Evaluation of openXDA for Automatic Fault Analysis

Ling Wu, University of Tennessee

Kyle Thomas, Dominion Virginia Power
Robert Orndorff, Dominion Virginia Power
Stephen Wills, Grid Protection Alliance

Presented at the Georgia Tech Fault and Disturbance Analysis Conference, April 27-28, 2015.

Abstract
The high efficiency of fault location calculation is one of the most crucial techniques which has significant influence on the
reliability of power systems. Once an event occurs on a power transmission line, it is followed by a series of automatic device
activities: protections triggered, breakers tripped, load outage, etc. The most stringent mission for the utility is to pinpoint
the fault location, and then to clear the fault, and restore power. The sooner the power supply recovery, the less power
outage, and the better the power system’s reliability. Typically, manual analysis is required to calculate a reasonable and
acceptable fault location result, including faulted line, fault type, and fault location distance. Based on the result, line crews
are targeting to find fault trace in the field. The currently used traditional fault location calculation method is analyzed in this
paper, and several problems have been exposed within it. In order to improve fault location analysis efficiency in utility, GPA
(Grid Power Alliance Co.) proposed a novel application, openXDA (open-source Extensible Disturbance Analytics) to solve
these problems. This project mainly purposes to evaluate openXDA, which targets to realize highly efficient calculations for
saving time, and to obtain a reasonable result which could be close enough to the actual event location for line crews’ work.
It is expected to accomplish a series of analytical processes by automatic calculation, and improve accuracy by using multiple
advanced algorithms and plentiful samples for calculating. In this paper, we use several real fault cases on Dominion Virginia
Power’s transmission network to evaluate openXDA.

Key words: openXDA, power system reliability, automatic fault analysis, fault location

I. Introduction is to pinpoint fault location, clean the fault, and

recover the power supply. The sooner the power
Power system reliability is one of the most
supply recovery, the less power outage, and the
significant indexes needed in order to evaluate the
higher power system reliability. So the high
ability of providing reliable and qualified power. For
efficiency of fault location calculation is one of the
some specific customers, it is strictly required that
most crucial techniques, which has a significant
the outage duration be limited to a certain time
influence on the reliability of power system.
period in the contract. The reliability index is closely
connected to the profit of a utility. A common In order to accomplish fault analysis after power
target of different utilities is the improvement of outage, IEDs (intelligent electric devices) are
power supply and transmission reliability. Most of installed in substations, like PMUs and DFRs. They
the time, short circuit events are the primary reason can be used for system monitoring under steady
for power outages. However, it is unavoidable in a states and provide disturbance records when faults
power system. Any type of fault occurrence on occur in the system. During a transient process
power transmission line is followed by a series of caused by a disturbance, they can store disturbance
automatic activities: protection operations, information, including system voltage and current
breakers tripping and reclosing, and, in some cases, waveforms during transience, and then they
an eventual power outage. Because power outage generate records in COMTRADE format for analysis
duration contributes to reliability impairment, the purposes. Currently, analysts do fault analysis based
utility should target to eliminate outage duration on these records and obtain fault information,
after any power outage. The most stringent mission including faulted line, fault type, and fault location.

Evaluation of openXDA for Automatic Fault Analysis Page 1 of 7

The traditional method to do fault analysis is using them with the analysis application. COMTRADE is an
vendor specific or third party applications along IEEE standard universal format for disturbance
with system models. However there are several records. The COMTRADE file contains voltage and
problems with it. The detailed process of this current data captured during the fault event.
traditional method is introduced in this paper, and
drawbacks are displayed. Meanwhile, we are
seeking a better way to fulfill higher requirements 2) Analyze the COMTRADE data to determine the
to improve the efficiency of fault analysis. The fault type and the faulted line
openXDA is an advanced application released by
GPA for providing an extensible platform for According to visualization of voltage and current
processing event records. Based on the waveform waveforms in the analysis application, the faulted
data and transmission line parameters provided by line and the fault type can be decided. For example,
customers, it determines the fault type and Fig 2.1 shows a group of voltage and current
calculates the line distance to the fault. The purpose waveforms during a fault.
of this project is to evaluate the openXDA
application for fault location calculation compared
with the traditional method. By using fault cases of
Dominion Virginia Power, another target is to verify
that openXDA can obtain convincible distance to
fault for line crews in utility.
This paper is organized as follows: The traditional
fault analysis method is discussed in section II,
openXDA application is displayed in section III, real
fault cases analysis comparisons are presented in
section IV, and the evaluation of novel openXDA Fig 2.1 transient voltage and current waveforms
application is concluded in section V.

In Fig 2.1, the upper three curves are the phase

II. The traditional fault analysis method voltages of line 183, and the lower four curves are
the phase current and the ground current of line
The traditional fault location method is based on a 183. When a short circuit fault occurs on line 183,
third party application. By using this application, the both voltage and current waveforms change
transient current and voltage waveforms of faulted correspondingly. In this case, the phase B and
line can be visualized. According to the waveforms, ground current change abruptly at the same time,
the faulted line and fault type can be decided. and the magnitude of phase B voltage decreases
Meanwhile, it can calculate the fault location on the gradually. It represents a phase B to ground fault.
samples in COMTRADE files. In this section, the After a period of oscillation, these current
process of fault location analysis in the traditional waveforms change to zero, which indicates the
method is introduced as are the problems we have breakers on line 183 tripped the circuit. The fault is
met. isolated.
Through the display of the transient waveforms in
1) Download disturbance record in COMTRADE Fig. 2.1, we can conclude that it is a phase B to
format from a digital device ground fault occurrence on line 183.

The disturbance records are generated once a fault

occurs and they are stored on a server. Analyzers 3) Calculate the fault location with one picked
download fault records remotely, and then analyze sample during the fault

Evaluation of openXDA for Automatic Fault Analysis Page 2 of 7

There are three algorithms for fault location Mainly on account of these three issues, the
calculation which are fixed in the analysis traditional method is not able to meet the
application. For each time, one sample during the requirement of providing credible fault location
fault is selected, which contains a completed result in a reasonable time.
period, and one fault location result can be
calculated by the application. Based on the
information of a faulted transmission line and a III. openXDA for fault analysis
specific algorithm, the fault location result can be
openXDA is an extensible platform for processing
shown as circled in percentage term. Then
event files from disturbance records. Based on
analyzers need to look up the total distance of the
transient waveform data and transmission line
transmission line to calculate fault location from the
parameters provided by customers, it determines
substation in miles. For each sample, there are
the fault type, faulted line and calculates the line
three fault location results at most with three
distance to the fault. The most attractive attribute
algorithms available.
of openXDA is the ability of automatic fault analysis,
for saving much time. It solves every problem of the
4) Repeat fault location calculations at different
traditional method with advance techniques.
timestamps Compared with the traditional method, openXDA
presents a more powerful ability on fault analysis.
Typically, at least four samples at different First, openXDA is able to read COMTRADE event
timestamps are needed to conclude a credible files in a user-specified folder, and parse these files
result. The same calculations must be repeated at into timestamp-value sets. With transmission
least 4 times, and 12 different results obtained. network parameters, it calculates fault location
automatically and produces a complete fault
analysis result in a few seconds. It has an
5) Conclude a best fault location based on all advantage on calculation efficiency comparing with
results manual process of traditional method. Secondly,
the performance of calculation can be extended
Generally, the best result is the average value of all
with new algorithms. For now, all five currently
results. Then analyzers report this fault location
available algorithms are used. Compared with only
result to help the line crews find it in the field.
three or fewer algorithms in traditional analysis
Through the process of the traditional fault analysis applications, the result of openXDA is more
method, we can find several drawbacks which comprehensive. Thirdly, openXDA calculates
apparently impact the efficiency of analysis. Firstly, frequency domain values for each cycle of
the whole process of traditional fault location measured data, which means the result of each
calculation is manual. Sequentially, analysis would algorithm is computed for every sample in a
take a lot time using this process. When a fault COMTRADE file. In contrast with the traditional
occurs in transmission, saving time means saving method, which uses a much smaller amount of
the economic loss of utility. However this manual samples selected manually for calculation, openXDA
calculation would cost us a lot of time. Second, it utilizes the complete information of fault records.
calculates the result with only three algorithms. For
openXDA is expected to help overcome the
now, there are available more advanced algorithms
weaknesses of traditional methods, improve the
which can be useful to provide more
fault analysis process, and provide a significant
comprehensive results. Third, the result is based on
contribution to power system reliability. As we can
countable randomly selected samples during the
see from the analysis of the traditional method, the
fault, which means it would result in a significant
most time consuming part is the process of fault
amount of information loss within the other
location calculation. Improvement on this point
thousands of samples which are not selected.

Evaluation of openXDA for Automatic Fault Analysis Page 3 of 7

would prompt work efficiency of the whole fault substation. openXDA confirms the fault location
analysis process most obviously. So in section IV, we based on results of all five currently available
mainly focus on the comparison of fault location algorithms. Five results are displayed in the plot
calculations. with different colors. For each curved line, when the
trend goes down, and then keeps a relatively flat
period for a while, it signifies a fault occurrence
IV. Comparison of multiple fault analysis tools during this period (x-axis) around this area (flat area
in y-axis). Through the plot, we can see that the
In order to verify the capability of openXDA to
results of different algorithms are very close. They
calculate fault location efficiently with a certain
have similar change trends, and some of them
error limit, six real fault cases that happened on
overlap with each other, which indicates that these
Dominion’s transmission network are analyzed in
five algorithms correlate very well. They can be
this section. With use of fault records generated by
evidence for each other to prove the results are
IEDs, openXDA produces fault location results.
credible. Based on five credible results calculated by
Based on the comparison with actual fault location
five available algorithms, openXDA concludes a
found in the field, we verify its calculation
comprehensive result, which is closer to the actual
capability. Generally in utility, the error of
location compared with the traditional fault
calculation result less than 1 mile is acceptable. On
location result which uses only three algorithms. In
the other hand, we also use the traditional method
this case, the fault occurred on line 126. The
to calculate the fault location as well. By
location can be estimated at approximately 17.75
comparison of two of these results, we verify the
miles from Earleys Substation through the
improvement of openXDA application.
corresponding average y-axis value of a relatively
The first case is a phase C to ground fault that flat period when the fault endures. Comparing to
occurred on transmission line 126. The line crews the actual fault location and result of traditional
found the actual fault location is 18.50 miles from methods provided, we can see that openXDA can
Earleys Substation. With the traditional fault calculate the fault location effectively, with only
location method, we can get a result, 17.70 miles 0.75 miles difference.
from Earleys substation. However by use of
There are five other fault cases used for calculating
openXDA to calculate fault location is about 17.75
the fault location results with openXDA. In order to
miles from Earleys substation, which is close to the
display the results compared with the traditional
traditional method result and matches the actual
method results and the actual fault locations, the
fault location found. We can conclude that the
openXDA results of each case is displayed in Fig 4.2,
result of openXDA in this case is effective. More
and the corresponding comparison is shown in
details about the result of openXDA are shown in
Table 4.1.
Fig 4.1.

Fig. 4.1 openXDA Result for the Fault on Line 126 a. openXDA Result for the Fault on Line 40

In Fig 4.1, the x-axis represents time, and y-axis

represents the distance of the fault from the

Evaluation of openXDA for Automatic Fault Analysis Page 4 of 7

 obvious descending trends and flat area for a
specific time period, which means that these
algorithms installed in openXDA are able to
calculate fault location for these real cases
effectively. The curves in each plot are close to
each other with some of them are overlapping. The
biggest difference between different curves is about
0.5 miles. This indicates that five algorithms in
openXDA correlate with other algorithms very well.
b. openXDA result of fault on Line 271
Based on those high correlated results, it is
convincing that the result of openXDA on fault
location calculation is credible.

Faulted Fault type Actual Result of Result of

Line fault Traditional openXDA
location Method
40 Phase A- 0.90 0.91 0.90
c. openXDA result of fault on Line 2034 Phase B
271 Phase A- 0.83 1.00 0.70
Phase B
2034 Phase B- 1.39 1.56 1.60
Phase C
247 Phase A- 0.58 0.40 0.65
Ground
2118 Phase C- 4.24 3.37 3.41
Ground
Table 4.1 Comparison of Fault location results
d. openXDA result of fault on Line 247

These five cases are real faults that occurred in the

utility transmission network. The actual location
that line crews found can be used as a reference for
evaluating the results from openXDA. The result of
the traditional manual calculation method is also
compared in this paper. The comparison of results
between openXDA and the traditional method is
shown in Table 4.1. These faults occurred in the
e. openXDA Result for the Fault on Line 2118 230kV transmission network. Fault records
Fig. 4.2 Fault location results of openXDA for different cases
generated by DFRs on one side of each faulted line
are used for calculation. The data in Table 4.1 shows
for distance between the fault location and the
In Fig 4.2, the results of openXDA for each case are corresponding substation. There are mainly two
shown respectively. The calculated result of each fault types tested in this paper, phase-to-phase and
algorithm is displayed in terms of a curve in phase-to-ground. For phase-to-phase fault cases,
different colors. In each plot, the curves have the largest difference between openXDA results and

Evaluation of openXDA for Automatic Fault Analysis Page 5 of 7

the actual fault location is 0.21 miles, compared to VI. Further research
the largest error of the result of traditional method
We plan to use the results of the various fault
is 0.17 miles. For phase-to-ground fault cases, the
location methods to further classify which methods
largest error is 0.83 miles versus 0.87 miles for the
perform best for various types of faults. For
results of openXDA and traditional method,
example, we may find that some algorithms may
respectively. The calculation result is less than 1.0
perform better than others on high impedance
mile different from the actual location. In practical
ground faults.
utility work, this calculation error is acceptable. It
can prove that openXDA can calculate acceptable Through visualization and statistical analysis we
fault location effectively. Furthermore, the results hope to determine which of the many calculated
of openXDA for some cases are better than that of locations is most likely to be accurate.
traditional method. To some extent, it means that
openXDA is able to improve the accuracy of fault
location calculation. VII. Conclusion
In this paper, comprehensive analysis and
evaluation of openXDA on fault analysis
V. Visualization performance is presented. It is an extensible
In addition to providing multiple fault locations application for processing event records
based on different algorithms, this tool allows for automatically and determining the fault results
visualization of the fault location. Traditional single efficiently. Through an all-sided explanation of the
point locations do not offer as much detail about process of the traditional fault analysis method,
the nature of the fault as the graphs of the several problems are revealed. By contrast,
calculations provide. We wanted a simple way to openXDA is designed for not only implementation
graph these locations without having to import the of the fault analysis function, but also targeting to
data in a spreadsheet and create a graph. The solve these problems, and ultimately to realize
simple solution to this was to provide a COMTRADE better advancement of power system reliability. In
file of the associated channels and include traces order to furnish a convincing comparison between
from each fault location calculations. This output the calculation ability of openXDA, a series of fault
COMTRADE file is produced whenever a reasonable cases in Dominion’s transmission network are
location is calculated on a line. analyzed by both openXDA and the traditional
method. Based on the results and comparison with
In addition to the analog traces of fault location,
the actual fault locations found in the field, the
there are also digital status traces indicating a fault
conclusion of openXDA performance evaluation is
condition as well as fault type. One example is
exhibited as follows.
shown in figure 5.1.
The results of openXDA match the actual fault
locations very well for both phase to phase and
phase to ground fault types. The distance errors are
less than one mile. openXDA is capable of
calculating fault location correctly with an
acceptable error.
Furthermore, five algorithms correlate with each
other. For each case, the final fault location result
given by openXDA is more credible, because it is
based on all five currently available algorithms. It
Fig 5.1 Visualization of fault traces in Comtrade calculates and analyzes every sample during the
whole fault period, not just several random samples

Evaluation of openXDA for Automatic Fault Analysis Page 6 of 7

as the traditional method does. Sequentially, in Stephen Wills began as an intern at the Tennessee
most of the cases, results of openXDA can be closer Valley Authority in 2009, working on open source
than those of the traditional method with less error. software for synchrophasor systems. Since
graduating from the University of Colorado at
Finally, the calculation abilities of openXDA can
Colorado Springs in 2011, he has continued work on
improve the analytic efficiency because it is
open source software for electric power systems at
performed automatically. For each case, the process
the Grid Protection Alliance. He has been the
only takes several seconds to give us the result.
primary developer for openXDA since the project
Compared with the traditional method, which
began in 2013.
repeats the process manually with just a few
selected samples, openXDA can save us much time
on fault analysis. Plus it provides a more accurate
Kyle Thomas received his M.S. degree in Electrical
result, line crews can find the fault, and restore
Engineering from Virginia Tech in 2011 and is
power faster.
currently pursuing his Ph.D. while working for
Dominion Virginia Power’s Electric Transmission
Operations Research group. He has technical
expertise in power system protection/control, wide-
About the authors area measurements, fault analysis, cascading
analysis/physical security, and system simulations.
Ling Wu is currently a Ph.D. student in the Min H.
Kyle is a technical lead of Dominion’s synchrophasor
Kao Department of Electrical Engineering and
installations, applications, and training, and is
Computer Science at the University of Tennessee,
actively involved in the North American
Knoxville. She received her B.S and M.S from Xian
Synchrophasor Initiative (NASPI), IEEE, and Cigre
Jiaotong University, China, in 2008 and 2011,
organizations.
respectively. Her research interests involve wide-
area power system monitoring and power system
dynamic analysis. Prior to pursuing her Ph.D., she
Robert Orndorff has worked at Dominion since
was a substation engineer in the State Grid
1984. He earned an A.A.S degree in Electronics in
Corporation of China.
1986 and spent 11 years as a field relay technician
and in 1997 transferred to the Fault Analysis
department where he currently works. His current
responsibilities include maintaining and configuring
Dominion’s Digital fault recorders, event retrieval
and analysis from smart relays and DFRs. Robert is
an IEEE member and has been a member of the
Transient Recorder’s User Council (TRUC) since
2002.

Evaluation of openXDA for Automatic Fault Analysis Page 7 of 7