Fault location using the high frequency travelling waves

measured at a single location on a transmission line

M Aurangzeb P.A. Crossley P. Gale

UMIST, UK UMIST, UK Hathaway, UK

Summary The validity of the proposed single ended ‘fault location

technique is assessed by comparing the fault location
The paper describes how the high frequency travelling decision obtained when analysing the current signals
waves caused by a fault on a transmission line can be captured by a high frequency disturbance recorder [ 11 with
observed in the secondary of a conventional protection the actual location of the fault. The actual location was
current transformer [I]. Information extracted from the determined using information from a double-ended GPS
secondary currents, measured at one location, is used to synchronised travelling wave fault locator [1,3] and if
determme the phases mvolved in the fault and its location. available information from a maintenance crew.

The fault locator measures the magnitude of the first wave Introduction
in each phase signal and using their relative size
determines which phases are involved in the fault. The Power systems must provide continuous, economic and
high frequency current mformation on a faulted phase is reliable electrical energy at a quality that satisfies customer
processed using a technique based on auto-correlation that or regulatory requirements. Faults on a transmission line
determines where the fault is most likely to be located. temporarily effect the quality, but normal service is rapidly
Effectively a “short” section of the signal that includes the restored either by auto-reclosure or by the removal of the
initial wave travelling from monitoring point to the fault is faulted line and the transfer of power to an altemative
stored and compared against a new section of the complete route. The latter often involves considerable economic cost
signal. The process is repeated numerous times, starting and may occasionally risk the integrity of the network
initially at the stored section and moving through the particularly if a second line is lost before the first is
signal by shifting each section in time by one sample restored to service.
period. The first section is identical to the stored section
and consequently the auto-correlation function is at a +ve Fault locators must determine the location of all types of
peak. As the section continues through the signal the faults, including high resistance faults, with an accuracy
similarity between sections start to get less and the and reliability trusted by the engineers and maintenance
magnitude of the auto-correlation function reduces. The crews who need to repair the damage. Ideally, the fault
reduction conhnues until the correlated section includes a should be located to the nearest tower and the accuracy
new travelling wave similar in shape to the initial wave, should be *(half a tower span). This ensures that after a
this then produces a local maximum. The paper will show permanent fault, the maintenance crew can immediately be
that the largest local maximum occurs when the correlated dispatched to the tower nearest the point of damage.
section includes a wave reflected from the fault. The time Inspection is also required after a transient fault, since any
required to move from the first +ve peak (the initial damage at the point of fault may cause a future fault [1,3].
section) to the largest local maximum can be directly
related to the distance from the measuring point to the What is a travelling wave fault locator?
fault. To confirm that this location decision is correct, a
fault at this position and involving the indicated phases is When a fault occurs on a transmission line, the voltage at
applied to an EMTP simulator that models the network in the point of fault suddenly reduces to a low value. This
the vicinity of the locator. The current signals generated by sudden change causes a transient that propagates away
the simulator and measured at the same location as the from the fault in both directions along the line. The effect
fault locator are cross-correlated against the actual signals is normally explained in terms of travelling waves that
measured on the real network. A high degree of similarity, initially travel away from the fault at a velocity almost
as defined by a large +ve cross-correlation value, indicates equal to the speed of light. When the initial wave reaches
that the simulated fault position is the actual fault position. an impedance discontinuity, such as a substation, a part of
If the cross-correlation value is small the simulated fault the wave is reflected back along the line towards the fault.
position is unlikely to be the actual fault position. The The remainder is transmitted through the substation into
process of simulation and cross-correlation is repeated the network. These waves continue to be divided into
with the next most likely fault location, i.e. the location reflected and transmitted waves and their, amplitude
corresponding to the next highest auto-correlation peak. attenuates with time until finally a new equilibrium is
reached [2].

Developments in Power System Protection, Conference Publication No.479 0 IEE 2001 . 403
Travelling wave fault location is accomplished by time- (01-19-1995) at 17 hours, fifty-six minutes and eleven
tagging each wave front as it passes a known location and seconds (17:56:11) and the GPS time reference was
then using various signal processing techniques to decide 01832743. The corresponding event, record 032 at station
which front corresponds to which wave [4, 51. The B occurred on the 19" January 1995 (01-19-1995) at 17
proposed fault locator detects the first fault initiated wave hours, fifty-six minutes and seven seconds (17:56:07) and
in the measured current signal. A section of the signal that the GPS reference time was 01831353. (note:- the clock at
describes this wave is stored as a correlation reference. The A was incorrectly set 4s faster than the one at B). The three
reference is then moved through the current signal and a phase transient current signals recorded at both ends of the
new auto-correlation value calculated at each sample point. line and the associated analysis records are displayed. The
Effectively the shape of the signal in the reference section post-fault data values are the instantaneous current at the
is compared against the shape of each wave that amves at point in time when the fault was detected. For record 003:-
the measuring location. The wave which corresponds to phase A = 7 quantisation levels, phase B = -91 and phase C
the reflection of the reference wave from the fault is = 73. The high values for B and C and the low value for A
similar in shape to the reference wave and consequently indicates the fault involves the B and C phases. The
results in a high positive auto-correlation value. existing double-ended GPS synchronised locator indicated
that the fault was 66.89km from INKI4 [1,6,7].
Single ended analysis of transient data
Fault location using only information at one line end is the
An algorithm was developed, based on travelling wave next step in the analysis process. The captured transient
theory, whch provides for the analysis and location of a fault data was converted and imported into Microsoft
fault on a transmission line. The algorithm decides which Excel. The Excel programme measured the time delay
phases are "involved in the fault and then uses auto- between the first fault-initiated wave and the first fault
correlation to evaluate the location. The transients captured reflected wave. These two waves were known to be similar
by two high frequency recorders located at the ends of a in shape and hence the first received wave was taken as the
transmission line and operating within a GPS synchronised reference and subsequent waves were compared against it
double ended fault location scheme are shown in Fig.1. using auto-correlation. A correlation peak indicated the
The records are from INKI4 (station A) and STHA4 best match. Table 1 lists the results obtained when using
(station B) on the Scottish Power 92.2km Inverkip to auto-correlation windows of 35, 70, 90 and 140 samples
Strathaven 400kV feeder. When an event is selected from (where 100 samples = 80p) to predict the fault location
the list at either station; the same event is highlighted at the for six different faults. For 003 the average "single ended"
other station. This is determined by a comparison of the auto-correlation prediction (66.6km) is almost identical to
event date, time and GPS reference. For example, record the double-ended result (66.8km).
003 at station A (INKI4) occurred on the 19* January 1995

ha&A PostFault Data-Value 7 PhaleA PostFault Data-Value -3 I

hass8 PostFault Data-Value -91 Phase6 PostFault Data-Value -86;
hassC PostFault Data-Value 73 PhaseC PostFault Data-Value 93 ,

-Correction - 0
!
1
004.DAT

Fig. 1 Information captured by GPS synchronised double ended fault locator

404
 simulated current signals cross-correlated with the original
NO F-T 2-E 35 70 90 140 signal. The effect on the cross-correlation factor of
003 b-c 66.8Km 66.7Km 66.6km 66.6Km 66.6Km changing the auto-correlation prediction point from 66.0
006 b-c 8.5 Km 7.92Km 7.8 Km 7.8 Km 7.8 Km
Km to 67.9 Km is shown in Figure 4. The cross-correlation
041 a-gnd 7.7 Km 8.4Km 8.28Km 8.28Km 8.28Km
050 b-c 66.1Km 66.4Km 66.4Km 66.4Km 66.4Km factor peaks at the predicted auto-correlation result,
059 b-c 71.3Km 66Km 71.8Km 71.9Km 71.9Km validating the accuracy of the original result.
124 a-gnd 7.8 Km 8.4 Km 8.4 Km 8.4 Km 8.76Km
150 - - fault sional
Drininal
U - - I

100
Simulation & Cross-Correlation 0

'c 50
m

Cross-correlation between the transient signals generated P

by a simulator and the recorded signals is used to validate -50
the results obtained from auto-correlation. EMTP is used to Time I
simulate the effect on the measured current of a fault on the
transmission network at the location predicted by auto-
correlation, see Fig.2. The current signals generated by the
simulator are cross-correlated against the signals captured
by the fault locator. The output of the cross-correlation
process, the cross-correlation factor, will be a high value if
the simulated fault position is the actual fault position. I
I
' ..
n.

For case 003:- a (b-c) fault was simulated on the 92.2-Km Fig3 Recorded and simulated signals for case 003
line between INKI4 and STHA4 at the predicted location of
66.6km from INKI4. The recorded and simulated fault I __._ Cross-CorrelationPeak
transients were cross-correlated and a cross-correlation
factor of 0.7 was calculated. This is a high value, hence the
location predicted by auto-correlation (66.6km from INKI4)
is considered correct. Close similarity between the recorded
and simulated fault transients can also be seen by visually
inspecting the signals shown in Fig.3. 65 5 66 66 5 67 675
Distance Wm)

To codinn that the cross-correlation factor peaks at the Figure 4 Effect on cross-correlation of changing
correct fault distance, different distances in the auto-correlation prediction from 66 - 67.9km
neighbourhood of 66.6 Km were simulated and the

DEVM4 INK14 FAULT STHA4 HAKB4

RES12 RES14
44.8 KM 66.1 KM 26.1KMl E 119KM
OIC

NEIL4
I 62.6 KM
/ 95.5 K M
Where OIC is OPEN
CIRCUIT and the line
lengths are not to
scale

Fig.2 EMTP Simulator model for section of Scottish Power network

405
Results:- single ended fault location recorded on a 92.2lun 400kV line between STHA and
INKI on the Scottish Power 400kV network.
The transient signals generated by various phase-phase At present most of the data recorded on the Scottish
and phase-ground faults on the line between INKI4 and Power network, including 16 of the 20 examples
STHA4 are analysed in this section. Auto-correlation was described here, have involved double phase faults
used to predict the probable location of the fault and resulting from conductor clashing. Further data from
cross-correlation was used to validate the result. The single-phase ground faults is urgently required, this will
single-ended auto-correlation results in km (l-E) and be used to confirm if the locator can operate correctly on
their corresponding cross-correlation factor (C-F) are what is normally the most common fault type.
summarised in Table 4 for the data captured by recorder
E at STHA4 and recorder D at INKI4 on the STHA-INKI
Acknowledgements
feeder. The table also includes the results generated by
the double ended travelling wave fault locator (2-E) and
The work described in this paper is part of a research
the error in metres between the single and double ended
decision. The double ended locator is accurate to a few project at UMIST sponsored by Hathaway Corporation.
The authors also acknowledge the crucial role of Scottish
hundred metres so the single ended error is bktween 0
Power who allowed UMIST access to the transient data
and 500m. The column TWS refers to the reference
recorded on their network.
number used to label the record in each transient
recorder.
References
The cross-correlation factors for the two sets of results
P.F. Gale, J. Stoke, P.A. Crossley, “Practical experience
are between 0.61 and 0.77, These are all considered high with travelling wave fault locators”, 1997 IEE DPSP
values and indicate that the locations predicted by the Conf., UK, pp. 192-196.
auto-correlation technique are correct. Note:- the L.V. Bewley, Travelling Waves on Transmission Systems,
maximum possible cross-correlation factor is 1.0, this Second Edition, John Wiley & Sons, 1963.
occurs for two waveforms that are identical. A. Burden, and A. MacGregor, “Power quality initiatives
at Scottish Power,” Transmission and Distribution World,
Conclusions February 1998, pp. 22-28.
P.A. Crossley, P.G. McLaren, “Distance protection based
on travelling waves,” IEEE Trans. PAS-102, no.9,
The paper has shown that fault location using the high
September 1983, pp. 2971-2983.
frequency travelling waves recorded at a single location L. Jie, S. Elangovan, J. B. X. Devotta, “Adaptive travelling
on a transmission line is a viable altemative to impedance wave protection algorithm using two correlation
based fault locators and GPS synchronised double ended functions,” IEEE Trans. Power Delivery, vol. 14, no. 1,
fault locators. This statement is supported by comparing January 1999, pp. 126-131.
the results generated by the proposed single ended fault M. Aurangzeb, P.A. Crossley, P.F. Gale, “Fault location
location algorithm with results obtained from a on transmission lines using high frequency travelling
commercially available double ended travelling wave waves”, IEEEPES winter Meeting, Jan 2000.
fault locator. The data used to validate the algorithm was P.F.Gale, P.A.Crossley, “Fault location based on travelling
waves”, 1993 IEE DPSP conf., UK, pp 54-59.

Table 4 Comparison between single ended fault location decisions and GPS synchronised double ended results.

406