You are on page 1of 12

Customer Cost of Electric Service

Interruptions
lnvited Paper

There is an increasing interest in the quantitative assessment of


power system relrability worth and its application to cost-benefit
evaluation in power system planning. An approach often used to
estimate reliability worth is to determine consumers monetary
losses resulting from service interruptions, i.e., the cost of unreliability. Numerous studies have been conducted to provide estimates of customer interruption costs and a wide range of methodologies has evolved. There is no universal agreement on the
appropriafeness of methodologies to particular situations nor on
the interpretation of the results obtained, but some appear to be
more acceptable and useful to the industry than others. This paper
presents a survey of the techniques available for estimating customer interruption costs, discusses the rationale of those which
are currently popular, and explores the application of such cost
data in creating a composite customer damage function.

INTRODUCTION

The basic function of a modern electric power system i s


to provide an adequate electrical supply to its customers
as economically as possible and with a reasonable level of
reliability. I n this context, the term reliability has a broad,
general meaning. It includes load or demand-side measures such as quality and continuity of service as understood bythecustomer. It also includes utilityor supply-side
concerns such as present and future energy reserves and
operational constraints, like equipment ratings and system
stability limits, which are not directly seen by the customers. As a consequence of this broad meaning, some
researchers have approached the reliability issue primarily
from the demand or customer viewpoint with little regard
for system imperatives, while others have viewed the situation primarily from the system and utility vantage point
with little regard for customer considerations. Since the primary purpose of the system i s to satisfy customer requirements and since the proper functioning and longevity of
the system are essential requisites for continued satisfaction, it i s necessary that both demand- and supply-side considerations are appropriately included.
Manuscript received March 1 5 , 1988; revised October 13, 1988.
This work was supported by the Canadian Electrical Association
and the Natural Sciences and Engineering Research Council.
The authors are with the Department of Electrical Engineering,
Power Systems Research Group, University of Saskatchewan, Saskatoon, Saskatchewan, Canada S7N OWO.
I E E E Log Number 8928070.

From the customer viewpoint, the issue of service reliability is, for many types of customers, simply a question
of whether the supply i s available or not. Other customers,
though fewer in number, have quality requirements more
stringent than normal utility-allowed voltage or frequency
variations and momentary interruptions, which might be
considered as a state of partial availability.Voluntarycurtailment by customers in response to utilityappeals also fits
this category. Basically, customers have come t o expect
electrical supply t o be continuously available on demand.
While most consumers would accept that this i s not realizable in practice since equipment failures will occur,
nevertheless the expectation remains and, t o many, it i s
considered almost a right. This is due, at least i n part, t o the
high levels of reliability enjoyed i n most service areas, and
it has been exacerbated by escalating rate increases during
the last two decades. These factors, along with the inherent
characteristics of electrical supply systems such as their
monopolistic nature, virtually universal clientele, pervasiveness into a l l areas of society and typical large size, result
in a major impediment to the determination of reliability
worth. Customers have little or no choice in terms of rates
versus quality, nor do they have experience or background
to choose if they were given that option. If this i s coupled
with the notion that electrical supply i s almost a social
right,thequestion What istheworth of electrical supply
reliability? becomes an exceedingly difficult one to
address. Unable t o assess reliability worth directly, many
researchers have turned their attention to evaluating the
impacts or losses resulting from electrical supply interruptions, that is, the societal cost of unreliability. It i s generally
recognized that interruption costs are not equal t o reliability worth but rather only indirect assessments thereof,
perhaps a lower bound. Avarietyof methods which attempt
to assess interruption costs have evolved and are discussed
later in this paper.
The power utility has continually attempted t o respond
to societys expectations regarding service reliability. Planning, design, and operating strategies and criteria have
evolved over many decades with the objective of rationalizing and optimizing the reliability, economic, and operational considerations. Improvements in reliability, which
are measured using various reliability criteria and indices,

0018-9219/89/ObO0-0919$01.OO

PROCEEDINGS OF T H t IEEE. VOL. 77, NO. 6, l U N t 1989

C
:

1989 IEEE

919

Downloaded from http://www.elearnica.ir

areaccompanied by increases i n expenditure. Although the


cost of improved service i s initially borne b y t h e utility, subsequent rate changes transfer this cost t o the consumer.
Therefore, in the simplest sense, if the aggregate total of
all customers interruption costs i s assumed t o be a measure of the worth or benefit o f service reliability t o society,
then an optimal target reliability level i s one in which the
marginal cost of incremental improvements in service reliability w o u l d result i n equal marginal reductions i n societal
interruption cost.
The criteria and techniques first used b y system planners
t o address reliabilityconcernswere deterministic i n nature.
The major weakness of these approaches i s that they d o not
reflect the probabilistic o r stochastic nature of system
behavior, customer demands, o r component failures. A
wide range of probabilistic techniques has evolved during
the last several decades. These include reliabilityevaluation
techniques, probabilistic load flow, and probabilistic transient stability calculations. Enhanced computational capability, an increased understanding and acceptance of probabilistic techniques, and the collection and computation o f
appropriate performance data made these approaches possible, although economic considerations were a primary
motivator. Present-day practice suggests that the worst-case
conditions (and their attendant infrequent occurrence)
should not be utilized as design limits o r criteria because
of economic considerations. I n contrast, appropriate probabilistic techniques not only recognize the likelihood of
occurrence of an event but also predict its severity and its
impact o n system behavior and operation. Such approaches
enable system planners and operators t o achieve practical
systems in which the overall cost t o society of providing
quality and continuity of electrical supply i s more closely
related t o the societal worth o r benefit o f having that quality
and continuity.
The conceptual objective of undertaking reliability costbenefit analysis makes it necessary t o independently assess
the cost of providing reliability and the w o r t h of having it.
Most of the techniques currently available for reliability
evaluation focus o n generation and transmission facilities
since inadequacies here can have widespread implications.
Reliability assessments and attendant costs of implementation are becoming well established [I].I n contrast, worth
assessments are immature procedures and have been
undertaken mainly at the customer load points, w i t h most
evaluations based o n losses d u e t o interruptions. Hence,
the costs of providing reliable service and the losses arising
from unreliability are being assessed at different locations
in the system. It i s clear that considerably more research
is required i n this entire area; nevertheless, the current
approaches represent an initial attempt t o optimize the
operational reliability of power systems.

RELIABILITYWORTHCONSIDERATIONS
The w o r t h o r value of electrical service reliability is not
particularly easy t o define and more difficult t o evaluate.
Yet the need for its evaluation i s becoming more important
in planning and operating power systems as outlined above;
therefore, an attempt t o define it i s essential. To an economist, value i s determined using the appropriate demand
function so that, at the margin, the price at which the product i s traded in the marketplace establishes its value. There

920

is a concern, however, that customers willingness to pay


may not be an entirely appropriate method t o evaluate electric service reliability worth.
Reliability, dependability, durability, etc., is a marketable
entity and has established market values i n areas such as
automobiles, business services, and home appliances. That
is, there are manufacturers o r suppliers whose products
have become established i n the marketplace as having
superior reliabilityattributes and thereforecommand prices
higher than similar products which d o not have the high
reliability reputation.An interesting anomaly isthatthe reliability/durability may be real o r it may only be perceived,
which, some w o u l d argue, can result in an erroneous market value. Additionally, there are products o r services for
which the use of a demand function may not be appropriate, such as, for example, matters related t o social o r personal safety orwell-being. If willingness t o pay methods are
used in such markets byoffering t h e c l i e n t a c h o i c e o f alternatives at different costs, the resulting evaluations may b e
flawed. For instance, if potential clients perceive the
approach t o b e inappropriate, their reaction might be t o
withdraw from the market, take an emotional position
rather than a logical one, o r indicate a high willingness t o
pay which they cannot o r d o not intend t o pay. Such
responses, if offered covertly, sabotage t h e process and
generate incorrect marginal market values. It should be
noted that customers need only perceive the approach t o
be inappropriate, not that it i s actually so. Similarly, even
in more conventional markets, one hears arguments that
certain goods o r services are overvalued o r undervalued by
current trading levels. This again suggests that the product
o r service has a true o r intrinsic value which may differ from
its market value.
Anomalies described above arise largely as a result of particular circumstances at work w i t h i n the social and political
context. It is probable that electric supply and its reliability
create such circumstances, and there are a number of
inherent characteristics which make this so. O n e i s that virtually all members of society are customers of the local
power utility. Another characteristic i s that usuallyonlyone
utility serves a particular area, creating a monopolistic situation. Utilities have long been awareof their universal and
monopolistic position, and have garnered public support
by promoting t h e notion that their primary purpose i s t o
serve the public. Governments have also recognized the
economic and social importance o f electrical supply and,
often i n response t o public pressure, have established public review boards w i t h various regulatory functions and
powers which usually include rate setting. W i t h i n this
framework, then, it is not surprising that electrical service
has come t o be viewed almost as a social right; public attitudes and perceptions play a significant role. Therefore, if
electric service customers are asked questions regarding
their willingness t o pay for various aspects of service reliability, their responses may be governed more b y their concern for (and reaction against) potential rate increases than
by an attempt t o address the issue. (Lest readers dismiss the
above argument as simple conjecture, t h e authors suggest
that merely the existence of public regulatory agencies in
almost all jurisdictions attests t o the publics concern for
these matters.)
There are other concerns regarding the definition and
evaluation of service reliability. O n e is t o be aware of the

PROCEEDINGS OF T H E IEEE,

VOL. 77, NO. 6,

JUNE 1989

distinction between the value of serviceperseand thevalue


of the reliability of that service. Another is that the value of
reliability is undoubtedly a function of the reliability level
itself. A third is the definition of what constitutes reliable
service. Notwithstanding any of the above concerns, it is
recognized that, as markets for service reliability become
established through innovative approaches such as voluntary curtailment or remote energy management methods, any marginal costs derived using these approaches
provide assessments of service reliability.

OUTAGE
COSTEVALUATION
A variety of methods has been utilized t o evaluate customer impacts due t o interruptions [2]-[4]. These methods
can be grouped, based o n the methodological approach
used, into three broad categories, namely: various indirect
analytical evaluations, case studies of blackouts, and customer surveys. These categories and variations of approach
within them are identified and briefly discussed below. The
intent i s not t o provide an exhaustive comparative analysis
of the methods used or results obtained; instead, only
selected works are briefly outlined and broad comparisons
are offered in an attempt t o identify the current state of the
art. While a single approach has not been universally
adopted, utilities appear t o favor customer surveys as the
means t o determine specific information for their particular
purposes. Therefore, a major portion of this paper i s
devoted t o survey methodologies, with the authors experiences and results providing the primary example.
A necessary preliminary step in the determination of
interruption costs i s an understanding of the nature and
variety of customer impacts resulting from electric service
interruptions. Impacts may be classified as direct or indirect, economic or otherwise (social), and short-term or longterm. Direct impacts are those resulting directly from cessation of supply while indirect impacts result from a
response t o an interruption. Hence, direct economic
impacts include lost production, idle b u t paid-for resources
(raw materials, labor, capital), process restart costs, spoilage of raw materials or food, equipment damage, direct
costs associated w i t h human health and safety, and utility
costsassociated with the interruption. Direct social impacts
include inconvenience due t o lack of transportation, loss
of leisure time, uncomfortable building temperatures, and
personal injury or fear. Indirect losses usually arise as spinoff consequences and it may be difficult t o categorize them
as social or economic. Examples of such costs are civil disobedience and looting during an extended blackout, or failure of an industrial safety device i n an industrial plant
necessitating neighboring residential evacuation. The final
distinction between short-term and long-term impacts
relates t o the immediacy of the consequence. Specifically,
long-term impacts areoften identified as adaptive responses
or mitigation undertaken t o reduce or avoid future outage
costs. Installation of protective switch gear, voltage regulation equipment, and cogeneration or standby supplies
would be included in this category, aswould the relocation
of an industrial plant t o an area of higher electric service
re1iabi Iity.
Broadly speaking, the cost of an interruption from the
customers perspective is related t o the natureof and degree
t o which the activities interrupted are dependent o n elec-

trical supply. In turn, this deperidency i s a function of both


customer and interruption characteristics. Customer characteristics include type of customer, nature of the customers activities, size of operation and other demographic data,
demand and energy requirements, energy dependency as
a function of time of day, etc. Interruption characteristics
include duration, frequency, and time of occurrence of
interruptions; whether an interruption is complete or partial; if advance warning or duration information is supplied
by the utility; and whether the area affected by the outage
i s localized or widespread. Finally, the impact of an outage
i s partially dependent o n the attitude and preparedness of
customers, which in t u r n is related t o existing reliability
levels.
Basic Evaluation Approaches
Many of the approaches devised t o evaluate interruption
costs can be broadly categorized as indirect analytical
methods which infer interruption cost values from associated indices or variables. Examples of such substitution
or proxy approaches include the following:

i) Electrical supply rates or tariffs are used t o derive


value of service reliability estimates [5]. The minim u m estimate of customers willingness t o pay i s
based o n electrical rate structures and the maximum
is based o n cost of standby plant.
i i) The value of foregone production i s determined taking the ratio of the annual gross national product t o
the total electrical consumption ($/kWh) and ascribing it t o the value of service reliability [6]. A similar
value-added approach has been used t o evolve an
analytical model which, w i t h appropriate adjustments, was applicable t o different customer categories [7]. The approach made use of detailed and
specific data (sales data, value-added data, employee
data) and numerous assumptions and derivations of
average consumption, price, and price elasticity.
iii) The value of foregone leisure time based o n customers wage rates has been used in several residential interruption cost assessments. This is based
o n the notion that consumers can and d o make optim u m laborlleisure time tradeoffs so that marginal
values of leisure and earnings are equal. Some derivations are based o n estimates, mini-surveys or discussions, and worked examples, and are presumed
t o include actual losses, household activities and leisure time [8].Others make simplifying assumptions
and base their results principallyon lost leisure time

~91.
iv) The hourly depreciation rates of all electrical household appliances unavailable because of an outage
have been used as the basis of residential outage
costs [IO].
The advantages of these and other similar methods i s that
they are reasonably straightforward t o apply, make use of
readily available data, and consequently are inexpensive t o
implement. Their disadvantages are that most are based o n
numerous and severely limiting assumptions. Most generate global rather than specific results and consequently
d o not reveal variations in cost with specific parameters as
required by the utilities. Therefore, the usefulness of the

WACKER AND BILLINTON: CUSTOMER COST OF ELECTRIC SERVICE INTERRUPTIONS

921

results t o the utilities for planning purposes i s significantly


reduced.
A second category of outage cost assessment i s t o conduct an after-the-fact case study of a particular outage. This
approach has been limited t o major, large-scale blackouts
such as the 1977 New York blackout [Ill. The study
attempted t o assess both direct and indirect short-term
costs. Direct costs included food spoilage, wage loss, loss
of sales, loss of taxes, etc. Indirect costs included emergency costs, losses due t o civil disorder (looting, rioting,
and arson), and losses of governments and insurance companies resulting from social disorder. The study also considered a wide rangeof societal and organizational impacts.
Such impacts are significant b u t difficult t o evaluate i n monetary terms. While specific data obtained were based o n
assumptions and were incomplete in many respects, some
important conclusions resulted. I n particular, the results
indicated that the indirect costswere much higher than the
direct costs.
The third methodological approach that has been used
t o assess direct, short-term customer interruption costs i s
that of customer surveys [12]-[16J W i t h this method, customers are asked t o estimate their costs o r losses due to
supply outages of varying duration and frequency, and at
different times of the day and year. The strength of this
method lies in the fact that the customer is probably in the
best position t o assess the losses. Direct costs are relatively
easy t o determine for some customer categories (e.g.,
industrial), but users opinions are particularly important
i n assessing less tangible losses, such as inconvenience, for
other categories (e.g., residential). Another advantage is that
the method can readily be tailored to seek particular information as related to the specific needs of the utility.
Obviously, this method i s beset with all the problems of
questionnaire surveys, and the cost and effort of undertaking surveys is significantly higher than using other
approaches outlined earlier. Nevertheless, this approach
appears to be the method favored by utilitieswhich require
outage cost data for planning purposes. Therefore, the
remainder of this paper deals w i t h the survey approach and
outlines the authors experiences and results derived from
its applications.

C O ~ OF
T INTERRUPTION
SUKVEY~
Cost of interruption surveys are usually undertaken with
specific objectives in mind, such as system expansion/
upgrading or major rate revision. Typically, the customer
pool is broken d o w n into appropriate major customer categories or sectors, such as residential, industrial, commercial, agricultural, etc., so that category-specific survey
instruments can be used. The Standard Industrial Classification (SIC) system of customer identification i s commonly utilized because of its wide general acceptance by
industry and government, and often it has already been
adopted by the utility for other reasons. Development of
survey instruments for each of the customer sectors i s a
major and important step in the process. Questionnaire
preparation and the attendant survey procedures require
an understanding of the many difficulties which can be
encountered in conducting surveys, such as representative
sample selection, questionnaire bias, non-response bias,
compromising questionnaire content w i t h length t o ensure

922

satisfactory response rates, etc. Additionally, the nature and


approach of the survey instrument should reflect a sound
theoretical basis and a clear statement of objectives. A n u m ber of these aspects, both practical and theoretical, are discussed below.
An early decision t o be made i s whether a mail or telephone survey i s t o be conducted. Despite numerous benefits of telephone surveys, interruption cost questionnaires
tend t o have considerable detail and require calculations
and respondent reflection and therefore are not particularly suited for telephone surveys. Another decision i s
whether the questionnaire should ask respondents t o consider actual past interruptions or hypothetical future interruptions. The theoretical distinction i s that the first provides hard data while the second i s based o n perception.
It i s the authors experience that, unless the interruption
was a recent occurrence or especially disruptive o r costly,
respondents tend not t o remember actual details but
respond from their recollection and perception of what
the situation probably was like. In effect, there i s little difference between this reconstruction and their response
concerning a future hypothetical interruption. I n essence,
their perception concerning future interruptions i s based
o n their experiences with past interruptions. Furthermore,
use of actual past interruptions makes it impossible t o vary
the interruption scenario, i.e., duration, time of occurrence, etc. Consequently, the actual experience approach
i s not particularly feasible unless a particular interruption
was t o be studied. This would require that the questionnaire be prepared i n advance so that it could be mailed a
few days after the interruption. A limitation c o m m o n t o any
approach is that long or frequent interruptions have not
been experienced by most respondents so the consequences of these extreme situations are u n k n o w n t o them
and subject t o their conjecture.
Cost Valuation Methods
Perhaps the most important questionnaire design consideration is thechoiceof interruption cost valuation methodology, since determining this cost is the primaryobjective of the survey. I t i s i n this regard that the greatest
variations in approach exist. If one accepts that indirect analytical methods are inadequate and that the customer is the
best source of the desired information as discussed earlier,
the problem still remains: I n what manner is the information solicited from the customer? There appears t o be
concurrence that some methods are more suitable than
othersfor particular sectors, b u t there i s n o universal agreement as to what those methods are. A brief discussion of
the various methods and their apparent suitability follows.
The most obvious approach i s a direct solicitation of the
customers interruption costs for given outage conditions.
Guidance can be offered as t o what should and should not
be included i n the cost estimate so that the meaning of the
result i s not ambiguous. This approach provides reasonable and consistent results i n those situations where most
losses tend to be tangible, directly identifiable and quantifiable. Independent researchers have derived valuations
which are reasonably similar i n magnitude [14], [16]. The
approach i s applicable for the industrial sector, most large
users, and for the commercial sector (retail trades and services). It has also been used for large institutions and office

PROCEEDINGS OF THE IEEE, VOL. 77, NO. 6, JUNE 1989

buildings [12]. Its major weakness lies in those areas where


the impacts tend to be less tangible and the monetary loss
is not directly identifiable.
Another approach i s to ask respondentswhat theywould
be willing to pay to avoid having the interruption, or conversely what amount they would be willing t o accept for
having to experience the outage. The basis of this approach
isthat incremental willingnessto pay(wil1ingnesstoaccept)
constitutes a valuation of corresponding marginal increments (decrements) in reliability. Theory would suggest that
incremental willingness to pay amounts should be nearly
equal towil1ingness to acceptva1uations. However, actual
valuations consistently yield willingness to pay values significantly less than willingness to accept values. This result
is believed to support the earlier argument that electric service and i t s reliability do not perform as normal markets,
though other factors may be a t work. Nevertheless, valuations based on willingness t o pay and accept are worthwhile measures, possibly as outside bounds if the limitations are recognized.
A third approach to be discussed is that of indirect worth
evaluation. If direct valuation i s not possible, customerselected alternatives or responses t o indirect method questions may be used to derive a value. The intent is to devise
a suitable approach so as t o decrease the problems associated with rate-related antagonism and the lack of experience in rating the worth of reliability. This is achieved by
asking questionswhich the respondentscan relate to in the
context of their experience. A limitation is the possibility
that the derived value i s not an estimate of the worth but
some other entity associated with the indirect approach.
Possible question forms that have been used or considered
include the following:

i) Cost of hypothetical insurance policies t o compensate for possible interruption effects, and the appropriate compensation payable in the event of an interruption claim
i i) Respondents opinions as to the appropriate interruption cost figures utilities should use in planning
iii) Respondents predictions of what preparatory
actions they might take in the event of recurring
interruptions
iv) Respondents selection of interruptible or curtailable options with reduced rates, which are, in effect,
self predictions of willingness to accept decreased
rates for reductions in reliability
V) Respondents rank ordering of a set of reliabilitylrate
alternatives and choosingan option that is most suitable to their needs
It should be noted that several of theoptions listed above
use a form of substitution either in services or in monetary
terms. While the substitution concept i s similar to that discussed earlier in this paper, the difference here i s that the
substitution is reasonablydirect and, more importantly, the
selection is being made by the customer rather than by the
analyst. A matter of concern for most of the approaches
cited above is the question: How closely would customers
actions match with their prior prediction of their actions?
Put another way, how valid is the customers perception?
This issue was discussed brieflyearlier in this paper, but the
question remains. Perhaps the strongest rejoinder to this
issue i s that it i s the customers perception that i s sought,

and that there are markets where customers selections are


based more on their perception than o n factual evidence.
Anasideatthis juncture isthat mostoftheapproachescited
above attempt to establish a market or at least an inferred
market for reliability. It i s believed that the market
response for small variations around the current reliability
value are reasonable, though obviously it can only be as
accurate as the particular substitution can accomplish.
Attempting considerable variations from the current reliability level, however, may not yield meaningful results,
mainly because service reliability may not respond as a true
market as discussed earlier. Additionally, the customers
perception i s doubtful in extreme situations because of lack
of experience, and the useful range for most of the substitutions i s questionable. Consequently, the authors would
caution against a rigorous application of market-based
methods such as a consumer surplus approach [4] in such
instances.
The authors have made considerable use of the preparatory action approach when a direct evaluation was not
suitable, and believe that it secures reasonable results. The
approach has been used in major Canadian residential and
agricultural surveys [17], [18]. With this approach, the
respondent is presented with a list of actions that one might
conceivably take in preparation against recurring interruptions. A reasonable cost figure for purchase and application
of each action is assigned and included in the list. The list
ranges from making no preparations through t o buying a
self-starting standby generator capable of supplying the
entire load. Respondents are then asked to indicate what
action or actions they would take for different failure scenarios. During analysis, the cost(s) of the chosen action(s) are
used as an estimate of the expenditure respondents are
willing to undertake o n their own behalf so as to prevent
or nullifythefull effectsofthe interruption.This represents
an indirect estimate of reliability worth in that the derived
expenditures are considered to be the users perception of
the value of avoiding the interruption consequences.
Respondent interviews during questionnaire development
are essential t o ensure that respondents accept the overall
approach and that they consider the choice of actions adequate and quoted costs reasonable.
Questionnaire Content and Survey Procedures
It i s possible and perhaps desirable to include more than
one method of interruption cost valuation in a survey. In
addition, the questionnaire should seek interruption cost
variations as functions of interruption characteristics to the
extent possible and consistent with the particular objectives of the study. Other information such as customer
demographics, principal uses of electric supply, availability
and nature of standby, and the possibility of creating hazardous situations due t o interruptions might be included.
Finally, customer energy and demand information must be
secured from utility files, and actual interruption statistics
are required if cost estimates based o n unserved energyare
being sought.
Sample size identification and sample selection are major
considerations in conducting a survey. Expected response
rates and intended breakdown of the sector in question
intoSICsubgroups,aswelI asanygeographical distinctions
to be made, must be included in decisions regarding sam-

WACKER AND BILLINTON: CUSTOMER COST OF ELECTRIC SERVICE INTERRUPTIONS

923

ple sizes. The objective of sample size and sample selection


'decisions i s to secure representative and statistically meaningful responses in all SIC categories and geographical or
regional divisions t o the extent possible. Generally, fixed
developmental and other costs of conducting a survey tend
to be high, while incremental costs per customer surveyed
are relatively low if reasonably large samples are used.
Practical problems that arise include multiple service
accounts for a given customer, uncertainty of customer
classification by SIC, branch (retail)outlets whose activities
are managed by a regional office, and the identification and
follow-up of customer responses or consumption data
which appear to be extreme.
EXAMPLES

OF

DETAILED
QUESTIONNAIRE DESIGN

Thus far, this paper has attempted to present a broad general review of thevarious approaches which have been used
to evaluate interruption costs along with the authors'commentary regarding the ability of the methods to deliver
results which are meaningful to the utility. I n this section,
summaries of various surveys conducted by the University
of Saskatchewan under the authors' direction will be presented. The intent i s t o illustrate methodology, the nature
of the details involved, and the type and order of magnitude
of the results obtained. Finally, the use of outage cost information to derive a composite customer damage function
i s outlined.
The studies discussed below were conducted by the
Power Systems Research Group at the University of Saskatchewan under contract with the Canadian Electrical
Association (CEA) [17], [18]. Residential, commercial (retail),
small industrial, and large user surveys were conducted in
1980and an agricultural survey was performed in 1985. The
surveys involved the participation of thirteen service areas
across Canada. A summary of sample size, effective
response rates and the number of SIC categories surveyed
by sector are shown in Table 1. The residential and agriTable 1

Survey Information
Sector
~

Total sample
Usable responses
Response rate
No. of utilities
No. of SICS

13359
4740
37.0%
10

3624
1001
30.7%
2
55

2311
425
24.0%
3
25

16

15
94%

16470
6020
36.6%
10
26

Headingcodes: R-Residential, C-Commercial, I-Industrial, L-Large User,


A-Agricultural.

cultural surveys were most extensive and provided reasonable Canada-wide coverage. The commercial and small
industrial surveys were restricted to the Canadian prairies
and some maritime areas. In the large user category, only
a small Saskatchewan-based survey was conducted. Users
were grouped according to Statistics Canada's Standard
Industrial Classification (SIC) categories and major group
categories. Samples were chosen by selecting a quota of
users from each SIC category, with the quota size related
to the number of users in that category, geography and
other factors. In categories with less than 25 users, all users
were included to obtain as large a response as possible. The

objective was to obtain meaningful resultsfor each SIC, but


with quota sizes broadly representative of the makeup of
customers within the survey area.
Thequestionnaires used i n the surveys underwent extensive development. This involved an iterative approach consisting of the identificationof factors t o be included, design
and development of the questionnaires, and small-scale
testingwhich included interviews with sample users. Questions developed by other researchers were used unchanged
or adapted. Consultants skilled in questionnaire techniques were used throughout.
Only summary information i s presented i n this paper.
Full reports of the results [17], [18], expanded summaries
[13]-[15], and related factors [19]-[21] have been published
elsewhere.
Questionnaire Content and Approach
While itwould bedesirable to investigateall possiblefactors which might affect the cost of interruptions, the length
of a questionnaire i s limited by the degree of effort which
respondents are willing to engage in. This limitation i s particularly relevant to the residential sector where a significant portion of the cost i s related to less tangible impacts.
A brief description of the sector questionnaires is presented below. This indicates the general approach used and
the factors finally selected for inclusion in each of the questionnaires.
The residential questionnaire used attitudinal, power
rationing preference, electric heat dependence, and experience with past interruption questions to set the stage.
These were followed by two sets of questions which request
qualitative assessments using a relative scale of undesirability to describe the severity of interruption impact. One
set of questions addressed a rangeof residential household
activities and end uses. The other posed hypothetical interruption scenarios, varying one interruption characteristic
at a time, with a "four hour monthly failure after 4:OO p.m.
on winter weekdays" serving as the "base" scenario. Variations with duration, frequency, and time of occurrence
(day, week, season) were requested. Quantitative (monetary) evaluations were obtained by means of an indirect
worth assessment and two rate change questions. The indirect worth approach requested respondents t o predict the
preparatory actions they would take in the event of particular anticipated recurring interruptions. Respondents were
asked t o choose one or more of the following six actions:
make no preparation; purchase and use a candle at $0.25
per hour, an emergency lantern at $0.50 per hour, an emergency stove at $1.50 per hour; purchase or rent and use a
small generator at $5.00 per hour, or a larger generator at
$20.00 per hour. During analysis, the costs of the chosen
action($ were summed to provide the "preparatory action"
cost estimates. Two rate change questions (one willingness
to pay and one willingness t o accept) sought respondents'
opinions concerning electrical rate adjustments appropriate for particular changes in reliability. One question suggested that the normal electric supply had become subject
to a specific reduced reliability level and asked what premium customers would be willing to pay for an alternative
assured supply if one was available. The second asked
respondents t o indicate the minimum reduction i n rates for
them tochoose a specific reduced reliability. Demographic
information such as sex and education of the respondent,

PROCEEDINGS OF THE IEEE, VOL. 77, NO. 6, JUNE 1989

924

number and ageof household members, and typeof dwelling was solicited, and users were asked whether a business
was operated from the residence.
The agricultural questionnaire followed an approach
similar to the residential questionnaire because it was realized that, for many types of farming activity, the household
impact of an outage might be as important as the impact
on the farming activity. In any event, the survey was
intended to evaluate both components. The introductory
questions were essentially identical, except that the range
of end useoptionswasexpanded t o includefarming related
categories. Qualitative assessments regarding the series of
interruption scenarios was deleted. The quantitative
assessment using preparatory actions was retained, and a
third generator was added t o the list of possible options
with the size and cost of the generators adjusted t o properly
relate to the farming situation. Increased severity outage
scenarios were added to be more consistent with rural
experience in some service areas. Respondents were asked
toindicatewhich month or monthsandwhich timeortimes
of day cause them the most inconvenience and financial
loss. All months the same and All 24 hours the same
were options. A subsequent question sought variation with
season. Users were requested to indicate whether interruptions primarily affect their farming operations as
opposed to their household activities or vice versa. Those
with primarily farming consequences were asked t o provide a direct evaluation of worst-case costs at the worst
time(s) and month(s) indicated earlier. These costs were
requested under various headings such as damage to stored
farm goods, loss of production, loss of livestock, etc., and
for various interruption durations. A third and final cost
evaluation used a rate change question. Respondents were
asked to indicate the minimum percentage decrease in their
rates for them to choose a specified reduced reliability
rather than the present supply. A question on the availability, size, type, and purpose of a backup supply was
included. Demographic information, questions to ensure
proper customer categorization by SIC, and permission t o
request consumption information from their utiity completed the questionnaire.
The commercial, industrial, and large user questionnaires attempted to qualitatively assess user dependence
on electrical supply according to end use. Power rationing
preference was requested. Quantitative assessment was
achieved using the d irect-worth evaluation approach.
Respondents were requested to estimate the costs t o their
company for various interruption scenarios. The interruptions were to occur without warning o n a Fridayat 1O:OO a.m.
near the end of January. Commercial respondents were told
to include lost business or sales, wages paid to staff who
were unable to work, equipment or goods damaged, etc.,
but not to include sales or business that could be made u p
after the interruption ceased. Small industrial and large
users were instructed to include plant and equipment damage, raw material and finished product spoilageor damage,
and the cost of special procedures to restart production
(e.g., extra cleanup, maintenance, checkups, etc.). Production lost during the failure and restart time was to be evaluated as the estimated revenue (sales price) of product not
made less the expenses saved in labor materials, utilities,
etc. If production could be made u p later during slack time
or overtime, that portion was not to be included. Other costs

such as the cost of operating standby equipment or of special procedures t o prevent damage could be listed as well.
Availability, size, and purpose of standby were t o be identified. An innovative method t o obtain cost estimate variations with time of day, day of the week, and month of the
year was developed. The approach makes use of Friday a t
1O:OO a.m. near the end of January as the base case for
which the respondent provides detailed dollar value cost
estimates. A tabular format in subsequent questions enables the respondent to readily provide comprehensive differences in costs for other situations relative to the base
case. Users were asked to indicate the possibility and
amount of cost saving that could be effected if advance
warning or interruption duration was provided. Demographic information o n the nature and size of the companys operation was requested: number of employees,
shifts, sales volume, etc.

COMPILATION
OF COST-OF-INTERRUPTION
DATA
Factors or variables which affect the composite customer
damage function can be broadly classified as customer
related or interruption related. Collection and analysis of
outage costs within customer sectors and subgroups (SICS)
tacitly assumes that the user group i s a primary customerrelated variable. This i s based o n the assumption that, as
the customer category becomes more homogeneous, there
should be less cost variation within the groups. Similarly,
duration and frequency of interruptions are inherently
accepted as principal interruption-related variables. Consequently, customer losses due to power interruptions have
typically been collected, compiled, and reported for various customer sectors and sector subcategories as functions of interruption duration and frequency. Analyses of
costsasfunctionsof theseand other interruption-and userrelated characteristics have been the subject of considerable ongoing activity with a view to increased understanding of customer costs.
At best, results of customer surveys provide an accurate
reflection of customers actual or perceived costs associated with electric power supply interruptions. However,
simple average (or median) values of users interruption
costs may not properly represent the cost incurred by that
user group. This is because a few extreme values can contribute inappropriately to the average cost per interruption.
More importantly, average costs per interruption are difficult if not impossible to use for utility planning purposes
since most relevant planning criteria and calculations are
based o n either demand or consumption or both. Average
values (even if they are meaningful) are therefore of little
value. Consequently, customer reported costs are usually
normalized with respect to the customers annual energy
consumption ($/kWh) or annual peak demand ($/kW).
Such normalization presents a number of problems. One
of these i s that normalization inherently gives credence to
the notion that maximum costs occur in coincidence with
maximum power, whereas in fact, worst case or maximum costs may be more appropriately related to the time
of day, season, activities interrupted, etc. Another major
difficulty related to the normalization procedure i s the lack
of load factor information for individual users. Actual data
are available only for large customers. Averages of these
values, by customer category, are often calculated and

WACKER AND BILLINTON: CUSTOMER COST OF ELECTRIC SERVICE INTERRUPTIONS

925

applied to all customers in that category regardless of their


size. A final dilemma, though perhaps the most important
one, i s that normalized $/kWh values are often misunderstood t o be the costs of energy not served, which i s simply
not the case. This would be true if it could be assumed that
the interruption occurred at the time of peak demand and
that this demand would have been sustained for the duration of interruption. There are numerous instances i n the
literature where interruption costs normalized as described
above have been inadvertently (and incorrectly) quoted and
used by others as being the cost of energy unserved. Yet
other reports of interruption costs make n o distinction and
the meaning of their results are uncertain. Clearly, the
implications of this distinction appear not to be fully appreciated by all researchers. To derive cost of unserved energy
from normalized cost values requires knowledgeand application of time-of-day load curves and frequency and duration distributions of time of occurrence of outages. The
question as t o which type of costs are appropriate i s dependent o n the application. For many planning applications,
calculations and decisions are based on a normalized power
or energy basis. I n other instances, for example if global
costs vs. monetary losses are required, the cost of energy
unserved i s appropriate.

1980 Canadian dollars except for agricultural costs which


are in 1985 dollars. (In real applications, costs would have
t o be updated using appropriate economic indices to reflect
current values.)
Table 2 lists the average interruption cost estimates
obtained as a function of interruption duration for each
Table 2 Interruption Cost Estimates (Cost Estimates in
1980 Canadian Dollars)

Interruption

Residential

Commercial

Industrial

Large
User

21
131
340
919
3418

2748
6185
11 385
19241
42 259

30812
37308
47976
101 125
161 098

Cost per Interruption ($)


1 min

20 min
I h
4h
8h

0.22
1.18
11.87

Cost per Interruption per Annual Energy Consumption ($/kWh)


1 min
20 min
I h
4h
8h

0.000028
0.000156
0.001566

0.000106
0.000707
0.002046
0.007533
0.019523

0.000215
0.000862
0.001830
0.005179
0.009956

0.000538
0.000881
0.001758
0.003356
0.005966

Cost per Interruption per Annual Peak Demand ($/kW)


Example of Cost Data Compilation
The compilation of data from the University of Saskatchewan surveys [17], [I81 is used t o illustrate the above process. The survey responses were subjected t o standard statistical analyses which included calculation of mean values,
standard derivations, and correlations among variables
(both univariate and multivariate) and their F-test significance levels. Such analyses were conducted for each customer category for the entire sample, for Canadian geographical regions, for each service area, and, i n some
instances, for regions within service areas. The objectives
of these analyses were t o obtain cost data and t o identify
variations, correlations and interrelationships of costs with
other variables; in short, to begin t o understand customer
interruption costs. The objective of this paper, however, is
not so much t o present or describe these variations, but
rather t o observe the nature of the costs and outline their
application and limitations i n describing customer damage
functions.
In an actual application, the analysis would establish the
service area in question and the composition of the load
within sectors and among them. The application might be
the entire service area or some portion of it as seen from
a particular generation location, a point within the transmission network, a bulk power load point, or a location
within thedistribution system. Inanycase,the relevantcost
data and composition of the load served from the point i n
question would be necessary t o create the applicable composite damage function. To illustrate this approach, acompilation of cost data i s presented i n this section of the paper
and i s used in the next section togeneratea n u m b e r o f c o m posite customer damage functions. The compilation of data
i s not entirely a consistent set, and presentation of the data
illustrates some of the practical aspects and difficulties i n
data collection. Most of the costs presented are representative of the entire Canadian scene as derived from surveys
conducted by the authors [17], [18]. Costs reported are i n

926

1 min
20 min
I h
4h
8h

0.06

0.31
3.16

0.28
2.05
5.88
21.51
63.06

0.70
2.88
5.19
13.87
27.60

1.80
2.22
3.19
6.89
10.47

Agricultural Cost Estimates in 1985 Canadian Dollars


20 min
I h
4h
8h

$/kWh

$/kW

1.58
7.98
66.02
185.55

0.000032
0.00016
0.00135
0.00378

0.068
0.34
2.82
7.88

customer sector. The table lists average values for costs per
interruption ($), costs per interruption normalized with
respect to the users annual energy consumption ($/kWh),
and costs per interruption normalized w i t h respect t o the
users annual peak demand ($/kW). Annual energy consumption (kWh) for each customer was collected from participating utilities, and so was annual peak demand (kW)
wherever available. Where demand readings were unavailable, k n o w n or estimated load factors for each SIC category
were used i n conjunction w i t h the users annual energy
consumption in determining$/kWvalues.An assumed load
factor of 23 percent was used for residential users. Demand
values provided as peak kVA were assumed t o be peak k W
in theabsenceof reliable power factor data. Costs were estimated using the direct evaluation approach in the commercial, industrial, and large user sectors, while the costs
i n the residential and agricultural sectors are based o n the
indirect preparatoryaction method. I t is believed that these
are theoretically sound and suffer less from rate-related
antagonism than the rate change questions. Costs listed for
each sector are obtained by combining costs of i t s constituent SIC categories, appropriately weighted, so the
result is representative of a particular service area. This
aggregated weighting was achieved by summing all the
component SIC groups dollar costs and dividing this total

PROCEEDINGS OF THE I E E E , VOL. 77, NO. 6 , JUNE 1989

cost by the total of the peak demands to obtain $/kW costs


and bythetotal oftheannual consumptionstoobtain $ / k w h
costs. Only respondents for which both the cost estimates
and the consumption figures were available were included
in the calculations. The commercial and industrial sector
cost estimates are based on a user composition similar to
that in the Manitoba Hydro service area. Agricultural estimates are representative of the entire Canadian sample
using weights from Statistics Canadas annual users costs
of electrical service by SIC in lieu of their respective energy
consumptions. Residential and large user responses were
notweighted. Residential results represent theentirecanadian sample considered as a single unit, while large user
costs are representative for the SaskPower service area. I n
addition to the listing of Table 2, the $/kW estimates for the
various sectors are presented graphically in Fig. 1. These

mates. There i s considerable variation of costs among SIC


groups ineach sector andwithin SICgroups. However,variations within groups are considerably less than between
groups. Analysis of variance has consistently indicated that
the SIC group i s the customer variable which correlates
most stronglywith cost estimates. The primary interruption
variables include duration and frequency of interruptions.
Another variable with observed correlations is the number
of interruptions experienced by the user. This variable i s
perhaps both interruption and user related, and might be
considered as an indication that interruption impacts (or
users perception of them) are a function of reliability level.
Large variations, including the significant number of
respondents who indicated negligible costs, must be kept
in mind when thesevaluesareused togenerateacomposite
customer damage function or for any other purpose.
GENERATING

1 mtn

2 0 min

1 hour

4 hrs 8 hrs

Interruption Duration
Fig. 1. Interruption cost estimates for various sectors.

cost functions are often called sector customer damage


functions.
Although sector cost functions are derived from an
appropriate weighting of constituent SIC groups costs as
described above, they can also be broken down into various
factors to the extent possible from the original data. For
example, residential costs vary by type of dwelling, i.e., single home, apartment, or mobile home [21]. Similarly, commercial and industrial factors include availability and type
of standby, and variations with seasons and day of theweek
[19], [20]. The usefulness of such distinctions depends o n
how well the customer composition and characteristics are
known in the service area under investigation and the
degree of finesse desired in the derived sector costs.
The values shown in Table 2 and Fig. 1 are mean cost esti-

CUSTOMER DAMAGE
FUNCTION

Conceptually, the generation of a composite customer


damage function (CCDF) for a particular service area i s an
attempt to define the total customer costs for that area as
a function of interruption duration. The customer mix for
the area must be known so that the costs for the various
customer categories can be proportionally weighted to their
respective energyconsumptions within the area. Some utility planners suggest that for short interruptions (say, less
than one hour), weighing by peak demand is more appropriate since losses are related t o a power shortage rather
than an energy shortage [22]. Weighted costs are summed
to provide the total cost for the area for each duration.
A number of example applications are presented below
to illustrate the derivation of composite customer damage
functions and the nature of the results. The first requirement in determining a CCDF i s t o prepare a listing of sector,
SIC, or subcategory customer damage functions which are
deemed applicable to the service area under study. These
may beobtainedfromacustomersurveyintheareaof interest, or from another utility or other source which has
obtained cost data from customerswith characteristics similar to those of the target service area. It may be necessary
to update costs t o account for inflation, and other assumptions and approximations may need to be made. Table 3
presents such a listing of costdata.Thesecond requirement
in establishing a CCDF i s t o know the customer composition of the service area in question, which i s obviously
dependent o n the particular application. To illustrate the
approach, composite customer damage functions are
determined for several hypothetical but exemplary situations. The costs presented in Table 3 are assumed to apply
for all examples, while the load compositions and the general location of the application within the system are outlined below. The calculation of the CCDF is accomplished
by acomputer program developed for the purpose [23]. The
resulting composite customer damage functions are presented in Table 4.
The first example considered is a situation in which several city blocks have only apartment buildings o n them and
are served from a common supply point. Most medium and
larger sized cities have such situations. The interruption
cost data of Table 3 can be applied directly to these service
areas as shown in entry 1 of Table 4. The next example is
again a residential service area but with a mix of types of

WACKER ANL) BILLINTON: CUSTOMER COST OF ELECTRIC SERVICE INTERRUPTIONS

927

Table 3 Typical Cost-of-Interruption Data Base Entries (Consumption-Normalized


Costs in 1985 Canadian Dollars per kWh)
Interruption Duration
User Type

1 min

20 min

I h

4h

8h

Apartment
Residential
Commercial
Agricu Itural
Industrial
Large User

0.0000000
0.0000000
0.0001509
0.0000000
0.0003193
0.0007703

0.0000672
0.0000486
0.0012561
0.0000296
0.0012393
0.0012603

0.0003949
0.0002622
0.0029216
0.0001569
0.0026178
0.0025154

0.0035484
0.0025639
0.0107603
0.0014417
0.0074116
0.0048015

0.0106594
0.0076696
0.0275177
0.0042685
0.0142708
0.0085364

residential users. Suppose that the mix of residential customers according t o type of residence i s similar t o the Canadian distribution, then the residential data shown in Table
3 i s directly applicable. These values are therefore listed as
entry 2 in Table 4. These data might apply to a service area
similar in size t o that of the first example, or they might be
more applicable t o a bulk distribution load point serving
a much larger service area.
Townsand small-to medium-sizedcitiescanoften be represented bya mix of commercial enterprises and residential
customers. Such a service area is now considered t o illustrate how interruption costs are affected by the presence
of the commercial sector. Suppose that the composition of
the service area i n question, perhaps a particular portion
of a small city, is known to be approximately one-half commercial and one-half residential by electrical energy consumption. Using the cost data for the residential and commercial sectors shown i n Table 3, the weighted damage
function shown as entry 3 in Table 4 i s the result. (This
assumes that the weightings used in generating the data
base costs adequately represent the residential and commercial customer compositions in the area under consideration.) Notice that the composite costs are significantly
higher than the wholly residential situation of the previous
example.
The next example illustrates a residential, commercial,
and agricultural mix of customers, each assumed to represent one-third of the energy requirement. This might be
the situation with a small town located in a farming community, and the customer damage function i s desired at a
common bulk distribution load point. The equal mix of residential and commercial customers of the town can be represented by the cost function of the previous example.
Combining this function with the agricultural sector costs

shown in Table 3 yields the composite cost function shown


i n entry 4 in Table 4.
A large city is the object of the next example. The composition of the service area i s assumed to be equally represented by the residential, commercial, and small industrial sectors. The cost data for these three sectors as listed
in Table 3 are assumed to be applicable. These entries differ
from those i n the second group of Table2 because the data
i n Table 2 are, as noted, largely representative of the Canadian scene while the data in Table 3 are from a given example. The resulting composite damage function is listed as
entry 5 in Table 4. A larger city similar to the preceding one
but with a large user component i s considered next and
yieldstheresultshownasentry6inTable4. Eachofthefour
sectors are assumed to account for one-quarter of the electrical energy consumption.
The final example illustrates a five-part composite of agricultural, residential, commercial, small industrial, and large
user, each accounting for20 percent of the electrical energy
consumption. This might be the load as seen from a bulk
transmission load point in the system serving this composite load. Entry 7 in Table 4 l i s t s the resultant CCDF for
this example.
The examples cited above were created to demonstrate
the approach used in generating a composite customer
damage function and to illustrate the order of magnitude
of expected costs as a function of interruption duration.
The examples also indicate that a CCDF can be generated
at many points within the electric power network depending o n the intended application. The examples used rather
simplistic situations t o reveal the conceptual framework of
the method; real situations would tend to be more complex.
Similarly, obtaining service area customer composition for
the creation of a CCDF may present a significant problem

Table 4 Composite Customer Damage Functions for Example Service Areas


(Consumption-Normalized Costs in 1985 Canadian Dollars per kWh)
Interruption Duration

928

Example Service Area

1 min

20 min

I h

4h

8h

1) Apartment buildings
2) Residential
subdivision
3) Small- to medium-size
city
4) Town and surrounding
farms
5) Large city
6) Larger city
7) Large city and
surrounding farms

0.0000000

0.0000672

0.0003949

0.0035484

0.0106594

0.0000000

0.0000486

0.0002622

0.0025639

0.0076696

0.0000754

0.0006524

0.0015919

0.0066621

0.0175936

0.0000503
0.0001567
0.0003101

0.0004448
0.0008480
0.0009511

0.0011136
0.0019339
0.0020792

0.0049220
0.0069119
0.0063843

0.0131519
0.0164860
0.0144986

0.0002481

0.0007668

0.0016948

0.0053958

0.0124526

PROCEEDINGS OF THE IEEE, VOL. 77, NO. 6 , JUNE 1989

in many situations. Finally, no attempt has been made to


describe the specific applications of t h e damage function
in t h e various situations.
APPLICATIONS AND LIMITATIONS
The primary applications o f a composite customer damage function are related to utilities' planning and operating
activities. The underlying principle i s to consider t h e aggregated cost o f interruptions (customer damage function) as
a measure, perhaps a lower bound, o f t h e benefit o f uninterrupted supply, i.e., t h e w o r t h of reliability. Hence, t h e
inclusion of reliabilityworth a s o n e o f t h e parameters in any
related costlbenefit assessment is possible. For example,
when considering system generating capacity evaluation,
t h e magnitude, frequency and duration of t h e expected
negative margin states in t h e system's capacity margin
model can b e used with t h e CCDF for t h e entire service area
to generate an interrupted energy assessment rate [24]. The
system's annual total expected unserved energy and its
related expected interruption cost are useful i n evaluating
reliability w o r t h for t h e system. Or, w h e n system planning
o r expansion situations call for n e w o r modified generation
o r transmission equipment, t h e w o r t h o f reliability can b e
included in rationalizing t h e decisions and justifying t h e
cost o f physical plant [22]. Similarly, w h e n considering various transmission o r distribution alternatives in a system
renovation program, t h e costs of providing service at various levels of reliability as a function of system configuration can b e weighted against t h e reliability w o r t h defined
bytheapplication ofthecustomerdamagefunction. In such
circumstances t h e customer mix will essentially b e k n o w n
as a function o f geographical location and various system
configurations in order to apply interruption costs along
w i t h configuration-specific failure statistics.
Reliability w o r t h also has applications in t h e adoption o f
operating policies and strategies. It might b e included, for
example, as o n e o f t h e criteria in prescribing t h e sequence
of adding o r d r o p p i n g generating units f r o m the system,
o r in devising maintenance policies o r schedules. An
obvious application i s in devising o r modifying load shedding or load restoration sequences, operating policies and
emergency strategies. In essence, since a primary objective
o f a modern power system i s to supply customer loads as
reliably and economically as possible, t h e inclusion o f reliability w o r t h along w i t h other criteria i s justifiable, particularly since i t i s easily recognized as a user-observable
parameter.
The major limitations in t h e application o f t h e customer
damage function are based largely on t h e inaccuracies o r
limitationsof t h e function. Principal among these i s t h e limitation o f t h e cost-of-interruption data itself. Limitations
included here are: that t h e estimates are customer predictions o f their losses or responses; that most respondents
have little experience w i t h longer interruptions; that costs
o f interruptions are likely socioeconomic/demographic/
geography specific; etc. The difficulties resulting f r o m variations o f cost estimates have been identified previously,
and so has t h e question of transferability. M o r e specifically,
h o w can o n e k n o w whether t h e data obtained i n another
utility's survey i s applicable to one's o w n customers, and
can these customers b e considered average?The variations
w i t h i n and among customer groups i s largely unexplained

and a subject for further research. Finally, t h e matter o f service area size becomes an issue: it is likely that global o r
large service area customer makeup w i l l b e assumed, w h i l e
the interruption cost (and t h e interruption itself) has a random, local, small-areacharacteristic. Comparisonswith data
obtained by other researchers attest t o these and other limitations o f t h e approach.
REFERENCES
R. Billinton and R. Allan, Reliability Assessment of Large /ectric Power Systems. Norwell, MA: Kluwer Academic Publishers, 1988.
"The value of service reliability to consumers," EPRl EA-4494,
Res. Project 1104-6, Proceedings, May 1986.
R. Billinton, G. Wacker, and E. Wojczynski, "Comprehensive
bibliography on electrical service interruption costs," /E
Trans. Power App. Syst., vol. 102, pp. 1831-1837, 1983.
A. P. Sanghvi, "Economic costs of electricity supply interruptions: US and foreign experience," nergyconomics,July
1982.
M. G . Webb, "The determination of reserve generating
capacity criteria in electricity supply systems," Appl. Economics, vol. 9, pp. 19-31, Mar. 1977.
R. Shipley, A. Patton, and J. Denison, "Power reliability cost
vs worth," I Trans. Power App. Syst., pp. 2204-2212, 1972.
D. Myers, "The economic effects to a metropolitan area of
power outage resulting from an earthquake," Earthquake
Engineering Systems Inc., San Francisco, CA, Feb. 1978.
L. Lundberg, "Report of the group of experts on quality of
service from the customer's points of view," UNIPEDE, Rep.
601D.1, 1972.
M. Munasinghe, "Thecosts incurred by residential electricity
consumers due to power failures,"/. Consumer Res., vol. 6,
1980.
L. Markel, N. Ross, and N.Badertscher, "Analysis of electric
power system reliability," California Energy Resources Conservation and Development Commission, Oct. 1976.
J. Corwin and W. Miles, "Impact assessment of the 1977 New
York City blackout," U.S. Department of Energy, Washington, DC, July 1978.
L. V . Skof, "Ontario Hydro surveys on power system reliability: summary of customer viewpoints," Ontario Hydro
Rep. R&MR 80-12, EPRl Seminar, Oct. 11-13, 1983.
G. Wacker, E. Wojczynski,and R. Billinton,"lnterruptioncost
methodology and results-A Canadian residential survey,"
I Trans. Power App. Syst., vol. 102, pp. 3385-3391, 1983.
E. Wojczynski, R. Billinton,and G. Wacker, "Interruption cost
methodoloevand results-A Canadian commercial and small
industrial survey," / E Trans. Power App. Syst., vol. 103, pp.
437-443, 1984.
G. Wacker and R. Billinton, "Farm losses resulting from electric service interruptions-A Canadian survey," I Trans.
Power Syst., vol. 4, no. 2, pp. 472-478, May 1989.
IEEE Committee, "Report on reliability survey of industrial
plants, part I I , cost of power outages, plant restart time, critical service loss, duration, time and type of loads lost versus
time of power outages," I Trans. PowerApp. Syst., pp. 236241, Mar.lApr. 1974.
R. Billinton, G . Wacker, and E. Wojczynski, "Customer damage resulting from electric service interruptions," Canadian
Electrical Association, R&D Project 907 U 131 Rep., 1982.
G . Wacker, R. Billinton, and R. Brewer, "Farm losses resulting
from electric service interruptions," Canadian Electric Association R&D Research Project 309 U 403, May 1987.
R. K. Subramaniam, R. Billinton, and G. Wacker, "Factors
affecting the development of an industrial customer damage
function," / E Trans. Power App. Syst., vol. 104, pp. 32093215, 1985.
R. Billinton, G. Wacker, and R. K. Subramaniam, "Factors
affecting thedevelopment ofacommercialcustomerdamage
function," / E Trans. PowerSyst.,vol. 1, pp. 28-33, Nov. 1986.
-, "Factors affecting the development of a residential customer damage function," / ETrans. Power Syst., vol. 2, pp.
204-209, Feb. 1987.
"I

151
161

171
[I81
1191

[20]
[21]

WACKER AND BILLINTON: CUSTOMER COST OF ELECTRIC SERVICE INTERRUPTIONS

929

[22] D. C. Coushleff, "Use of interruption costs in regional supply


planning," Ontario Hydro Rep. 80-301-K.
I231 R. Billinton, B. J. Hall, and G. Wacker, "A program and Canadian data base for the determination of individual and composite customer damage functions," presented at the 14th
Inter-ram Conf. forthe Electric Power Industry,Toronto, Canada, May 1987.
[24] R. Billinton and J. Oteng-Adjei, "Utilization of composite customer damage functions in the development of an interrupted energy assessment rate," presented at the 15th Interram Conf.for the Electric Power Industry, Portland, June, 1988.

Carry Wacker (Member, IEEE) received the


BSc. degree in engineering physicsand the
M.Sc. degree in electrical engineering from
the University of Saskatchewan, Saskatchewan, Canada.
He is a Professor of Electrical Engineering
at the University of Saskatchewan. Present
research activities includean ongoing study
of the customer losses or costs resulting
from electric service interruptions, and the
aDplication
of such cost data to power sys. ,
tem planning and operation.

930

Mr. Wacker i s past chairman of the local section of the IEEE, a


member of the American Society for Engineering Education, and
a member and Councillor of the Association of Professional Engineers of Saskatchewan.

Roy Billinton (Fellow, IEEE) received the


B.Sc. and M.Sc. degrees from the Universityof Manitoba, Canada, and the Ph.D. and
D.Sc. degrees from the University of Saskatchewan, Canada.
He worked for Manitoba Hydro i n the
system planning and production divisions.
In 1964, he joined the University of Saskatchewan. Formerly Head of the Department of Electrical Engineering, he i s now
the C. J. MacKenzie Professor of Engineering and Associate Dean, Graduate Studies, Research and Extension
of the College of Engineering. He i s the author or coauthor of five
books on reliability.
Dr. Billinton i s a Fellow of the EIC and the Royal Society of Canada, and a Professional Engineer in the Province of Saskatchewan.

PROCEEDINGS OF THE IEEE, VOL. 77, NO. 6, JUNE 1989

You might also like