Power Cable On-line Diagnosis Using
Partial Discharges Ultra Wide Band
Techniques
V.R. Garcia-Col&,
IEEE Member
Abstract. A non-invasive ultra-wide band (UWB) partial discharge technique was employed in cable diagnosis of
Mexico City underground distribution feeders. The technique presented here was applied to over 2000 joints, 400
terminals and thousands of meters of installed paper insulated impregnated cables rated at 23kV. The technique
allowed circuit classification according to electromagnetic field levels produced by partial discharges and recorded on
each manhole with a joint along the circuits evaluated in 12 substations. The PD detection technique employs near field
sensors working within the 80MHz range and whose information is digitized and fed to a conventional PD digital
detector, where a PD pattern is obtained. This pattern is correlated to the phase current by using an inductive sensor:
therefore corona and interference problems could' be disregard during measurements. To attain repetitive values,
calibration was performed using a very fast pulse (less than 50ns front) injected in a small cable loop. Results
demonstrate a great signal to noise ratio, high selectivity and great noise discrimination. The large amount of
measurements allowed the determination of limit values to differentiate areas in acceptable state from those requiring
attention. Those results include detection of bad joints, cable deformation due to ducts collapse and damaged terminals,
beside of aged cable detection caused by overheating operation conditions. The results obtained had been confirmed by
subsequent faults following the detection of very high levels of PD.
labor and more important, it will help in the task
X. INTRODUCTION of keeping failure rate low.
The population growth in large cities like Setting off with the assumption that faults
Mexico City demand a steady increment in evolved from aged cables and that those near its
electric load, that has reached up to 6% in recent life end would exhibit high dielectric losses,
years [i]. This increment, together with a difficult intense PD activity and degraded insulation, a
economic situation that restrains inversion, novel UWB method was implemented, as
forces the distribution system usage to its limits. discussed next.
Together with this load growth, there has been an
increment in failures at underground sections of XI. BACKGROUND
distribution feeders in Mexico City due to The study was developed for a traditional feeder
several causes [ii]. Beside overloading, normal arrangement used in Mexico City, where each
ageing was considered the main cause of failure, distribution substation has twelve feeders. Three
as most of underground cable network dates PlLC cables, 240mm' Cu conductor form each
from early 1960's. Therefore a major substitution feeder. Those cables are used for connection
program was launched two years ago, to replace between the substation bus and the aerial section
old 23kV Paper Insulated Lead Cover (PlLC) of the feeder. Being the main subject in this
cable with new polymeric insulation cable. study, a typical cable layout is depicted in the
Due to the large number of cables involved in the single line diagram shown in Fig. 1. The length
replacement program, only minor advances were of each feeder underground section varies from
obtained following a scheme where cables were 70m to 1500m.
substituted according to operational and financial
availability, while failure rate was kept
unchanged.
This paper presents a new approach for
scheduling cable replacement (if required), but
focuses on a predictive maintenance approach,
with the support of on-line measurements, that
would help in cables replacement prioritizing
20 1 0-7803-7107-0/01/$10.00 02001 IEEE
induced heating due to circulating harmonic
currents.
C. UWB-PD detection principles
For incipient fault detection, the main diagnostic
tool employed was UWB-PD detection [iv]. The
detection principle resides on a high signal to
noise ratio obtained when PD activity is
measured at MHz range frequencies, named
UWB, according to international standards [VI.
Fig. 1 Typical feeder arrangement for a At those frequencies, PD pulse is rapidly
distribution substation distorted due to dielectric losses and conductor
self-inductance that attenuate high order
The modification of aged cable substitution frequency components. Those high frequency
original plan involved two key parameters, components find lower impedance traveling
named extensive inspection along the cable path, trough the air. In equipment with distributed
and the use of three on-line diagnosis techniques, parameters, like cables and machines, PD pulse
focused on incipient fault detection. Those attenuation occurs after only a few meters [vi].
techniques named "Live detection" include On- This natural low pass filter formed in the cable,
Line PD detection using Ultra Wide Band implies that high frequency components can only
(UWB) techniques, load balance measurements be measured if sensors are placed nearby the
and hot spot detection by IR images. A brief location where PD takes place. Using this
description of those techniques is given below, principle, electromagnetic detection of PD pulse
while emphasizing the UWB-PD technique. energy in the high frequency region (above
SOMHz), would allow PD detection and location.
A. Mechanical damage detection
Furthermore, under those circumstances, PD
In a 30-year-old network with a high failure rate, pulse shape is not distorted, as cable length is
maintenance work has been performed on almost longer than PD wavelength, preventing
each installed circuit; therefore there has been a reflections and allowing PD wave shape analysis
lot of intrusive works performed on each if a short time constant circuit detects PD
manhole. Maintenance work may induce activity.
mechanical damage on installed cable, when the
cable is pulled to adjust length reduction during D. UWB Nearfield sensors
joint replacement. Other important source of Performing measurements above SoMHz in a
mechanical damage is produced during short 60Hz designed equipment represent a great
circuit, due to expansion forces originated during challenge if distortion and reflection in PD
transient fault current heating that produces large pulses are to be avoided. One solution offered to
cable movement [iii]. Considering the inspection maintain pulse linearity is the use of near field
on each manhole performed along the cable path, techniques, where PD electromagnetic field is
incipient fault of mechanically damaged cables detected near the evaluated cable or its accessory
can be found. Evidence of a mechanically using an active or passive sensor denominated
damaged cable involves changes in cross section UWB near field sensor, as depicted in Fig. 2.
and leakage, detectable by manhole floor
inspection.
B. Overload and hot spot delection
During manhole inspection, IR images were
obtained from each cable joint and cable duct, to
correlate cable temperature and circuit layout.
Temperature information was related to instant
load measured with a C T clamp. I n this way it
was easy to detect sections where the cable
exceeded its thermal capability even when
measurements were performed off the main load
UWB
peak. Furthermore, analysis of load balance on near held sensor
each phase provided information regarding Fig. 2 Capacitive n e a r field sensor princple
202
Those sensors have a response that attenuates cable section between two manholes, however it
with the cube of the distance and respond to the was considered that joint areas were much more
electric and magnetic field generated by a PD. susceptible to this type of damages than the rest
Because attenuation is larger for distant sources, of the cable, and those were easily covered by
and distance in a SOMHz circuit means a few the measurements performed here and reported
meters, calculated as: next.
1= s (1)
xlil. CALIBRATION TECHNIQUE
f When performing on-line measurements, it is not
Where: possible to follow the standard calibration
v Corresponds to wave speed inside the method according to [v]. However, for the
dielectric, calculated us: measuring frequency employed, the cable length
1 (21 is much longer than the calculated wavelength,
v= . .
:.
--
-:
-.
r>
=-
-.
::
::
-
as explained in eq. ( I ) . Considering a cable as an
-,./E()&, / f Q p p
infinite transmission line, the pulse ends before
Considering that: any reflection occurs.
co = 107/(4nC2) F/m, Absolute permitivity in If the same arrangement is performed on a h/4
vacuum
cable loop and the calibration pulse is injected on
po = 4~c.10.’ H/m, Absolute permeability in one end, then a PD pulse could be detected on
vacuum. the circulating current measured in the loop
E= Relative permitividad of cable insulation
using a UWB sensor. Although this arrangement
=3.
have pulse reflections and distortion, if the
p = Relative permeability of cable insulation calibration loop is maintained constant in all the
=I.
measurements, the PD detection system would
C = Light speed in vacuum = 3x1 08m/s have the same correction factor in all instances,
f = Frequency, given as Il4.r allowing repetitive measurements and
T= PD front time, 1-lOns range. comparison between different sites. Due to pulse
Then:
distortion, absolute values obtained through
~0.577C (3)
Therefore measurements cannot be directly related to
h=3.46m (4) apparent charge according to [VI o r induced
Experiments showed that the system responded charge according to [vii], but arbitrary units(u)
can be used for comparison purposes.
to signals generated in the region of 10h.
Therefore, PD pulses injected in any direction of
the cable length within a 40m radio could be XIV. RESULTS
detected by the UWB near field sensor. On the At best, a cable in good conditions generates low
other hand, signals, which were not generated ” _ - ~
inside the cable, were rapidly attenuated if its
source exceeded this distance. Considering that
in this case, the cables lay under the street, there,
were no other sources of interference in the
vicinity, allowing the system to offer sensitivity
better than 40pC.
XII. DISTRIBUTED MEASUREMENTS
Having a 40m measuring range, a so-called
distributed measuring approach was employed;
were a PD measurement on each manhole and
terminal along cable path was performed, this
during normal cable operation. Considering that PD levels along its whole length and leaving a
the distance between manholes varied between 5 few meters undetected causes no harm. On the
and loom, almost all the cable length was other hand a faulty cable with a serious damage
covered if measured on adjacent manholes. Only caused by ageing, leakage or mechanical damage
punctual damage caused by imperfections could would have tracking across some paper layers
be left undetected if located near the middle and would generate very high PD levels,
203
detectable by this method when occurring in the Fig. 5 PD pattern obtained in a cable classified as
manhole vicinity. Being accessories like joints high or requiring maintenance (> IOOOunits,
and terminals the main source of defects, the 500u/div).
distributed measurement approach was Although the technique may give apparent
considered as satisfactory and very easily charge values if attenuation due to position is
performed with a clamp-on sensor technique, as considered, the main goal of this method is the
depicted in Fig. 3. reduction in failure rate by focusing
Fig. 3 General view of a manhole during on- maintenance efforts on those cable sections,
line PD measurements. joints o r terminals that require labor. This
The study was conducted on each terminal and incipient fault detection scheme is simplified if
manhole with a joint, for each phase in all .... _. . .. .... -.
- .
feeders of 12 substation selected as those with
,higher failure rate. A typical plot Q-4 for a low
level cable classified as normal (<3OOu), is
presented in Fig. 4.
Manhole
-
- Phase
-
A - Phase
- ___-Phase 6__ .-__ C
. __ .
PD values a r e plotted vs manhole position,
taking 0 as substation terminal, as depicted in
Fig. 6.
Fig. 6 PD level vs. inanhole position indicating
;\
IEXI..i i
t ;
...-...... ..........*...-......i......-...+.
f /
Cl
....I.. .........
joints requiring maintenance (above 1OOOu)
Very high levels of PD were detected on dry,
leaking or broken oil terminals. Those levels
Fig. 4 PD pattern obtained in a cable classified as reached 8 , 0 0 0 ~in some cases, inducing a high
normal (<300 units, 5Ou/div). reading on the first manhole. Hot spots in
Defects located during measurements on each terminals also produced a high characteristic EM
phase found in a manhole along 75 feeders, interference when contact was almost lost.
allowed to classify the results in three large However those thermal problems were better
groups, one having less than 300 units, detected by IR images.
considered as satisfactory. A second group Leaking joints inside manhole produced medium
including measurements with intermediate readings, and visual inspection was enough to
levels, ranging between 300 and 1000 units demand for maintenance work to correct leakage.
and t h e third one considered as those areas On the other hand, those joints that leaked for
with a high electromagnetic field, giving long time went dry, and gave readings above
readings above 1000 units. A typical PD map 10,000 units.
for a high reading is presented in Fig. 5. Regarding cable measurements obtained in 30-
I. '
. .:,. .i ".......
...........I...
. 4.-...... +.
;,**. 1 " 4
... :..
4
: : I
..-. ......"..+ ..........:...........:.....-
1
:
..I.-
*
V..
-
year-old cables installed and which operated
according to specification, gave readings within
low to medium levels, indicating that those
cables were in very good conditions considering
its age. Those cables sections which ampacity
had been exceeded for several years, produced
high readings, reaching 20,000 units in some
cases. Those exceptionally high readings
produced a cable failure before any maintenance
work was carried out.
All measurements performed in 75 feeders, high
readings of PD or IR images indicated that
maintenance works should be carried out in 21
joints, 23 terminals and 15 cable sections
204
between manholes. Only those 15 cable sections
are considered for replacement, and the rest of iv V. R. Garcia-Colon, A. Nava-Guzman, L.A.
the circuits were satisfactory for their use under Salgado-Salgado, R. Canales-Ruiq “Location and
normal conditions. assessment of partial discharge on substation
equipment using ultra wide band techniques”, EPRI
!SubstationEquipment Diagnostics Conference VIII,
XV. CONCLUSIONS New Orleans, February 2000
A new diagnostic tool for on line evaluation was v International Electrotechnical Commission, Standard
presented and confirmed with field IEC60270 High Voltage Testing - Partial Discharge
measurements of over 2000 joints, 400 terminals Measurements, Third Edition. Switzerland, 1996-06.
and thousands of meters of installed paper vi A.R. Von I-lippel. Dielectrics and Waves, MIT
Press, Massachusetts, 1954.
insulated impregnated cables rated at 23kV, with
excellent sensitivity, selectivity and ease of use. vii Pedersen, “On the Electrodynamics of Partial
The results obtained by U WB technique Discharges in Voids in Solid Dielectrics”. 3rd Int.
indicated that high failure rate was associated to Conf. on Conduction and Breakdown, Torndheim.
punctual or isolated defects occurring along the Norway, 1989, pp. 107-1 16.
cable path. The main difference between the
U WB method and the original complete
substitution plan, resides on the incipient fault
detection criterion. This incipient fault detection
criterion considers that the cable is not aged
evenly along its length, but it has weak points
located on high stress areas where incipient
faults are evolving. When those incipient fault
areas are evenly distributed along the cable path,
then substitution of the whole cable length is
recommended in short time. On the other hand, if
only a few or none incipient fault points are
detected, minor maintenance work should be
carried out and the cable may remain in service
until a new UWB evaluation is performed.
This predictive maintenance approach centered
in cable reliability makes economic sense and
reduces cable outages and unnecessary
substitution.
XVI. ACKNOWLEDGMENTS
The author wants to acknowledge the facilities
and support in the development of this work to
Luz y Fuerza del Centro.
XVII. REFERENCES
i Luis Ttllez K., “Propuesta de Cambio Estructural de
la Industria Eltctrica en MCxico”, Secretaria de
Energia. Mtxico 1999.
ii R. Espinoza y Lara, D. Bautista Ortiz, A. Lira
Meneses, “Metodologia para el abatimiento del tieinpo
de interrupcion por usuario (TIU) parte II: Cables dt:
salida“, Reuni6n de verano dcl Capitulo dc Potencia
del IEEE Secci6n MCxico RVP-A1/2000, Acapulco,
Mtsico. julio 2000, pp. 60-65.
iii S.Y. King, N.A. Halfer. Utidt.r.groimdPower
Cables. Ch. 9, Longman Group LTD.. Essex, U.K.,
1982.
205