See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/269271707

Capacitive Slot Coupler for on-line partial discharge monitoring of stator

winding insulation

Conference Paper · June 2014

DOI: 10.1109/EIC.2014.6869383

CITATION READS

1 101

5 authors, including:

Yong Joo Kim Kim Geun Ju

Chung-Ang University Korea Electrotechnology Research Institute-KERI
49 PUBLICATIONS   234 CITATIONS    46 PUBLICATIONS   222 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Yong Joo Kim on 08 July 2015.

The user has requested enhancement of the downloaded file.

Capacitive Slot Coupler for On-line Partial Discharge
Monitoring of Stator Winding Insulation
Yong Joo Kim, Geun Ju Kim and Jung Il Kim, Hee Dong Kim and Tae Sik Kong
Ansan Research Division Technology Expert Center
KERI KEPCO Research Institute
Ansan, Kyonggi-do, Republic of Korea Daejeon, Republic of Korea
[email protected] [email protected]

Abstract — Authors introduce a novel on-line partial characteristic and sensitivity. CSC WL (Double Copper Sheet
discharge monitoring sensor, which frequency band is in 100 kHz Layers in Epoxy Plate, W: 30 mm, L: 500 mm, D: 3mm) is
- 20 MHz. The novel sensor called CSC (Capacitive Slot Coupler) designed to insert between turbine generator stator bar and
is very compact in size and could be inserted between wedge and wedge, and CSC WS (Single Copper Sheet Layer in Epoxy
stator bar. In this paper, an equivalent partial discharge Plate, W: 15 mm, L: 400 mm, D: 2mm) for large motor/hydro
detection circuit is introduced considering the various PD sources. generator.
And its capability in PD source identification is compared with
the conventional PD sensors in sensitivity and frequency domain.

Keywords—partial discharge, stator bar, on-line monitoring,

internal discharge, slot discharge, end winding discharge

I. INTRODUCTION
Nowadays on-line PD monitoring system becomes the
prevailing solution to protect stator winding insulation of large
rotating machine against an incipient failure. Generally there
are two types of such systems available in market. One is to
detect PD in frequency band of higher than tens of MHz which
has high signal to noise ratio [1], [2]. But the system is known
to detect PD only several stator bars away from PD source
which are under the highest voltage stress. The other one is to Fig.1. Prototypes of CSC WL and CSC WS
monitor PD in low frequency band (50 kHz-20 MHz) which B. Equivalent Partial Discharge Detection Circuit
has low signal to noise ratio [3]. Therefore, it has been
claimed that the latter system could identify PD sources such CSC is generally installed on outer semiconducting layer of
as internal discharge, end winding discharges and slot stator bar as shown in Fig.2 (a). Considering surface discharge,
slot discharge, and internal discharge, an equivalent PD circuit
discharges [4]. Also, the previous works [5], [6], [7], [8] have
could be drawn as shown in Fig.2 (b) [9]. In addition, phase-to-
already introduced the similar scheme to judge insulation
phase capacitive coupling and airborne radiation could be
condition of large rotating machine due to low attenuation included in the equivalent PD circuit [10].
characteristic of PD signals in frequency range below 500 kHz
under low noise environment. C. Test set-up and Conventional PD pulse Calibration
In this context, the authors introduce a novel on-line partial As shown in Fig.3 (a), the test set-up was arranged with one
discharge monitoring sensor, which frequency band is in 100 6.6 kV diamond stator winding of hydro generator. In the first
kHz-20 MHz. The novel sensor called CSC is very compact in experiment, four sensors (CSC WL/50Ω, CSC WS/50Ω,
size and could be inserted between wedge and stator bar. In 1,000pF/50Ω, 110pF/680Ω) were connected to the winding.
the paper, an equivalent PD detection circuit for slot discharge Two sensors (CSC WL and WS) were fixed on semiconducting
and end winding discharge is introduced considering the layer of the winding center and the other two sensors
design structure and location of the sensor. Furthermore, the (1,000pF/50Ω, 110pF/680Ω) were connected to conductor of
laboratory experiment on CSC explains its capability to the winding end. Regarding cable connection, the sensor
identify the various PD sources. (1,000pF/50Ω) was connected to high voltage cable of 30 m
II. EXPERIMENTAL SET UP AND LABORATORY TEST long (about 250 n seconds delay), and the other one
(110pF/680Ω) has coaxial signal cable of 10 m long (about 60
A. CSC Partial Discharge Sensors n seconds delay).
As shown in Fig.1, two types of CSC PD sensors were After test set-up, PD calibration pulse was injected into the
fabricated in the laboratory to compare their frequency end copper conductor of the winding by a conventional PD
calibrator. In Fig.3 (b), it could be seen that the sensor
(1,000pF/50Ω) received the highest signal level
(250mV/1,000pC) without any time delay. Investigation shows
that the initial fast pulse of the wave shape was affected by
radiation. In sensitivity comparison of the sensors, both CSC
WL and 110pF/680Ω PD sensors show the level of about
96mV/1,000pC, and CSC WS indicates 66.8mV/1,000pC.

(b) Response from 1,000 pC calibration pulse injection

(Horizontal axis: 400ns/div, Vertical axis: 200mV/div)
Fig.3. Test set up with diamond stator winding and five PD sensors
(1) CSC WL/50 Ω (1-1) Conventional VHF PD Detector/ 50 Ω
(2) CSC WS/50 Ω (3) 1,000pF/50 Ω (4) 110pF/680 Ω

D. Sensitivity for Sine wave Injection

After conventional PD pulse calibration, sine waves (1 V,
(a) Location of CSC PD sensor and PD activities in stator bar 100 kHz - 30 MHz) were also injected into the end copper
conductor of the winding, and the responses of the four sensors
are illustrated in Fig. 4.

140
130
120 CSC(WL)
110 CSC(WS)
1000pF
100
110pF
90
Voltage [mV]

80
70
60
50
40
30
20
10
(b) PD equivalent circuit for CSC PD sensor 0
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32
C0: capacitance of stator winding insulation Frequency [MHz]
C1: capacitance in series with a discharge cavity
R0: resistance of semiconducting layer
L0: inductance of stator copper conductor
R1: surface resistance between CSC PD sensor and discharge point
R2: contact resistance between CSC PD sensor and semiconducting layer Fig.4. Response of four sensors for sine wave injection
Ls: inductance of CSC PD sensor
Cs: capacitance of CSC PD sensor
Rs: resistance of CSC PD sensor
Below 800 kHz, four sensors show low response level of
less than 40 mV. But around 1 MHz, the response level of two
Fig.2. PD activities and respective equivalent circuit for CSC PD sensor sensors (1,000pF, 110pF) sharply increased to about 80 mV -
100mV. And they show two more peaks around 10 MHz and
20 MHz respectively. However, CSC WS/50 Ω shows the peak
response level of 50 mV - 130 mV in frequency range of 8
MHz - 12 MHz.
E. PD and corona measurement under high voltage stress
For corona (generated by artificial steel needle, confirmed
by buzzing noise and spark light under dark environment) and
PD signals, the laboratory experiment were made to detect their
respective wave shapes and response levels of four sensors
from the winding as illustrated in Fig.5 and Fig.6 respectively.
Fig.6 (a) (b) shows that the wave shapes were detected from
sensor (1-1) conventional VHF PD detector instead of CSC
(a) Test set-up with stator winding WL/50 Ω. In Fig.5 (a), (b), corona responses of four sensors
which occurred near sensor (4) and sensor (3) are shown
respectively. Fig.5 (b), corona response of sensor (3) shows
delay time of about 250 n seconds due to high voltage cable of
30 m long. However, in Fig.5 (a), no time delay was detected
in wave shape of sensor (3) due to direct airborne radiation
effect from corona occurred near sensor (4). In Fig.6 (a), (b),
corona and PD wave shapes from four sensors are shown
respectively. Fig.6 (a) shows the responses of four sensors
from corona occurred near sensor (3) and the responses of PD
occurred near sensor (1-1) is shown in Fig.6 (b) respectively.
From the laboratory experiment, it has been confirmed that
sensor (2) exhibits better response than sensor (1-1) both in
corona and PD measurement.

(b) Response from PD occurred near sensor (1-1) (Horizontal axis: 1 μ sec/div, Vertical axis: 200mV/div)
Fig.6. Response of four sensors from PD, (1-1) Conventional VHF PD
Detector/50 Ω (2) CSC WS/50 Ω (3) 1,000pF/50 Ω (4) 110pF/680 Ω

F. PD measurement with CSC WL/50 Ω and Conventional

VHF PD Detector/50 Ω
(1), (2) CSC WL/50 Ω and (3), (4) conventional VHF PD
detectors/50 Ω were mounted on the winding as shown in Fig.
7 (a), and PD measurement was performed under 8 kV stress.
From Fig. 7 (b), (c) and (d), a sensor which obtains the highest
frequency component and the highest peak magnitude in its
wave shape could be judged as the one that is near PD location.
a) Response from corona occurred near sensor (4) (Horizontal axis: 400ns/div, Vertical axis: 200mV/div) Therefore, investigating the signals in Fig. 7 (b) (1), (c) (3) and
(d) (4), it could be said that their PD locations are near sensor
(1), sensor (3) and sensor (4) respectively. Also It could be
seen clearly that all signals contain VHF component (rise time
less than 5 n seconds) at their wave front. In addition, in Fig.7
(b), time delay of about 20 n seconds between signals (3) and
(4) was also associated to phenomena of airborne radiation.
And as expected, it is also confirmed that the sensor contains
the opposite polarity of HF component to the other sensors
could be defined as the one near PD source. As shown in Fig.7
(b), the conventional VHF PD detector shows very low
response level to PD event occurred near CSC WL (1).
Regarding arrival time delay, about 20 n seconds between CSC
WL (1) and CSC WL (2) as shown in Fig.7 (1), (2) was
(b) Response from corona occurred near sensor (3) (Horizontal axis: 400ns/div, Vertical axis: 500mV/div)
observed in VHF signal component but not in HF signal one.
Fig.5. Response of four sensors from PD
(1) CSC WL/50 Ω (2) CSC WS/50 Ω (3) 1,000pF/50 Ω (4) 110pF/680 Ω

(a) Response from corona occurred near sensor (3)

(Horizontal axis: 1 μ sec/div, Vertical axis: CH1.2: 200mV/div, CH3, 4: 500mV/div)
(a) Test set-up for (1), (2) CSC WL /50 Ω and (3), (4) Conventional VHF PD Detector/50 Ω
 detector/50 Ω were mounted at the same location as shown in
Fig.8. And the additional sensors (1) CSC WL/50 Ω and (4)
110pF/680Ω were also connected to the winding copper end.
Again conventional calibration pulse was injected at the
winding cooper end, and the result is shown in Fig.9 (a). As
expected, conventional VHF PD detector/50 Ω obtained the
lowest signal level due to its intrinsic characteristic of micro
strip antenna. But 110pF/680Ω shows slightly higher response
than CSC WL/50 Ω PD sensor.

(b) Response from PD occurred near sensor (1) (Horizontal axis: 40n sec/div, Vertical axis: 100mV/div)

Fig.8. Test set-up for (1), (2) CSC WL/50 Ω, (3) Conventional VHF PD
Detector/50 Ω and (4) 110pF/680 Ω

In Fig.9 (c), the signals obtained from senor (2) and (3)
represent the response to PD pulse occurring directly under
sensor (3). Both signals contain several peaks of sharp rising
VHF component and tailed by opposing polarity in HF
(c) Response from PD occurred near sensor (3) (Horizontal axis: 80n sec/div, Vertical axis: 200mV/div) component. In Fig.9 (b), (2) and (3) show the wave shapes
explaining that PD occurred near sensor (3). At a glance
investigation, it could be found that peak level of VHF
component shown in Fig.9 (b) (3) decreases with lower rising
time and its wave shape is tailed by relatively higher HF
component showing polarity reversed to the wave shape in
Fig.9 (b) (2). Finally, the wave shape of PD signal occurred
near sensor (2) is shown in Fig.9 (d) (2). However, both
conventional VHF PD detector/50 Ω and 110pF/680Ω show
negligible response to the PD.

(d) Response from PD occurred near sensor (4) (Horizontal axis: 80n sec/div, Vertical axis: 100mV/div)

Fig.7. Response of four sensors from PD

(1) (2) CSC WL/50 Ω, (3) (4) Conventional VHF PD Detector/50 Ω

G. PD measurement with CSC WL/50 Ω, Conventional VHF

PD Detector/50 Ω and 110pF/680Ω
To investigate propagation/radiation characteristic of PD
signal, (2) CSC WL/50 Ω and (3) conventional VHF PD (a) Response of calibration pulse injection (Horizontal axis: 200n sec/div, Vertical axis: 1 V/div)
 occurred longer than several meters away from PD
source.
2) Two types of signals such as corona and internal PD
could be identified by CSC PD sensor.
3) CSC WS shows the higher PD sensitivity in 8 MHz-12
MHz than the conventional PD sensors.

However, further experiment is essential to make more

reliable PD diagnosis system.

ACKNOWLEDGMENT
Many thanks are due to Dr. Seok Gy Jeon and Dr. Seung
Chae Jeon (Korea Electrotechnlogy Research Institute) for
their contribution in fabrication of CSC PD sensors. Also
(b) Response from PD occurred near sensor (3) (Horizontal axis: 40n sec/div, Vertical axis: 200mV/div)
authors really appreciate KEPCO Research Center for
supporting our laboratory experiment.

REFERENCES
[1] H.G. Sedding, S.R. Campbell, G.C. Stone and G.S. Klempner, “ A new
sensor for detecting partial discharges in operating turbine generator,”
IEEE Transactions on Energy Conversion, vol.6, no. 4, Dec. 1991, pp.
700-706.
[2] M. Kurtz, and G.C. Stone, “Diagnostic testing of generator insulation
part III - the partial discharge analyzer and coupling system,” CEA
Research Report Contract RP 76-17, Dec. 1980.
[3] W. Hutter, “Partial discharge detection in rotating machines,” IEEE
Electrical Insulation Magazine, vol. 8, no. 3, May/June 1992, pp. 21-32,
[4] L. Fornasari, G.C. Montanari, and A. Cavallini,”Alarm management in
permanent PD monitoring for generators,” IEEE ISEI, San Juan, Puerto
Rico, June 2012, pp. 571-575.
(c) Response from PD occurred under sensor (3) (Horizontal axis: 40n sec/div, Vertical axis: 100mV/div) [5] A. Wilson, R.J. Jackson and N. Wang, “Discharge detection techniques
for stator windings,” IEE Proceedings, vol. 132, Pt. B, No. 5, Sept. 1985,
pp. 234- 244.
[6] H. Zhu, I.J. Kemp, “Pulse propagation in rotating machines and its
relationship to partial discharge measurements,’ Conference Record of
the 1992 IEEE International Symposium on Electrical Insulation,
Baltimore, MD USA, June 7-10, 1992, pp.411-414.
[7] A.J.M. Pemen, P.C.T. van der Laan, and Wout de Leeuw, “ Propagation
of partial discharge signals in stator windings of turbine generators,”
IEEE Transactions on Energy Conversion, Vol. 21, No. 1, March 2006,
pp. 155-161.
[8] Y.J. Kim, D.H. Hwang, D.S. Kang, and D.Y. Park, “On-line condition
monitoring system for large turbine generator,” CIGRE 2006, A1-303.
[9] Y. Tian, L. Lewin, A.E. Davies, S.J. Sutton and S.G. Swingler, “Partial
discharge detection in cable using VHF capacitive couplers,” IEEE
Transactions on Dielectrics and Electrical Insulation, Vol. 10, No. 2,
April 2003, pp. 343-353
[10] J. Fuhr, M. Haessig, P. Boss, D. Tschudi and R.A. King, “ Detection and
location of internal defects in the insulation of power transformer,” IEEE
(d) Response from PD occurred near sensor (2) (Horizontal axis: 40n sec/div, Vertical axis: 200mV/div) Transactions on Electrical Insulation, Vol. 28, No. 6, Dec. 1993,
pp.1057-1067
Fig.9. Response of four sensors from PD, (1), (2) CSC WL/50 Ω,
(3) Conventional VHF PD Detector/50 Ω and (4) 110pF/680 Ω [11] W. Koltunowicz, and R. Plath, “ Synchronous multi-channel PD
measurements,” IEEE Transactions on Dielectrics and electrical
Insulation, Vol. 15, No. 6, Dec. 2008, pp. 1715-1723
III. CONCLUSION
The laboratory experiment on CSC PD sensor along with
the conventional PD sensors could deliver us the following
results.
1) CSC PD sensor shows the better characteristic than the
conventional PD sensors in detecting PD signals

View publication stats