Optimal Electric Power Distribution System Reliability Indices Using

Binary Programming

Rosawan Bupasiri • Provincial Electricity Authority • Bangkok

Naruemon Wattanapongsakorn • King Mongkut’s University of Technology Thonburi • Bangkok
Jamnarn Hokierti • Kasetsart University • Bangkok
David W. Coit • Rutgers University • Piscataway

Key Words: SAIFI, SAIDI, Reliability, System cost, Optimization, Power system, Protective device

customers (referred to as load points). Most existing research

SUMMARY & CONCLUSIONS tends to use optimization techniques to improve efficiency in the
operational field. Alternatively, very few papers focus on the
optimization of reliability indices and outage costs. Soudi et al.
At present, most electric distribution utilities measure their (Refs. 2-3) presented a technique to minimize the total cost
reliability performance using reliability indices such as considering only SAIFI and SAIDI indices. They used reclosers
System Average Interruption Frequency Index (SAIFI) and and fuses as protective devices to minimize the number of
System Average Interruption Duration Index (SAIDI). interrupted customers.
However, using SAIFI and SAIDI as performance indices is In this paper, we present an optimization technique to
insufficient to measure the outage cost of utilities and minimize the total cost including the outage cost, the LCC and
customers. The outage cost reflects actual damage efficiently. the investment cost. We identify optimal SAIFI and SAIDI
Additionally, life cycle cost (LCC) and investment cost of indices and determine device allocations of breakers,
protective devices are important factors, which the utility reclosers, disconnecting switches and occurring cost.
providers need to consider. Adding protective devices in an
electrical distribution system can decrease the outage cost by 3. ELECTRICAL DISTRIBUTION SYSTEM
protecting public customers from local faults, but it may also 3.1 Protective Devices Configuration
increase the LCC and investment cost of the protective A typical configuration of a distribution system is
devices. In this paper, we propose an optimization technique illustrated in Figure 1, having one main feeder, and multiple
to identify types and positions of protective devices to branches. The locations and quantities of circuit protection
minimize the outage cost, the LCC and the investment cost devices are design variables which need to be selected to
according to system requirement constraints. This research optimize system performance for a defined objective function.
aims to help decision-maker in providing appropriate
protective device allocations in electrical distribution system. Load point
Main feeder No. of customer
We apply our optimization technique with a non-linear binary No. of kW

programming tool. Fuse

Breaker Recloser Switch

2. INTRODUCTION
Branch
There are many reliability indices used to evaluate electrical
distribution systems, known as electric power distribution
reliability indices (Ref. 1). The most common indices used by
Figure 1. Protective Device Configuration
electrical utility providers are System Average Interruption
Frequency Index (SAIFI) and System Average Interruption This research focused on the electrical distribution system,
Duration Index (SAIDI), to measure the impact of the utility starting from the part after the first breaker, in Figure 1, at the
outage, in terms of the number of interrupted customers and first substation through the main feeder. We consider three
interruption durations, respectively. The outage cost, which is types of protective devices which are reclosers, disconnecting
the main factor to be minimized in this paper, is difficult to switches and fuses. On the main feeder, there are many load
evaluate and this value varies according to type and location of points providing electric energy to a number of customers. At

0-7803-7717-6/03/$17.00 © 2003 IEEE

556 2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium
each load point, there will be a protective device to protect 1) Initial procurement, which is the total cost through
customers from electric power outage. process of purchasing devices including price, warehouse
The breaker, recloser and fuse, shown in Figure 1, are fee, transportation cost, managing cost etc.
automatic protective devices installed at substation, 2) Installation cost
distribution feeders and branches, respectively. Disconnecting 3) Testing cost
switches are manual protective devices used to separate faults 4) Operation cost
from the network. Moreover, a recloser can trip and reclose to 5) Maintenance cost including labor cost, device or tool cost.
clear momentary fault while a fuse can only perform ‘open- 6) Disposal cost, considered when the protective devices
circuit’ and not be able to clear the momentary fault by itself. are eliminated or sold out. It can be either returned cost
or expense cost.
3.2 SAIFI and SAIDI indices 7) Interests on a loan of investment cost.
Both SAIFI and SAIDI indices are frequently used indices From the above, the utility providers can calculate the LCC,
to measure the system reliability and performance. SAIFI and and approximate the average LCC per year assuming that the
SAIDI are used by most electric utility providers. Definitions lifetime of devices is known. As a result, the utility providers can
of the indices are mathematically written as select appropriate manufacturers of protective devices to reduce
Total number of customer interruptions cost per benefit by comparing the LCC of each manufacturer.
SAIFI =
Total number of customer served

=
∑ λi Ni 4. RESEARCH METHODOLOGY
NT (1) 4.1 Model Formulations for Main Feeder and Branches
Total customer interruption durations
SAIDI = To simplify the optimization problem, we separately
Total number of customer served
consider the main feeder of the distribution system and its
=
∑U i Ni branches. We divide the branches that have no local fault from
NT
(2) the branch that disturbs the main feeder by using a fuse on the
where Ni is the number of customers in section or load point i, NT tab point of the branch. Similar to Ref.2, the feeders are
represents the total number of customers on the feeder, λi and Ui divided into branches as shown in Figure 2. Each labeled
represent failure rate and down time of load point i, respectively. block represents a load point. It is a possible location to install
a protective device. These blocks contain information of
3.3 Outage Cost and Life Cycle Cost power (kW), number of customers, and failure rates of
There are two types of outage cost: interruption outage cost permanent fault and momentary fault.
and duration outage cost. Outage cost is considered from both
customer and utility sides. Interruption outage cost of customers 21 41 51
reflects loss or damage due to the interruption (but not the
duration) of the production process of customers. Duration 11 12 13 14
outage cost of customer reflects loss of customer production
during the downtime duration. Duration outage cost of utility is 31
the energy sale loss. For instance, direct and indirect effects of
the outage cost (Ref. 9) include the cost of product which would Figure 2. Divided Feeders for Optimization Approach
be produced if the outage events do not occur, labor cost with no
products, cost of damaged products, cost of rebooting customer’s 4.2 Objective Function
system, and cost from damage of system functions or parts. The selected objective function includes both the
Different types of customers have different outage cost interruption cost and duration cost, which are evaluated from
characteristics. For example, the outage cost of industrial both the utility providers and customers and LCC of the
customers is greater that the outage cost of general customers. protective devices, as presented in equation (3).
Consequentially, the customer types are classified by Thai
Standard Industrial Classification (TSIC). Electricity producers Total cost= Interruption Cost of customers + Duration Cost of customer +
Duration Cost of the utilities + LCC of protective devices
in other countries have analogous standards. However, the
utility providers always attempt to divide the zone of the = IER fc Ap + (IER dc +IER du ) Ae + LCC
(3)
customers such as industrial customers or residential customers. where IERfc represents interrupted energy rate of customers,
Generally, outage costs are estimated by calculating IER IERdc represents interrupted energy rate of customers caused by
(Interrupted Energy Rate), which is the total interruption cost per duration. IERfu represents interrupted energy rate of the utilities
energy not supplied (Ref. 5) or outage cost Baht/kWh. (Baht is and LCC represents life cycle cost of protective devices. Ap
the currency in Thailand) represents total power interrupted by outage events (kW) and
Life cycle cost (LCC) is the total cost that exists during the Ae represents total energy not supplied (kWmin). Ap and Ae
lifetime of protective devices. It consists of: are expressed in equation (4) and (5).
α +1
Ap = ∑ Apq
q =1
(4)

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium 557

 α +1 can be taken out of outage event by a disconnecting switch.
Ae = ∑ Aeq This term calculates energy interruption with switching time.
q =1
(5) The fourth and fifth terms of Adq are similar to the third and
where α is the number of branches where a fuse is installed at fourth terms of Afq, while their outage durations are switching
tap-point locations. Fuses can protect the main feeder from time.
local faults. Branches in the network with no fuse will be 4.3 SAIFI and SAIDI Analysis
included as load points in the main feeder. “α+1” represents The SAIFI, and SAIDI indices can be calculated according
the number of branches plus one main feeder. Thus, if we to equations (8), (9), (10) and (11) respectively. Both
have i main feeder, the term would be changed to “α + i”. q = equations are similar to the equations for Apq and Aeq.
1 represents the installation at the main feeder. The first load α +1

point (block 11) or a breaker position has q=1 only, since a ∑λ N = ∑ A i i fq

breaker has to be installed. The model formulations of Apq and

q =1
(8)
α +1
Aeq are presented in equations (6) and (7), respectively.
qn qn qn i −1 i
∑U N = ∑ A i i dq

Apq = ∑ λqi ∑ Wqj + ∑ λqi ∑ Wqj ∏x x

qk 1 qk 2
q=1
(9)
qn qn qn i −1 i
i =1 j =i i=2 j =1
Afq = ∑ λqi ∑ N qj + ∑ λqi ∑ N qj ∏ xqk1 xqk 2
k = j +1
qn qn qn i −1 qn i
+ ∑ γ qi (1 − xqi 2 )∑ Wqj + ∑ γ qi ∑ (1 − xqj 2 )∑ Wqk ∏ xql 2 i =1 j =i i =2 j =1 k = j +1
qn qn qn i −1 qn i
i =1 j =i i =1 j =1 k= j l = j +1
(6) + ∑ γ qi (1 − xqi 2 )∑ N qj + ∑ γ qi ∑ (1 − xqj 2 )∑ N qk ∏ xql 2
qn qn qn i −1 i
Aeq = ∑ λqi rr ∑ Wqj + ∑ λqi rr ∑ Wqj ∏x (10)
i =1 j =i i =1 j =1 k= j l = j +1
x x
qk 1 qk 2 qk 3
i =1 j =i i =2 j =1 qn qn qn i −1 i
k = j +1
qn i −1 i qn qn
Adq = ∑ λqi rr ∑ N qj + ∑ λqi rr ∑ N qj ∏x x
qk 1 qk 2 qk 3 x
+∑ λqi rs ∑ Wqj ∏x (1 − xqk 3 ) + ∑ γ qi rr (1 − xqi 2 )∑ Wqj
i =1 j =i i=2 j =1 k = j +1
x
qk 1 qk 2 qn i −1 i qn qn
+ ∑ λqi rs ∑ N qj (1 − xqk 3 ) + ∑ γ qi rr (1 − xqi 2 )∑ N qj
i=2 j =1 k = j +1 i =1 j =i
qn i −1 qn i
∏x x
qk 1 qk 2

∑ γ r ∑ (1 − x )∑W ∏ x
i =2 j =1 k = j +1 i =1 j =i
+ qi r qj 2 qk ql 2 qn i −1 qn i
i =1 j =1 k= j
(7) l = j +1
+ ∑ γ r ∑ (1 − x )∑ N ∏ x
qi r qj 2 qk ql 2

where qn is the number of load points in the main feeder or i =1 j =1

(11) k= j l = j +1

branches and qi represents the branch index i. For example, in where Afq and Adq represent total number of customer interruptions
Figure 2, block “14” represents the 4th load point at the branch and total customer interruption durations, respectively.
q =1, which is the main feeder. The first load point of main Apq and Aeq can be simplified because reclosers and
feeder is reserved for breaker only. Wqi represents average disconnecting switches are devices on the main feeder only.
power consumption of load point i, λqi is the permanent failure Moreover, reclosers or breakers can clear momentary fault by
rate and γqi is the momentary failure rate for load point i, and trip/reclose function. We assume that no sustain fault caused
Nqj is the number of customers at load point j including all of momentary fault in the main feeder. Therefore we revise
branches connected to that section. Note that, if there is a Apq and Aeq for the main feeder as shown in equations (12) and
three-phase device (recloser) at location qk, the variable xqk1 = (13), and for branches as shown in equations (14) and (15).
qn qn qn i −1
0, or otherwise xqk1 = 1. Here, the subscript 1 is used to Apq = ∑ λqi ∑ Wqj + ∑ λqi ∑ Wqj
i

represent a three-phase device or recloser, the subscript 2 i =1 j =i i=2 j =1

∏x
k = j +1
qk 2

(12)
represents a fuse and the subscript 3 represents a qn qn qn i −1 i
disconnecting switch. For outage duration notations, rr is the Aeq = ∑ λqi rr ∑ Wqj + ∑ λqi rr ∑ Wqj ∏x x
qk 1 qk 3
repair time and rs is the switching time. Protective device i =1 j =i i =2 j =1 k = j +1
qn i −1 i
failure rate is assumed to be neglected when compared to the +∑ λqi rs ∑Wqj ∏x qk 1 (1 − xqk 3 )
failure rate at a load point. i=2 j =1 k = j +1
(13)
For Apq , shown in equation (6), the first term represents qn qn qn i −1 i
Apq = ∑ λqi ∑ Wqj + ∑ λqi ∑ Wqj ∏x
power interruptions caused by permanent faults in load point i i =1 j =i i=2 j =1 k = j +1
qk 2

which are counted from load point i through qn. The second qn qn qn i −1 qn i

term represents the power interruptions caused by permanent + ∑ γ qi (1 − xqi 2 )∑ Wqj + ∑ γ ∑ (1 − x )∑ W ∏ x

qi qj 2 qk ql 2

faults before load point i through section having recloser or

i =1 j =i i =1 j =1 k= j l = j +1
(14)
qn qn qn i −1 i
fuses. The third term considers the power interruptions cause Aeq = ∑ λqi rr ∑ Wqj + ∑ λqi rr ∑ Wqj ∏ xqk 2
of momentary faults in load point i that has fuse(s) installed. i =1 j =i i=2 j =1 k = j +1
qn qn qn i −1 qn i
The last term considers all power interruptions caused by +∑ γ qi rr (1 − xqi 2 )∑ Wqj + ∑ γ qi rr ∑ (1 − xqj 2 )∑ Wqk ∏ xql 2
momentary faults in the load points before load point i. i =1
(15)
j =i i =1 j =1 k= j l = j +1

For Aeq, in equation (7), the first term represents energy 4.4 Optimization Techniques
interruptions with repair time caused by permanent faults in We use an optimization software provided by LINDO
load point i, which are counted from load point i through qn. Software Inc (Ref. 6), called LINGO. The LINGO software is
The second term considers energy interruptions with repair a linear, non-linear, and integer programming solver with a
time caused by permanent faults before load point i through mathematical modeling language.
the load point that has recloser or disconnecting switch or fuse Our objective function is to minimize the total cost as shown
installed. The third term covers energy interruptions which in equation (3), using the information of Ap, Ae, Apq, and Aeq

558 2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium

as presented in equations (4), (5), (6), and (7), respectively. Recloser
Variables “xqi, xqj” in the objective functions are calculated 1 728,341.22 12.9 56,460.56
using LINGO to obtain the global optimum solution. The 2 733,243.80 14.8 49,543.50
LINGO reports the value of the variables x including their 3 871,283.40 13.4 65,021.15
three subscripts equalled “1” or “0”. The value “1” means not-
4 930,383.64 15.1 61,614.81
installing and the value “0” means installing. The type of
Disconnecting switch
protective devices can be identified by the third subscript. For
example, x142=0 means that recloser is installed at the fourth 1 654,475.37 13.7 47,771.92
load point on the main feeder (q=1). The constraints are the 2 709,356.45 16.1 44,059.41
number of reclosers and disconnecting switches. For example, 3 639,147.15 15.5 41,235.30
the meaning of x121+ x131… x151 = 4 is that the second position 4 818,258.83 16.4 49,893.83
through the fifth position can be installed using only one
recloser. Moreover, we can define a forbidden position such as Table 2. Data of Main Feeder (q=1 and qn=13), with Repair
not-install fuse at the third postion; meaning x232 =1. We time = 58.5625 min. and Switching time = 25.2438 min.
reserve the restricted position for the design of protective qi λqi Wqi Nqi
device co-ordination.
11 0.40533 142 34
5. CASE STUDY 12 0.53333 215 53
13 2.12267 237 103
All system and device information was based on actual 14 1.06667 212 28
data collected during 1998-2001 by Provincial Electricity 15 0.50412 246 62
Authority (PEA), Thailand (Ref. 8). Outage costs of customers 16 0.12143 188 32
were estimated from past research, which classified types of 17 0.25576 172 84
customers according to the TSIC standard. We assume that the 18 0.17812 67 18
main feeder provides electric energy for a small general 19 0.55123 254 79
business sector. The outage of customers (Ref. 9) was 110 0.86256 213 47
estimated as shown in Figure 3. 111 0.17234 183 24
Baht
Outage cost of Small General Business 112 0.63256 98 6
1200
113 0.82556 136 57
y = 2.1627x + 34.043
1000
qi : ith load point, qth main feeder or branches, λqi :permanent
800
failure rate (interruptions/year), γqi: momentary failure rate
600
(interruptions/year), Wqi :average power consumption (kW), Nqi :
400
number of customers.
200

0
minute
Table 3. Data of 1st Branch(q=2 , qn=5), Repair time= 38.5351min.
0 100 200 300 400 500 600

qi λqi γqi Wqi Nqi

Figure 3. Outage Cost of Small General Business
21 0.521394 0.639414 81 43
The trend line of outage cost in Figure 3 is approximately
22 0.200035 0.357431 47 23
linear. The interruption cost (IERfc) is estimated to be 34.043 23 0.694322 0.481284 39 9
baht/kW/interruption from the y-intercept in Figure 3. The 24 0.499022 0.885314 49 12
outage duration (IERdc) is 2.1627 baht/kW/min, which is 25 0.600958 0.223843 21 16
estimated from the slope of the trend line in Figure 3. The
outage duration of the utilities (IERdu) estimated from the Table 4. Data of 2nd Branch(q=3,qn=4), Repair time= 45.5771 min.
average profit from this customer types and energy sale (Ref. qi λqi γqi Wqi Nqi
11) is 0.04811 baht/kW/min (where 1 US$ = 45-50 baht).
31 0.402173 0.116384 77 9
Four manufacturers of protective devices; reclosers and
32 0.389024 0.831144 43 18
disconnecting switches were considered as shown in Table 1,
33 0.709822 0.569382 65 14
with their approximated LCC (Ref. 10). From the table, we
34 0.820032 0.687477 28 6
selected the devices from the manufacturers with the device
lowest average yearly cost; reclosers from the second Complexity of this problem depends on many parameters
manufacturer and the disconnecting switches from the third including the size of the feeder (number of load points), types
manufacturer. The system data for this case study is shown in of allocated components and the number of choices for
Tables 2, 3 and 4 including permanent failure rate, momentary component allocations. In this case study, there is one main
failure rate, average kW and the number of the customers feeder (q= 1) and two branches (q= 2,3). Each of the load
served. points can have a recloser or a disconnecting switch installed,
Table 1. Estimated LCC of Reclosers and Disconnecting Switches or else neither of the devices are installed.
Manu- LCC Lifetime cost/year From the data in Tables 1-3 together with the cost function
facturers (baht) (years) (baht) from equation (3), we can establish the total cost equation of

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium 559

 main feeder (q=1) as shown in equation (16), and the total cost interruptions/year SAIF I
of each branch (q= 2,3) are shown in equations (17) and (18), 9 .0

respectively. 8 .0
0 SW
7 .0 1 SW
2 SW
3 SW
6 .0 4 SW
13 13 13 i −1 i 5 SW

Total cost1 =[34.043∑ λ1i ∑W1 j + ∑ λ1i ∑W1 j ∏x 1k 2 ]+ 5 .0

i =1 j =i i=2 j =1 k = j +1
13 13
[58.5625(2.1627+0.048108)∑ λ1i ∑W1 j +
4 .0
no. of recloser
0 1 2 3 4 5 6 7 8 9
i =1 j =i
13 i −1 i
58.5625∑ λ1i ∑ W1 j ∏x x
1k 1 1k 3 + (16) Figure 5. SAIFI of Main Feeder from Total Cost Optimization
i=2 j =1 k = j +1
qn i −1 i m in/y e a r S A ID I
25.2438∑ λ1i ∑ W1 j ∏x 1k 1 (1 − x1k 3 )] + 500
i=2 j =1 k = j +1
13 13
[49,543.5∑ (1 − x1i1 ) + 41, 235.3∑ (1 − x1i 3 )] 450
0 SW
i =2 i =2 1 SW

Total cost 2 = (34.043 + 38.5351(2.1627+0.048108) ) × (17) 400

2 SW
3 SW
4 SW
5 5 5 i −1 i
[ ∑ λ2i ∑W2 j + ∑ λ2i ∑W2 j ∏x
5 SW
2k 2 + 350
i =1 j =i i=2 j =1 k = j +1
5 5 5 i −1 5 i 300
∑ γ 2i (1 − x2i 2 )∑W2 j + ∑ γ 2i ∑ (1 − x2 j 2 )∑W2k ∏ x2l 2 ]
i =1 j =i i =1 j =1 k= j l = j +1
250 no . o f reclo ser

Total cost 3 = (34.043 + 45.5771(2.1627+0.048108) ) × 0 1 2 3 4 5 6 7 8 9

(18)
4 4
[ ∑ λ3i ∑W3 j + ∑ λ3i ∑W3 j ∏ x3k 2 +
4 i −1 i
Figure 6. SAIDI of Main Feeder from Total Cost Optimization
i =1 j =i i =2 j =1 k = j +1
4 4 4 i −1 4 i

∑γ 3i (1 − x3i 2 )∑W3 j + ∑γ ∑ (1 − x )∑W ∏ x

3i 3 j2 3k 3l 2 ] Table 5. Optimal Positions of Protective Devices with 6
i =1 j =i i =1 j =1 k= j l = j +1
Lowest Total Costs
Positions of
Positions of
Note that the LCC of fuses can be neglected in the ‘Total cost2’ disconnecting Total cost SAIFI SAIDI
reclosers
and ‘Total cost3’ because fuses are not recycled devices and their switches
costs are very cheap. The corresponding SAIFI and SAIDI indices 1) 3,4 8,11 2,142,261 6.37 289.40
can be determined using equations (7) and (8), according to the 2) 3,4,8,10,12 - 2,145,931 4.76 278.81
component allocations with the optimal total costs. 3) 3,4 7,9,12 2,147,148 6.37 285.34
4) 3 4,8,12 2,148,062 7.22 293.82
6. OPTIMIZATION RESULT 5) 3 5,8,10,12 2,149,846 7.22 291.29
We evaluate parameters λ, γ and W and use popular 6) 3,4,9 7,12 2,153,395 5.28 284.94
nonlinear programming software (LINGO) to optimize The six lowest total costs of the optimal positions of
equations (11), (12) and (13). We assume protective devices protective devices are listed in Figure 5, where their cost
have zero failure rate, and therefore, they have no effect on the values are very close. The corresponding SAIFI and SAIDI
total cost. The number of disconnecting switches is fixed, and indices are also displayed. The second choice is interesting,
the number of reclosers is varied to determine the optimal total with minimum SAIFI(=4.76 interruptions/year) and SAIDI
cost. The optimization results are shown in Figures 4, 5 and 6. (=278.81 min/year) compared to all other choices. This result
The optimal device positions are shown in Table 5. identifies that to limit local faults, reclosers are more desirable
T o ta l c o st
than disconnecting switches. Interestingly, the third choice has
B a ht
3 ,2 0 0 ,0 0 0 total cost less than the total cost of the fourth choice, while the
3 ,0 0 0 ,0 0 0 third choice use more reclosers. This can be explained because
adding an extra recloser can prevent more local faults from the
0 SW
2 ,8 0 0 ,0 0 0 1 SW
2 SW

2 ,6 0 0 ,0 0 0
3 SW
4 SW
5 SW
distribution system, even though the device cost and its LCC
2 ,4 0 0 ,0 0 0
are considered. With appropriate device allocations the total
2 ,2 0 0 ,0 0 0
cost can be less (comparing the costs of choice 3 with choice 4).
2 ,0 0 0 ,0 0 0
0 1 2 3 4 5 6 7 8 9
no . o f re c lo se r With many reclosers installed, the total cost in Figure 4
tends to increase because of adding the disconnecting switch.
Figure 4. Total Optimization Cost of Main Feeder (q=1)
This is because the cost of a disconnecting switch is higher
and All Branches (q = 2,3)
than its returned benefit. However, also shown in Figure 4,
when the number of reclosers is less than 3, adding
disconnecting switches can obviously reduce the total cost. In
summary, these results can help the utility providers make
decision in order to decrease their SAIFI and SAIDI indices.

560 2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium

Both of graphs in Figures 5 and 6 imply the minimum values From equation (12) and (13), the results are shown in
of SAIFI and SAIDI, which the utility providers can obtain. Figures 7, 8 and 9.
Baht Total cost interruptions/year SAIFI of branch min/year SAIDI of branch
160,000 5.5 250

140,000 5.0

4.5 200
120,000
q= 2 4.0 q= 2 q= 2
100,000 q= 3 q= 3 150
3.5 q= 3
80,000 3.0
100
60,000 2.5

2.0 no.of fuse no.of fuse

40,000 no.of fuse
1 2 3 4 5 50
1 2 3 4 5 1 2 3 4 5

Figure 7. Total Optimization Cost of Figure 8. SAIFI of Branches q= 2, 3 from Figure 9. SAIDI of Branches q=2, 3 from
Branches q= 2 and q= 3 Optimal Total Costs Optimal Total Costs

The graph in figure 7 demonstrates that adding fuse in

branches can increase total cost efficiently. Although, SAIFI BIOGRAPHIES
and SAIDI are slightly decrease after two fuses are installed at Rosawan Bupasiri, B.E., EE
Reliability Analysis Section, System Operation Division
each branch. The utility providers are satisfied with the Provincial Electricity Authority
obtained SAIFI and SAIDI indices while their customers are 200 Ngamwongwan Rd., Chatuchak District, Bangkok 10900 Thailand
effected by not only the indices but also the cost from the
Internet (email): [email protected]
outage event.
Table 6. Optimal Positions of Fuses for All Solutions. Rosawan Bupasiri is a graduate student in Computer Engineering at King
Mongkut’s University of Technology Thonburi (KMUTT), Thailand. She
Positions of fuses Total cost SAIFI SAIDI received the B.E. degree in Electrical Engineering (1999) from Mahidol
q=2 University. Her research interests include power system design and reliability
optimization. She is working for the Provincial Electricity Authority of Thailand.
1 144,206.70 5.10 196.65
1,3 92,547.70 2.93 113.07 Naruemon Wattanapongsakorn, PhD, EE
Department of Computer Engineering
1,3,4 82,274.71 2.74 105.63 King Mongkut’s University of Technology Thonburi,
1,2,3,4 76,890.60 2.51 96.66 91 Suksawasd 48, Ratburana, Bangkok 10140 Thailand
1,2,3,4,5 72,071.62 2.41 92.96 Internet (email): [email protected]
q=3 1 129,941.20 4.53 206.26 Naruemon Wattanapongsakorn is a faculty member in Computer Engineering
1,3 84,861.60 2.92 133.29 at King Mongkut’s University of Technology Thonburi (KMUTT), Thailand.
1,2,4 66,402.01 2.73 124.51 She received the B.S. degree in Computer Engineering and the M.S. degree in
Electrical Engineering, both from The George Washington University, and
1,2,3,4 58,986.95 2.24 102.18 Ph.D. degree in Electrical Engineering from the University of Pittsburgh. Her
research interests include distributed system dependability analysis,
optimization algorithms, real-time system modeling, software fault-tolerance,
REFERENCES and statistical analysis of system reliability. She is a member of the IEEE.
1. IEEE Working Group on System Design, “Trial Use Guide for Electric
Power Distribution Reliability Indices”, Report Draft #14, pp 1366. Jamnarn Hokierti, Ph.D., Associate Professor, EE
2. F. Soudi and K. Tomsovic, “Optimized Distribution Protection Using Department of Electrical Engineering
Binary Programming”, IEEE Trans.Power Delivery, vol. 13, No. 1, 1998 Jan, Kasetsart University
pp 218-224. 50 Phahon-Yotin Rd., Chatuchak District, Bangkok 10900 Thailand
3. F. Soudi and K. Tomsovic, "Optimal Trade-offs in Distribution Protection
Design," submitted to IEEE Transactions on Power Delivery., Vol 16, Internet (email): [email protected]
2001April., pp 292-296
4. R. Billinton, R.Allan, Reliability Evaluation of Power System, Pitman, Jamnarn Hokierti is a faculty member in Electrical Engineering Department at
1990. Kasetsart University, Thailand. He received the B.S. degree in Electrical
5. R. Billinton, R.Allan, Reliability Evaluation of Engineering System-Concept Engineering from Chulalongkorn University, Thailand, the M.Sc. degree and
and Techniques, Pitman, 1990. the Ph.D. degree in Electrical Engineering, both from the University of
6. Y-H. Song and M. R.Irving, “Optimization Techniques for Electrical Power Missouri-Rolla. His research interests include electric power system design,
Systems Part 2 Heuristic Optimization Methods”, Power Engineering and energy conservation.
Journal, 2001 June, pp 151-160.
7. LINGO, LINDO System Inc., www.lindo.com David W. Coit, Ph.D., Assistant Professor, IE
8. Reliability Field Data referenced from the Reliability Analysis Section, Department of Industrial Engineering
System Operation Division, Provincial Electricity Authority, 1998-2001. Rutgers University
9. Outage Cost Research, Energy Research Institute, Chulalongkorn 96 Frelinghuysen Rd. Piscataway, NJ 08854-8018, USA
University, Bangkok, Thailand, pp 199-200.
10. Estimation cost report for protective device, data referenced from the Internet (email) :[email protected]
Power System Analysis Section, Research Division, PEA, 2001
11. Average baht per energy sale of small general customers (baht/kWh), data David Coit is a faculty member in Industrial Engineering Department at
referenced from the Statistics Section, Power Economic Division, Provincial Rutgers University, NJ, USA. He received the B.S. degree in Mechanical
Electricity Authority (PEA), 2001. Engineering from Cornell University, the MBA degree from the Rensselaer
12. R. Bupasiri, N. Wattanapongsakorn, J. Hokierti, “Protective devices Polytechnic Institute, the M.S. and Ph.D. degrees in Industrial Engineering
allocation optimization for electrical distribution system”, Proc. International both from the University of Pittsburgh. His research interests include system
Technical Conference on Circuits, Systems, Computer and Communication, reliability engineering, applied operations research, and evolutionary
2002 July, pp 433-436. algorithms. He is a member of the IEEE.

2003 PROCEEDINGS Annual RELIABILITY AND MAINTAINABILITY Symposium 561