Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

A Binary Programming Model for SAIFI Considering

Protective Device Failure


Al Osiris F. Ingking, Russel John C. Gallano, Adonis Emmanuel DC. Tio
Electrical and Electronics Engineering Institute
University of the Philippines Diliman
Quezon City, Philippines
angelopongkol@yahoo.com, russel.gallano@eee.upd.edu.ph, adonis.tio@eee.upd.edu.ph

Abstract — Reliability indices such as SAIFI (System Average involves minimizing or maximizing an objective function such
Interruption Frequency Index) are useful not only in assessing as cost or some reliability metric. While there are a variety of
the reliability of a power distribution system but also in placement methods already available in the literature, a
optimizing the placement of protective devices such as reclosers common denominator among the cited works is the use of a
and fuses along a distribution feeder. In this paper we present a binary equation as a direct representation of the objective
binary formulation for SAIFI that is expressed as an explicit function. That is, the objective function, e.g. reliability metric
function of the placement and failure rate of recloser, fuse-blow such as the System Average Interruption Frequency Index
fuses, and fuse-save fuses along a given feeder. The inclusion of (SAIFI), is expressed directly as a function of the decision
device failure rate in the formulation is more intuitive than the
variables, i.e. the placement of the protective devices
existing model because a lookup table is no longer necessary and
laterals need no longer be classified as one of three types.
considered. Expressing the objective function as a direct
function of the decision variables allows for more flexibility in
We used this equation in a sample test system to show how it the optimization process since deterministic or heuristic
can be used to get the placement of protective devices that results optimization methods can be used. This is as opposed to using
in the least value for SAIFI. Results show that when protective a procedural or iterative method in computing the objective
device failure is considered, the reliability of a system worsens as function, in which case, deterministic optimization methods
expected. But while optimal device placement may change as can’t be used.
device failure rate increases, the optimal solutions obtained are
limited to a small subset of device placement configurations.
The binary models available in the literature have been
improving since Soudi’s and Tomsovic’s work in 1998 [1] with
Keywords — distribution system reliability, optimal protective newer models addressing the limitations of earlier works.
device placement, protective device failure Initial models had the following limitations: (1) only the
placement of reclosers and fuse-blow fuses were considered,
(2) laterals needed to be categorized in one of three types, (3)
the model needs to be linearized using linearization constraints,
I. INTRODUCTION
and (4) only SAIFI is modeled as a binary equation.
Succeeding works tried to add to this initial work by
Power interruptions are inevitable due to a wide variety of considering fuse-save fuses [2,3], switches [4,5] tie-lines [4],
supply-, network-, and load-related issues. In the side of the and even distributed generation [6]. Binary models for other
distribution network, system operators can lessen the number indices such as MAIFI [3] and SAIDI [4,7] are now also
and duration of interruptions by placing protective devices such available. Furthermore, lateral classification is no longer
as fuses and reclosers in strategic locations in the distribution necessary in the more recent formulations [3,4,6].
feeder. When protective devices are installed optimally, we can
expect less customer interruptions as less customers are These works, however, still do not consider device failure
disconnected from the supply for the most frequent line failure in the optimization process. While [8] tried to incorporate
events. Less frequent interruptions not only benefit the device failure in their binary equation, their formulation still
customers in terms of minimized losses and increased categorizes feeder segments in one of three types of laterals
productivity, but also benefit the distribution utility in terms of which limits the network topology that can be considered.
increased revenues and minimized operational cost in servicing Furthermore, their formulation can be a bit complicated which
the fault. necessitates the use of a lookup table in the expansion of the
binary formulation.
Optimizing the placement of protective devices allows the
utility to harness the full benefits of their investment. It usually In this paper, we present a binary formulation for SAIFI
that is expressed as an explicit function of the placement and
We would like to thank the Philippine Department of Science and failure rate of recloser, fuse-blow fuses, and fuse-save fuses
Technology – Engineering for Research and Development for Technology along a given feeder. The equation provides a closed form
(DOST-ERDT) for providing us with research dissemination grant. representation of SAIFI as a function of the placement and

978-1-5386-3917-7/17/$31.00 ©2017 IEEE


failure rate of recloser, fuse-blow fuses, and fuse-save fuses
along a given feeder. The inclusion of device failure rate in the
formulation is more intuitive than the existing model because a
lookup table is no longer necessary and laterals need no longer
be classified as one of three types. We then used this equation
in an illustrative example to show how it can be used to get the
optimal placement of protective devices in a feeder that results
in the least value for SAIFI.

II. DISTRIBUTION SYSTEM PROTECTION AND RELIABILITY

A. Distribution Reliability Indices


Billinton [9] defines reliability as some sort of probability
or expectation that an equipment or system will perform Fig. 1. Single line diagram of the test system
satisfactorily for at least a given period of time when used
under stated conditions. In the context of distribution systems,
this could be interpreted as a probability or expectation that Fuses may be coordinated with an upstream recloser via
electricity is satisfactorily delivered to the customers. either the fuse-blow or fuse-save scheme. Fuse-blow fuses are
Depending on the industry standards and load sensitivity coordinated to blow for any kind of downstream fault. On the
considered, unsatisfactory electricity supply can be any sort of other hand, fuse-save fuses are coordinated to allow the
deviation from the clean sinusoidal supply at the nominal upstream recloser to perform its reclosing operations but to trip
voltage and frequency. In this paper, we are only concerned first before the recloser locks out.
with a small subset of these unsatisfactory power supply
characteristics namely power interruptions. Specifically, we are Deciding where to install these devices is generally non-
only concerned with power interruptions resulting from the trivial because we need to take into account which sections are
opening of protective devices due to line and device failures, more prone to failure and which sections have more connected
e.g. system faults and device failure. Furthermore, we’ll only customers. Ideally, we want to isolate the most number of
look at sustained power interruptions, i.e. interruptions lasting customers from sections that fail frequently. This, however, is
more than five minutes [10]. easier said than done.

The more common measures of distribution reliability Adding to the complication in the design process is the
concerning sustained power interruptions are given by the issue of device failure, that is, when the primary protective
System Average Interruption Frequency Index (SAIFI) and the device fails to perform its expected protective action. Examples
System Average Interruption Duration Index (SAIDI). SAIFI is include reclosers that fail to operate or fuses that fail to blow
the expected frequency of sustained interruptions while SAIDI due to random failure or gradual wear and tear. When the
is the expected duration of sustained interruptions. These primary protective devices fail, the upstream devices are
indices are largely dependent on the placement and operation usually designed to trip as some sort of backup or redundant
of protective devices. In this paper, we present how SAIFI can protection, but this would mean that a larger number of
be modeled as an explicit equation involving these parameters. customers would be affected.
Obviously, the placement of protective devices and their
failure to operate significantly affects the reliability of service
B. Protectve Devices and Distribution Reliability provided by the utility. In the succeeding section, we will
A simplified diagram of a distribution feeder is shown in present how we can model the placement and failure of these
Fig. 1. The feeder can be divided into sections defined across devices as an explicit function of SAIFI, the expected
two nodes. Customers may or may not be connected to each frequency of sustained interruption, of a distribution feeder.
section and protective devices may or may not be installed at
the start of each section. The numbered boxes in the figure
indicate where the devices may be installed. In this paper, we III. SAIFI EQUATION CONSIDERING DEVICE FAILURE
consider three types of protective devices, namely, reclosers,
fuse-save fuses, and fuse-blow fuses.
The basic formula for SAIFI is given by the ratio between
Reclosers are devices that can be set to perform fast the expected number of customers affected by sustained
reclosing operations before locking out. The fast reclosing interruptions and the total number of customers connected to
operations are intended to test for momentary faults so that the feeder. It is usually defined for a given period, usually for a
service is immediately restored after the fault clears by itself. If year.
the fault does not clear within the set reclosing period, the
recloser goes on lockout and the utility must service the fault Eq. (1) shows the basic equation for SAIFI. Eq. (2) shows
before re-closing the recloser. how the numerator can be expanded as a function of the
placement of protective devices and device failure.
Aq
SAIFI = (1)
Nt

Aq = ¦λ N ' + ¦λ ¦ (N '
i∈L
i i
i∈L
i
j∈F ( i )
j − N 'h ( j , i ) ) ∏( ) x
k∈G j ,i
k 1 xk 2 xk 3

+ ¦γ
i∈L
i (1 − xi2 ) N 'i + ¦ γ ¦ (1 − x ) N ' ∏
i∈L
i
j∈F ( i )
j2 j
k∈G ( j ,i )
xk 1 xk 2

+ ¦λ ¦
i∈L
i
n∈{1, 2 ,3}
βn ( 1 − xin ) ¦ (N ' − N ' ( ) ) ∏
j∈F ( i )
j h j ,i
k∈G '( j ,i )
xk 1 xk 2 xk 3
(2)
+ ¦γ β
i∈L
i 2 (1 − xi2 ) ¦ (1 − x )( N '
j∈F ( i )
j2 j − N 'i ) ∏
k ∈G '( j ,i )
xk1 xk 2

+ ¦Ȝ ¦ ¦
i∈L
i
j∈F ( i ) n∈{1, 2 , 3}
(
βn 1 − x jn )¦
k ∈F j ( )
( N' −N' ( )) ∏ x
k
( )
h k, j
r∈G j ,i
r1 xr 2 xr 3 ∏
s∈G '( k , j )
xs1 xs2 xs3

+ ¦ γ ¦ β (1 − x ) ¦
i∈L
i
j∈F (i )
2 j2
k ∈F ( j )
(1 − xk 2 )( N 'k − N ' j ) ∏
r∈G ( j ,i )
xr1 xr 2 xr 3 ∏
s∈G '( k , j )
xs1 xs2

Eq. (2) uses the same notation as in [3] and [4] as follows: sustained interruption provided that there are no devices
Nt is the total number of customers served by the feeder; L is installed between the upstream device and the faulted section
the set of all line sections in the system; Ȝi is the permanent and (2) customers downstream of a fuse-blow fuse with a
failure rate of section i; Ȗi is the temporary failure rate of downstream temporary fault will experience a sustained
section i; ȕn is the probability that device type n will fail to interruption provided that there are no devices installed
operate, where n is equal to 1 for reclosers, 2 for fuse-blow between the upstream fuse-blow fuse and the faulted section.
fuses, and 3 for fuse-save fuses; xin is the decision variable
where i is the line section number and n is the device number The fifth and seventh terms model the following case: when
where 1 is for reclosers, 2 is for fuse-blow fuses, and 3 is for the nearest upstream device of a line section with a permanent
fuse-save fuses; Ni is the total number of customers fault fails, customers upstream of the failing device up to the
downstream of section i; h(j,i) is a function that returns the line upstream functioning protective device will also experience a
section downstream of j in the path to i; F(i) is a function that sustained interruption in addition to those already accounted
returns the set of line sections in the path from the substation to for by the first two terms.
line section i, including the line section nearest the substation The sixth and eighth terms model the following case: when
and excluding line section i; G(j,i) is the set of line sections in the nearest upstream device of a line section with a temporary
the path from line section j to line section i, excluding line fault is a fuse-blow fuse that fails to operate and if the nearest
section j but including line section i; G’(j,i) is the set of line functioning upstream device is a fuse-blow fuse, customers
sections in the path from line section j to line section i, upstream of the failing fuse-blow fuse up to the functioning
excluding both line sections j and i. The decision variables, xin, upstream fuse-blow fuse will also experience a sustained
can only either be “0” or “1”. It is “0” if device type n is interruption in addition to those already accounted for by the
installed in location i and “1” otherwise. third and fourth terms.
As an example using the test system in the next section, Eq. (2) can be expanded for any given distribution feeder
N’1 = Nt = 5,000; N’5 = 800; N’6 = 300; h(1,3) = 2; h(2,3) = 3; for predictive assessment or optimal placement of devices. If
F(3) = {1,2}; F(9) = {1,2,3,4}; G(1,2) = {2}; G(3,9) = {4,9}; used in an optimization model, binary constraints similar to
G’(1,4) = {2,3}; G’(3,9) = {4}. There are cases where these that defined in [4] also apply. The constraints may be used to
functions are undefined, e.g. h(3,2) and G(3,2) since the first account for the following: (1) limit on the number of reclosers
index must be upstream of the second index. The indices under and fuses, (2) design constraints, e.g. a device is required in a
the summation terms ensure though that these cases do not given location and at most one device may be installed at any
come up. given location, (3) device coordination constraints, (4) banned
The first four terms are interpreted similar to the configurations such as reclosers banned downstream of fuses
interpretation in [4], that is, (1) customers downstream of any and fuse-blow fuses banned downstream of fuse-save fuses,
device with a downstream permanent fault will experience a and (5) linearization constraints if using linear optimization,
among other sets of constraints.
IV. ILLUSTRATIVE EXAMPLE: OPTIMAL PLACEMENT OF DEVICES highlight the changes in the optimal placement of the devices
as well as the worsening of the value of SAIFI as devices
become more prone to fail. Note that the optimization process
Fig. 1 shows the single line diagram of the test feeder with
does not recommend the installation of fuse-blow fuses (as
nine sections. Table I shows the corresponding reliability data
expected) since the fuse-blow scheme tends to increase SAIFI
of the test system in terms of permanent and temporary failure
when fuses are coordinated to trip even for temporary faults.
rates and number of customers connected per section.

TABLE II. OPTIMAL DEVICE LOCATION RESULTS*


TABLE I. RELIABILITY DATA OF THE TEST SYSTEM

From To Ȝi Ȗi, Case Ⱦrec Ⱦfuse Rec FBF FSF SAIFI


Section N
Node Node f/year f/year
1 1 2 0.8 1.2 800 0 0.0 0.0 3,4 --- 5,6,7,8,9 2.346
2 2 3 0.8 1.4 1200 1 0.4 0.1 3,4 --- 5,6,7,8,9 2.827
3 3 4 0.9 1.6 800 2 0.2 0.9 3,5,7 --- 6,8,9 3.482
4 4 5 0.7 1.0 600 3 0.6 0.6 3,4,5 --- 6,7,8,9 3.941
5 2 6 0.9 2.0 500
*Rec – Optimal recloser locations; FSF – Optimal fuse-save fuse location;
6 6 7 0.7 1.7 300 FBF – Optimal fuse-blow fuse locations
7 3 8 0.8 2.8 400
8 4 9 1.0 3.2 200
To further illustrate the effect of device failure on the
9 5 10 0.5 0.7 200 optimal solution and on SAIFI, we swept the values of ȕfuse and
ȕrec from 0 to 1, while maintaining the same constraints and
We impose the following constraints for this test system: other pertinent data. Fig. 2 displays the result in a graph of the
(1) there are only three available reclosers, (2) there is no limit optimal SAIFI as a function of the device failure rates and
to the number of other devices that can be installed, (3) a fuse Table III specifies the corresponding optimal device placement
or a recloser must be installed at all tap points, (4) a recloser is for the points in the graph. While we can observe that the
installed at the start of the feeder, (5) fuses cannot be installed optimal device configuration changes as device failure
in the main feeder, and (6) proper protection coordination increases, the optimal configurations are limited to a small
cannot be achieved if reclosers are installed at section 2 and 3. subset of only three possible device configurations. These
Eq. (3) shows the resulting equation when Eq. (2) is observations could have been a result of the limited scale of the
expanded and simplified for ȕ = 0.1 for all †‡˜‹ ‡•. In the test system used. It would be interesting to replicate this
equation, we use the following notation: x’in = 1 – xin. experiment to a larger and more varied test system.

Aq = 360 + 100x '52 + 25x '62 + 225x '72 + 675x21 + 562.5x21 x31
+600x31 + 500x21 x31 x41 + 250x31 x41 + 56.25x '31 x21
+50x '41 x21 x31 + 25x '41 x31 + 13.875x '51 x21 + 7.5x '51
+13.875x '52 x21 + 7.5x '52 + 13.875x '53 x21 + 30x '53
+52.5x '61 x21 + 30x '61 + 52.5x '62 x21 + 30x '62 + 52.5x '63 x21
(3)
+15x '63 + 12.5x '71 x21 x31 x41 + 6.25x '71 x31 x41 + 9.5x '71 x41
+15x '71 + 12.5x '72 x21 x31 x41 + 6.25x '72 x31 x41 + 9.5x '72 x41
+15x '72 + 12.5x '73 x21 x31 x41 + 6.25x '73 x31 x41 + 9.5x '73 x41
+50x '31 x41 x21

The system has the following constraints: (1) Only three


reclosers are available: x21+x31+x41+x51+x61+x71+x81+x91 > 5,
(2) only one device per location: x61+x62+x63 > 2, (3) locations
that require an installed fuse or recloser: x51+x52+x53 = 2, Fig. 2. Optimal SAIFI versus failure rate
x71+x72+x73 = 2, x81+x82+x83 = 2, x91+x92+x93 = 2, (4)
coordination not possible between locations: x21+x31 > 1, (5)
reclosers disallowed at fuse load-side: x52+x53+x61 > 2, and (6) TABLE III. LEGEND FOR THE OPTIMAL DEVICE LOCATIONS
fuse-save fuses not allowed at fuse-blow load side: x52+x63 > 1.
Symbol Recloser Loc. FSF Loc. FBF Loc.
Table II shows the optimal device locations and the
corresponding SAIFI obtained using binary programming for TRIANGLE 3,4 --- 5,6,7,8,9
three cases of device failure rates. The values are validated CIRCLE 3,4,5 --- 6,7,8,9
using the traditional method of computing for SAIFI using a
load-point table. These illustrative examples were chosen to SQUARE 3,5,7 --- 6,8,9
V. CONCLUSION AND FUTURE WORK REFERENCES

[1] F. Soudi and K. Tomsovic. “Optimized distribution protection using


In this paper, we have presented a binary equation that binary programming,” IEEE Transactions on Power Delivery, Vol. 13,
expresses SAIFI as an explicit function of the location and No. 1. 1998. Page(s) 218-224.
failure rate of reclosers, fuse-blow fuses, and fuse-save fuses [2] F. Soudi and K. Tomsovic. “Optimal trade-offs in distribution protection
installed in a given distribution feeder. This formulation can be design,” IEEE Transactions on Power Delivery, Vol. 16, No. 2, 2001.
used in predictive reliability assessment or device placement Page(s) 292-296.
optimization, using either deterministic or heuristic [3] A.E.D.C. Tio, I.B.N.C. Cruz, B.M. Malquisto, and R.D. del Mundo. A
binary programming model for reliability optimization considering fuse-
optimization methods. This formulation is more general than blow and fuse-save schemes. In TENCON 2012 - 2012 IEEE Region 10
the existing binary model with device failure because laterals Conference, pages 1-6, Nov 2012.
need no longer be categorized in one of three types and that [4] A.E.D.C. Tio and I.B.N.C. Cruz. A binary formulation of SAIDI for the
device failure rate is more seamlessly incorporated in the predictive reliability assessment of radial distribution systems with tie-
formulation without the use of a lookup table. lines. In Power and Energy Engineering Conference (APPEEC), 2013
IEEE PES Asia-Pacifc, pages 1-6, Dec 2013.
We presented an illustrative example on how the binary [5] L.G.W.da Silva, R.A.F.Pereira and J.R.S.Mantovani. "Optimized
formulation presented can be used to optimize the device Allocation of Sectionalizing Switches and Control and Protection
placement in a simple distribution feeder. Results obtained Devices for Reliability Indices Improvement in Distribution Systems."
from the illustrative example show the expected effects of IEEE/PES Transmission and Distribution Conference and Exposition:
Latin America (2004): 77-84.
increased device failure rates: the value of SAIFI worsens and
[6] M. T. G. Donalvo, A. E. D. Tio and W. R. D. Tarnate, "Maximizing
a different optimal device configuration may be obtained. It is reliability by optimal siting of distributed generation and protective
interesting to note that the optimal solution is just drawn from a devices," Environment and Electrical Engineering (EEEIC), 2015 IEEE
small subset of key device placement schemes as the device 15th International Conference on, Rome, 2015, pp. 455-460.
failure rate is varied. This can be due to the small scale of the [7] R.Bupasiri, N.Wattanapongsakorn, J.Hokierti, and D.Coit. "Optimal
test system used. It would be interesting to explore this further Electric Power Distribution System Reliability Indices Using Binary
in a future work using a larger test system. We are also looking Programming." Annual Reliability and Maintainability Symposium
(2003): 556-561.
at extending the formulation for other indices such as MAIFI
and SAIDI as well as considering other components such as [8] J.M.Sohn, S.R.Nam and J.K.Park. "Value-based radial distribution
system reliability optimization." IEEE Transactions on Power Systems
switches, tie-lines, and distributed generation. 21.2 (2006): 941-947.
[9] R. Bil linton and R.N. Allan. Reliability evaluation of power systems.
Pitman Advanced Publishing Program, 1984.
[10] IEEE Guide for Electric Power Distribution Reliability Indices. IEEE
Standard. New York: The Institute of Electrical and Electronics
Engineers, Inc., 2004.

You might also like