Sie sind auf Seite 1von 11

This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.


IEEE TRANSACTIONS ON POWER SYSTEMS

Outage Management of Distribution Systems


Incorporating Information From Smart Meters
Yazhou Jiang, Student Member, IEEE, Chen-Ching Liu, Fellow, IEEE, Michael Diedesch, Erik Lee, and
Anurag K. Srivastava, Senior Member, IEEE

AbstractA critical function in outage management for distribution systems is to quickly detect a fault and identify the activated
protective device(s). With ongoing smart grid development, numerous smart meters and fault indicators with communication capabilities provide an opportunity for accurate and efcient outage
management. Using the available data, this paper proposes a new
multiple-hypothesis method for identication of the faulted section
on a feeder or lateral. Credibility of the multiple hypotheses is determined using the available evidence from these devices. The proposed methodology is able to handle i) multiple faults, ii) protection
miscoordination, and iii) missing outage reports from smart meters
and fault indicators. For each hypothesis, an optimization method
based on integer programming is proposed to determine the most
credible actuated protective device(s) and faulted line section(s).
Simulation results based on the distribution feeders of Avista Utilities serving Pullman, WA, validate the effectiveness of the proposed
approach.
Index TermsDistribution automation, fault diagnosis, fault indicator, multiple hypotheses, outage management, smart meter.

I. INTRODUCTION

ELIABILITY is a critical performance index for a power


system. According to customer outage statistics, more
than 80% of service interruptions originate from a fault or
component failure in power distribution systems [1]. Advanced
technologies for distribution automation such as automated
(remote-controlled) switches contribute to faster fault detection and service restoration, enhancing distribution system
reliability. Distribution automation includes the installation of
Supervisory Control And Data Acquisition (SCADA) systems,
Outage Management Systems (OMS), and Distribution Management Systems (DMS) into distribution operating centers.
Identication of the fault location and protective device activated by the fault is an important task for outage management. Traditionally, trouble call handling helps to determine the
areas affected by the outage [2]. The accuracy of the result relies
highly on the availability of trouble calls from customers. The
Manuscript received August 03, 2015; revised October 15, 2015; accepted
November 17, 2015. This work was supported by the U.S. National Science
Foundation under the award number CNS-1329666. Paper no. TPWRS-011022015.
Y. Jiang, C. -C. Liu, and A. K. Srivastava are with Washington State University, Pullman, WA 99163 USA (e-mail: yjiang1@eecs.wsu.edu; liu@eecs.wsu.
edu; asrivast@eecs.wsu.edu).
M. Diedesch and E. Lee are with Avista Utilities, Spokane, WA 99252 USA
(e-mail: michael.diedesch@avistacorp.com; erik.lee@avistacorp.com).
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TPWRS.2015.2503341

deployment of Automated Meter Reading (AMR) systems provides a supplementary way to access the customer information
for outage management. AMR based outage management relies
on polling of AMR meters. When an outage occurs, AMR meters are polled and the response is used to determine if the meters experience an outage. Methods of polling AMR meters and
the escalation method for outage management are reported in
[3][5]. Once the fault is detected, isolation of the faulted section for service restoration can be achieved quickly by use of
remote controlled switches [6], [7].
Existing methods for fault diagnosis can be classied into two
categories: 1) estimating the precise fault location [8][13]; 2)
identifying the faulted line section [2][5]. The exact fault location can be estimated using the voltage and current information from computer relays at the substation. For the purpose of
outage management, this paper is focused on identication of
the faulted line section and reconstruction of the outage scenario
for fault isolation and service restoration.
As part of the effort in smart grid development, intelligent devices have been deployed along feeders as well as on customers'
premises. These devices include automated switches, automatic
reclosers, smart meters, and fault indicators with communication capability. These intelligent devices are valuable in identifying the fault location [14][19]. A method to locate a fault
is proposed using information provided by intelligent devices
[16]. A fault location method based on voltage sags is reported
in [17]. [18] presents an algorithm to locate a fault using smart
meters on a feeder. Identication of the faulted section using
fault indicators is proposed in [19].
Although more than 50 million smart meters have been installed in the U.S. by 2013, there is not a systematic method
that enables the use of outage notications from numerous smart
meters for outage management. In contrast with AMR meters,
smart meters have the embedded function of automatic outage
reporting. When a smart meter experiences the loss of power,
it sends the outage notication. Due to limitations of communication systems, evidence from smart meters and fault indicators
can be missing or delayed. Moreover, multiple faults and protection miscoordination may occur to make it a challenge to determine the outage scenario. To address these issues, this paper
proposes a systematic method based on multiple hypotheses
[20] for outage management. Outage reports from smart meters
serve as data input for the proposed multiple-hypothesis analysis. This method can be used to quickly determine the most
credible outage scenario using data from smart meters and fault
indicators. The contributions of this paper are:

0885-8950 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
2

IEEE TRANSACTIONS ON POWER SYSTEMS

1) An optimization model based on integer programming is


proposed to determine the most credible outage scenario
of (an) actuated protective device(s) and (a) faulted line
section(s) for each hypothesis. Linearity of the proposed
optimization model ensures that the solution is optimal.
2) The new multiple-hypothesis method can handle complex
outage scenarios with multiple faults, failure of fault indicators, protection miscoordination, and missing outage
notications from smart meters.
3) A method to calculate the credibility of each hypothesis is
proposed. The credibility represents how strongly the most
credible outage scenario under this hypothesis is supported
by the available evidence from smart meters and fault indicators.
4) An Extended Protection Tree (EPT) is proposed to keep
track of notications from smart meters.
The remaining of this paper is organized as follows:
Section II describes the problem of using evidence from fault
indicators and smart meters to reconstruct the outage scenario. Section III introduces the proposed Extended Protection
Tree that keeps track of smart meter notications. Outage
management with complete outage evidence is presented in
Section IV. Section V is concerned with the proposed optimization model to determine the most credible outage scenario.
Section VI presents the simulation results based on the distribution system in Pullman, WA, served by Avista Utilities. The
conclusions and future work are stated in Section VII.
II. PROBLEM STATEMENT
Distribution systems are installed with well-coordinated protective devices. These protective devices include the breaker at
substations, reclosers, sectionalizers, and fuses along feeders or
laterals. When a fault occurs, some protective devices are designed to isolate the fault and minimize the outaged area. Actions to isolate a fault include opening of a breaker, recloser, or
sectionalizer as well as melting of a fuse.
As shown in Fig. 1, the radial topology of a distribution
system starts from the breaker at the substation downstream
through feeder sections to smart meters on the customer sites.
Upon the occurrence of a permanent fault, Recloser 1 opens
to isolate the fault. In this scenario, fault indicator 1 sends an
overcurrent ag to the distribution operating center and some
smart meters downstream of Recloser 1 report the power outage
by transmitting outage notications. Outage management is
aimed at pinpointing the faulted line section (the section represented by the dashed line) and the actuated protective device
(R1) by utilizing the outage evidence. This evidence for outage
management includes overcurrent ags from fault indicators
(FI1) and outage notications from smart meters (downstream
of R1).
Although more data for outage management can be acquired
by smart meters and fault indicators, the accuracy of fault diagnosis depends on the completeness of outage evidence. A fault
indicator may fail to send the overcurrent ag, and outage notications from smart meters may be missing. In this study, outage
notications from smart meters that should have been but have
not been received at the distribution operating center are dened as missing outage reports. A smart meter processes the

Fig. 1. Schematic of a distribution system with smart meters.

Fig. 2. Causal relationship for fault diagnosis.

sampled data. Data transmission depends on device-level communication technologies, communication towers, data collection servers, and software to process the data at the operating
center. If there are issues such as rmware problems, data tampering, software failures, and/or communication network issues,
outage notications from smart meters can be missing at the operating center. Furthermore, multiple faults and protection miscoordination may occur, making it challenging to determine the
outage scenario. In an environment where the evidence from
fault indicators and smart meters is incomplete or inaccurate,
multiple outage scenarios may be supported by the same set of
outage evidence. The multiple-hypothesis method is proposed
to deal with incomplete or inaccurate evidence. Each hypothesis
may involve multiple faults, failure of fault indicators, protection miscoordination, and missing smart meter reports.
The relationship among the cause, effect, and evidence for
an outage scenario is illustrated in Fig. 2. For each scenario of a
faulted line section and an actuated protective device, fault indicator ags and smart meter notications consistent with this scenario are acquired. By comparing the evidence with the effect,
the credibility of each scenario is quantied. Thus the most credible outage scenario consisting of the faulted line section and
actuated protective device is determined by ranking the level of
credibility.
III. EXTENDED PROTECTION TREE
Upon occurrence of a fault followed by resulting actions of
protective devices to isolate the fault, numerous notications
from smart meters reporting a power outage are collected by
the control center. To organize these smart meter notications,
an Extended Protection Tree is proposed. An EPT contains information on how well each actuated protective device is correlated with smart meter notications.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
3

JIANG et al.: OUTAGE MANAGEMENT OF DISTRIBUTION SYSTEMS INCORPORATING INFORMATION FROM SMART METERS

Fig. 3. Extended Protection Tree.

A. Extended Protection Tree


Based on the radial topology of a distribution system, the EPT
can be applied to model the spatial connection of protective devices. Customers are served from feeders and laterals that are
protected by devices such as reclosers, sectionalizers, and fuses.
Thus, the EPT can be extended to incorporate smart meters representing customers. Some customers may be served from multiple phases, e.g., three phase customers fed by three phase laterals, and each phase is protected by a fuse. To handle the case of
one customer served from multiple phase fuses, phase fuses at
the same location can be grouped into a fuse set. Then the EPT,
as shown in Fig. 3, that incorporates smart meters can be applied
to determine the faulted sections regardless of the faulted phase.
B. Interpretation of the Extended Protection Tree
In an EPT, vertices represent protective devices and smart
meters, and edges indicate the spatial connection (upstream and
downstream relationship) of devices. If a smart meter reports
an outage, the status of the corresponding vertex is 1. Otherwise, the smart meter status is 0. The Escalation method [3]
is applied to determine the number of smart meter outage notications correlated with a protective device. Three indices,
, are associated with each vertex that
represents the protective device.
: number of smart meters downstream of the
protective device reporting a power outage
: total number of smart meters downstream of the
protective device
: number of protective devices on the path to the
breaker at the substation
Generally, the topology of a distribution system is radial.
The EPT, therefore, serves as a universal method to model the
spatial connection of protective devices and customer meters.
EPT is utilized in the proposed multiple-hypothesis method for
outage management. For the given fault location in Fig. 1, the
three indices associated with the vertex representing R1 are
,
, and
as shown in Fig. 3.
IV. OUTAGE MANAGEMENT WITH COMPLETE
OUTAGE EVIDENCE
After a fault occurs in a distribution system, smart meters report the outage by sending notications to the distribution oper-

ating center. Fault indicators detect the fault and transmit ags.
If the outage evidence is complete without failures of fault indicators or missing outage reports from smart meters, the outage
scenario with credibility index 1 is identied.
Taking the outage scenario in Fig. 1 as an example, R1 is
opened to isolate the fault. Without any abnormalities, such as
failures of fault indicators, missing outage reports from smart
meters, or protection miscoordination, the distribution operating
center would receive the complete outage evidence, i.e., the ag
from FI1 and outage notications from SM4-SM15.
1) Outage evidence from fault indicators to determine the
faulted line section
The line section, locating right downstream of the fault
indicator which sends the ag and has the high level in
EPT, is determined to be the faulted section. In the given
example,
a ag from FI
the faulted line section is between FI and of FI
Based on the evidence from fault indicators, the credibility
of the hypothesized faulted line section is

2) Outage evidence from smart meters to determine the actuated protective device
The protective device corresponding to the maximum
number of outage notications and no smart meter downstream of which fails to report the outage is the actuated
device. In this example,
notications from SM

SM
actuates to isoalte the fault

Based on the evidence from smart meters, the credibility


of the actuated protective device is

3) Credibility of the outage scenario with complete outage


evidence
The ag from the fault indicator is used to determine the
faulted line section, and outage notications from smart
meters indicate the actuated protective devices. In the
given example, the hypothesized outage scenario is that
R1 is actuated to isolate the faulted line section represented by dashed lines. Thus, the credibility of the outage
scenario with complete outage evidence is

V. MULTIPLE-HYPOTHESIS ANALYSIS
In practice, the fault indicator ags and smart meter notications can be missing or delayed due to hardware failures, software/rmware issues, and communication network problems.
Incomplete outage evidence increases the uncertainty for hypothesized outage scenarios. Moreover, outage data from unrelated outage scenarios, multiple faults, and miscoordination of

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
4

IEEE TRANSACTIONS ON POWER SYSTEMS

protective devices can occur. Therefore, in this work, a multiple-hypothesis method is proposed to handle both simple and
complex outage scenarios.
A. Multiple Hypotheses
To handle an incomplete or inaccurate evidence set, a multiple-hypothesis method is proposed to determine how well each
outage scenario is supported by the available evidence. Hypotheses are generated automatically by various combinations
of the number of faults, number of fault indicator failures, and
number of protection miscoordination pairs. For each hypothesis, the optimization model based on integer programming is
developed to determine the most credible outage scenario supported by the evidence. The result of the most credible scenario
includes the actuated protective device(s), faulted section(s),
failure of a fault indicator, and smart meters' failure to report
a power outage. Credibility based on the available evidence is
quantied for each hypothesis using the proposed method.
B. Optimization Model
1) Objective Function: The objective is to identify the actuated protective device, which is correlated with the maximum
smart meter notications. As observed from the EPT, two protective devices with an upstream and downstream relationship
may correlate with the same number of notications from smart
meters. If these two protective devices both satisfy the constraints, the one downstream is judged to be more credible as
the actuated device. The reason is that fewer smart meters are
correlated with the protective device downstream compared to
the one upstream. Thus, the objective function, as expressed in
(1), is set to nd the protective device collecting the most smart
meter notications with the highest level in the EPT. Note that
the level is higher for a downstream node.
Maximize

(1)

where
represents the number of smart meter notications correlated with a protective device;
indicates
the level of a protective device in the EPT;
is a large number
so that the inequality
.
If two protective devices correlate with different numbers
of smart meter notications, the minimum difference of the
number of collected smart meter notications is one. Therefore,
by imposing
, the objective to identify
the protective device correlated with the maximum smart meter
notications is not compromised by
.
The binary decision variable
is used to represent
the actuation of the
protective device due to the
fault.
Otherwise,
. With the binary decision variables , the
values of
and
are given by
(2)
(3)

is the level of the


protective device in the EPT.
is chosen
to be
, where
is the maximum number
of concurrent faults in the system, and
is the number of
protective devices. In this way,
is satised.
Substituting
and
into (1), the objective function becomes
Maximize
(4)
and
are determined from the EPT;
and
are constants depending on the system. The
objective function, therefore, relies only on the binary decision
variables .
2) Constraints: For each hypothesis, the actuated protective
device(s), faulted section(s), and status of fault indicators
determined from the optimization model must satisfy logical
constraints. These logical constraints indicate that the actuated
protective device(s) can only be upstream of the faulted line
section(s) and fault indicators upstream of the faulted sections(s) should be agged. These underlying logical constraints
are modeled as follows:
a) For each fault, only one protective device is actuated to
isolate the fault.
where

(5)
b) For each fault, only one line section can be faulted. The
binary decision variable
is 1 if the
line section is
the location of the
fault. Otherwise,
is 0.
is the set of line sections.
(6)
c) For each fault, the fault indicators upstream of the faulted
section should detect the fault current and be agged.
The binary decision variable
is 1 if the
fault indicator is upstream of the
fault. Otherwise,
is 0.
indicates the set of line sections downstream of the
fault indicator.
is the set of fault
indicators.
(7)
d) Since a fault indicator may be upstream of multiple
faulted sections, it should be agged if it is upstream of a
fault. The binary decision variable
is 1 if the
fault
indicator is upstream of a fault. Otherwise, is 0. is a
large number, which is chosen as any number larger than
.
(8)

is the set of faults;


denotes the set of protecwhere
tive devices;
represents the number of smart meter notications correlated with the
protective device;

e) The number of fault indicator failures is the summation of


the fault indicators that should not detect a fault current

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
JIANG et al.: OUTAGE MANAGEMENT OF DISTRIBUTION SYSTEMS INCORPORATING INFORMATION FROM SMART METERS

but send overcurrent ags, and the ones that should detect
a fault current but do not report it.
is the status
of the
fault indicator collected by the control center.
is 1 if the
fault indicator sends an overcurrent
ag. Otherwise, it is 0.
is the number of fault
indicator failures.
(9)
In order to linearize (9),
associated constraints:

where
fails. With
formulated as:

is introduced with its

indicates whether the


replaced by

fault indicator
, (9) is re-

(15)
In the proposed optimization model, the objective function
and constraints are linearly dependent on the binary decision
variables. For each hypothesis corresponding with the number
of faults, number of fault indicator failures, and number of
protection miscoordination pairs, the optimization problem is
solved to determine the most credible outage scenario based
on smart meter notications. Linearity of the proposed model
ensures the optimality of the solution for each hypothesis. For
different hypotheses, the credibility levels are calculated. The
credibility of hypotheses is intended to quantify the strength of
outage evidence to support the hypothesized outage scenario.
Multiple hypotheses are ranked by the credibility index. As
a result, the most credible hypothesis and the corresponding
outage scenario are determined.
C. Credibility of Hypotheses

(10)
is utilized to indicate
f) The binary decision variable
whether the
protective device is upstream of the
faulted section where the
fault locates. If it is,
is
1. Otherwise, it is 0.
is the set of line
sections downstream of the
protective device.
(11)
g) The decision variable
is used to indicate whether the
protective device is upstream of the actuated protective device because of the
fault or status of the actuated protective device itself.
is the set of
protective devices that are downstream of the
protective device.
(12)
h) For each fault, the corresponding actuated protective device should be upstream of the faulted section.
(13)
i) The number of protection miscoordination pairs
is expressed as the number of protective
devices that detect a fault current and are downstream of
the actuated protective device(s).
(14)
j) Multiple faults should be isolated by protective devices
in different branches (one fault should not be upstream of
any other fault).
is the set of protective
devices upstream of the
protective device.

For each hypothesis, it is likely that the most credible outage


scenario is supported by only some of the outage evidence. The
evidence from smart meter notications and fault indicators are
taken into account to quantify the credibility of each hypothesis.
1) Evidence of Smart Meter Outage Notifications: The credibility associated with smart meter evidence is dened by
(16)
where
is the number of smart meters reporting
a power outage correctly.
is the number of
smart meters that should not report an outage but they do.
indicates the number of smart meters that should
report an outage but they do not.
, representing the evidence from smart meters, is negative
if
. If all the smart meter notications comply with the identied outage scenario,
falls
within the range from 0 to 1.
2) Evidence of Fault Indicator Flags: The credibility associated with fault indicator evidence is given by
(17)
is the number of fault indicators that should
where
be agged to comply with the outage scenario determined from
the optimization model.
is the number of available fault indicator ags. If there is a fault indicator that should
not send a ag but it does, then
.
On the other hand, if a fault indicator fails to detect the fault
current, then
.
3) Credibility of Each Hypothesis: The credibility of each
hypothesis is expressed as the weighted sum of
and
with
weighting factors
, and
. That is
(18)
are chosen based on operators' experience with the
where
reliability of smart meters reporting a power outage and fault

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
6

IEEE TRANSACTIONS ON POWER SYSTEMS

Fig. 4. Simplied one-line diagram of FD1.

TABLE I
SYSTEM PARAMETERS OF FD1.

Fig. 5. Simplied FD1 conguration for a single fault scenario.

TABLE II
SMART METER NOTIFICATIONS FOR A SINGLE FAULT SCENARIO.

indicators sending overcurrent ags. If some devices are more


reliable and trustworthy, a larger number for the weight
can
be chosen.
VI. CASE STUDY
Through a grant from the U.S. Department of Energy, Avista
installed 13,000 smart meters for all electricity customers in
Pullman. The project is intended to demonstrate the safety, reliability, and efciency of energy delivery enhanced by smart
grid technology [21]. With two-way communication via a secure wireless network, smart meters deployed across Pullman
can automatically detect outages at a home or business. However, a systematic method to utilize smart meter data for outage
management is not available at present. The proposed methodology helps operators quickly detect an outaged area and accurately reconstruct the outage scenario.
A. Pullman System FD1
The feeder FD1 in Pullman serves as the test case to validate the proposed methodology for outage management. Based
on the actual system data provided by Avista Utilities, the simplied one-line diagram of FD1 is shown in Fig. 4. The base
voltage of the feeder is 13.2 kV, and 1,320 customers are served
by the feeder with an annual energy consumption of 2,558,673
kWh.
The feeder-relevant information is given in Table I. Sectionalizers are not deployed in the Pullman distribution systems.
Three automatic reclosers and one automated switch along FD1
sectionalize the entire feeder into segments. These automated
devices as well as the digital relay can send overcurrent ags
when a fault is detected. Thus these devices are regarded as fault
indicators with communication capability. The breaker at the
substation, 3 reclosers along the feeder, and 33 fuses at the lateral are the protective devices protecting the 579 line sections.
In this study, hypotheses are generated automatically based
on the assumptions that at most two concurrent faults, one fault
indicator failure, and one pair of protection miscoordination
can occur in a scenario. Each hypothesis corresponding to the
number of faults, number of fault indicator failures, and number
of protection miscoordination pair is modeled in the constraints

of the proposed optimization model. To solve the optimization


problem, the IBM ILOG CPLEX is used as the optimizer.
Although the case studies are based on the assumption of two
concurrent faults, outage scenarios with multiple faults in general can be handled using the proposed method. Hypotheses
are automatically generated based on the possible maximum
number of faults, number of possible fault indicator failures,
and pairs of miscoordination of protective devices. These maximum numbers are determined for the proposed method based
on the experience of distribution operators' about outage scenarios and operating conditions of distribution systems. The
maximum numbers can be changed as needed and applied to
the proposed analysis.
1) A Single Fault Scenario: To demonstrate the proposed
method for outage management, the simplied system conguration for a single fault scenario is shown in Fig. 5. A link in
Fig. 5 is dened as the set of consecutive line sections protected
by the same protective devices. The true outage scenario is that
Link4 is faulted and Fuse20 serves to isolate the fault. Note that
the scenario of interest is the one in which smart meters report
a permanent fault, thus only the last gasp from smart meters
is utilized for outage management. In this paper, the last gasp
refers to the smart meter outage report after the switching sequence associated with the fault is settled.
After the fault takes place as shown in Fig. 5, the distribution
operating center collects the outage evidence: 1) fault indicator
ags from Brk, R1, and R2; 2) smart meter notications given
in Table II.
A total of 50 out of 87 smart meters connected with Link4,
Link5, Link6, and Link7 report the loss of power. The EPT is
constructed to organize the smart meter notications in Fig. 6.
Although Brk, R1, R2, and Fuse20 are all correlated with 50

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
JIANG et al.: OUTAGE MANAGEMENT OF DISTRIBUTION SYSTEMS INCORPORATING INFORMATION FROM SMART METERS

TABLE IV
SMART METER NOTIFICATIONS FOR A MULTIPLE-FAULT SCENARIO.

Fig. 6. Extended Protection Tree for a single fault scenario.


TABLE V
SMART METER NOTIFICATIONS FOR FD2.
TABLE III
HYPOTHESES FOR MULTIPLE-HYPOTHESIS ANALYSIS.

smart meter notications, Fuse20 has the highest level in the


EPT. Meanwhile, no smart meter notications come from the
link downstream of Fuse21. With the power outage data and
system topological data imported into the optimization model,
the multiple-hypothesis analysis is launched to determine the
most credible outage scenario for each hypothesis.
Under the assumption that at most two concurrent faults, one
fault indicator failure, and one pair of protection miscoordination can occur in a distribution system, a total of 8 hypotheses
are automatically generated as shown in Table III. For instance,
in hypothesis 1, there is one fault without any fault indicator
failures or protection miscoordination for the outage scenario.
The result of the multiple-hypothesis analysis for a single
fault scenario is given in Table VI. For each hypothesis, the optimization problem is solved to determine the most credible actuated protective device(s) and faulted section(s). To comply with
the scenario, the proposed method determines the correctness of
smart meter outage reports as well. This correctness information
includes the number of smart meters that report a power outage
correctly, the number of smart meters that should report a power
outage whereas they do not, and the number of smart meters that
should not report a power outage whereas they do. Moreover,
the abnormality of a fault indicator failure and protection miscoordination can be identied. In this case, both
and
are
chosen to be 0.5. After
and
are calculated based on the
evidence from smart meters and fault indicators, the credibility
of each hypothesis is determined as shown in Table VI.
Based on the multiple-hypothesis analysis, hypothesis 1 and
hypothesis 2 are the most credible outage scenarios with the

highest credibility of 0.787. In hypothesis 1, without any failures


of fault indicators or miscoordination of protective devices, the
most credible scenario is that the fault takes place at Link4 and
the upstream protective device, Fuse20, is blown to isolate the
fault. R1 and R2 send the outage ags whereas R3 does not. If all
protective devices are well coordinated, the actuated protective
device is located upstream of the hypothesized faulted line section without protective devices in between. For this hypothesis
without miscoordination of protective devices, the most credible outage scenario is determined by the proposed method.
Both hypothesis 1 and hypothesis 2 identify Fuse20 as the actuated protective device, but different line sections are faulted.
If Fuse14/16-Fuse20 is miscoordinated, Fuse20 needs to actuate
due to the fault at Link6/5. In this case, even though the outage
scenarios are different in hypothesis 1 and hypothesis 2, both
hypotheses collect the same outage evidence. Therefore, these
two hypotheses are ranked at the same level of credibility. The
phenomena result from insufcient fault indicators deployed in
the system. If a fault indicator is installed at the end of Link4, the
fault indicator would help to distinguish the outage scenarios in
hypothesis 1 and hypothesis 2. Most importantly, the true scenario with Fuse20 blown and Link4 to be faulted is successfully
captured by the multiple-hypothesis analysis.
2) A Multiple-Fault Scenario: Fig. 7 shows the simplied
FD1 conguration for a multiple-fault scenario. The true outage
scenario is that two faults occur at Link4 and Link10, respectively, and Fuse20 and Fuse19 serve to isolate the two faults.
After the two fuses blow, the distribution operating center receives the power outage data: 1) overcurrent ags from Brk,

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
8

IEEE TRANSACTIONS ON POWER SYSTEMS

TABLE VI
MULTIPLE-HYPOTHESIS ANALYSIS FOR A SINGLE FAULT SCENARIO.

TABLE VII
MULTIPLE-HYPOTHESIS ANALYSIS FOR A MULTIPLE-FAULT SCENARIO.

Fig. 7. Simplied FD1 conguration for a multiple-fault scenario.

Fig. 8. Extended Protection Tree for a multiple-fault scenario.

R1, R2, and R3; 2) notications from smart meters shown in


Table IV.
A total of 66 smart meters from Link4, Link5, Link6, Link7,
Link8, Link10, and Link11 report the power outage. Compared
to the true outage scenario, an abnormality of smart meters is
that two smart meters at Link8 send outage notications as well,
which may result from cyber attacks or outage data from unrelated outage scenarios.
The EPT is shown in Fig. 8. The same hypotheses as for a
single fault scenario are generated in Table III. From the result
of the multiple-hypothesis analysis presented in Table VII,
hypothesis 5 is the most credible outage scenario with the
highest credibility 0.742. Fuse20 and Fuse19 are determined as
the actuated protective devices. In hypothesis 5, with Fuse20
and Fuse19 blown, 64 of 128 smart meters report the power
outage correctly. However, 2 smart meter notications from
Link8 contradict with the outage scenario. Such an abnormality
of smart meters reporting a power outage can be successfully
captured by the multiple-hypothesis analysis. Furthermore,
the data from the 64 smart meters, which should report the
power outage whereas they do not, may be lost during the data
transmission. In summary, the proposed method successfully
identies the true outage scenario that Fuse19 and Fuse20 blow
to isolate the two faults at Link4 and Link10, respectively.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
JIANG et al.: OUTAGE MANAGEMENT OF DISTRIBUTION SYSTEMS INCORPORATING INFORMATION FROM SMART METERS

TABLE VIII
MULTIPLE-HYPOTHESIS ANALYSIS

FOR

FD2.

Fig. 10. System congurations.


Fig. 9. Simplied one-line diagram of FD2.

B. Pullman System FD2


Feeder FD2 in Pullman, as shown in Fig. 9, is used to demonstrate how the multiple-hypothesis analysis can be used to identify miscoordination of protective devices. On FD2, three automatic reclosers are available to sectionalize the feeder into segments. In addition, two automated switches are deployed. FD2
is serving a total number of 1,975 customers. Including the automatic reclosers, a total of 73 protective devices are installed
on the system. The true outage scenario is that R1 is actuated
to isolate the fault on Link8 in Fig. 9. The collected outage evidence is: 1) ags from Brk, R1, R3, and AS2; 2) smart meter
outage notications in Table V.
The same hypotheses as for FD1 are generated in Table III.
Results from the proposed method for outage management are
summarized in Table VIII. From the table, the most credible
outage scenario is that R1 actuates to isolate the faulted line section of Link8. A total of 534 out of 1178 smart meters report the
outage correctly. Based on the available evidence, R3 and R1
are miscoordinated. Note that the credibility of hypothesis 6 is
close to that of hypothesis 2. Furthermore, no smart meter downstream of Fuse11 sends the power outage notications. Thus, it
is concluded that only one fault occurs in the outage scenario.
The multiple-hypothesis analysis successfully captures the true
outage scenario that R1 is actuated to isolate the faulted line
section, Link8. Moreover, miscoordination of devices R1-R3 is
identied by the proposed method.
C. Relationship Between the Result of the Case Studies and
Constraints in the Optimization Model
This section serves to illustrate how the results of the cases
satisfy the constraints in the proposed optimization model.

Constraints a) and b) show that, for each fault, one protective


device should be actuated.
Constraints c)-e) require that only the fault indicators upstream of the fault(s) send ags and the number of fault indicator failures is counted.
Constraints f)-i) impose the condition that, for each fault, the
actuated protective device is upstream of the faulted line section,
and the protection miscoordination pair(s) is(are) counted.
Constraint j) indicates that, for multiple-fault scenarios, a different fault should be isolated by different protective devices.
Taking the result of hypothesis 1 in the single fault scenario
of FD1 as an example, there is one fault, no protection miscoordination, or failure of fault indicators. The hypothesized outage
scenario is that Fuse20 melted to isolate the faulted line section,
Link4.
The scenario satises Constraints a) and b) as one protective device is actuated to isolate the fault. In the hypothesized
outage scenario, the devices upstream of Link4, R1 and R2, send
the fault current ags, whereas other devices, such as R3, do
not. Therefore, the number of fault indicator failures is zero,
and the result satises Constraints c)-e). In the outage scenario,
the actuated protective device, Fuse20, is right upstream of the
faulted line section, Link4, without protective devices in between. Therefore, no protection miscoordination exists and the
result is consistent with Constraints d)-i). In the outage scenario
that Fuse20 is actuated to isolate Link4, Constraint j) is satised. The same analysis is applicable to other cases.
D. Existence of the Solution in the Optimization Model
For some hypotheses, the optimization model may not have a
feasible solution. An example of this is shown in Fig. 10(a), in
which FI1, FI2, and FI3 send ags. The given hypothesis is that
one fault occurs without a fault indicator failure or protection
miscoordination.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
10

IEEE TRANSACTIONS ON POWER SYSTEMS

Based on the evidence from fault indicators, there are at least


two faults downstream of FI2 and FI3, respectively. However,
the condition of one fault is given in the hypothesis, which results in a conict with the fault indicator evidence. Thus, for
some hypotheses, a feasible solution may not exist.
E. Sensitivity With Respect to Protective Devices, Fault
Indicators, and Smart Meters
Placement of the devices, such as fault indicators, automatic
reclosers, and smart meters, has an impact on identication of
the faulted line section. Take, for example, the addition of a
new smart meter into the system. For an outage scenario, if the
added smart meter should report the outage whereas it does not,
the credibility of the outage scenario would be lower than the
one without the new smart meter. Indeed, the credibility after
the additional smart meter is installed is reduced since the new
meter does not perform as expected:

On the other hand, if new devices to be installed add to the


outage evidence, it will improve the result of the proposed multiple-hypothesis analysis. The same argument can be made for
addition of automatic reclosers and fault indicators with communication capability.
The allocation of these devices does not directly impact the
outage evidence for the proposed multiple-hypothesis analysis.
However, identication of faults on line sections depends on
the allocation of protective devices and fault indicators. Optimal placement of these devices in distribution system planning
needs to take outage management into account.

research is needed to use smart meter data for accurate fault


location estimation.
APPENDIX
UNIQUENESS OF THE OPTIMAL SOLUTION
The proposed optimization model is designed to determine
the most credible outage scenario using the evidence from fault
indicators and smart meters. The uniqueness of optimality is
discussed as follows:
If two protective devices both achieve optimality of the optimization model, then the two protective devices meet the following two conditions. That is
Condition 1: both protective devices are correlated with the
same number of smart meter notications. That is,

Condition 2: both protective devices have the same levels


in the EPT, i.e.,

Without loss of generality, the hypothesis of one fault without


fault indicator failures or protection miscoordination is considered. The typical system congurations, satisfying Condition 2,
are shown in Fig. 10(b) and (c). In the case of Fig. 10(b), protective device is upstream of protective device and protective
device . The proposed optimization model will determine protective device as the most credible actuated protective device,
and uniqueness of the optimal solution is achieved. For the conguration in Fig. 10(c), if protective device
and protective
device are correlated with an identical number of smart meter
notications, then both protective devices would lead to the optimal solution. In this case, the optimal solution is not unique.
However, this conguration of a 3-terminal junction without a
protective device on each terminal is rare. Taking the Pullman
system as an example, there is no such conguration in the entire system with 13 feeders.
If the rare cases occur, more fault indicators may need to be
deployed to assist distribution operators in decision making of
the most credible outage scenario.

VII. CONCLUSIONS AND FUTURE WORK


This paper presents a novel technique based on multiple-hypothesis method and EPT for outage management in
distribution systems. The proposed multiple-hypothesis method
can handle complex outage scenarios involving multiple faults,
missing outage reports from fault indicators and smart meters,
and protection miscoordination. For each hypothesis, a linear
optimization model based on integer programming is proposed
to determine the most credible actuated protective device(s)
and faulted line section(s) supported by the evidence. Linearity
ensures optimality of the solution. The most credible outage
scenario can be quickly reconstructed using the proposed
method. Simulation results based on distribution systems in
Pullman validate the method.
Future work includes extension of the proposed method
to handle meshed distribution systems with micro-grids and
distributed resources for outage management. Furthermore,

REFERENCES
[1] A. A. Chowdhury, Power Distribution System Reliability: Practical
Methods and Applications. Hoboken, NJ, USA: Wiley, 2009.
[2] W. R. Cassel and M. Corporation, Distribution management systems:
Functions and payback, IEEE Trans. Power Syst., vol. 8, no. 3, pp.
796801, Aug. 1993.
[3] Y. Liu and N. N. Schulz, Knowledge-based system for distribution
system outage locating using comprehensive information, IEEE
Trans. Power Syst., vol. 17, no. 2, pp. 451456, May 2002.
[4] S. T. Mak, A synergistic approach to using AMR and intelligent electronic devices to determine outages in a distribution network, in Proc.
Power Syst. Conf., Clemson, SC, USA, 2006.
[5] R. A. Fischer, A. S. Laakonen, and N. N. Schulz, A general polling
algorithm using a wireless AMR system for restoration conrmation,
IEEE Trans. Power Syst., vol. 16, no. 2, pp. 312316, May 2001.
[6] Y. Xu, C. -. Liu, K. P. Schneider, and D. T. Ton, Placement of remote-controlled switches to enhance distribution system restoration capability, IEEE Trans. Power Syst., to be published.
[7] Y. Xu, C. -C. Liu, and H. Gao, Reliability analysis of distribution
systems considering service restoration, in Proc. IEEE PES Conf. Innovative Smart Grid Technol., 2015, pp. 15.

This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
JIANG et al.: OUTAGE MANAGEMENT OF DISTRIBUTION SYSTEMS INCORPORATING INFORMATION FROM SMART METERS

[8] M. S. Choi, S. J. Lee, D. S. Lee, and B. G. Jin, A new fault location


algorithm using direct circuit analysis for distribution systems, IEEE
Trans. Power Del., vol. 19, no. 1, pp. 3541, Jan. 2004.
[9] M. S. Choi, S. J. Lee, S. I. Lim, D. S. Lee, and X. Yang, A direct
three-phase circuit analysis-based fault location for line-to-line fault,
IEEE Trans. Power Del., vol. 22, no. 4, pp. 25412547, Oct. 2007.
[10] J. Zhu, D. L. Lubkeman, and A. A. Girgis, Automated faulted location and diagnosis on electric power distribution feeders, IEEE Trans.
Power Del., vol. 12, no. 2, pp. 801809, Apr. 1997.
[11] Y. Liao, Generalized fault location methods for overhead electric distribution systems, IEEE Trans. Power Del., vol. 26, no. 1, pp. 5364,
Jan. 2011.
[12] J. Kim, M. Baran, and G. C. Lampley, Estimation of fault location on
distribution feeders using PQ monitoring data, in Proc. IEEE Power
Eng. Soc. General Meeting, Tampa, FL, USA, Jun. 2428, 2007.
[13] R. H. Salim, M. Resener, A. D. Filomena, K. R. Oliveira, and A. S.
Bretas, Extended fault-location formulation for power distribution
systems, IEEE Trans. Power Del., vol. 24, no. 2, pp. 508516, Apr.
2009.
[14] B. D. Russell and C. L. Benner, Intelligent systems for improved reliability and failure diagnosis in distribution systems, IEEE Trans. Smart
Grid, vol. 1, no. 1, pp. 4856, Jun. 2010.
[15] M. Kezunovic, Smart fault location for smart grids, IEEE Trans.
Smart Grid, vol. 2, no. 1, pp. 1122, Mar. 2011.
[16] Y. Gong and A. Guzman, Integrated fault location system for power
distribution feeders, IEEE Trans. Ind. Appl., vol. 49, no. 3, pp.
10711078, May/Jun. 2013.
[17] R. A. F. Pereira, L. G. W. Silva, M. Kezunovic, and J. R. S. Mantovani,
Improved fault location on distribution feeders based on matching
during-fault voltage sags, IEEE Trans. Power Del., vol. 24, no. 2, pp.
852862, Apr. 2009.
[18] F. C. L. Trindade, W. Freitas, and J. C. M. Vieira, Fault location in
distribution systems based on smart feeder meters, IEEE Trans. Power
Del., vol. 29, no. 1, pp. 251260, Feb. 2014.
[19] J. -H. Teng, W. -H. Huang, and S. -W. Luan, Automatic and fast
faulted line-section location method for distribution systems based
on fault indicators, IEEE Trans. Power Syst., vol. 29, no. 4, pp.
16531662, Jul. 2014.
[20] J. Jung, C. -C. Liu, M. Hong, M. Gallanti, and G. Tornielli, Multiple
hypotheses and their credibility in on-line fault diagnosis, IEEE Trans.
Power Del., vol. 16, no. 2, pp. 225230, Apr. 2001.
[21] Avista Utilities's Pullman Smart Grid Project, Avista Utilities [Online]. Available: https://www.smartgrid.gov/document/
utility_scale_smart_meter_deployments_plans_proposals.html

11

Yazhou Jiang (S'12) received the B.E. degree from Huazhong University of
Science and Technology, Wuhan, China, in 2011. He is currently pursuing his
Ph.D. degree at Washington State University, Pullman, WA.
His research interests include power system restoration, distribution automation, and optimization methods for power system applications.

Chen-Ching Liu (F'94) received his Ph.D. from University of California,


Berkeley, CA.
He is Boeing Distinguished Professor at Washington State University,
Pullman, WA. He was Palmer Chair Professor of EE at Iowa State University,
Ames, IA, and a Professor of EE at the University of Washington, Seattle, WA.
Dr. Liu received an IEEE Third Millennium Medal in 2000 and the Power
and Energy Society Outstanding Power Engineering Educator Award in 2004.
He was recognized with a Doctor Honoris Causa from University Politehnica
of Bucharest, Romania.

Michael Diedesch received his B.S. in Electrical Engineering from Washington


State University in 2008.
Since joining Avista Utilities in Spokane, WA, he has worked in various engineering roles including Distribution Engineering, SCADA and System Protection. He is currently Manager of Avista's Electric Meter Shop. Mr. Diedesch
is a registered Professional Engineer in Washington.

Erik Lee received his B.S. in Electrical Engineering from Washington State
University in 2007. He currently works in Distribution System Operations. Mr.
Lee is a registered Professional Engineer in Washington.

Anurag K. Srivastava (S'00M'05SM'09) received his Ph.D. from Illinois


Institute of Technology (IIT), Chicago, in 2005.
He is an Associate Professor at Washington State University, Pullman, WA.
He worked as an Assistant Research Professor at Mississippi State University
from 20052010. His research interests include power system operation and
control using smart grid data.
Dr. Srivastava is Past-Chair of the IEEE PES Career Promotion Subcommittee, Chair of the IEEE PES Student Activities Subcommittee, Co-Chair of
Micro-grid Working Group, and past-chair of IEEE ICAP synchrophasor conformity assessment. He is an editor of the IEEE TRANSACTIONS ON SMART GRID
and a distinguished lecturer of the IEEE PES.

Das könnte Ihnen auch gefallen