2, MAY 2002
451
I. INTRODUCTION
452
Fig. 1.
power system, such as unit commitment, power flow and stability analysis, which require large numerical programs, outage
management requires logic-based manipulation like a human
mind. An AI technique can provide a way to capture the experience of human operators and engineers.
Although the previous work tried to deal with system outages
depending on the trouble calls or AMRs, no single data source
can provide enough accurate outage information. For example,
a midnight power outage may not have enough customer calls.
The wireless communication devices on the automatic meters
may not send correct information during bad weather or during
a low battery condition. Distribution system SCADA is more
reliable compared with the previously mentioned two information sources. The distribution SCADA has been primarily used
for system analysis and operation purposes instead of outage
management. Moreover, due to the vast size and complexity of
the distribution system, monitoring lower levels of the system
may not be affordable, especially for smaller companies. Consequently, distribution SCADA equipment may not be available to
provide information at lower levels of the distribution system including the customer level. A data filter was developed to combine all of this available information and provide higher quality
outage information. The data filter is introduced in [5].
The objective of this project is to use all the available outage
information and apply an AI method to provide a more effective distribution outage-locating algorithm. The overview of the
whole project is shown in Fig. 1.
The outage-locating algorithm involves two major parts: primary outage locating and outage location confirmation by meter
polling. Fig. 2 shows the flowchart of this outage-locating algorithm.
B. Primary Outage Locating
The primary locating procedure used the traditional outage
escalation method to provide the outage region and the question
points. Escalation is roughly defined as raising the level of a
job from a downstream device to a device upstream [1]. This
operation is based on the assumption that it is more likely for
one device to fail than several. With this assumption, the outage
escalation tool searches the outage reports for common points
of connectivity. The primary locating procedure provides the
questioned devices that need to be checked.
LIU AND SCHULZ: KNOWLEDGE-BASED SYSTEM FOR DISTRIBUTION SYSTEM OUTAGE LOCATING
453
454
Fig. 4.
Fig. 5.
LIU AND SCHULZ: KNOWLEDGE-BASED SYSTEM FOR DISTRIBUTION SYSTEM OUTAGE LOCATING
455
TABLE I
METER-POLLING PROCEDURE OF TEST CASE 1
Fig. 7.
TABLE III
TEST RESULT OF OUTAGE SCENARIOS OF CASE 3
Fig. 6.
TABLE IV
SUMMARY OF THE TEST SCENARIOS
identification in each scenario, it shows that the proposed algorithm has over 60% better performance than the traditional
escalation method. The total processing time is 3 s, excluding
meter-polling time. As addressed in [11], average meter-polling
time is less than 12 s/m.
456
V. CONCLUSION
In this project, an outage locating and confirmation knowledge-based system was designed, developed and tested. The
purpose of this study was to find an efficient method to use the
comprehensive outage information and provide a fast and effective way to locate the distribution outages. This work represents
an extension of some preliminary work that involved using automated meter reading in the distribution system outage handling
[4]. A knowledge-based system (KBS) was developed to provide an integrated approach to solving the challenge of outage
locating. It creates a list with the possible outage locations and
the additional checking suggestions. The KBS can take the filtered outage information from the fuzzy data filter [5]. It performs the outage escalation to find the primary outage locations
and then conduct the meter polling to confirm the locations.
Some major contributions of this work include the following:
1) using the comprehensive outage information from trouble
calls, wireless automated meter readings (AMR) and distribution SCADA for the outage locating;
2) studying the heuristics and rules for outage locating;
3) designing and developing a meter-polling scheme for the
outage location confirmation;
4) using the modern expert system shell G2 to provide advanced operation methods and graphical interface.
The on-demand read feature of the AMR system provides
utilities the convenience of indirectly communicating with the
customers. The meter-polling scheme designed in this project
provides an effective way to confirm the outage locations by
reading the meter reports in the meter controller.
While our test cases used specific systems for outage information, the algorithm was designed to use general data from
various outage information sources. With further development
and testing, this intelligent outage handling system can be integrated into daily distribution system operations for locating
outages. It will help improve the quality of the outage handling,
reduce outage-locating errors, and reduce the cost of the outages.
REFERENCES
[1] E. P. Laverty and N. N. Schulz, A improved algorithm to aid in post-heat
storm restoration, IEEE Trans Power Syst., vol. 14, pp. 446450, May
1999.
Yan Liu (M01) received the B.S.E.E. and M.S.E.E. degrees from Zhejiang University, Hangzhou, China, and the Ph.D. degree from Michigan Technological
University (MTU), Houghton, in 1995, 1998, and 2001, respectively.
Currently, she is with the GE Global Research Center, Niskayuna, NY. Her
research interests are computer applications in power system analysis, distribution management systems, and distributed generation.
Noel N. Schulz (S93M96SM00) received the B.S.E.E. and M.S.E.E. degrees from Virginia Polytechnic Institute and State University (Virginia Tech),
Blacksburg, and the Ph.D. degree in electrical engineering from the University
of Minnesota, Minneapolis, in 1988, 1990, and 1995, respectively.
Currently, she is an Associate Professor with the Electrical and Computer
Engineering Department, Mississippi State University. She has prior teaching
experience at Virginia Tech; the University of North Dakota, Grand Forks; and
Michigan Technological University, Houghton.
Dr. Schulz is an NSF Career award recipient. She is a member of PES and
ASEE.