2016 International Conference on Condition Monitoring and Diagnosis - Xi'an - China

Comparison Study of Partial Discharge

Detection Methods for Switchgears

Wenbo Fanl, Shilei Guanl, Jinwei Ful, Lisheng Li2, Jianxiu Li2 ,Wei Wang3*, Xiaoyu Yan3
I. China electric power research institute, State grid corporation of China, Beijing, China

2. Shandong electric power research institute, State grid corporation of China, Jinan, China

3. Beijing high voltage and EMC key laboratory, North China electric power university, Beijing, China
ww790326@163.com

Abstract-Metal enclosed switchgears are widely used in occur fmally [I, 2]. According to the above mentioned fact, it is

distribution networks. According to the investigation, most of an extremely important work in the grids that monitors the

power outages are caused by switchgears thermal or partial switchgears insulation real time, especially the PD detections.

discharge failure directly. So, more attentions are focused on the

Fairly numerous methods can be used for PD detection of
detection of switchgear cabinets. For the partial discharge
metal enclosed switchgears today, includes ultrahigh
detection, there are many PD detection methods can be used for
frequency (UHF), radiofrequency scanning (RFS), transient
switchgears, such as acoustic emission (AE), ultrahigh frequency
earth voltage (TEV), acoustic emission (AE) and high
(UHF), transient earth voltage (TEV), high-frequency current, et
frequency current (HFCT), etcY' 4]. These methods have
al. However, the comparative analysis of the sensitivity and
similar principles which detect the mutations of the electric
locating effect of the above methods is not enough. In this paper,
signals or non-electric signals from the inside of switchgears
a real switchgear partial discharge test platform with three
to determine the amplitude of PD, and locate PD by
cabinets was set up, partial discharges from three kind of typical
comparison of signals detected from the sensors in different
models in different locations of the switchgear under various
locations [4]. They are all widely used in the on-site detection;
voltages were measured by the four methods, and the sensitivity
however, it is fairly rare that these methods are used at the
of detection and locating for the four methods were compared.
same time, so the comparison study of the accuracy of these
The test results show that the AE method has the best sensitivity
methods is now rarely done, and the conclusions are also few.
on the switchgear PD detection, TEV method has largest

amplitude of PD signal; the four methods can locate the partial

In this paper, PO models which based on real switchgear

discharge all. The union of AE and TEV is the most effective

were set up, and the PO signals were detected by four

measure to detect the PD of switchgears in the on-site detection

different detecting methods (UHF, HFCT, TEV and AE) at the

same experimental condition. With the voltage on the models

environment with a high background noise.
increasing, the background noise and valid signals were
Key words-Switchgears, PD detection, PD locating
recorded by oscilloscope and PO detector, curves of signals

I. INTRODUCTION changing were drawn to compare the sensitivity, and the PO

Metal-clad switchgears are one kind of the major electric locating function was compared by changing the location of

equipment in the lOkV distribution grids; the grids' operation sensors and comparing the PO signals from different

safety is deeply affected by the running stability of the locations.

switchgears. However, the switchgears perhaps lose part of

II. BASTS OF FOUR PD DETECTING METHODS
insulation capability, because of the insulation aging,
A. Ultrahigh Frequency (UHF)
operating errors of the operation personnel or erosion from the
Transient pulse current will appear when PO occurs, and
terrible operating environment, and partial discharge (PD)
ultrahigh frequency electromagnetic waves will be excited by
would be caused by the above reasons. Then the insulation of
the transient pulse current and the radiation spread outwards
switchgears will get worse as PD affection, and accident may

978-1-5090-3396-6/16/$3\.00 ©20161EEE 319

in some speed. These electromagnetic waves normally have a in pattern recognition area [1, 9].

frequency from 300MHz to 30Hz, because the frequency of

!:e&!:!� or i r'L 6a.l lated .u �.!.r
the on-site interference from coronas are often lower than � l i l\.e r l=larh

300MHz, UHF method has several advantages such as high

sensitivity and strong resistance of interferences[5, 6].

B. High Frequency Current (HFCT) e let: t romAgl'L.e tie waVe

HFCT normally means pulse current signals which have a

frequency of 3MHz-30MHz, as strong energy as the HFCT

has, it will spread to the ground wires of switchgears when partial d i !;e t-Jarge

PD occurs inside the switchgears. Signals can be detected on

the ground wires by the HFCT sensors which are made based Fig. 1. Principle of TEV in PD detection of switchgears

on principle of Rogowski coil, induced voltage will be excited III. DESIGNING OF PD DETECTING EXPERIMENTS

by the HFCT passing through the center of core and transport A. Set up of the Experiment Platform
to the amplifier of the sensor [7]. HFCT method is kind of The experiment platform had been set up based on real

non-invasive methods, simple and reliable. switchgear, and the PD models had been designed on the

insulators of switchgear directly. Three models were designed

C. Acoustic Emission (AE)
included point-plate, surface and floating; Fig. 2 shows the
High frequency mechanical vibration can also be caused
circuit connection of the experiment equipment, UHF and
by the PD appearance. This acoustic emission can be detected
HFCT signals were read from the oscilloscope; AE and TEV
by special acoustic sensor; therefore, PD signals are
signals were read from the detector. Signals from the
calculated by the AE detector. As non-electric signals this
detecting impedance were used as the PD reference signal.
method detecting, it has very strong resistance from the

electromagnetic interferences. These acoustic signals have a

frequency of about 20 kHz-2MHz, and solid medium will

make the signals a severe attenuation, so the sensor must have Test Transfonner Limiting Resistor

a special amplifier to detect acoustic signal outside the case of �II Coupling Capacitor
switchgears [6]. Now the PD locating of AE method is based 220V AC power
lIFeT Sensor

on time difference of the AE signals from sensors installed in Detecting Impedance

different locations.

D. Transient Earth Voltage (TEV)

Transient earth voltage is a small transient signal (duration Fig. 2. Circuit Connection of the Experiment Equipment

time only several nanoseconds, amplitude about dozens of

In order to compare the PO locating function of these
millivolt), which appears on the metal case of switchgear
methods, after locating the PO model, these sensors would be
when PO occurs in it, and caused by the radiation of high
installed in different places to detect the signals' difference at
frequency electromagnetic waves [2, 8]. The PO strength can
the same voltage condition as Fig. 3 shown. In Fig. 3, the
be measured indirectly by detecting TEV signals using special
numbers are as the different installation places and detecting
TEV sensors. The principle of TEV method is shown by Fig.
points, and different types of sensors are displayed as different
1. Signal amplitude, number of PO times and characteristic of
colors.
PO phases are three major references which be recorded of

TEY. Advantages such as easy to operate and larger detecting

range does TEV have, it is widely used in low-voltage grids.

However, less data than other methods makes its shortcoming

320
 DETECTING rOlNTS OF TEV as 2.7kV ; under this voltage, PD strength was 400pC. After
DETCETING POINTS OF AE
DETECTING POINTS OF HFCT
DETECTING POINTS OF UHF
the voltage increasing, the increasing PD strength has been

recorded as Table 2 as followed.

Table II PD strength of point-plate model under increasing voltage

"v Voltage/kV 2.7 3.0 3.3

HVPOWERSUPPLV PD Strength/pC 400 500 640

Voltage/kV 3.7 4.2 4.4

PD Strength/pC 850 1000 1200

From 2.7kV to 4.4kV, under every voltage, signals

Fig. 3. Different Detecting Installation Places of the Sensors
detected by each sensor have been recorded and drawn into

curves as Fig. 4. In this figure, the left Y-axis corresponds to

B. Process o{the Experiment
the signal amplitude of UHF and HFCT; the right Y-axis to
After the platform set up, the four sensors were all
the signal amplitude of AE and TEY.
installed in their own No.1 detecting points, and read the

background noise signals before the power supply turning on.

40
Then, high voltage around lOkV would be increased to the PD rr - --' -- --'--'
- l - -- - . '
I' I -- n -
3 00 � Am p-i Ide O f V I F S ig al
�-

model, recorded the inception voltage of PD and the PD - Amplitude of AE signal

n
- Amplitude ofTEV sig al
30
strength and signals from different sensors at the higher - Amplitude of HFCT signal

voltages. This process should be stopped after the voltage

increased to some level (almost twice as the inception

voltage), then another same procedure carried on after

changing the locations of the sensors as No.2/3 in Fig. 4. The

same variable which must be controlled was the PD strength

o 2 3 4
detected from the detecting impedance. YoltagelkY

The reflected parameters of the sensors used in this Fig. 4. Curves of Different Signals Following Voltage (Point-plate Model)

experiment were shown as Table 1. Because of the sensitivity

As the figure shows, at the voltage of 2.7kV, AE and TEV
difference of the sensors, deviation may exist, but it will not
signals have been increased, however, the signals of UHF and
affect the qualitative results.
HFCT are still none. Therefore, AE and TEV have better
Table I Parameter of Sensors Used in this Experiment
minimum measuring ranges; if comparing the relative
Frequency Detection
detecting value, AE is much better. While comparing the
Sensor Sensitivity
Band Range
signal and noise ratio (SNR), as UHF has a quite small
UHF 300M-2GHz 1O-50pC 0- 10000pC background noise less than 3mV, according to the equation to
HFCT 3M-30MHz 50-100pC 0-25000pC calculate SNR:
AE 20k-2MHz <ldBuV -7-65dBuV

TEV 10M- 150MHz <ldBmV 0-60dBmV (1)

IV . ANALYSIS OF THE RESULTS OF EXPERIMENT

UHF has SNR larger than 40dB under the voltage of
A. Comparison of sensitivity between the above methods
4.4kV and larger than TEV and AE at the same time. UHF has
Three PD models were all under voltages one time. All the
much better effect when PO is stronger. However, in this
PD inception voltages and PD strength at other different
paper only one PO source exists, UHF noise would be much
voltages have been recorded. Point-plate model was as the
more when detect PD on site, so its SNR may decrease either.
example and the PO inception voltage of it has been recorded
B. Comparison ofPD Locating Function and Features

321
 Measuring results of floating model would be the example detecting location, the largest gaps would be found in signals

to demonstrate the comparison of PD locating function. In this from AE sensor. Therefore, if the PD locating way which

experiment, under 11.2kV, PD strength and signal amplitudes based on spatial geometry by the difference from signals of

from the sensors of floating model have been recorded and different detecting location is taken, AE would be the best

displayed as Table 3. recommendation. Moreover, TEV signals have a large gap,

but the absolute peaks are also larger than AE, so the relative
Table III Comparison between Different Detecting Locations
difference is smaller.
Voltage/kV 1 1.2

PD Strength/pC 1000 V. CONCLUSIONS

Signal t of UHF Sensor/mV 50 In this paper, experiment platform based on real

Signal 2 of UHF Sensor 1mV 40 switchgear has been set up. PD signals have been detected by

Signal 3 of UHF Sensor 1mV 50 four different methods: UHF, HFCT, AE and TEV at the same

Signal 1 ofHFCT Sensor 1mV 40 time. Signals under increasing voltages and from different

detecting locations have also been recorded and compared by

Signal 2 ofHFCT Sensor 1mV 40
math analysis. According to the results from the experiment of
Signal 3 ofHFCT Sensor ImV 37
this paper, conclusions can be summarized as: under the same
Signal t ofTEV Sensor Id8 33
condition of voltage, the signal amplitude of TEV is the
Signal 2 ofTEV Sensor Id8 32
largest one; in better environments, UHF has a better SNR;
Signal t ofAE Sensor Id8 17
AE has the best locating function on spatial geometry locating
Signal 2 ofAE Sensor Id8 15
method. Another factor which needs to be noticed is that in
Number of signal is from relative number of detecting
many distribution grids, the ground wires of switchgears are
point location in figure 4. After a series of voltage increasing,
closed inside, so the HFCT method has using limits to some
a series of signal data were recorded, and had been
extent. In summary, AE with TEV will have the best effect in
transformed to curves in Fig. 5. In this figure, different signals
on-site PD detecting of switchgears.
from relative sensors are distinguished by different colors;

signals from every detecting location No.1 are displayed as ACKNOWLEDGMENT

solid lines, and No.2 as dashed lines, NO.3 as dotted dash

Thanks for the support from state grid corporation of
lines.
China for the work in this paper.

REFERENCES
400 '-�------�--rT 50
-UHF
> [I] WEI Zhen, ZHANG Qiang, QI Bo, et al. Analysis of TEV Caused by
��
<:C
"0 Partial Discharge of Typical Faults in HV Switchgear [1]. High Voltage
�
--UFCT
300 40 Apparatus,20 14,02:60-67.
"
" "
tI)
<....
..,
tI) [2] ZHANG Tao, WANG .Iunbo, LI Guowei. Research on the Partial
o Discharge Detection Technology for 10 kV Switchgear [1]. High
.g 200 30 � Voltage Apparatus,20 12, 10:100- 104.
,e
"0
0. .� [3] Cheng Xue, Du, B.X., Tao Han, Yang, B.X. Comprehensive detection
E 0.
-< 100 20 � and analysis of partial discharge in switchgear[C).lnnovative Smart Grid
0; Technologies - Asia (ISGT Asia), 20 12 IEEE , vol., no., pp. 1-6, 2 1-24
� May 20 12.
en
[4] P. Mulroy, A. Hurtado and D. Badetz. On-line partial discharge
1000 1500 monitoring system for distribution networks[C]. Condition Monitoring
PO AmplitudelpC and Diagnosis (CMD),20 12 International Conference on, Bali,20 12, pp.
542-545.doi: 10. 1 1 091CMD.20 12.6416200.

Fig. 5. Curves of Signals from Different Sensor Locations under Increasing [5] ZENG Xiong-jie, JIANG Jian-wu, HOU Jun. Application of TEV and
UHF in PO Detection for the Live 10 kV Switchgear [.I]. High Voltage
Voltages Apparatus,20 12,0 1: pp.4 1-47.
[6] S. Tenbohlen, D. Denissov, S. M. Hoek and S. M. Markalous. Partial
discharge measurement in the ultrahigh frequency (UHF)
According to the testing results, if evaluating the PD
range[C).IEEE Transactions on Dielectrics and Electrical Insulation, vol.
locating function with the relative differences between every 15, no. 6, pp. 1544-1552, December 2008.doi:
10. 1 109/TDEl.2008.47 12656.

322
[7] GUO Chanxin,ZHANG Lianhong, YAO Linpeng, et al. Application of Insulation (ISEI), Conference Record of the 20 12 IEEE International
HF/UHF Joint Partial Discharge Analysis to On-site Power Cable Symposium on , vol.,no.,pp.439-442, 10- 13 June 20 12.
Terminal Detection [J]. Electric Power Automation Equipment, 20 10, [9] L. J. Wang, et at. Experimental investigation of typical partial discharge
05: pp.92-95. signals in medium-voltage switchgear[C]. Discharges and Electrical
[8] Reid, A.J, Judd, MD, Duncan, G. Simultaneous measurement of partial Insulation in Vacuum (ISDEIV), 20 14 International Symposium on,
discharge using TEV, IEC60270 and UHF techniques[C].Electrical Mumbai,20 14,pp. 393-396.doi: 10. 1 109/DEIV.20 14.696 1702.

323