Prediction of customer load point service reliability

worth estimates in an electric power system

L. Goel
R. Billinton

Indexing terms: Power system optimisation, Reliability test systems

refers to the value of continuous electrical service. It is

Abstract: The basic concepts associated with more easily estimated and expressed, however, as a func-
quantitative reliability evaluation are reasonably tion of the costs resulting from the lack of such service.
well accepted in the electric power industry. The This implies that the losses and damages resulting from
justification of new facilities and system modifi- interruptions in electrical supply might represent the
cations now normally includes specific reference to amount customers and utility management would be
reliability and one approach that is receiving con- willing to pay to prevent such power outages. The worth
siderable attention is the assessment of the societal of reliability can therefore be expressed in terms of cus-
worth or benefit associated with power system tomer outage costs.
reliability, and conversely the costs to consumers A wide range of procedures have been used to assess
of power supply interruptions. The paper presents reliability worth. One approach which is finding con-
three different methods for evaluating system cus- siderable acceptance uses the cost of electric supply inter-
tomer load point reliability worth factors desig- ruptions incurred by customers due to power failures
nated interrupted energy assessment rates (IEAR). [l, 21. Customer interruption costs can be related to the
These factors consider the influence of outages in calculated reliability indices used in system planning and
all parts of the electric power system. The devel- operation and provide a practical tool for reliability
oped IEAR can be used with the available ade- worth evaluation.
quacy indices to assess the severity associated with The basic objective of a modern electric power system
unsupplied energy owing to supply interruptions. is to satisfy the system load requirements as economically
The IEAR can be used as customer related indices as possible and with a reasonable degree of continuity
in making decisions regarding load curtailment and quality. Power system reliability is usually defined as
philosophies and reliability related rate setting. the ability of the system to provide an adequate supply of
The concepts of using a customer damage func- electrical energy [3]. System reliability can be divided
tion in association with customer reliability into the two distinct categories of system security and
indices are presented and illustrated using a reli- system adequacy [4] in which security relates to dynamic
ability test system. system analysis and adequacy is confined to steady-state
long-term assessment. This paper is confined to adequacy
assessment of an overall electric power system in regard
to measuring the worth of service reliability. Overall ade-
1 Introduction quacy evaluation of an electric power sytem involves a
There is an increasing interest in economic optimisation comprehensive analysis of the three principal functional
approaches in power system planning and expansion zones of generation, transmission and distribution. These
studies. Evaluation of the costs associated with different zones can be combined to obtain three distinct hierarch-
system configurations and the corresponding customer ical levels (HL) [4] and adequacy assessment can be con-
reliability worth is generally termed as reliability cost/ ducted at each HL. HLIII adequacy assessment involves
worth assessment. A basic objective in these evaluations the consideration of all three functional zones to assess
is to determine the optimum level of service reliability. electric service reliability at specific customer load points
The evaluation of reliability cost through the identifica- [SI. This paper pertains to adequacy assessment at
tion and analysis of criteria and methods used to predict HLIII.
and quantify reliability levels has progressed significantly To assess the worth of electric service reliability it is
during the past two decades. By comparison, the assess- necessary to link the calculated adequacy indices to the
ment of reliability worth is not as well developed with consumer costs of interruption. Two sets of reliability
most approaches providing only an indirect evaluation. indices can be defined at HLIII, namely the customer
This is due to the fact that the assessment of the societal load point indices and system performance indices [3, 41.
worth of electric service reliability is an extremely The individual load point indices are the average failure
complex and subjective task. The worth of reliability rate 1, the average outage duration r and the average
~~~ ~
annual unavailability U. The most commonly used
0IEE, 1994 system indices are those of system average interruption
Paper 9946C (B), received 5th October 1993
frequency index (SAIFI), system average interruption
L. Goel is with the School of Electrical and Electronic Engineering, duration index (SAIDI), customer average interruption
Nanyang Technological University, Singapore 2263 duration index (CAIDI), energy not supplied (ENS),
R. Billinton is with the Power Systems Research Group, Department of average service availability index (ASAI) and its com-
Electrical Engineering, University of Saskatchewan, Canada S7N OW0 plimentary average service unavailability index (ASUI).

390 IEE Proc.-Gem. Transm. Distrib., Vol. 141, No. 4, July 1994
The basic definitions and mathematical formulations of tomer classes incur different monetary losses due to
these indices are given in the Appendix. power supply interruptions. The cost coefficients shown
This paper presents a contingency enumeration in Table 1 reflect these differences. The cost coefficients
method for obtaining the customer load point expected for each class are also dependent on whether the inter-
energy not supplied (EENS) index. This index is utilised ruption occurs during an on or off peak period, etc. It
together with customer outage cost functions to evaluate can be easily appreciated from the data given in Table 1
a factor designated as the interrupted energy assessment that the interruption cost coefficients vary in a nonlinear
rate (IEAR) at HLIII [SI. The IEAR aggregates the costs manner with the outage duration. The SCDF given in
incurred by a customer for each unit of unsupplied Table 1 are the basic data used in this paper for the
energy due to power interruptions. IEAR have been evaluation of IEAR at specific customer load points.
developed [7, 81 at both HLI and HLII. This paper
describes a method for IEAR evaluation at HLIII using 3 IEAR evaluation a t HLlll
the concepts applied to HLI and HLII. Two other
methods [9] using the final load point indices of 1,r and 3.1 Method 1 :Contingency enumeration method
U and the system indices SAIFI, SAID1 and C A l D l in The estimation of IEAR at HLIII involves all the cus-
conjunction with appropriate cost functions are also pre- tomer load points associated with a given load centre or
sented. The three methods are then compared in terms of service area. Three basic models are required for cus-
their ability to provide a valid IEAR. The concepts pre- tomer load point IEAR assessment, viz. the cost model,
sented are illustrated by application to a small but practi- the load model and the system generation, transmission
cal test system designated as the RBTS [lo, 111. and distribution model. The cost model is in the form of
the sector type and its associated SCDF. The load model
2 Customer damage functions involves the loads connected at each customer load point.
The combined system model consists of relevant tech-
Actual or perceived costs of customer supply inter- nical and reliability parameters of all major components
ruptions can be used to determine the worth of electric from the system generating units down to the laterals
service reliability. A variety of approaches have been used serving the individual customers. The HLIII adequacy
to investigate these costs [l, 23. One method which has assessment presented in this paper includes independent
been used to establish acceptable reliability worth estim- outages of generating units and transmission lines and
ates is to survey electrical consumers to determine the also outages owing to station related failures.
monetary losses associated with supply interruptions. The method [6] is divided into three distinct steps:
The data compiled from these surveys leads to the formu- (i) At HLII, a composite generation and transmission
lation of sector cost functions, which are referred to as system contingency evaluation program termed
sector customer damage functions (SCDF). The SCDF COMREL was used to obtain the expected load cur-
presents the sector interruption costs as a function of the tailed, probability, frequency and duration of each con-
interruption durations. The customer costs associated tingency that leads to load curtailment for each system
with an outage at any point in the system involves the bus. The contingencies considered include outages of up
combination of the costs associated with all customers to four generating units, up to three transmission lines,
affected by that outage. A composite customer damage up to two generators with one line, up to two lines with
function (CCDF) can be created which aggregates the one generator, all the four outages owing to station
composite costs of the customers in the service area to related failures and the isolation of up to two load buses
the interruption duration [Z]. owing to station related failures.
Table 1 shows interruption cost estimates in 1987 (ii) At the subtransmission system level, the impact of
C$/kW for five interruption durations for the selected all outages were obtained at each distribution system
customer sectors [l, 123. The data presented in Table 1 supply point. The contingencies considered include all
first-order active failures [3], all first-order permanent
Table 1 : Cost of interruptions in S/kW for various sectors [1, outages and all second-order overlapping permanent
121 outages.
Durn. Sector NDE
(iii) At the radial distribution level, the effects owing to
outages of system components such as primary main/
(min) Lg.lnd. Sm.lnd. Comm. Agric. Resid. G&l Office
laterals/low voltage transformers etc. were considered.
1 1.005 1.625 0.381 0.060 0.001 0.044 4.776
20 1.506 3.866 2.969 0.343 0.093 0.369 9.878 The detailed procedure is given in Reference 6 and is
60 2.225 9.085 6.552 0.649 0.482 1.492 21.065 summarised below :
240 3.966 25.163 31.317 2.064 4.914 6.558 68.630
480 8.240 55.808 83.008 4.120 15.690 26.040 119.180
Step i
(a) At HLII, for a specified load level, the following
Resid. = residential users
Comm. =commercial users
parameters are obtained at each major load centre (bus):
G&l =government and institutionalusers Lkj = load curtailed (MW) at bus k owing to contin-
Lg.lnd. = large industrial users (peak demand 2 5 MW) gencyj
Sm.lnd. =small industrial users (peak demand <5 MW) fkj = frequency (occ/yr) of contingency j at bus k, and
Agric. =agricultural users D, = duration (h) of contingency j at bus k.
Office =office space users
(b) For contingency j, the load curtailed L , at bus k is
assumed to be shared in proportion to the average load
are compiled from two major postal customer surveys connected at each individual customer load point at that
conducted by the power systems research group of the bus. The load curtailed at each customer load
University of Saskatchewan. The first one was conducted point p of bus k owing to contingency j at HLIII is there-
in 1980 for the residential, commercial and industrial fore evaluated.
sectors [l], while the second one for the agricultural (c) Outage duration Dkj is used to evaluate the cost of
sector was undertaken in 1985 [12]. The various cus- unsupplied energy owing to the interruption seen by all
1EE Proc.-Gener. Transm. Distrib., Vol. 141, No. 4, July 1994 391
customers at bus k. The impact of this duration is differ- and ECOST from each contributing event obtained as in
ent for different load points due to the unique customer step 2.
characteristics and SCDF. Corresponding to this dura- The results from the three steps are combined to
tion, the cost Ck,j , in $/kW is obtained using the SCDF obtain the HLIII indices. The overall ECOST, EENS
of the respective customer class at the load point. The and IEAR at load point p associated with bus k are
costs of interruptions longer than 8 h duration are evalu- obtained using eqns. 5, 6 and 7, respectively, where i
ated assuming the SCDF to have the same slope as that denotes the step number.
for the 4 to 8 h duration.
(d) The unsupplied energy EENS,, j , is obtained using
eqn. 1:
EENSk. j . p Lk, j . p fkj D k j (MWh/yr) (1)
(e) The expected cost of unsupplied energy ECOST,,,,.
owing to this contingency is obtained using eqn. 2:
ECOST,, j , p = Lk,j , p f k j ck, j , p (MW-WW-Yr) (2) HLIII ECOST,, e
HLIII IEAR,. =
HLIII EENSk.,
($/kwh) (7)
(j)The total EENS and its total cost ECOST at load
point p owing to N k load curtailment events are obtained Eqns. 1-7 are formulated for a constant load level. In
using eqns. 3 and 4, respectively. practice the load does not stay constant throughout the
Nk period of study but is best represented by a load duration
EENSk, = EENSk,j , (MWh/yr) (3) curve. A seven-step load model [6] was used to incorpor-
I =1
ate the effect of varying load levels. The unsupplied
ECOST,, = c ECOST,,
Nt

j= 1
j, (MW-$/kW-yr) (4)
energy owing to HLII contingencies and its cost for load
point p associated with bus k are evaluated using eqns. 8
and 9, where Nkm) is the number of contingencies
Step ii leading to load curtailment at bus k for load step m and
Any outage event that causes the isolation of the dis- prob(m) is the associated probability.
tribution supply point involves all customers connected
at that supply point. Continuity of supply is assumed to
be the sole criterion and therefore all the loads at a
supply point are completely isolated due to outage of the
EENSk, p = 1[
m=l
7

7
zl
Ndm)
EENSk,

Nt(m)
j.p] prob (m) (8)

supply point. The following procedure is followed for

estimating the total contribution from this step to the
unsupplied energy EENS and its cost ECOST at each
E C 0 S T . p = Jl
[l: ECOST,, j , p ] Prob (m) (9)

The contributions from the subtransmission and distribu-

system load point.
tion networks are based on continuity and the average
(a) For each supply point connected to the sub-
load levels. and are therefore not affected.
transmission networks at bus &, the outage events con-
tributing to its isolation are evaluated in terms of the
average failure rate At, j , average outage time rk, and
annual unavailability U t ,j .
32 Method 2: basic indices method [9]
The three-step contingency enumeration method for
(b) Corresponding to outage time rk, owing to event j,
IEAR assessment at HLIII described previously can be
for each of the load points isolated by this event, the cost considered to be reasonably comprehensive as it involves
of interruption, Ck,j , using the SCDF of customer cate-
consideration of every possible contingency/outage con-
gory p are evaluated. dition which causes load curtailment at customer ter-
(c) If La,!, is the average connected load at load p, the minals. The expected frequency and duration of each
corresponding ECOST and EENS owing to event j are outage, in combination with the expected load curtailed
evaluated using the following equations :
and the appropriate SCDF are used to obtain the HLIII
ECoST,, j . p = ck, j. p pAkj
IEAR.
This Section describes a method that uses the average
E E N S k . 1. p = p ukj or expected HLIII adequacy indices and sector cost func-
Steps (a) to (c) are repeated for each outage event contrib- tions. The three basic adequacy indices of average failure
uting to a supply point. If N , is the number of sub- rate A, average outage duration r and average annual
transmission outage events contributing to customer load outage time U evaluated at each HLIII customer load
point p connected to bus k, then the total EENS and point are used together with the SCDF associated with
ECOST to individual load points is evaluated using the the customer type at each individual load point.
following equations : The HLIII average outage duration index rk. at load
point p of bus k is used to obtain the cost ck,q in $/kW
from the SCDF of the customer type at load point p . The
average load curtailed is the average load connected
L,, p . The unsupplied energy EENS and its cost ECOST
at each load point are then formulated as follows:
EENSk.p = Lao, p I k , p 'k. p (10)
Step iii
The final contribution to an individual customer load ECoS&, p = p Lk, p ch. p (11)
point comes from the radial distribution network. For
a given configuration, the outage events contributing to IEAR = E = %Z ($/kWh)
the isolation of load point p are aggregated and EENS EENSk, p rk.p

392 IEE Proc.-Gem. T r a m Disnib., Vol. 141, No.4, July 1994

T
Eqns. 10 to 12 can then be used to evaluate the IEAR at the nonlinear nature of the individual load point cus-
each load point in the configuration under study. The tomer damage functions (SCDF) to be incorporated in
individual load point IEAR can be multiplied by their the event outage cost analysis. In method 2, designated as
respective per units loads and added to obtain a bus the basic indices method, the individual contingency
IEAR. events are aggregated to provide the average load point
indices of 1, r and U . These final indices, and not the
3.3 Method 3: system indices method [9] actual individual event data, are then used in conjunction
This method combines the system indices at the load bus with the associated SCDF to generate the cost factors, i.e.
with the CCDF applicable at that bus. The system a single outage duration representing the average inter-
CAIDI represents the average duration of outage per ruption duration of the load point is used in conjunction
system customer interruption. The cost in $/kW from the with the SCDF. This is therefore an approximation of the
CCDF using this CAIDI duration value can be obtained approach used in method 1 where the individual event
and used to obtain the system cost of unsupplied energy. average costs are aggregated. The process of aggregation
The unsupplied energy is obtained by combining the is further extended in method 3, in which the basic load
system SAID1 and the average connected bus load. The point indices are aggregated with the customer (number
overall IEAR reflected at the load centre is then evalu- and load) data to provide the system indices of SAIFI,
ated using eqn. 15. This method, however, cannot be used SAIDI, CAIDI, etc. The primary approximation in
to obtain individual load point reliability worth factors. method 1 is the use of average event frequencies and
system EENS = (SAIFI)(average bus load)(CAIDl) durations in obtaining the cost factors. Methods 2 and 3
are successive approximations of method 1.
= (SAIDIXaverage bus load) (13)
system ECOST = (SAIFIXaverage bus load) 4 Application t o the RETS
x (cost of CAIDI) (14) A single line diagram of the RBTS with complete sub-
cost of CAIDI transmission and distribution facilities at two system
system IEAR = (%/kWh) (15) buses (2 and 4) is shown in Fig. 1. The RBTS has a peak
CAlDJ
load of 185 MW and an installed capacity of 240 MW
The contingency enumeration method designated as consisting of two generating stations (buses 1 and 2) and
method 1 uses a comprehensive failure mode and effects five major load centers (buses 2 to 6). The detailed com-
approach which recognises the average frequency and posite generation and transmission system data, load
duration associated with each contingency event. This is data and cost data are presented in Reference 10. The
not the same as recognising each distinct event, as the subtransmission and distribution system data for two
average frequency and duration are used. Recognition of system load buses 2 and 4 are provided in Reference 11.
the average duration of each contingency event permits HLIII analysis has been conducted for these two system

LP17 LP19 LP21

LP16 LP18 LP20 LP22

Fig. 1 Complete

I E E Proc.-Gener. Transm. Distrib., Vol. 141, No. 4, July 1994 393

I

buses in all the studies reported in this paper. The basic Methods 2 and 3 use eqns. 10-15. Eqns. 12 and 15 are
sector cost data required to evaluate reliability worth fundamentally the same and simply require the average
indices are given in Table 1. Fig. 1 presents the overall outage duration and a cost look-up table to evaluate the
test system, which can be used for a wide variety of func-
tional zone and hierarchical level studies. Table 3: Sector HLlll /EAR for RETS buses 2844: method 1
A computer program was developed to evaluate the Sector OHL case Cable case
HLIII l E A R using the three methods described in this type
paper. The subtransmission/distribution networks for EENS ECOST /EAR EENS ECOST /EAR
buses 2 and 4 are shown in Fig. 1. There are three supply MWh/yr 1000$/yr $/kWh MWh/yr 1000$/yr $/kwh
points SP1, SP2 and SP3 in the network at bus 4 and Bus 2:
only the load points connected to the individual supply Sm.lnd. 1.7333 12.52864 7.23 4.5789 38.39766 8.38
points are affected owing to isolation of that supply Comm. 8.8393 178.47649 20.1 9 11.9719 289.06064 24.14
point. There is, however, only one supply point SP at bus Resid. 17.4534 107,63470 6.1 7 23.7385 196.43318 8.27
G&l 13.2136 233.06996 17.64 17.8608 496.98282 27.82
2 and therefore all customers are disconnected due to an
outage of this supply point. This paper illustrates a single Bus 4:
Smhd. 7.2356 56.64048 7.83 14.5261 117.63302 8.10
case study for the radial distribution system config- Comm. 11.0620 230.13928 20.80 13.4099 331.91846 24.75
uration, i.e. presence of the disconnects, 100% reliable Resid. 44.5979 276.89380 6.21 54.2726 464.21317 8.55
fuses in the laterals, 100% available alternate supply and
repair of low-voltage transformers. Two subcases are
illustrated, as the distribution system is assumed to HLIII IEAR values. The IEAR values obtained for the
consist of either underground cables or overhead lines RBTS using method 2 are shown in Table 4 for bus
(OHL). 2(OHL). The weighted bus HLIII IEAR values are also
A single aggregate l E A R value for each load bus at shown in this table. A comparison of IEAR values
HLIII is evaluated using a weighting procedure i.e. obtained for load buses 2 and 4 of the RBTS using
individual HLIII IEAR are weighted in proportion to the methods 1 and 2 is shown in Table 5. The percentage
fraction of bus load and. summed. Reference 10 specifies
the customer compositions at the individual load points Table 4: HLlll /EAR at bus 2 OHL case: method 2
in the distribution configurations connected to buses 2 ~ ~~~

and 4. The individual load point IEAR for the various Load Failure Outage Outage /EAR
Dt rate duration cost
load points at RBTS bus 2 (OHL case) using the contin-
gency enumeration method are shown in Table 2. The Vvr h $/kW $/kWh
1 0.3189 12.4481 32.90 2.64
HLIII l E A R can also be obtained for each sector class 2 0.3319 12.1564 31.62 2.60
associated with the system buses. The overall sector 3 0.3319 12.1564 31.62 2.60
HLIII IEAR for buses 2 and 4 for the two cases of OHL 4 0.3189 12.4481 62.75 5.04
and cable (using method 1) are shown in Table 3. The 5 0.3319 12.1564 59.86 4.92
6 0.3287 12.2273 150.74 12.33
7 0.3319 12.0389 147.48 12.25
Table 2: HLlll /EAR at bus 2, OHL case: method 1 8 0.2195 4.2735 27.15 6.35
9 0.2195 4.0962 25.86 6.31
Load EENS ECOST /EAR %Load Wt. /EAR
10 0.3222 12.3327 32.39 2.63
point
11 0.3320 12.1564 31.62 2.60
MWh/yr 1000$/yr $/kWh $/kwh 12 0.3352 12.0871 31.32 2.59
1 2.0648 13.5959 6.5846 0.0435 0.2866 13 0.3320 12.0000 58.34 4.86
2 2.0996 12.2270 5.8235 0.0435 0.2535 14 0.3352 11.9320 57.68 4.83
3 2.0996 12.2270 5.8235 0.0435 0.2535 15 0.3222 12.3327 152.57 12.37
4 2.1845 44.0519 20.1661 0.0460 0.9286 16 0.3319 12.1564 149.51 12.30
5 2.2212 37.2683 16.7781 0.0460 0.7726 17 0.3222 12.3731 32.57 2.63
6 1.7743 35.5858 20.0557 0.0369 0.7408 18 0.3222 12.3327 32.39 2.63
7 1.7640 35.0365 19,8620 0.0369 0.7337 19 0.3352 12.0484 31.15 2.59
8 0.8271 5.9489 7.1928 0.0814 0.5852 20 0.3352 12.0484 58.81 4.88
9 0.9063 6.5798 7.2603 0.0936 0.6793 21 0.3320 12.0000 58.34 4.86
10 2.0666 13.5967 6.5794 0.0435 0.2864 22 0.3352 11.9320 145.64 12.21
11 2.0996 12.5251 5.9654 0.0435 0.2597
12 1.7733 10.2950 5.8054 0.0366 0.2125 Aggregate: 5.68
13 2.1 919 38.6277 17.6232 0.0460 0.81 15
14 2.2010 37.2232 16.9116 0.0460 0.7788
15 1.7537 37.1 549 21.1867 0.0369 0.7826 Table 5: Comparison of HLlll /EAR (S/kWh) using methods 1
16 1.7817 35.6500 20.0084 0.0369 0.7391 and 2
17 1.7441 11.4464 6.5630 0.0366 0.2403
18 1.7382 11.4365 6.5794 0.0366 0.2409 Bus (case) Method 1 Method 2 Ohdifference
19 1.7675 10.2851 5.8190 0.0366 0.2130
20 2.2231 37.271 1 16.7653 0.0460 0.7720 2 (OHL) 12.11 5.68 53.09
21 2.1919 38.6277 17.6232 0.0460 0.8115 2 (cable) 16.63 8.17 50.88
22 1.7655 35.0494 19.8524 0.0369 0.7333 4 (OHL) 8.59 5.24 39.05
4 (cable) 10.28 5.91 42.56
Total 41.2395 531.7098 12.1155

difference between the results obtained using the enumer-

individual customer sector l E A R values shown in Tables ation method and method 2 with respect to the enumer-
2 and 3 provide valuable information on sector reliability ation method are shown for the weighted bus HLIII
worth at specific points in the system. This information IEAR values in Table 5. The results shown in this Table
can be used in managerial assessment of reliability worth clearly indicate that method 2 does not provide indices
and in any consideration of assigning customer rates comparable to those provided by the basic contingency
associated with reliability levels. enumeration approach for the test system studied.
394 IEE Proc.-Gener. Tmnsm. Distrib., Vol. 141, No. 4, July 1994
 Table 6 gives the system performance indices SAIDI, IEAR values obtained using eqn. 15, the CCDF and
SAIFI, CAIDI, etc. for the RBTS load buses 2 and 4. The the system indices given in Table 6 are shown in Table 9.
values shown in Table 6 were obtained by using the This table also shows the percentage difference with
respect to the enumeration method values. As in the case
Table 6: HLlll system indices at buses 2 and 4 of method 2, the results obtained by method 3 do not
compare well1 with those obtained by the enumeration
Bus SAIFI SAID1 CAlDl ASAl ENS
method.
int./cus-yr hlcus-yr h/cus-int. kWh/yr The results shown in Tables 5 and 9 clearly indicate
2 (OHL) 0.3279 4.0077 12.222 0.999542 42601.8 that IEAR values obtained using either method 2 or 3 for
2 (cable) 0.2382 5.4114 22.719 0.999382 5951 1.3
4 (OHL) 0.4025 3.8864 9.655 0.999556 64675.3
RBTS buses 2 and 4 are not comparable to those
4 (cable) 0.2930 4.7145 16.089 0.999462 83986.8 obtained using the enumeration method for the test
system studied. The IEAR from methods 2 and 3,
however, are quite comparable to one another. This can
equations given in the Appendix and customer data given be explained from the fact that these two methods use
in Reference 11 in conjunction with the individual load ‘single event’ parameters in conjunction with a single
point failure rates and annual unavailabilities. These nonlinear cost curve. Method 3 uses CCDF and CAIDI
indices reflect the overall system performance at the load which are directly aggregated from the SCDF and indi-
buses and therefore could be used in conjunction with an vidual load point indices, respectively. It can therefore be
appropriate composite customer damage function at that concluded that these two methods, in general, do not
bus. The CCDF for any bus within the RBTS can be provide correct estimates of the system IEAR and that a
obtained by suitably combining the customer types and detailed contingency enumeration technique should be
their SCDF with the peak demand and energy consump- used to obtain practical estimates. The main advantage
tions at that load centre. of method 2, however, is that if the average outage dura-
To obtain the CCDF for load centres 2 and 4 of the tion at a given customer load point and its associated
RBTS, the sectors allocated at these buses and their peak cost function are known, the IEAR can be readily evalu-
demand and average load values were used to obtain the ated using eqn. 12. Similarly, if the average system
percentage peak demand and energy consumptions. CAIDI and the CCDF are available, the system IEAR
These values are given in Table 7. The CCDF at each of can be easily obtained using method 3 (eqn. 15). The
results obtained from methods 2 and 3 may be used as
Tabla 7: Percentage peak demand and energy consumptions
starting values to make planning decisions in the absence
for buses 2 and 4 of more accurate values. The error in these results may
also be system specific and therefore detailed studies
Sector Bus 2 Bus 4 should be conducted before deciding which method is
Energy Demand Energy Demand appropriate for a given configuration.
Resid. 36.4087 36.25 47.4980 47.4985
Srn.lnd. 17.4925 17.50 40.6835 40.7500 5 Conclusions
G &I 27.6300 27.50 - -
Cornrn. 18.4700 18.75 1 1 .8185 11.7495 Owing to emerging market pressures and the regulatory
Total (%) 1OO.OOOO 100.00 100.0000 100.oooO environment, and changes in the cost structure of the
power industry, the traditional system planning practices
are being expanded to facilitate an explicit examination
these buses is obtained by weighting the SCDF from of the cost and benefit implications from the customer’s
Table 1 with the peak demand (for durations less than perspective. This paper concerned the assessment of the
30 min) or the energy consumption (for durations over societal worth of electric service interruptions at the
30min) values given in Table 7. The CCDF for RBTS overall system level (HLIII). Three methods have been
buses 2 and 4 are shown in Table 8. Method 3 uses these. presented for evaluating reliability worth in an electric
data in evaluating IEAR cost factors. power system. The first method uses a contigency enu-
meration technique and a failure mode and effects
Table 8: Compoaite customer damage functions for b u m 2
analysis procedure in association with appropriate cus-
and 4 tomer cost functions to evaluate the bulk system load
point IEAR and customer load point IEAR values. This
Duration Interruption cost procedure involves a comprehensive analysis of all major
Bus 2 Bus4 component outages in an electric power system. The
$/kW $/kW second method uses the average HLIII adequacy indices
1 min 0.3683 0.7074 at each customer load point together with the sector cus-
20min 1.3688 1.9692 tomer cost characteristics, and the third method makes
l h 3.7564 4.9357 use of the HLIII system performance indices in conjunc-
4h 13.7866 16.2724
8h 38.0000 39.9674
tion with appropriate composite customer cost character-
istics. All three methods are illustrated by application to
Table 9: Comparison of HLlll E A R using method. 1 and 3
Bus (case) CADI Cost of IEAR IEAR %Difference
CAlDl (method 3) (method 1 ) in IEAR
h/cus-int $/kW $/kWh $/kWh
2 (OHL) 12.22 70.63 5.78 12.11 52.30
2 (cable) 22.72 174.92 7.70 16.63 53.71
4 (OHL) 9.65 51.00 5.28 8.59 38.54
4 (cable) 16.09 98.88 6.14 10.28 40.25

IEE Proc-Gem. Tramm.Djstrib.., Vol. 141, No. 4, July 1994 395

 I

the RBTS, a small but comprehensive test system. The 7 Appendix

results indicate that methods 2 and 3 can be used to
SAIFl is defined as the average number of interruptions
obtain approximate customer interruption costs in the
absence of a detailed study. To obtain accurate values,
per customer served per time unit :
however, a detailed procedure such as the contingency K
enumeration method must be used. C
-liNi
The IEAR values obtained at both HLII and HLIII SAIFI = (interruptions/system-customer)
can be used to relate customer interruption costs with the
worth of electric service reliability in an electric power 1Ni
i= 1
system [13, 141. The individual customer load point
(16)
IEAR values and the customer sector IEAR values at
each bulk system load point can be used to make deci- SAlDI is defined as the average interruption duration for
sions on preferred load curtailment strategies or in customers served per time unit:
studies considering reliability based electric utility cus-
K
tomer rates.
1 UiNi
SAlDI = (h/system-customer) (17)
8 References
1Ni
1 BILLINTON, R., WACKER, G.,and WOJCZYNSKI, E.: ‘Cus- i= 1
tomer damage resulting from electric service interruptions’. Cana-
dian Electrical Association, R&D project 907 U 131 Report, 1982 CAIDI is defined as the interruption duration for CUS-
2 BILLINTON, R., and WACKER, G.:‘Customer cost of electric tomers interrupted per time unit:
service interruptions’, Proc. IEEE, June 1989,77, (a), pp. 919-930
3 BILLINTON, R., and ALLAN, R.N.: ‘Reliability evaluation of K
power systems’(Plmum, New York, 1984)
4 BILLINTON, R., and ALLAN, R.N.: ‘Reliability evaluation of large
1UiNi
elecuic power systems’ (Kluwer, Boston, MA, USA, 1988)
CAlDI = (h/interruption)
5 BILLINTON, R., and GOEL, L.: ‘Overall adequacy assessment of Ai N i
an electric power system’, IEE Proc. C, 1992,139, (l),pp. 57-63 i=l
6 GOEL, L., and BILLINTON, R.: ‘A procedure for evaluating inter-
rupted energy awxsment rates in an overall electric power system’, ASAI is defined as the ratio of the total number of cus-
IEEE Trans., 1991,PWRS6, (4), pp. 1396-1403 tomer hours that service was available during a year to
7 BILLINTON, R., OTENG-ADJEI, J., and GHAJAR, R.: ‘Compari-
son of two alternate methods to establish an interrupted energy the total customer hours demanded. The complement to
assessment rate’, IEEE Trans., Aug. 1987, PWRS2, pp. 751-757 this index is the system ASUI:
8 OTENG-ADJEI, J., and BILLINTON, R.: ‘Evaluation of inter-
rupted energy assessment rates in composite systems’, IEEE Trans.,
1990, PWRSS, (4), pp. 1317-1323
9 GOEL, L., and BILLINTON, R.: ‘Evaluation of interrupted energy
aSScSSment rates in distribution systems’, 1EEE Trans., 1991, PD-6,
(4), pp. 1876-1882 f 8760Ni
10 BILLINTON, R., KUMAR, S., GOEL, L., CHOWDHURY, N., i=l
cnu, K.DEBNATH, K., KHAN, E., KOS, P., NOURBAKHSH, ENS is defined as the energy not supplied due to supply
G., and OTENG-ADJEI, J.: ‘A reliability test system for educa-
tional purposes - basic data’, IEEE Trans., 1989, PWRS-4, (3), pp. unavailability :
1238-1244
11 ALLAN, R.N., BILLINTON, R., GOEL, L., SJARIEF, I., and SO, K
K.S. : ‘A reliability test system for educational purposes - basic dis- ENS = 1U i L i
i= 1
(MWh/yr)
tribution system data and results’, IEEE Trans., 1991, PWRS-6, (2),
pp. 813-820
12 WACKER, G.,and BILLINTON, R.: ‘Farm losses resulting from The symbols used are defined as
electric service interruDtions - a Canadian survey’. 1EEE Trans.,
1989, PwRsq (2), pp.b72-478 li= average failure rate of load point i
13 BILLINTON, R., and OTENG-ADJEI, J.: ‘Utilization of inter- U i = average unavailability of load point i
~ p t e denergy assessment rates in generation and transmission N i = number of customers connected to load point i
system planning’, IEEE Trans., 1991, PWRS-6, (3), pp. 1245-1253
14 GOEL, L., and BILLINTON, R.: ‘Utilization of interrupted energy
Li= average load (MW)connected to load point i,
assessment rates to evaluate reliability worth in electric power and
systems’, IEEE Trans., 1993, PWRS8,(3), pp. 929-936 K = number of load points in the system.

396 IEE Proc.-Gener. Transm. Distrib., Vol. 141, No. 4, July 1994

I
--