Académique Documents
Professionnel Documents
Culture Documents
Abstract - This paper describes the use of wavelet transform for Among the limitations of the traveling wave methods, the
analyzing power system fault transients in order to determine the requirement of high sampling rate is frequently stated. Other
fault location. Traveling wave theory is utilized in capturing the stated problems include the uncertainty in the choice of sampling
travel time of the transients along the monitored lines between the window and problems of distinguishing between traveling waves
fault point and the relay. Time resolution for the high frequency reflected from the fault and from the remote end of the line.
components of the fault transients, is provided by the wavelet Recent developments in optical current transducers technology
transform. This information is related to the travel time of the enabled high sampling rate recording of transient signals during
signals which are already decomposed into their modal compo- faults [7]. Availability of such broad bandwidth sampling capa-
nents. Aerial mode is used for all fault types, whereas the ground bility facilitates better and more efficient use of traveling wave
mode is used to resolve problems associated with certain special based methods for fault analysis.
cases. Wavelet transform is found to be an excellent discriminant
for identifying the traveling wave reflections from the fault irre- The correlation based fault location method introduced in
spective of the fault type and impedance. EMTP simulations are [ 2 ] , is very effective as long as the width of the time window
used to test and validate the proposed fault location approach for to save the forward moving wave is properly selected. Since
typical power system faults. this selection depends on the fault location which is unknown,
Keywords: Power System Faults, Electromagnetic Transients, the window width selection remains an unresolved issue for the
Wavelet Transform, Fault Location, Traveling Waves. practical implementation of this method. Combined use of a
short and a long window has been proposed as one solution for
this problem in [4].
1 Introduction
In this paper, a different approach, based on the wavelet
Transmission line fault location has long been one of the primary transform of the fault transients, is presented. Wavelet transform
concerns of the power industry. Methods of locating power sys- possesses some unique features that make it very suitable for this
tem faults introduced so far, can be broadly classified under two particular application. It maps a given function from the time do-
categories: one based on the power frequency components, and main into time-scaling domain. The wavelet, the basis function
the other utilizing the higher frequency contents of the transient used in the wavelet transform, has bandpass characteristics which
fault signals. The latter is also referred to as traveling wave or makes this mapping similar to a mapping to the time-frequency
ultra high speed fault location method, due to its use of traveling plane. Unlike the basis functions used in Fourier analysis, the
wave theory and shorter sampling windows. wavelets are not only localized in frequency but also in time.
The use of traveling wave theory for fault detection was This localization allows the detection of the time of occurence
initially proposed by Dommel and Michels in [I], where a dis- of abrupt disturbances, such as fault transients. Fault generated
criminant was defined based on the transient voltage and current traveling waves appear a:; such disturbances superposed on the
waveforms in order to detect a transmission line fault. McLaren power frequency signals recorded by the relays. Processing these
et al. have later developed a correlation based technique where signals using the wavelet transform reveals their travel times
the cross correlation between stored sections of the forward and between the fault and the relay locations.
backward traveling waves were used to estimate the travel times
of transient signals from the relays to the fault point [2,3,4]. An
overview of traveling wave based fault location methods can be The potential benefits of applying wavelet transform for anal-
found in [ 5 , 6 ] . ysis of transient signals in power systems have been recognized
in the recent years. Robertson et al. present a comparative
overview of Fourier, short time Fourier and wavelet transforms,
PE-303-PWRD-0-12-1997 A paper recommended and approved by give examples of applying wavelet transform to analyze power
the IEEE Transmission and Distribution Committee of the IEEE Power system transients and extraction of their particular features in [8].
Engineering Society for publication in the IEEE Transactions on Power A similar overview along with application of wavelet transform
Delivery. Manuscript submitted July 28, 1997; made available for to detect and classify power quality disturbances, are given in [9].
printing December 12, 1997. Advantages of using wavelet transform for analyzing transients
and solution of linear time-invariant differential equations using
wavelet transform is demonstrated in [ 101. In this paper, another
useful application of the wavelet transform in solving the prob-
lem of fault location, will be presented. A brief introduction to
wavelet transform will be given before formulating the problem
and presenting the proposed solution method.
where, a and b are the scaling (dilation) and translation (time is used. The phase signals are transformed into their modal com-
shift) constants respectively, andQ,I is the wavelet function which ponents by using this transformation matrix as follows:
may not be real as assumed in the above equation for simplicity.
The choice of the wavelet function (mother wavelet) is flexible Sinode = TSphase (4)
provided that it satisfies the so called admissibility conditions where, Smo& and Spha.ve are themodal and phase signals (voltages
P11. or currents) vectors respectively.
Wavelet transform of sampled waveforms can be obtained by Clarke's transformation is real and can be used with any
implementing the discrete wavelet transform which is given by: transposed line. If the studied line is untransposed, then an eigen-
vector based transformation matrix, which is frequency depen-
dent, will have to be used. This matrix should be computed at
a frequency equal or close to the frequency of the initial fault
transients.
where, the parameters a and b in Eq.( 1) are replaced by a; and Recorded phase signals are first transformed into their modal
ka?, k and m being integer variables. In a standard discrete components. The first mode (mode l), is usually referred to as the
wavelet transform, the coefficients are sampled from the contin- ground mode, and its magnitude is significant only during faults
having a path to ground. Hence, this component can not be used
uous WT on a dyadic grid, a0 = 2 and bo = 1, yielding a: = 1, for all types of faults. The second mode (mode 2), also known as
U;' i,
= etc. b = k x 2-',i being an integer variable. the aerial mode, however is present for any kind of fault. Accord-
Actual implementation of the discrete wavelet transform, in- ingly, the fault location problem is formulated based essentially
volves successive pairs of high-pass and low-pass filters at each on the aerial mode, making occasional use of the ground mode
scaling stage of the wavelet transform. This can be thought of signal for purposes of distinguishing between certain peculiar sit-
as successive approximations of the same function, each approx- uations, which will be discussed in the next section. Depending
imation providing the incremental information related to a par- on the existing communication scheme between the two ends of
ticular scale (frequency range), the first scale covering a broad the line, fault location problem can be solved in two different
frequency range at the high frequency end of the spectrum and the ways described below.
higher scales covering the lower end of the frequency spectrum
however with progressively shorter bandwidths. Conversely, the
first scale will have the highest time resolution, higher scales will Two ended sync ronized recording
cover increasingly longer time intervals. While, in principle any
admissible wavelet can be used in the wavelet analysis, we have Fault signals are recorded simultaneously at both ends of the
chosen to use the Daubechies4 [9],[12] wavelet as the mother line by two separate channels both of which are using the same
wavelet in all the transformations. time reference synchronized using Global Positioning Satellite
(GPS) receivers. The recorded waveforms will be transformed
into modal signals, after which the modal signals will be analyzed
using their wavelet transforms. Let t~ and t g correspond to the
lem times at which the modal signal wavelet coefficients in scale 1,
show their initial peaks for the signals recorded at bus A and
Consider a single phase lossless transmission line of length B respectively. Assuming that the recorded signals at the two
e, connected between buses A and B, with a characteristic ends of the line are fully synchronized, the delay between the
impedance 2, and traveling wave velocity of U . If a fault occurs fault detection times at the two ends, i.e. t d = t g - t A , can be
1477
determined [13]. The distance between the fault point to bus A Insignificant coefficients will imply that the fault is in the remote
will then be given by: half of the line, and vice versa.
If the fault is determined to be in the near half ofthe line, then
X =
e - v, td td in Eq.(6) will simply be the time interval between the first two
2 peaks of the scale 1 WTCs for the aerial mode.
If the fault is suspected to be in the second half of the line,
where, then td in Eq.(6) will be replaced by:
e is the length of the line, x is the distance to fault from bus A,
and U, is the speed of the traveling waves for mode m. td = 2t - tx (7)
where:
3.2 Single ended recording t is the travel time for the entire line length, and e, is the time
interval between the first two peaks of aerial mode WTCs in
A more robust configuration that does not require remote end scale 1.
synchronization is when the fault location is determined based Figure 1 shows the flowchart for the proposed fault location
solely on the recorded signals at one end of the line. However, in algorithm based on the wavelet transform coefficients. Next
such a case, due to the lack of any other time reference, all time section contains results of simulations used to test this proposed
measurements will be with respect to the instant when the fault algorithm for various fault types and line configurations.
is first detected. Therefore, fault location calculations will be
based on the reflection times of the traveling waves from the fault Transducer
point. Unfortunately, for faults involving a ground connection, output
not only those reflections from the fault point, but also from the
remote end bus will be observed at the sending end of the line.
Proper algorithms should therefore be devised in order to dis-
tinguish between close-in and remote faults which may produce Transformation
similar reflection patterns for the grounded faults. The following
sections describe our proposed approach to accomplish this task.
Wavelet
tions from the remote end bus during the fault transients. Thus,
by measuring the time delay between the two consecutive peaks
in the wavelet transform coefficients of the recorded fault signal
at scale 1, and taking the product of the wave velocity and half of
this time delay, the distance to the fault can easily be calculated
for these kinds of faults. The fault distance will be given by the YE
equation:
Ungrounded Fault
345 KV 1-x
9
:t
X
I
I( D-c------h 345 Kv
* ,
I I .** . .
time (ms)
A T1 T1+2TZ 3T1 T1+4T2
*Ilj The arrival time of the first transient peak depends on the
velocity of the line and the fault distance, it is independent of
the type of fault, hence the method applies to all type of faults
Figure 3: Lattice diagram for a single phase to ground fault at 120 provided the two terminal recordings are synchronized in time.
miles from A.
4.2 Single ended recording
4.2-1 Ungrounded faults:
4. ronized recording
Figure 6 shows the WTCs for an example of a phase to phase
Assuming synchronized recording of fault transients at both ends fault at 30 miles from A. It can be seen from the figure that mode
of the line, a three phase fault is simulated at 20 miles away from 1 (ground mode) signals are zero, therefore mode 1 WTCs can
bus A. Mode 2 (aerial mode) voltage signals are used only. Figure be used to identify this as an ungrounded fault.
1479
In this case, td will be directly obtained by measuring the time a)Scale 1, mode 2 c)Scale 1, mode 1
2 1.5r I
difference between the first two peaks in Figure 6(a) and using
Eq.(6) of Approach I described in section 3.2.1.
A4
30 21 22 23 0
7
time (ms) time (ms) x 3
b)Scale 2, mode 2 d)Scale 2, mode 1 cu
2, I I
< 2
6
z1 0
20 21 22 23
time (ms) time (ms)
345 KV 345 Kv
[7] J. Blake, P. Tantaswadi, and R.T. de Carvalho, In-Line
Sagnac Interferometer Current Sensor, IEEE Trans. on
I \ PowerDelivery, Vol.11, No.1, January 1996, pp.116-121.
345 KV I 345 Kv
[SI D. C. Robertson, 0. I. Camps, J. S. Mayer, and W. B. Gish,
-
I
100 miles
I
100 miles
I
c
Wavelets and Electromagnetic Power System Transients,
IEEE Transactions on Power Delivery, Vol.11, No.2, pp.
1050-1058, April 1996.
Figure 9: Circuit diagram of the simulated system with mutually [9] S. Santoso, E. Powers, W. Grady, and P. Hoffmann, Pow-
coupled lines. er Quality Assessment via Wavelet Transform Analysis,
IEEE Transactions on Power Delivery, Vol.11, N0.2, pp.
924-930, April 1996.
5 [lo] 6. T. Heydt, and A. W. Galli, Transient Power Quality
Problems Analyzed Using Wavelets, IEEE Transactions
This paper presents a new, wavelet transform based fault loca- on Power Delivery, V01.12, No.2, pp. 908-915, April 1997.
tion method. Using the traveling wave theory of transmission [ 111 I. Daubechies, Ten Lectures on Wavelets, SIAM, Philadel-
lines, the transient signals are first decoupled into their modal phia, Pennsylvania, 1992.
components. Modal signals are then transformed from the time
domain into the time-frequency domain by applying the wavelet [12] MATLAB Users Guide, The Math Works Inc., Natick,
transform. The wavelet transform coefficients at the two lowest MA.
scales are then used to determine the fault location for various
types of faults and line configurations. The proposed fault loca- [ 131 A. Phadke, J. Thorp, ComputerRelayingfor Power System-
tion method is independent of the fault impedance and is shown s, John Wiley & Sons Inc., New York, 1988.
to be suitable for mutually coupled tower geometries as well as
series capacitor compensated lines. The method can be used both [ 141 Alternative Transients Program, Bonneville Power Admin-
with single ended and synchronized two ended recording of fault istration, Portland, Oregon.
transients. The fault location estimation error is related to the
sampling time used in recording the fault transient. Furthermore, [ 151 J. R. Marti, Accurate Modeling of Frequency Dependent
for grounded faults near the middle of the line, mode 1 signals Transmission Lines in Electromagnetic Transient Simula-
from the fault and from the far end become comparable increas- tions, IEEE Trunsactlons on Power Apparatus and Sys-
ing the error of the fault location algorithm. Simulation results tems, Vol. PAS-101, no. l,pp.147-155, Jan. 1982.
are given to demonstrate the performance of the method. Fernando N. Magnago obtained the B.S. degree from UNRC,
Argentinain 1990 and his M.S. degree from Texas A&M Univer-
sity, College Station, TX in 1997. He is currently a Ph.D. student
at Texas A&M University.
References
Ali Abur (SM90) received the B.S. degree from M E W , Turkey
H. W. Dommel, and J. M. Michels, High Speed Relaying in 1979, the M.S. and Ph.D. degrees from The Ohio State Uni-
using Traveling Wave Transient Analysis, IEEE Publica- versity, Columbus, OH, in 1981 and 1985 respectively. Since
tions NO. 78CH1295-5 PWR, paper no. A78 214-9, IEEE late 1985, he has been with the Dept. of Elect. Eng. at Texas
PES Winter Power Meeting, New York, January 1978, A&M University, College Station, TX, where he is currently an
pp.1-7. Associate Professor.