CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0024

METHODOLOGIES FOR THE ANALYSIS OF BOTH

SHORT-TERM AND LONG-TERM NETWORK RISK

Simon BLAKE Philip TAYLOR Alan CREIGHTON

Durham University - UK Durham University - UK CE Electric UK - UK
[email protected] [email protected] [email protected]

understanding of the risk inherent in existing distribution

ABSTRACT networks, to develop a means of quantifying this risk, and
to develop methodologies to model and analyse that risk.
The level of electricity distribution network risk in the UK is
perceived to be increasing in the short term, and an asset In the longer term, a number of additional factors could be
management approach seeks to optimise the balance expected to impact on levels of network risk, including:
between capital expenditure and the risk to customer • Increased awareness of global warming leading to
supplies. Further drivers that could lead to increased a tighter regulatory environment and to increased
demands on the system;
network risk can be anticipated in the longer term. Against
• Increased levels of connected generation, leading
this background, the paper seeks to model, analyse and to a requirement to proactively manage power
quantify both short-term and long-term network risk, and flows and voltages across the network;
has developed methodologies for doing so. These • Possible national energy supply shortages if aged
methodologies are illustrated by a case study based on the generators are not replaced as quickly as they are
anticipated rebuilding of a key 132 kV switching station on retired;
part of the UK network. • The development of active network management,
using sophisticated control technology, including
INTRODUCTION deployment of demand side management;
• Possible further unbundling and restructuring of
The annual rate of growth of the power distribution network the industry e.g. introducing further competition
in the UK has decreased from around 10% before 1970 to and other system operators such as Energy
just over 1% in recent years. This decrease, as well as new Services Companies.
environmental, financial and regulatory pressures, has The project also seeks to quantify and model these longer
tended to create an environment where the need for and term, more uncertain factors.
benefit arising from capital investment in the network are
subjected to increased scrutiny. Research Background
Many investigative studies concentrate on optimizing the
Where in the past low-risk network configurations with high performance of feeders operating at medium voltages such
levels of built-in redundancy were preferred, today’s asset as 11 kV [1-3]. This may be because most of the consumer
management emphasis on delivering good value for money
for electricity consumers has meant that it is appropriate to unavailability occurs at these voltage levels, and because
re-examine the balance between system reliability and such feeders tend to be similar to one another and can
capital expenditure. therefore be represented generically. At voltages, in the 33
kV to 132 kV range, there tends to be rather less research
There are drivers that are tending to increase the level of material, and it is with networks operating at these higher
risk to the continuity of supply to customers (or Network voltage levels that the present study is mainly concerned.
Risk). These drivers include:
• Increasing utilisation of distribution networks; The growth of an asset management approach has also
• Increasing frequency of severe weather conditions; focussed attitudes. Instead of being risk-averse, utilities
• Increasing levels of both accidental and malicious increasingly identify prudent performance targets and
damage, leading to network vulnerability;
actively manage risk [4]. With this focussed outlook, asset
• Ageing of the asset base;
• Loss of engineering experience and expertise as management is a complex process. Norgard et. al. list 31
older engineers retire. distinct challenges to distribution utilities, grouped into six
Set against these drivers is increasing expectations from categories: legal, organizational, technical, environmental,
customers and increasing regulatory focus on understanding reputation and societal. They state that the electricity
and managing levels of network risk. distribution industry increasingly regards the concept of risk
assessment as an important tool in distribution network asset
Concern about these issues has led to the creation of a joint management [5].
research project between CE Electric UK, who are
responsible for electricity distribution over a large area of For any network risk studies, a range of approaches are
northern England, and the School of Engineering at Durham
University. This research seeks to gain a deeper available and the most appropriate must be chosen. Brown

CIRED2009 Session 5 Paper No 0024

CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0024

and Hanson, in a seminal paper, compare ‘four common Figure 1 – Case Study Network Configuration
methodologies used for distribution reliability assessment’ There are a number of options for carrying out the project,
[6]. Standard textbooks describe all four in detail, and apply such as whether to rebuild in situ or on an adjacent green
them to power systems [7]. The choice of methodology for field site. The in situ solution costs less, but could involve
the present research was described in detail in a previous longer circuit outages and consequently a greater risk.
paper by the present authors [8]. The decision was made to
use Analytical Simulation, supplemented where appropriate The rebuilt NP would have an expected lifetime of at least
by the other three approaches (Network Modelling, Markov 40 years, and this raises questions as to whether the
Modelling, and Monte-Carlo Simulation). The approach was straightforward replacement of the existing configuration is
illustrated in that paper by three different case studies taken the best solution to those challenges to network reliability
from within the CE Electric UK network. that are likely to arise over the next 40 years. One option
concerns a separate 132 kV double circuit, not shown in
132 KV CASE STUDY Figure 1, which feeds two other loads from a different GSP,
but which passes within a few metres of the NP. It would be
Figure 1 shows the network around a nodal point (NP), relatively straightforward to connect this double circuit into
which is a 132kV switching station whose aged assets are the NP by the installation of additional switchgear, but the
due for replacement. The network operates as a closed ring cost of doing so and the customer benefits delivered need to
and whilst load D (at Substation D) is clearly supplied via be understood. The additional interconnection would give
the nodal point, network studies indicate that a proportion of an extra supply option for all six loads, thereby increasing
load B is also supplied via the nodal point under normal network flexibility and possibly overall reliability, and has
operating conditions. Load A, load C and the remainder of been included as part of the construction project.
load B are supplied directly from Grid Supply Point (GSP).
Double circuits are used throughout this part of the network. Conversely, it could be possible to eliminate NP from the
Dotted lines indicate underground cable, solid lines indicate network altogether. Figure 2 shows a possible
overhead lines, and X indicates a circuit breaker. The reconfiguration which does this. This would cost
replacement of the assets at the nodal point is a major significantly less than the replacement of NP, and leave a
construction project, expected to last over two summer simpler network, although one with rather fewer operational
seasons (when UK demand is lowest), and to cost around options and possibly with an increased level of risk.
£8M. During construction works, principally when
transferring circuits from the old equipment to the new, one However, the biggest issue in rebuilding for the next 40
or more circuits will be out of service, and consequently the years concerns challenges which can perhaps only be
risk to customer supplies during these periods will increase. qualitatively perceived at present. These challenges could
include global warming, greater levels of distributed and
renewable generation, demand side management, smart
grids, active network management, and restructuring of the
industry. Any solution adopted for the rebuilding or
replacement of NP must at least attempt to address some or
all of these challenges.

ANALYSIS OF SHORT-TERM RISK

The level of network risk before, during and after the
construction period can be quantified in expected £K per
year, made up of three distinct elements:
1. The financial penalty imposed by the regulator if
the number of customer interruptions (CI) exceeds
a target level. (There is a corresponding reward if
the number is below target)
2. A similar financial penalty or reward based on
outage duration, the number of customer minutes
lost (CML).
3. The additional cost of unplanned repairs to assets
as a consequence of circuit failure compared to
undertaking repairs as part of a planned
programmed (whether or not such failures result in
a loss of customer supplied).

CIRED2009 Session 5 Paper No 0024

CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0024

within the revised configuration. This configuration was

modelled using the DINIS system, and a network analysis
was carried out to compute steady-state power flows and
voltage levels throughout the revised network. This
modelling confirmed that, under all (n-1) outage scenarios
(which in the UK require immediate restoration of supply
for this size of demand) the revised network was not
overloaded, and that voltage levels remained within
statutory limits.

A second consideration concerns the size of protection

zones resulting from this configuration. Figure 2 shows that
one circuit creates a protection zone with switching on 4
separate sites. This is the upper limit permitted by the UK
grid code for 132kV circuits, and is above the upper limit of
3 separate sites currently recommended at CE Electric UK.

A third consideration is the use of average values for model

input parameters such as failure rates and restoration times,
which produces the average or expected risk cost per year
over a number of years. However, CE Electric UK are also
concerned about their liability, an indication of the possible
risk cost in a particularly bad year, such as the worst in 10
. Figure 2 – Alternative configuration without NP years or even in 100 years. To assess the magnitude of the
liability, Monte Carlo Simulation was used, using the Risk
A model was constructed incorporating probabilities of Solver package [9].
component failure, and likely consequences of such failure,
for various scenarios. The base run of this model evaluated In this methodology, the frequency and severity of circuit
the network risk inherent in the present configuration, failures are not fixed at average values, but are sampled
across all six loads. The expected annual risk came to £67k from a representative probability distribution. Restoration
for CI, £63k for CML and £32k for repairs, a total of £162k. times are likewise sampled from an underlying distribution.
It is against these levels of risk that different construction A large number of possible years (50000 was the number set
and network configuration options should be assessed. in the model) were evaluated and ranked. The 90th
percentile of this output then gives the estimated network
For example, the additional risk incurred due to circuit risk incurred by configuring the network without NP in the
outages across the whole construction period is an expected worst year in 10 years. This came to £302k (as against
£42k if construction takes place on a green field site. If, £207k for the average year). The worst year in 100 years
however, construction has to take place in situ, then the (the 99th percentile) gave a risk of £439k. These are the
extended outage periods increases this risk by £66k, to figures that CE Electric UK management would consider in
£108k. This extra £66k of risk can be compared with the deciding whether the additional liability was acceptable
additional capital cost of constructing on a green field site. when set against possible capital expenditure savings.
The risk reduction once new equipment is installed proved
harder to assess, as the gas insulated switchgear (GIS)
ANALYSIS OF LONG-TERM RISK
technology used is still fairly new at this voltage level, and
so accurate expected failure rates are not known. It can be seen that the methodology for evaluating short-
term risk in expected annual £k is a versatile and powerful
Additional Considerations risk assessment tool, particularly when supported by power
The alternative configuration shown in Figure 2 was also flow analysis and operational considerations and when
evaluated using this short-term risk methodology. The extended by Monte Carlo Simulation to encompass
expected annual network risk increases by £45k over the variability and liability. However the modelling does require
base run, to £207k. This extra annual cost can be compared accurate input data such as failure rates. These are available
with the one-off capital cost of around £8M to rebuild the for present and foreseeable future risk. But any asset
NP. On this basis, the alternative configuration appears to replacement needs to perform a valuable network function
be a cost-effective option. However, there are other for at least 40 years to mitigate risks that could well be quite
considerations to be taken into account. Before evaluating different to those presently experienced. A methodology is
network risk, it is necessary to investigate power flows required to address these future risks.

CIRED2009 Session 5 Paper No 0024

CIRED 20th International Conference on Electricity Distribution Prague, 8-11 June 2009

Paper 0024

To do this, a separate modelling technique was devised, to possible intermediate solution, in terms both of capital cost
measure the robustness or resilience of the network to cater and of network risk. Scoring it by this methodology gives an
for a number of possible future scenarios. For initial trial of average vulnerability index of 52, an intermediate value.
this technique, a list of 10 such scenarios was used, as
shown in Table 1. Each of the six loads was assessed under CONCLUSIONS
each of the 10 future scenarios, and scored on a scale from 0
(no vulnerability) to 10 (extremely vulnerable). The level of risk inherent in electricity distribution networks
is perceived to be increasing in the short term and still more
Scen- Description in the long term. Several investment options are available to
ario mitigate that risk, including renewal or reinforcement of
assets, reconfiguration of the network, and more radical
1 Loss of two incoming circuits on one 132 kV tower
operating solutions including various kinds of active
2 Loss of two incoming circuits due to cable damage
network management. Choosing the most cost effective of
3 Loss of some or all busbars due to fire or flood
these options is the responsibility of managers, who need
4 Restricted supply from normal GSP (say to 50%)
appropriate ways of measuring and ranking network risk in
5 Increased failure rates of ageing assets order to make informed decisions. This paper has described
6 Political sensitivity of customers to loss of supply and illustrated by case study two such measurements, a
7 Location: vulnerability to weather or damage short-term one using expected annual £k, and a longer term
8 Amount and intermittency of connected generation one using a vulnerability index.
9 Amount and complexity of active network mgt.
10 Sensitivity to a tighter regulatory environment REFERENCES

Table 1 – Future scenarios assessed for vulnerability [1] L.A. Freeman, D.T. Van Zandt, L.J. Powell, 2005,
"Using a Probabilistic Design Process to Maximise
The results of such analysis have to be regarded as Reliability and Minimise Cost in Urban Central
provisional at this stage. For more reliable results, a greater Business Districts", Proceedings, CIRED 18th
amount and variety of expert knowledge would need to be International Conference on Electricity Distribution.
incorporated, to determine [2] J.G. Slootweg, P.M. Van Oirsouw, 2005, “Improving
• Whether the list of possible scenarios was realistic Reliability Calculations in Routine Network Planning:
and exhaustive Theory and Practice”, Proceedings, CIRED 18th
• Whether each scenario should be equally weighted International Conference on Electricity Distribution.
(on a score of 1 to 10), or whether some should [3] M. Oravasaari, J. Pylvaneinen, J. Jarvinen, A.
carry a greater weighting than others, and Maniken, P. Verho, 2005, “Evaluation of Alternative
• What the likely level of vulnerability would be, for MV Distribution Network Plans From Reliability
each load under each scenario Point of View”, Proceedings, CIRED 18th
The provisional results are, nonetheless, interesting. As a International Conference on Electricity Distribution.
base run, the existing network was evaluated for future [4] R.E. Brown, J.H. Spare, 2004, “Asset Management,
robustness. For each load, the 10 vulnerability scores were Risk and Distribution System Planning”, IEEE PES
summed to give a score out of 100. The range was from 37 Power System Conference and Exposition.
(for Load A, the most robust and least vulnerable) up to 52 [5] D.E. Nordgard, K. Sand, O. Gjerde, M.D. Catrinu, J.
for load C (the least robust), giving an average vulnerability Lassila, J. Partanen, S. Bonnoit, J. Aupied, 2007, “A
score of 45 out of 100. Risk Based Approach to Distribution System Asset
Management and a Survey of Perceived Risk
With the less robust configuration of Figure 2, these scores Exposure Among Distribution ompanies”,
increase to give a range from 51 (Load A) to 66 (Load D), Proceedings, CIRED 19th International Conference on
with an average of 56 out of 100. This increase in Electricity Distribution
vulnerability could be used by CE Electric UK management [6] R.E. Brown, A.P. Hanson, 2001, “Impact of Two-
to assess whether the long term risk incurred by simplifying Stage Restoration on Distribution Reliability”, IEEE
the network was acceptable. This would require familiarity Transactions on Power Systems, pp 624-629.
with using the vulnerability index, and confidence in its [7] R. Billinton, R.N. Allan,, 1996, Reliability Evaluation
reliability as an index of risk. of Power Systems, 2nd Edition, Plenum Press, New
York and London.
The vulnerability index could also be used to assess other [8] S.R.Blake, P.C. Taylor, A.M. Creighton, 2008,
possible configurations. For example, a network without “Distribution Network Risk Analysis”, MedPower 6th
NP, but with four circuits instead of two at Load B and an Conference 2008.
additional busbar on the 132 kV side of Load B, would be a [9] www.solver.com/risksolver

CIRED2009 Session 5 Paper No 0024