysed. After
correcting for some obvious input-errors, the failure statistics
were divided in surface equipment (i.e. substation component)
failures and subsurface component failures (cables and
joints). Note: It should be borne in mind that there are no
overhead lines in The Netherlands at MV distribution level;
only underground cables.
After the cost/benefit analysis, (only) the following two
mitigating measures appeared to be economically viable:
- Develop and implement a pro-active cable joint
replacement strategy, relating replacement priority
to joint type and soil conditions
- Develop and implement procedures for handling of
cables and joints at excavation sites
Again, further investigation with more accurate data would
have to prove whether these projects indeed are economically
as good as they appear. Some of these studies have been, and
others are currently being carried out.

In the analyses, estimates were made to distinguish between UNCERTAINTY

unknown slow degradation processes of cables (and joints)
for which no particular cure or prevention method was
available. The quantity thereof was estimated to be 30%. For
the remaining 70% of the cable and joint failures, some form
of prevention would have been possible. Figure 4 shows the
results of this analysis in graphical format.
Several mitigating measures were proposed for each of the
five component-types that had caused problems. Several of
the proposed measures actually complemented eachother or
would tackle more problems than just the one they were
proposed for. The list of mitigating measures therefore was
combined into a smaller list of measures. When a mitigating
measure mitigates multiple risks, the economical
attractiveness obviously increases, representing the
convergence of the methodology.
After assigning monetary values to each risk and mitigating
measure (with the L, M, H, VH categories, as described
above), a cost/benefit analysis was carried out.
To obtain insight in the sensitivity of the calculated
cost/benefit ratios that are close to ‘1’ (i.e: the mitigating
measure cost is comparable to the risked cost), statistical
uncertainty distributions were added to input parameters.
Therefore, probability distributions were added to estimated
costs, perceived benefits and anticipated service lives.
With a probabilistic software program (an add-on to MS
Excel), the cost/benefit ratios were calculated 10,000 times
with each time a (nearly) random value being picked from the
input distributions. These calculations resulted in a spread on
the outcomes that were initially calculated as fixed values.
An example of two resulting probability distributions can be
seen in figure 5 overleaf. The distribution on the left has a
certainty of only 60% that the mitigating measure would be
economical while the mitigating measure represented on the
right has a 92% chance of being cost-effective.

Failures that could

Unknown slow have been prevented
degradation process: no to a certain extent
30%
131 action possible
92 cable joints
joints 70% (78,752 cust.hrs.)
80%
Unknown slow
30% degradation process: no
Mitigation Measures

33 action possible
164 20% 23 cables
subsurface cables
70% (19,688 cust.hrs.)
equipment
failures
201 Outages 3 3 terminations
terminations (2,568 cust.hrs.)
(from failure
batabase) 37 7 7 circ. break.
surface equipment circ. breakers (5,992 cust.hrs.)
failures
(includes cable
4 4disconnects
terminations)
Disconnects (3,424 cust.hrs.)

23 Insufficient detail or low

total: 100% Other number of failures; no total: 64%
sensible action possible

Figure 4: Failure database analysis

CIRED2005

Session No 5
CIRED 18th International Conference on Electricity Distribution Turin, 6-9 June 2005

With regard to the contents, the following conclusions were

drawn:

x The QRA has proven that some perceived

economical solutions to existing problems were not
nearly as economical as was thought. Examples are:
Figure 5: Probability Distributions of Cost/Benefit ratios ‘increase the number of field repair staff’ and ‘install
automation systems on the existing substation
components to enable remote control’.
The results of the exercise for Eneco Utrecht yielded many
distributions with no red color at all (in black&white print: x Because of the relatively limited number of surface
the lightly shaded right-hand side of the distributions), (substation) component failures and the lack of
implying that those measures would be economical, correlation among them, it is not economical to take
regardless of the uncertainties of the input. any mitigating measures against them at this stage.

The same software application also provides sensitivity

analyses in the form of a tornado diagrams (see figure 6). In additon to achieving the primary objective, each
From these diagrams, it can be seen which factors are of most implemented mitigating measure also has a positive effect on
influence on the Kosten/baten 10,15&16 power quality and grid capacity. These benefits were not
cost/benefit ratios. In 0,4 0,5 0,6 0,7 0,8 0,9
included in the QRA, hence they are the ‘hidden extras’ that
essence, it indicates that Levensduur 15 47,86556687 31,91037792
will become available when implemented.
improvements made to the Kosten 10, 15 & 16 5.102 7.654

top items on the list yield 6-14% 12 8 Other ‘hidden extras’ are available because of the simplified
much more benefit than 2 - 6% 4,8 3,2 accounting method (linear depreciation and Money of the Day
improving the ones at the Kosten 22 72 [MOD]). When calculations are repeated with Net Present
bottom of the list. Each Kosten 26 1.440 Values (NPV) and realistic depreciation times, the projects
item in the list represents Baten 46 17,51265455
with a longer service life tend to become even more
an input parameter like Kosten 46 24.579
Downside economical.
costs, times, improvement Upside

assumptions etc.
Figure 6: Tornado Diagram
In summary, the QRA has proven to be a valuable tool in
achieving cost-effective SAIFI and SAIDI reduction. To
obtain more accurate budget figures for the proposed (and
CONCLUSIONS selected) mitigating measures, the process can be repeated for
each measure in more detail. In this manner, obtaining
detailed information is only required for the selected
Upon completion of the above exercises, the following solutions, not for all solutions as identified in the beginning
conlusions could be drawn with respect to the QRA process: of the QRA process.

x The QRA has clearly identified the hazards with the

highest impact on ‘customer minutes lost’
originating from component failures as well as REFERENCES
organisational imperfections.
[1] J.P. Hodemaekers, M. v.d. Meijden, 2003, “Multi-utility
x The first cycle of the QRA has proven that, even Asset Management in a business driver oriented
with very rough input estimates, some very useful
conclusions can be drawn. The differentiation concept”, Proceedings CIRED Barcelona.
between ‘most likley uneconomical’ and probably
economical’ mitigation measures could easily be [2] P. v.d. Aardweg, R.A. van Dijk, 2004, “An integrated
made. Even with significant refinement of the input approach of assessing the condition of power
parameter values at a later stage, none of the ‘most transformers”, Proceedings of MEDPOWER conference,
likely uneconomical’ measures became economical! Cyprus

x Because of the multi-disciplinary approach, this

exercise has proven to work as a catalyst in
achieving good communications between the
different participating departments.

CIRED2005

Session No 5