Evaluation of reliability worth and value of lost load

K.K. Kari uki

R. N.AI Ian

Indexing terms: Reliability, Reliabilit indices, Customer interruption costs, Composite customer damage fiction, Customer outage costs, Incremental
customer outage costs, Value of lost Lad

the overall behaviour of service areas or systems. In the

Abstract: Customers’ perceptions of reliability UK, nationwide fault statistics collected under a
may not always reflect the level of reliability national fault and interruption reporting scheme are
purported by traditional reliability indices. This used to calculate these indices. The most widely
has been recognised and efforts have been made reported performance indices, availability and security,
by the electricity supply industry (ESI) in the are defined as the minutes lost per customer and the
United Kingdom to relate ‘reliability investment’ number of supply interruptions per 100 connected con-
with customers’ marginal benefits obtained from sumers, respectively. These indices provide useful infor-
such investment. A difficulty encountered during mation to the utilities and the regulator on the
such earlier and recent efforts has been the lack performance of each public electricity supplier (PES).
of appropriate valuation of these benefits. With a From the customers’ perspective, their perceptions of
view to correcting this paucity, the authors have reliability are influenced by,
conducted studies, based on customer surveys, (a) the number of interruptions experienced
aimed at assessing the customer outage costs (b) the duration of these interruptions
(COC) due to electric service interruptions. The (c) the costs incurred as a result of the interruptions.
incremental values of these costs, ACOC,
following reliability investment are considered This perception of reliability may not always mirror the
proxies of reliability worth and customers’ level of reliability purported by the performance indices
marginal benefits. The results of these studies since these essentially represent the average perform-
have provided a very good insight into customers’ ance over the entire PES. This apparent discrepancy
concerns regarding supply interruptions, but most between the utility’s and customers’ perception of relia-
importantly a coherent method for evaluating the bility probably led to the consideration in the late
customer benefits (ACOC) has been developed 1970s of customers’ marginal benefits as a factor in
and the required generic data for such evaluation aspects of power system planning, design and operation
generated. Using these data, a consistent method [2, 31. Specifically, it was considered that in making an
for calculating the value of lost load (VOLL) is ‘incremental reliability investment’, the aim of the
also developed. industry should be to match the marginal cost of pre-
venting loss of supply to the marginal benefit conse-
quently obtained by customers [2]. A difficulty
encountered then was the lack of appropriate values of
1 Introduction customers’ marginal benefit. This difficulty has per-
sisted and 15 years later, the need for more appropriate
The ability of a power system to provide adequate and values of customers’ valuation of losses was highlighted
secure supply of electrical energy at any point in time is in the NGC’s review of transmission security standards
referred to as the reliability of a system. In this defini- [41.
tion, ‘adequacy’ refers to the static system conditions In the meantime, since October 1992 a research activ-
and the existence of sufficient facilities within the sys- ity, supported by Manweb, MEB and Norweb, aimed
tem to meet the system load demand while ‘security’ is at assessing the customer outage costs due to electricity
associated with the dynamic response of the system to supply interruptions, has been conducted by UMIST.
perturbations to which it is subjected [l]. This paper This activity centred on extensive surveys in these three
applies to the adequacy aspect and any reference to regional electricity company (REC) areas.
reliability should therefore be assumed to infer this A study of the factors affecting the perceived cus-
aspect, rather than the overall definition of reliability. tomer interruption costs (CIC) [5] has shown that, in
the event of supply interruptions, customers are only
Two sets of indices are used by utilities to measure
concerned with the inability to use their equipment and
the reliability of power systems. These are load point the likely damage to this equipment. This supports the
indices which reflect the behaviour of individual load
view expressed in [6] but is contrary to the concepts of
points, and system performance indices which reflect
‘implied cost per kilowatt-hour saved’ [2, 31 and value
0IEE, 1996 of lost load (VOLL) [4] which suggest that customers’
ZEE Proceedings online no. 19960191 marginal benefits of reliability worth are related to the
Paper first received 29th August 1995 and in revised form 27th November ‘energy not supplied’. In consequence, the authors have
1995 derived a consistent and coherent method for evaluat-
The authors are with the Manchester Centre for Electrical Energy, ing customer outage costs (COC) for any network,
UMIST, PO Box 88, Manchester M60 lQD, UK service area or PES, using the perceived values of CIC
IEE Proc.-Gener. Transm. Distrib., Vol. 143, No. 2,March 1996 171
 Table 1; GIG (f)values
Sector No. of Response CIC (f)for an interruption duration of:
Responses Rate(%) mom. I m i n 20min I h 4h 8h 24h
Residential 4014 19.1 - - 0.19 0.70 4.78 - -
Commercial 203 4.0 11.47 11.74 49.12 106 345 719 1.0k
Industrial 119 5.7 1.2k 1.5k 2.9k 4.3k 7.6k 12.0k 16.3k
Largeuser 19 29.2 216k 216k 219k 233k 329k 413k 581k

ed by these surveys. This is the primary purpose not required. Instead, this paper concentrates on devel-
paper since such COC are observed to provide oping the cost model.
objective and consistent measures of the worth of sup-
ply. In addition the authors have reconsidered VOLL 2.2 Developing the cost model
y and numerically using the cost data For a particular service area, the cost model is a series
resulting from these studies. of values referred to as the composite customer damage
In order to achieve this an understanding of the per- function (CCDF) [XI which represent the costs due to
customer interruption costs (CIC) and the cus- supply interruptions as a function of interruption dura-
outage costs (COC) is required. These are tion for the particular customer mix. The following
d as follows: steps are followed to develop a CCDF:
CIC: the perceived individual customer or average 2.2. I Step A. Values of the perceived costs of
sector customer costs resulting from electricity inter- interruption (f) for the various interruption
ruptions. They are therefore system independent durations are obtained: Mail surveys were
COG: the expected total costs incurred by all the employed in the UMIST studies to obtain estimates for
customers connected to a particular network or service the CIC per interruption for residential, commercial,
area. They are calculated from the CIC and take into industrial and large user (>8MW) sectors. The survey
consideration the network performance data and load- procedures, sample sizes and the number of responses
ing information. They are therefore customer mix and are given and discussed in [SI. This paper concentrates
system dependent. on how the perceived customer interruption costs,
whose average sector values are presented in Table 1,
u § ~ o r n eoutage
~ costs (COC) can be processed into COC.
In order to facilitate analyses, it is important that the
2. I Factors affecting COC CIC data is collated according to the sector and where
From their definition, the customer outage costs of a possible grouped according to the SIC classes or sub-
service area and time period are a function of: sectors. The following definitions apply:
(i) the number of interruptions experienced Sector: This is a broad category which describes the
(ii) the duration of these interruptions main usage of electricity. It refers to the commercial,
the perceived interruption costs of the customers industrial, large user and residential categories.
SIC class: This is a subcategory of a sector as classi-
the numbers and mix of customers. fied under the 3-digit code of consumers, e.g. code 620
Consequently to evaluate the COC for a particular represents hotels and motels classes of the commercial
service area, at least two models must be established; a sector. The 3-digit code covers all categories of con-
scribing the performance of the power sumers.
the area and a cost model representing Subsector: This term defines a group of SIC classes,
IC for the mix of connected customers. under a category, which have a common characteristic.
third model, the load model, becomes For example all residential sector consumers, irrespec-
A brief overview of the system model is tive of SIC class, supplied under Economy 7 tariff are
classified under the Economy 7 subsector. In the light
COC, knowledge of the average failure of these definitions, a sector is formed by a number
rate (h) and the average duration of interruption (r) is SIC classes or subsectors which are representative of
these indices are derived from data the make-up of its population.
fault reporting scheme mentioned
earlier. They can also be deduced from the availability 2.2.2 Step B. The estimates for each interrup-
. Unfortunately these indices are tion are normalised b y dividing the costs by
uced from collected data and are specific to the per- either the annual energy consumed or the
systems and may not be suitable peak demand: Since they are system independent,
for predicting future performance especially if the rele- CIC would neither be suitable for comparing historical
hilosophies alter in or future performance of alternative network configu-
prediction of the rations nor for crossutility comparison. In order to
s are to be used in make these comparisons on a similar and consistent
values of 1 and r basis, the CIC per interruption are normalised with
techniques which, respect to either the customers’ annual energy con-
sumption (MWh) or with respect to their peak demand
(kW). These normalised costs are given by eqns. 1 and
2,
(in L/MWh) (1)

172 IEE Proc.-Gener. Transm. Distrib., Vol. 143, Na. 2, March 1996
 each duration rr studied, the SIC classes’ normalised
(in L/kW) (2) costs are appropriately weighted to yield a composite
value for the overall sector, CE,y(r,) or CL&J.
where CIC,(r,) are the-costs incurred by customer x per Although some utility planners suggest that for short
interruption of duration Y, while E, and L, are the interruptions (say less than l h ) weighting by peak
user’s annual energy consumed and the peak demand, demand is more appropriate since losses are related to
respectively. a power shortage rather than an energy shortage [9], in
these studies weighting by both parameters is carried
2.2.3 Step C. If individual SIC classes or sub- out across the entire range of durations. Eqns. 7 and 8
sectors are sufficiently large, their normalised show the calculation of sector normalised costs for a
costs are first calculated before proceeding to general sector y consisting of ns subgroups with weight-
Step D, otherwise if retaining the SIC classes ing with respect to the annual energy consumed by SIC
or subsectors does not yield any meaningful class k and the SIC class peak demand, respectively:
results, the sector values are calculated at this
stage and Step D omitted:
(i) If SIC classes or subsectors are sufficiently large, C E , y (Tz) = kEns (in .Z/MWh) (7)
then the normalised CIC values for customers within 7 Ek
U -
the same SIC class are averaged to give the correspond- kEns
ing SIC class values. For a general SIC class k with y1
consumers these values, CE,k(rJand CL,k (YJ are given
by: CL,y(Tz) = kens (in L/kW) (8)
CE,x
-5-
U
LE’.
kEns
xEk
CE,k(Tz) = (in L/MWh) (3) Both sets of sector values (for the entire range of dura-
n
~

tions studied), CE,y(r,)and CL,?(r,),are referred to as a

sector customer damage function (SCDF) @I. This is
defined as the sector’s normalised costs due to supply
(in L/kW) (4) interruptions expressed as a function of interruption
duration for the customer mix supplied. In effect they
(ii) It is expected that there will be instances when represent the costs an average consumer in a sector
poor response rates will mean that sector sizes are would incur per megawatt-hour consumed annually or
insufficient for subdivision and prevent consumer sec- per kilowatt of peak demand.
tors being divided into their SIC classes. In such
instances the overall sector normalised costs, CE,,(r,) 7
and CL,y(rJ,depending on whether energy consumption
or peak demand are used in normalising, are obtained
by simply averaging the relevant sector’s individual
customer values. As indicated above, this makes step D
redundant.
For a sector y with y1 consumers, therefore, the nor-
malised values are given by:

CE,y(G) = ~
c
XEY
CE,x
(in L/MWh) (5)
n 01 1 10 100 lo00 lo000
duration ,min
Fig. 1 Ratios of C
, to C
,
Residential (-), c6mmerciaY(- - -), industrial ( ), large user (- - -)
(in L/kW) (6)
2.2.4 Step D. The SIC class or subsector nor- The SCDFs for the sectors surveyed, combined for
malised costs are weighted to give the sector the three RECs involved in the surveys, are given in
values: When responses are adequate to enable subdi- Table 2. An analysis of these SCDFs indicates there is
vision of a sector into its constituent SIC classes, for an approximately constant ratio (see Fig. 1) between
Table 2: SCDFs for sectors investigated
SCDFs (f/MWh) SCDFs (f/
Duration Res. Comm Ind. L. user Res. Comm Ind. L. user
mom - 0.46 3.02 1.07 - 0.99 6.15 6.74
lmin - 0.48 3.13 1.07 - 1.02 6.47 6.74
20min 0.06 1.64 6.32 1.09 0.15 3.89 14.27 6.86
Ih 0.21 4.91 11.94 1.36 0.54 10.65 25.26 7.18
4h 1.44 18.13 32.59 1.52 3.72 39.04 72.22 8.86
8h - 37.06 53.36 1.71 - 78.65 120.11 9.71
24h - 47.58 67.10 2.39 - 99.98 150.38 13.35

IEE Proc.-Gener. Trunsm. Distrib., Vol. 143, No. 2, March 1996 173
the corresponding SCDF values for each sector, i.e. by implementing steps B and C only. The combined
between C,,(rJ and CLJri). This suggests a relation- REC SCDFs for the residential, commercial, industrial
ship between the two types of SCDFs. For any sector and large user sectors are given in Table 2.
y , this relationship can be shown to be empirically Also, during the surveys the respondents were asked
expressed by eqn. 9, to estimate CICs for only a limited number of outage
C L >Y
durations. It is therefore suggested that the SCDF val-
(in L/MWh) (9) ues for any intermediate durations within the specified
LFy x 8.76
range should be computed by linear interpolation while
where LFy is the load factor for sector y. This relation- those outside this range should be calculated using lin-
ship is important since it enables the ‘derivation of a ear extrapolation methods. The only exceptions being
single cost model rather than the pair that would be for,
expected from the two SCDFs. This is shown in the (a) outages of less than Imin, whose momentary
next and final stage of developing a cost model. SCDF values, where computed, should be assumed;
2.2.5 Step E. Finally the SCDF values are (b) residential sector outages less than 20min: in the
appropriately weighted in proportion to give case of the residential sector, the option to assume the
CCDF values; Cost of interruption surveys should be 20min SCDF values for shorter durations is given as
undertaken with specific objectives, such as system there is no real basis for assuming otherwise.
expansion, reinforcement or refurbishment. Further- The need for load information has gradually emerged
more, in order for the costs of the surveys to be justi- in the process of developing the load model. Although
fied, the projects involved should be considered not truly a load model because it only refers to the
substantive. Such projects are likely to affect all cus- annual energy consumed and peak demands (for the
tomer sectors supplied by the network under review. In customer or sector), this consumption information is
order to moderate the costs incurred, each SCDF value nonetheless loosely referred to as the load model.
for the sectors supplied is appropriately weighted to
yield a corresponding composite value, C(ri), for the 2.3 Calculating customer outage costs using
customer mix being considered. The expression for cal- CCDF
culating this composite value, assuming weighting by For any system or part of, the calculation of COC
the sector annual energy consumption, is given by involves the convolution of the cost model, the load
model and the system model. Assuming busbarj to be
a load point in a network consisting of b busbars, the
annual COC due to supply interruptions at the busbar
(COCJ and in the network (SCOC) are given by:

(in L/MWh)
where Ey is the annual energy consumed by sector y
and ny is the number of sectors connected to the sys-
tem under review. Energy consumption is used as the
moderating factor due to its information being readily
SCOC = ccocj
b

j€ b
(in .F) (13)

available, otherwise contribution at peak demand could Fig. 2 shows schematically the steps for calculating the
also be used. In instances where difficulties in disaggre- customer outage costs.
gating the load and therefore obtaining t.he load factors
may be encountered, the following expression which 2.4 Example
uses the supply point’s load factor (LF) may be used To illustrate the calculation of COCs, the following
instead: example, based on part of an actual 33kV network,
considers some proposed changes in order to remove
some obsolete switchgear. Figs. 3 and 4 show the sys-
tem before and after the proposed decommissioning of
obsolete switchgear around busbars 1 and 2. The
arrows show the points at which the network is con-
(in L/MWh) nected to the rest of the 33kV system. The analysis is
confined to buses 3 and 4 which are likely to be the
The series of values, C(r,), for the range of durations
most adversely affected by the proposed changes. The
studied forms the composite customer damage function
load, cost and system models are shown in Table 3.
(CCDF) for the associated network. This CCDF is
defined as the normalised costs due to power supply It should be noted that:
interruptions expressed as a function of the interrup- (i) the CCDF given in Table 3 is calculated by
tion duration for the customers and mix in the particu- weighting the appropriate SCDFs in Table 2 using the
lar service area [SI. The CCDF constitutes the cost load model data in accordance with eqn. 10
model used in evaluating customer outage costs (COC). (ii) the system model indices are derived from a reli-
It should be noted that in the UMIST surveys, only ability assessment of the two buses using the RELNET
the residential sector had sufficient responses to justify program which has been developed at UMIST [lo].
classification of the respondents into their constituent This program uses typical component failure and repair
subgroups. Consequently, only this sector’s results have data to compute the indices. However, in order to sim-
undergone the entire weighting process, represented by ulate the superior reliability of contemporary compo-
steps B to D, in order to obtain the SCDF. In the non- nents and to allow for degraded performance due to
residential sectors, the SCDFs are therefore obtained age, the failure rate and repair times for switchgear in
174 IEE Proc.-Genev. Transm. Distrib., Vol. 143, No. 2, March 1996
 calculate normalised
CIC estimates, ie
CE& 1 and C&i 1
L

are survey responses

sufficient to enaMe sector
subdivision into SIC classes?
-no
simply average GJ(r, )
and CL~(G 1to give
Ccy(ri 1and C$i 1
c

group cE;A ri )and

CE &ri )and CJ,ri weighted

information to ‘give CE,y(ri arid C~L(’,

I
’
1 I

Fig.2 Calculating customer outage costs

Table 3: Load, cost and system models for example

Load model Cost model (CCDFs) System model
Sector, y E(MWh) L (MW) Load Duration C(rJ Index Bus 3 Bus4
Factor (rj) (f/MWh)
Residential 8700 2.43 40.90% mom. 0.50 Existing system
Commercial 14600 3.97 42.00% Imin 0.52 h 0.48 0.46
Industrial 9800 2.00 55.90% 2Omin 1.34 r 0.95 0.99
Largeuser 0 0 0.00% 1h 2.84 U 0.46 0.46
Total 33100 8.40 45.00% 4h 9.32 System after pro-
posedchanges
8h 17.2 h 0.70 0.53
24h 22.91 r 3.69 4.68
U 2.58 2.48
~~ ~ ~~~

Fig. 4 are assumed to be two thirds of those of the cor- (a) Although the failure rates do not seem to
responding components in Fig. 3. increase very much, the average outage durations
Using the three models and applying eqns. 12 and 13, change significantly. The availability and security indi-
the COCs, SCOCs and ACOCs can be calculated. ces would be expected to capture these effects. How-
These are presented in Table 4. ever, they would scarcely be able to reflect the impacts
on the customers in the way the ACOCs do.
2.5 Discussion (b) Whichever index is used to measure reliability,
The results obtained from the above analysis lead to ACOC represents the incremental customer benefits or
the following comments: costs associated with the increase or decrease in the
IEE Proc.-Gener. Transm. Distrib.,Vol. 143, No. 2, March 1996 175
 most likely the result of escalating the overall Ioad
weighted mean corresponding to l h (21.7lkWh) by
some index in accordance with:
1.7(1 + a)t = 2 (14)
This equation is satisfied for a = 4.4 and t = 4; where t
is interpreted to be the time (in years) between 1985
COCs suggests that the COC can be con- and 1989 whde a may be assumed to be the average
sidered as a supplementary reliability index. RPI over this period. The BlkWh has been increased
with the RPI since 1989 to the 22.345lkWh quoted in
the NGC review [4] and used by the Pool during 19921
93 (implying an average RPI of 4.06%); it is now
&2.389/kWh. No objective reasoning for adopting this
figure is available although the NGC review deemed a
value of around &2/kWh to be a reasonable ‘average’ or
central value. Despite this lack of justification, VOLL
bus 4 has had a very important role in the setting of Pool
payments. Lately there have also been proposals to
include VOLL as a factor in deciding the operational
Fig. 3 Network system before modpcution and planning security standards for the transmission
system. These applications are addressed below.

3.2 Applications of VOLL

There are two main areas in which VOLL is used; in
setting Pool payments and in planning operational and
future transmission security standards.
P 33/11kV ,
J >< bus4 3.2. I Setting Pool payments: Since vesting in
1989, the VOLL has been used in the setting of capac-
Fig.4 Network system after proposed changes ity payments to generators and also in setting the max-
imum purchase price for energy traded. Its prime
Table 4: COCs and ACOCs before and after proposed
purpose is to compensate expensive generation used to
changes mitigate against likely capacity and energy shortages in
the Pool. The likelihood of shortages is represented by
Existing network After proposed ACOC (f) the loss of load probability (LOLP).
changes price (PPP) and the relationship between the VOLL,
C o s t s ( f ) Bus3 Bus4 Bus3 Bus4 Bus3 Bus4 PPP and the system marginal price (SMP) are shown in
COC 43300 42900 200400 187000 157100 144100 eqns. 15 and 16.
SCOC 86200 387400 301200 PPP = SMP + CAPACITY ELEMENT (15)
(in L/kWh)
r valuation fs (March 1985
PPP = SMP + LOLP x (VOLL - SMP) (16)
Load/customer Outage duration (h) (in X/kWh)
category 1 3 6 12 24 24+
For LOLP = 1, PPP = VOLL, thereby setting the Pool
Industrial: minimum 0.1 0.1 0.1 0.1 0.1 0.1 maximum price, otherwise known as the Pool ceiling
mean 3.7 2.5 2.0 1.5 1.3 1.1 price. For LOLP = 0, PPP = S M P .
maximum 9.9 6.5 5.0 3.9 3.4 2.8
Commercial 5.9 6.8 6.1 5.1 4.7 4.1
3.2.2 Operational and planning aspects of the
transmission systern: The impetus for considering
Farm 1.5 1.4 1.6 1.8 2.0 2.1
VOLL as a factor in transmission system planning and
Domestic 0.8 0.7 0.5 0.4 0.4 0.4 operation was driven by OFFER who, in October
Overall load 1.7 1.5 1.2 1.1 0.9 0.8 1992, not only expressed concern about the high levels
weighted mean of constraint costs that arose during maintenance and
other outages, but also questioned whether such costs
3 ~ load (V
~ a l ofu lost were always justified by the values customers are likely
to place on additional security [12]. Consequently a
3.7 Derivation of VOLL currently in use in review of the transmission security standards was insti-
the ESI tuted by the NGC, who offered [4] the most detailed
VOLL is said to represent the value an average con- discussion to date of the concepts and apphcations of
sumer puts on an unsupplied kWh. Its iaitial value of VOLL in transmission system planning and operation.
&2/kWh adopted in 1989 by the Electricity Supply This report explicitly suggests that VOLL has potential
Industry was apparently derived from the results of a application in the cost/benefit assessment of alternative
Finnish survey conducted in 1977/8 [ll] and has since transmission security standards. VOLL is used to deter-
been pegged to the retail price index (RPI). mine the expected unreliability costs (X)(eqn. 17) and
Table 5 shows the equivalent results of the Finnish combined with other costs (eqn. 18) gives the overall
survey in 1985 prices [Ill. From this table it can be transmission system cost (TT).
shown empirically that the E21kWh adopted in 1989 is where T = total cost of transmission investment, Q =
176 IEE Proc -Gener Transm Dzstiib Vol 142, No 2 March 1996
 X = [ENS] x [VOLL] end consumers) and consultants identified the need for
(17) more work to review and determine a more appropriate
TT = T 0 X + + (18) VOLL [4]. This is timely since it coincided with the
operating costs associated with transmission (uplift and concluding stages of the UMIST project aimed at
losses) and ENS = energy not served. assessing the customer outage costs due to electricity
The desired result in both transmission planning and service interruptions.
operation is minimisation of TP In the operation Before seeking to develop a new method for calculat-
phase, however, T is fixed and the minimisation of ing VOLL some fundamental flaws of the current value
O+X is the target. based on intrinsic aspects of this concept are first high-
In order to perform cost/benefit assessment of alter- lighted:
native transmission standards, i.e. whether generation (i) The components of VOLL, the ‘worth of supply’
would be constrained so that the system is secure and ‘energy not supplied’, are not directly related or
against single circuit ( N - l), double circuit (N’ - D) functional [6]. Consequently, although VOLL can be
and high risk pairs of circuit (N’ - 2) outages, advan- calculated, as presently defined it does not and cannot
tage is taken of the fact that when LOLP = 1, PPP = reflect the value customers place on their unsupplied
VOLL. Consequently, VOLL is considered to be the energy.
maximum cost an average customer would be willing to
pay to avoid an interruption. Based on this premise (ii) The current value of VOLL is based on a 1977/8
VOLL was subsequently used to determine the appro- Finnish survey. Yet the applications of electricity and
priate security levels for the different weather states in the impacts of interruptions in electricity supply have
each par1 of the system [4]. The criterion used is cen- obviously changed. The authors are convinced that
tred on the comparison of the security cost (SC) with even if the original value of &1.7/kWhwas a fair reflec-
VOLL such that if for a region SC<VOLL, the higher tion of the value of lost load in 1985, escalation by the
RPI cannot aptly represent today’s values of lost load.
security level is upheld. SC is defined as the extra con-
straint costs required to sustain a higher level of relia- (iii) Although it is understandable that the adoption
bility divided by the expected energy loss per annum of a figure based on a foreign survey may have been
(E) due to faults, otherwise secured at the higher level, due to lack of UK data, a critical assessment of the
resulting from operating the system at a lower security definition of VOLL indicates that it is dependent on
level. i.e. climatic conditions, the composition of the load, availa-
bility of alternative sources of energy, etc. It is there-
(in L/kWh) (19) fore geographic area and customer mix specific. Some
weighting factor(s) may therefore be needed to escalate
where C, and COare the constraint costs at the higher the Finnish value(s) to reflect UK conditions.
and alternative security levels respectively. They are
dependent on the bid prices while the other component (iv) Although the definition of VOLL does not spec-
of SC, i.e. expected energy loss are a function of the ify when the kWh is unsupplied nor the circumstances
failure rate, outage duration and the size of load lost. leading to the curtailment, the factors affecting it are
The security costs for UK and Wales boundaries as expected to include:
defined in the NGC report [4] are shown in Fig. 5. 0 the activities curtailed by loss of supply and
1000, I
therefore time of day
0 the duration of interruption
1 0 the number of interruptions
0 the availability of advance warning
0 the cause of interruption, i.e. is it due to: forced
outage; planned maintenance outage; scheduled or con-
tractual load shedding
0 the prevailing weather conditions, and therefore
time of year, etc.
From the foregoing it is apparent that VOLL is also
time (including seasons)- and circumstance-specific.
Consequently, the best value of VOLL for system plan-
ning and operation is an expected aggregate of all its
possible values.
Although it is evident that the authors question the
validity of VOLL as a concept, it remains a fact that it
boundary is used in setting Pool payments and as a means of
Fig.5 Security costs jfkF/lcwh)f4] comparing alternative transmission standards. This
being the case, the authors have reconsidered the
3.2.3 Other applications: OFFER has also used numerical value of VOLL in the light of these studies
VOLL to value the energy potentially not supplied by and they have developed the following approach for its
the transmission system [12]. evaluation. The method is formulated in such a manner
that allows the distribution of outage durations to be
3.3 A new method for evaluating VOLL taken into consideration. With greater availability of
The significance of the current and potential roles of the required data, the principle may be extended to
VOLL was addressed during the NGC review, and consider other factors such as, the proportion of rural
NGC customers (RECS, direct load consumers and to urban networks, seasonal variations, etc.
IEE Proc -Gener Transm Distrib , Vol 143, No 2, March 1996 177
3.4 The proposed method not supplied for a 1-h interruption is the VOLL, as
The starting point for the evaluation of the VOLL implied when the &2/kWh was adopted in 1989, then
using the proposed method are the SCDFs for the sec- any of the three values, 518.25, L14.11 and 52.62lkWh
tors surveyed, expressed in 5IMWh consumed p.a. The (read off Fig. 6) can be claimed and justified as repre-
method uses two steps; the first establishes the VOLL senting the value an average consumer, in the assumed
values for the durations studied while the second calcu- mix, places on an unsupplied kilowatt-hour during a Ih
lates the expected values taking into account the distri- interruption period, thereby highlighting the weak-
bution of the outage durations. nesses indicated earlier. Although the value weighted
by the number of consumers compares favourably with
3.4.7 Step 7: This step is approached in two ways the presently used value, the authors believe that an
depending on the amount of information available; the energy weighted value is more appropriate because it
first assumes that individual sector load factor informa- reflects the importance and greater energy needs of the
tion is unavailable, i.e. only the overall load factor of larger consumers.
the customer mix is available while the second assumes
that sector load factors can be estimated or are known.
In the first approach, the SCDF values are first
appropriately weighted to give the CCDF values (refer
to eqns. 10 and 11) which are in turn used to calculate
the corresponding VOLL for each duration, ri (hours)
studied, as shown in eqn. 20,
C(ri)
VOLL(r,) = (in L/kWh) (20)
r, x LF
~

LF refers to the load factor of the customer mix con- 1

sidered.
When developing the CCDF in Section 2.1 (steps D
and E), the term ‘appropriately weighted’ was used and 0.1
weighting by energy consumption subsequently 0.1 1 10 100 1000 lo000
adopted. Although weighting by either the annual outage duration,mrn
energy consumed or the peak demand seems natural, Fig.7 Sector VOLL curves
Residential (-), commercial (- - -), industrial ( ), large user (- --)
nothing precludes weighting by the number of consum-
ers. In order to explore the implications of the weight- In the second approach, from each SCDF value
ing criteria, the VOLL curves for the combined RECs (given by eqn. 7) a value of lost load attributable to
surveyed using this first approach are shown in Fig. 6. each sector for the corresponding outage duration is
These curves were calculated for an assumed customer derived. The formula for the VOLL corresponding to a
mix and energy consumption similar to those of the duration ri (hours) for a general sector y is given by,
combined three RECs such that the residential, com-
mercial, industrial and large user proportion of:
VOLL,(?-i) = -
CL,Y( T i ) (in X/kWh) (21)
1000 LF, x ri
As in the Finnish survey mentioned earlier, a series of
VOELs can be derived, one for each sector expressed
as a function of time, as shown in Fig. 7. The VOLL
curves or functions presented in this Figure assume
load factors of 33l/,, 40, 50 and 66213%for the residen-
tial, commercial, industrial and large user sectors,
respectively. The corresponding composite curve for a
mixture of ny sectors is calculated using the following
2 62 equation:

1 VOLL(r,) = VOLLy(r,) x k, (in L/kWh) (22)

0.1 1 10 100 1000 1mo zEny
outage duration,min where ky is a weighting factor used to appropriately
Fig.6 VOLL curlies derived using dgerent weighting criteria - frst moderate the effect of sector y amongst the sectors
approach
Energy-weighted (-); demand-weighted (- - -); number-weighted (-....) considered. The values of ky may be the number of
consumers in sector y , their annual energy consumed or
(a) energy consumed is 35%, 25%, 30% and 10% - peak demand expressed as a fraction of the service area
(- 1 total. For the three types of ky, suffice it to mention
(b) simultaneous maximum demand (SMD) is 52%, that although the VOLL curves derived have similar
16%, 24% and 8% (assuming 80% residential, 40% com- characteristics with those shown in Fig. 6, their values
mercial, 65% industrial and 85% large user, sector peak are higher as shown in Fig. 8, e.g. weighting with
demands occur at SMD) - (- - -) respect to the energy consumed, SMD contribution and
with respect to the number of consumers yields VOLLs
(c) number of consumers is 91%, 8%, 1% and O%, - of E23.46, 517.14 and &4.6O/kWhfor a one hour inter-
respectively (......). ruption. As in the previous approach, the authors con-
A load factor of 60% is assumed. On the basis of these sider the energy weighted VOLL to be more
curves and if the average customer valuation in Elkwh appropriate.
178 IEE Proc -Gene? Transrn Distrrb , Vol 143, No 2,March 1996
 3.5 Likely impacts of the new method
The proposed method results in VOLLs significantly
different from the existing &2.389/kWh.Table 6 attests
to this. Considering the uses of VOLL, the effects of
these values would include:
(a) higher capacity payments to the generators
(b) higher unreliability costs, such that some
expected savings due to operation or planning of the
transmission system at alternative security standards
would be offset
1
(c) more boundaries would meet the SC < VOLL
01 1 10 100 1000 10000 criterion and therefore call for their operation at a
outage durationpn higher security standard, e.g. as shown in Fig. 6, only
Fig.8 VOLL curves derived using dflerent wezghting criterul - second one boundary, Manmidimp, meets this criterion for the
approach
Energy-weighted (-), demand-weighted ( ), number-wei$hted (- - -) then applicable VOLL (E2.345) considering all weather
fault rates [4]. With an EVOLL of &10.88/kWh (this
3.4.2 Step 2: Although there are many factors that figure corresponds to the combined value calculated for
affect VOLL, it is assumed in this paper that it is a three RECs), six boundaries would satisfy it.
function of the duration of interruption only. The
expected value, EVOLL, can therefore be calculated as 4 Conclusions
follows:
24 The discrepancy between the utility’s and customer’s
EVOLL = VOLL(ri) x p ( r i ) (23) perception of reliability has been highlighted and sug-
ri =O gested as the possible reason behind efforts to include
where p(ri) is the probability of an outage of duration reliability worth in aspects of power system planning,
ri. Since every PES in the UK is required to restore design and operation. However, ostensibly due to lack
supply after an interruption within 24h, this duration is of alternative methods and the requisite data, the past
used to set the upper limit. The only problem foreseea- and present values of reliability worth used in these
ble is the determination of a reasonable distribution of applications have been established using deterministic
the outage durations. Simulation techniques are likely methods. Despite the general acceptance that such val-
to be useful in this respect. In order to indicate what ues are inappropriate, they have nevertheless continued
and how VOLLS may be obtained for the RECs in to be used in historical applications of reliability worth.
whose areas the surveys were conducted, the actual dis-
Through this paper, the authors therefore make availa-
tribution of restoration stages by duration from 132kV
to LV were obtained from the RECS for the period ble a set of generic data which can be used in reliability
1993/4. These were assumed to represent the distribu- worth evaluation. Secondly, by presenting an objective
tion of the outage durations. EVOLLs were then calcu- procedure for evaluating incremental customer outage
lated (using the first approach) using the SCDFs and costs which are considered proxies for reliability worth,
CCDFs for each REC and weighted using the propor- a consistent method for evaluating reliability worth is
tions of energy consumption (calculated load factors made available. Finally, a procedure for evaluating the
are in the range of 59-66%) and customer mix given in VOLL is also established.
the privatisation prospectus [131. For the combined
RECs’ values the proportion of energy consumption 5 Acknowledgments
and the customer mix described previously are used.
The results are given in Table 6. For comparison pur- The authors express their gratitude to Manweb, MEB
poses, EVOLLs for 50% load factors and the respective and Norweb for providing the financial resources and
VOLLs corresponding to a 1-h interruption are also customer data that enabled these studies to be con-
given in this Table. ducted.

Table 6:VOLLs - f/kWh

Weighting LF VOLL RECA RECB RECC Combined
Weighting 59-66% EVOLL 9.36 12.30 11.53 10.88
w.r.t. energy VOLL’ 18.60 20.12 20.79 18.25
comsumption
50% EVOLL 12.30 14.54 13.81 13.29
VOLL’ 24.44 23.78 24.91 22.30
Weighting 59-66% EVOLL 2.10 2.36 3.29 2.68
w.r.t VOLL’ 2.10 2.59 3.19 2.64
number of
cOnSumerS 50% EVOLL 2.76 2.79 3.94 3.28
VOLLl 2.76 3.10 3.83 3.23
VOLL’ refers to values of lost load for 1h

IEE Proc-Gener. Transm. Distrib., Vol. 143, No. 2, March 1996 179
 7 BILLINTON, R., and ALLAN, R.N.: ‘Reliability assessment of
large electric power systems’ (Kluwer Press, 1988)
1 BILLINTON, R., and ALLAN, R.N.: ‘Reliability evaluation of 8 WACKER, G., and BILLINTON, R.: ‘Customer cost of electric
power systems’ (Plenum Press, 1984) service interruptions’, Proc. IEEE, 1989, 77, (6), pp. 919-930
2 Electricity Council: ‘Report on application of engineering recom- 9 GOUSHLEFF, D.C.: ‘Use of interruption costs in regional sup-
mendation P.215 security of supply’. A.C.E. Report No. 51, 1979 ply planning’. Ontario Hydro Report 80-301-K
3 Electricity Council: ‘Report of reliability investment in radial HV 10 ALLAN, R.N., SO, T.Y.P., and DA SILVA, M.G.: ‘Reliability
distribution systems with overhead lines’, A.C.E. Report No. 67, analysis of power system networks’. Users Manual for RELNET
1979 Version 4.1, October 1993
4 NGC: ‘A review of transmission security standards’, August 1994 11 The Electricity Council: ‘Guide to investment appraisal - Area
5 KARIUKI, K.K.: ‘Assessment of customer outage costs due to Boards’, December 1985
electric service interruptions’. PhD thesis, UMIST, October 1995 12 OFFER: ‘Report on constrained-on plant’, October 1992
6 ALLAN, R.N.: ‘VOLL fact or fiction’, Power Eng. J., 1995, 9,
~ 13 ‘The Regional Electricity Companies Share Offers - Main Pro-
(11, PP. 2 spectus’, November 1990

180 IEE Proc -Gener Transm Drstrrb , Vol 143, No 2, March 1996