You are on page 1of 9

Electrical Power and Energy Systems 75 (2016) 74–82

Contents lists available at ScienceDirect

Electrical Power and Energy Systems


journal homepage: www.elsevier.com/locate/ijepes

Discriminating transformer large inrush currents from fault currents


Hamed Dashti ⇑, Mahdi Davarpanah, Majid Sanaye-Pasand, Hamid Lesani
School of Electrical and Computer Engineering, College of Engineering, University of Tehran, Tehran 14395-515, Iran

a r t i c l e i n f o a b s t r a c t

Article history: One of the most widely used approaches for transformer protection is the low impedance differential
Received 24 August 2014 relay, which can be adversely affected by inrush currents. Although the conventional inrush current
Received in revised form 16 August 2015 detectors, i.e. the gap detection and second harmonic criteria, can block the differential relay in most
Accepted 28 August 2015
of such cases, they are severely prone to maloperation in the case of large inrush currents. This paper pre-
Available online 11 September 2015
sents a new strategy to enhance the security of the differential relay. The suggested approach exploits an
intrinsic feature of large inrush currents in three-limb three-phase transformers with the widely used
Keywords:
star-delta connection. Based on the feature, some novel criteria are proposed to improve the inrush cur-
Differential relay
Helping effect
rent detection scheme. To evaluate the performance of the proposed approach for both inrush current
Large inrush current and internal fault phenomena, a real 230/63-kV power transformer is modeled based on the time-
Power transformer protection based transient simulation. Extensive simulation studies and also evaluation using real data reveal that
Security the proposed approach results in a more secure inrush current discrimination method.
Ó 2015 Elsevier Ltd. All rights reserved.

Introduction ential relay should be blocked when the current harmonic ratio
exceeds a predefined set-value [3]. The current harmonic approach
Power transformer is one of the most important equipment in may fail to distinguish large inrush currents, which can occur dur-
electrical systems. Thus, it should be protected by fast and accurate ing energizing a transformer with a considerable remanent flux,
relays to prevent subsequent damages to the transformer due to especially for newly designed transformers [4]. Traditionally, the
internal faults. The low-impedance differential relay is one of the length of the time interval between the zero-crossing points of
most widely used approaches for transformer protection. However, the differential current, the so-called gap detection technique,
this relay may maloperate due to system disturbances, such as has been used by some relay manufacturers to discriminate inrush
over-excitation, CT saturation and especially magnetizing inrush and fault currents [5,6]. However, correct operation of the gap
current. detection technique is likely to be threatened by CT saturation
When a transformer is switched on, the core is likely to satu- which is probable due to a large dc component of the inrush cur-
rate. Under such a condition, the primary windings draw large rent [7].
magnetizing currents from the power system, known as inrush A variety of algorithms have been recently presented to over-
current, even when there are no secondary currents. This results come the inrush current detection problem. These include artificial
in a large differential current causing the transformer differential neutral network (ANN) [8], fuzzy logic [9,10] and wavelet analysis
relay to maloperate [1]. The ability of the differential relay to dis- [11–13]. However, some of these methods (1) need a large data set
tinguish magnetizing inrush currents from internal or external for training, (2) impose a high computational burden on the differ-
fault currents is a major concern when this relay is applied to pro- ential relay, (3) depend on the transformer parameters or initial
tect a transformer [2]. conditions, and (4) seem to be unpredictable against high-
To avoid the differential relay maloperation due to the inrush frequency noise [7]. Owing to these limitations, the methods have
current, it is a common practice to utilize the ratio of the second not reached a practical level yet, and the second harmonic and also
harmonic component of the differential current to its fundamental gap detection techniques are still widely used regardless of their
harmonic component. In this widely utilized approach, the differ- drawbacks to detect large inrush currents [14]. This paper
addresses a new strategy to improve the security of a differential
relay subjected to a large inrush current. In addition, unlike the
⇑ Corresponding author.
gap detection method, the approach is not sensitive to the inrush
E-mail addresses: h.dashti@ut.ac.ir (H. Dashti), m.davarpanah@ut.ac.ir
(M. Davarpanah), msanaye@ut.ac.ir (M. Sanaye-Pasand), h.lesani@ut.ac.ir
current measured by a saturated CT. Furthermore, the proposed
(H. Lesani).

http://dx.doi.org/10.1016/j.ijepes.2015.08.025
0142-0615/Ó 2015 Elsevier Ltd. All rights reserved.
H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82 75

Nomenclature

iA[j] instantaneous current of phase A at the HV-side NDF normalized difference function
iZ[j] zero-sequence current of the HV-side NDF mean value of NDF in each cycle
ia[j] LV-side current of phase-a DF difference function
ic[j] LV-side current of phase-c pu per unit
IN rated current of the HV-side Br residual flux density
In rated current of the LV-side CT current transformer
KCF correction factor equal to the HV-side CT ratio over the NO not operated (illustrating a poor performance)
LV-side one T sampling interval
N number of samples per cycle ms millisecond
np differential current peak value sample number

approach well detects internal faults contaminated with an inrush


current. 1.4
A set of experimental and comprehensive time-domain simula-
tion cases are investigated to demonstrate that the proposed tech- 1.2
nique can improve the differential relay security and
1
dependability, especially if the proposed approach is combined

Current (pu)
with the conventional methods. 0.8

0.6
Conventional inrush current detection approaches
0.4
Conventional approaches often utilize instantaneous differen-
tial current to discriminate between inrush and fault currents. This 0.2
section precisely reports how the differential current should be cal- dwell time
0
culated and how conventional approaches exploit such a current.
0 4 8 12 16 20 24 28 32 36 40
Furthermore, conventional approaches shortcomings to detect Samples
large inrush currents are scrutinized.
Fig. 2. A typical inrush current.
Calculating instantaneous differential current

The schematic diagram of a power transformer is illustrated in of the HV-side and LV-side, respectively. KCF is a correction factor
Fig. 1 with the vector group YNd11. Since the currents at the equal to the HV-side CT ratio over the LV-side one.
transformer low voltage (LV)-side lag those of the high voltage
(HV)-side by 30°, the current phase-shift due to the YNd vector Conventional approaches
group should be compensated to calculate the differential current.
Furthermore, the zero-sequence current should be eliminated from Most commercial differential relays are supplemented by the
the transformer HV-side whose winding is grounded [15]. To ratio of the second harmonic to the fundamental component of
achieve these objectives, the instantaneous differential current of the differential current to recognize inrush current conditions.
phase A is calculated by The ratio has been usually set at the range of 12–20%. Fig. 2
  illustrates a typical transformer inrush current waveform, which
iA ½j  iZ ½j ia ½j  ic ½j
iAdiff ½j ¼  þ K CF pffiffiffi  ð1Þ includes a fraction of time in each power system cycle, known as
IN 3I n dwell time as specified in Fig. 2, where the differential current is
almost zero. Therefore, an inrush condition can be identified where
where iA[j] is the instantaneous current of phase A at the HV-side.
current enters the differential relay operating characteristic
iZ[j] is the zero-sequence current of the HV-side calculated by
whereas the differential current is less than 5% of the rated current
adding up the phase currents. ia[j] and ic[j] are the LV-side currents
for about a quarter of the fundamental power system cycle [16].
of phase-a and phase-c, respectively. IN and In are the rated currents
This criterion is known as the gap detection approach.
CT saturation due to both of the short circuit fault and inrush
currents adversely affects both inrush current detection criteria.
iA
When CT saturation occurs after a short interval subsequent to
an internal fault, CT output currents may be distorted which
A ia
include large harmonic components. Consequently, second har-
a monic criterion may incorrectly operate and block the differential
b relay. However, the gap detection approach can usually distinguish
ib
such a case from inrush current conditions.
On the other hand, when CT saturates due to the transformer
C B c inrush current, the CT output current will shift down in respect
ic
iC to the non-saturated CT current as illustrated in Fig. 3, which
endangers correct operation of the gap detection approach.
iB However, the second harmonic criterion can well operate and
block the differential relay. Accordingly, to take advantage of both
Fig. 1. Schematic diagram of a YNd11 power transformer. methods while avoiding their drawbacks, combination of the
76 H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82

IA shown in Fig. 3a, second harmonic ratio becomes smaller than the
2 pre-defined threshold of 0.15 at 517 ms, i.e., 17 ms after the
Current (kA)

1 transformer energization. Such a condition lasts for 397 ms. Fur-


thermore, Fig. 3b shows that the gap detection approach cannot
0
block the differential relay until 123 ms after transformer ener-
-1 gization. Accordingly, in such a large inrush current condition,
450 500 550 600 650 700 750 800 850 900 950 1000
none of the conventional approaches is able to correctly recognize
(a) the inrush current and thus the differential relay maloperates for
this case.
Second harmonic criterion
1
Ratio

0.5 Proposed strategy


517 ms - 914 ms
0 Large inrush current asymmetries
450 500 550 600 650 700 750 800 850 900 950 1000
(b) Three-limb step-down power transformers commonly utilize
Gap detection criterion the YNd connection, which frequently are supplied from the
Dwell time (ms)

star-grounded winding. For such a winding connection, a large


4 inrush current waveform contains some asymmetries, which is
2 qualitatively justified by the helping effect theory introduced in
500 ms - 623 ms [20]. The asymmetries are utilized to reliably recognize large
0 inrush currents.
450 500 550 600 650 700 750 800 850 900 950 1000
Time (ms) Fig. 5 illustrates a typical set of inrush currents for energizing a
(c)
YNd transformer at the phase-A voltage peak instant. In this case,
Fig. 3. Large inrush current (a) waveform, (b) ratio of second harmonic to the inrush current is initially experienced by phase-C, from O to
fundamental component, and (c) dwell time. t1, while the core limbs are not still saturated in other two phases.
In such a condition, a zero-sequence current flows from the
star-grounded winding that causes a circulating current through
i the delta winding. Therefore, A and B phases simultaneously start
0max
Bm to draw equal currents to cancel out the ampere-turn generated by
their associated secondary side windings. These currents have an
i opposite polarity with respect to the phase-C current, as illustrated
V in Fig. 5, and add up to the phase-A and the phase-B currents.
Meanwhile, the circulating current also generates a flux in the
Br phase-C core limb, which has the opposite polarity with respect
to the phase-C main flux and reduces the phase-C inrush current.
Consequently, the delta winding circulating current has been
known as helping current. It should be noted that the studied
Switching-in instant transformer is switched-on at the phase-A voltage peak with the
Switching-off instant -B mp
zero remanent flux. Thus, the phase-A core limb is not subjected
to saturation and its current is only proportional to the helping
Fig. 4. Maximum inrush current due to a remanent flux.
current.

second harmonic and gap detection methods are used in some


commercial relays [17]. Even by applying such a combined 4
method, the commercial relay may maloperate in the case of large Phase-A current
Phase-B current
inrush current, which is scrutinized in the following section.
Phase-C current

Conventional approaches shortcomings 2 t1

When a transformer is switched off, the excitation current fol-


Current (pu)

t4
lows the hysteresis curve to zero, whereas the flux density may t2
be a non-zero value Br, i.e., remanent flux. This depends on the O
0
instant of the transformer switching-off and the power factor of
the load connected to the secondary side [18,19]. For a remanent
flux density of +Br, a maximum inrush current is drawn when a
transformer is switched on at the instant of the zero voltage with t3
the positive rate of change of the voltage, as shown in Fig. 4. In such -2
conditions, the flux density can theoretically reach the peak posi-
tive value of Br + 2Bmp, where Bmp the is peak of the flux density
which is produced by applying the rated voltage to the transformer
at the voltage peak with zero remanent flux [18]. This condition -4
0 10 20 30 40 50
drives the core into a heavy saturation and thus leads to large
Time (ms)
inrush currents.
Fig. 3 illustrates a simulated large inrush current for which both Fig. 5. Three phase inrush currents in a three-phase transformer with YNd
of the second harmonic and gap detection criteria are evaluated. As connection.
H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82 77

After t1, phase-B saturates and starts to experience its own I x-diff
inrush as the current magnitude begins to increase quickly and 1.5
thus affects on the helping current. From t1 to t2, phase-C current
is larger than phase-B current. Consequently, the helping-current
1 39
is mainly influenced by the phase-B current during the above men- 11 DF [14]≈ 0
tioned time interval. As such, phase-A current has an opposite rate
of change with respect to that of the phase-B current. This current 0.5

Current (pu)
helps to decrease the phase-C inrush current by generating the
opposite magnetic flux. 0 np=75
np=25
After t2, the phase-B inrush current is greater than that of the
phase-C in magnitude. Thus, phase-A current, which is propor- -0.5
tional to the helping current, helps to reduce the phase-B inrush
current by generating a magnetic flux with the opposite polarity. DF [14]≈ 0 61 89
-1
From t3 to t4, phase-C is dominated and determines the helping
current polarity and amplitude. This process is repeated in the next
cycle. -1.5
0 2 4 6 8 10 12 14 16 18 20
According to Fig. 5, the helping effect distorts phase-B and
Time (ms)
phase-C inrush current waveforms specifically at the t1 and t3
instants, respectively. These phenomena result in distortion of (a)
instantaneous differential currents, especially for large inrush cur-
Ix-diff
rent conditions. Phase-A differential current of a YNd transformer
as compared with that of a YNy transformer is illustrated in 3.5
data window
Fig. 6, when both are energized at phase-A zero voltage instant
with the remanent flux of 84%. The YNy transformer inrush current 3
is almost symmetrical with respect to its corresponding peak;
however, symmetry of the YNd transformer inrush current 2.5
waveform is affected due to the abovementioned phenomena. Current (pu)
It should be noted that a larger inrush current results in more 2
asymmetry of one or two transformer current waveforms. This is
exploited to extract an index for applying a novel inrush current 1.5
detection approach that can well operate under large inrush cur- 27
rent conditions and therefore can complement the conventional 1 DF [28]
approaches. The index is elaborated in the following section.
0.5
Difference function 83
np= 55
0
The proposed method is activated once the power transformer 0 4 8 12 16 20
is energized, which can result in a large inrush current unde- Time (ms)
tectable by the conventional approaches. Afterwards, a window
is constructed that begins at one sample after the zero-crossing
instant of the differential current. At the np sampling point, the
current extremum is reached whose current value is denoted by
ixdiff[np], where x represent one of the phase A, B or C. As DF [5]

IdiffA (YNd)
4.5 Peak
IdiffA (YNy)

3.5
Current (pu)

2.5
(b)
2
Fig. 7. DF calculation for the differential current of a typical large inrush event.
1.5
illustrated in Fig. 7a, the two sides of the current local extremum
1 is thoroughly symmetrical for a fault current neglecting the decay-
ing DC component. Fig. 7b illustrates an inrush current of the same
0.5 transformer. It can be seen that the samples are not symmetrical
anymore with respect to vertical line passing through the current
0
peak. To quantify such asymmetries, the difference function, i.e.,
0 2 4 6 8 10 12 14 16 18
DF, is defined as
Samples
DF½j ¼ ixdiff ½np  j  ixdiff ½np þ j; j ¼ 1; . . . ; np ð2Þ
Fig. 6. Phase-A inrush current waveform of YNy and YNd transformers.
78 H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82

where ixdiff [0] denotes the differential current sample after the Saturated CT current
zero-crossing instant, when the transformer is energized. It should 3 Non-saturated CT current
be noted that DF is computed in each cycle after transformer ener-
gization including the time interval when the calculated differential 2.5
currents enter the differential relay characteristic. When the inrush
current damps out, applying the proposed complementary 2
approach is not required anymore.

Current (pu)
1.5
The difference function might be sensitive to the differential
current extremum. Therefore, its associated threshold to recognize
1
the inrush current should be determined for each transformer. To
overcome such a problem, the normalized difference function
0.5
(NDF), i.e., DF/max(Ixdiff) is used and compared with the pre-set
threshold Th to eliminate the current extremum dependency. It
0
should be noted that Ixdiff is the rms value of the differential cur-
rent that is determined for each sample, and max(Ixdiff) is the -0.5
maximum value for each cycle. The next section reports NDF eval-
uation under various system disturbances and also explains how 0 2 4 6 8 10 12 14 16 18 20
Th should be selected. Time (ms)

Fig. 9. Comparison of phase-A differential current of saturated and non-saturated


Evaluating normalized difference function CTs.

The defined NDF is initially evaluated for large inrush currents. gap detection approaches until 366 ms after the transformer
Subsequently, to determine the appropriate threshold, NDF should energization.
also be determined under system disturbances including CT satu-
ration subsequent to the large inrush current, internal or external Inrush current distorted by CT saturation
fault current including maximum decaying dc and transformer Inrush currents usually contain a noticeable dc component,
internal faults that may occur after the transformer energization. which results in CT saturation. Fig. 9 illustrates phase-A differential
These are studied in the following. current where the CT output current saturates as a result of the
inrush current, 380 ms after transformer energizing, in comparison
with a non-saturated one. Symmetry of the differential current
Large inrush current waveform not only is adversely affected due to the helping effect
The normalized difference function is computed for inrush cur- phenomenon, but also is severely influenced by CT saturation
rent waveforms of a YNd transformer for the following extreme distortions. Fig. 10 shows NDF calculated for the inrush current
conditions, as illustrated in Fig. 8. subjected to CT saturation, as compared with the non-saturated
CT. Since CT saturation intensifies the degree of asymmetry of
Case 1: Energizing the transformer with 84% remanent flux at the inrush current, larger NDF is resulted that facilitates discrimi-
the voltage zero crossing instant (VZCI). nation between the inrush current and other disturbances.
Case 2: Energizing the transformer with 84% remanent flux 1 ms
after VZCI.
Fault current including decaying DC
Case 3: Energizing the transformer with 84% remanent flux 2 ms
Fault currents often include a decaying DC component in addi-
after VZCI.
tion to the fundamental frequency component. In the presence of
Case 4: Energizing the transformer with 84% remanent flux 3 ms
the decaying DC component, the symmetry of the current wave-
after VZCI.
form with respect to its associated peak value is slightly affected.
The DC component depends on the X/R ratio of the system.
Larger inrush current leads to more asymmetry of current
Fig. 11a illustrates a typical current with a decaying DC component
waveform due to the helping current phenomenon. As shown in
resulted by a short circuit fault in a system with X/R = 40. As shown
Fig. 8, a noticeable NDF is resulted for the largest inrush, case 1,
in Fig. 11b, NDF increases only to 0.2, which has a considerable
which cannot be detected based on the harmonic ratio and the
margin the corresponding NDF for large inrush currents.

0.6
Internal fault subsequent to transformer energization
Energizing at VZCI
To enhance the performance of the conventional approaches to
1 ms after VZCI
recognize inrush currents and prevent differential relay malopera-
)

0.4 2 ms after VZCI


x-diff

3 ms after VZCI tion during transformer energization, these approaches are


sometimes set very sensitive. Such strategy leads to the relay misop-
NDF=DF / Max ( I

0.2
eration where an internal fault occurs subsequent to the transformer
energization. To evaluate the proposed approach performance sub-
0
jected to such a condition, various cases are studied as follows.

-0.2 Case 5: Energizing the transformer at the voltage zero crossing


instant subjected to a terminal phase-to-phase fault.
-0.4 Case 6: Energizing the transformer at the voltage zero crossing
V_zero 2 4 6 8 10 12 14 16 18 20
instant subjected to a 15% high voltage winding internal fault.
Time (ms)
Fig. 12 shows NDF for these two cases. Furthermore, Fig. 13
Fig. 8. NDF evaluation for large inrush currents. illustrates the current waveform for case 5, which includes a sinu-
H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82 79

Saturated CT As shown in Fig. 8, the NDF crest exceeds 0.6 perunit for the lar-
0.9 Non-saturated CT gest inrush current whereas it is limited to 0.2 perunit for fault cur-
rents including a large decaying DC component. Furthermore, the
0.8 NDF crest is not beyond 0.2 for the studied internal faults following
0.7 the transformer energization. Accordingly, maximum NDF crest in
NDF = DF / Max ( Ix-diff )

one cycle, i.e., 20 ms for the system frequency of 50 Hz, is consid-


0.6 ered as the first criterion.
n
0.5 Criterion 1 ¼ Maxj¼1
p
fjNDF½jjg ð3Þ
0.4
NDF waveforms can also be exploited to differentiate between
0.3 large inrush currents and other disturbances, especially fault cur-
rents. Comparing NDF for large inrush currents and other distur-
0.2
bances reveals that the area bounded by the NDF waveform for
0.1 large inrush currents is larger than that of the other disturbances.
Consequently, the cumulative NDF from the current zero crossing
0 point until the current peak instant is utilized as the second
-0.1 criterion.
0 2 4 6 8 10 12 14 16 18 20
Time (ms) X
np
Criterion 2 ¼ jNDF½jj ð4Þ
Fig. 10. NDF evaluation for a large inrush current with saturated CT as compared j¼1

with non-saturated CT.


Extensive simulation studies show that NDF experiences notice-
able variations for large inrush currents as compared with that of
soidal fault current and a slowly damped inrush current. It should the other disturbances. Thus, the standard deviation of DF is used
be noted that the second harmonic approach blocks the differential as the third criterion.
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
relay for about 220 ms for this faulty case. u
u 1 X np
Criterion 3 ¼ t
2
As discussed in this section, NDF for large inrush currents is ðNDF½j  NDFÞ ð5Þ
clearly different from fault currents including a large decaying np  1 j¼1
DC component, in terms of both NDF amplitude and its waveform.
In addition, inrush current resulting in CT saturation is more easily where the NDF is the average of first to npth NDF samples.
distinguishable using the proposed NDF, contrary to the conven- Comprehensive case studies based on time-domain simulations
tional gap detection approach. Furthermore, fault current subse- reveal that none of the mentioned three criteria is able to individ-
quent to the transformer energization can be detectable using ually distinguish all studied large inrush currents from other
NDF approach. Therefore, the conventional approaches can be disturbances. However, combining these criteria can help to recog-
complemented with the proposed approach, especially for large nize large inrush currents for all of the studied cases. The final
inrush currents. To apply the proposed approach, a proper index criteria (FC), can well discriminate large inrush currents from fault
should be extracted from the obtained NDF, which is discussed currents contaminated with large decaying DC and even fault
in the next section. currents subsequent to the transformer energization.
Criterion 1  Criterion 2  Criterion 3
FC ¼ ð6Þ
Proposed index T

Evaluating the normalized difference function for various dis-


turbances reveals that NDF waveforms have some recognizable
features to discriminate large inrush currents. To highlight NDF
waveform differences between the large inrush current and other
disturbances, three criteria are introduced that are combined as
0.15
the final index.
NDF=DF / Max ( I x-diff )

idiff-A 0.1
Current (pu)

40
20
0 0.05
-20
(a)
0
0.2
NDF

0.1
Phase-to-phase fault
0 Turn-to-turn fault
-0.05
500 520 540 560 580 600 620 V_zero 2 4 6 8 10 12 14 16 18 20
(b) Time (ms)

Fig. 11. (a) Fault current including a noticeable decaying DC component, and (b) Fig. 12. NDF evaluation for energizing the YNd transformer subjected to internal
NDF evaluation for the fault current. faults.
80 H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82

Current (pu)
idiff-A 4
4 2
Current (pu)

0
220 240 260 280 300 320 340 360 380 400 420 440 460 480 500
2 (a)
0.4

Criterion1
0 0.2
500 520 540 560 580 600 620
0
Time (ms)
(b)
Fig. 13. Fault current waveform for case 5.

Criterion2
5

where T is the sampling interval which is utilized to compensate the 0


effect of sampling frequency. Various inrush current conditions and (c)
other system disturbances should be studied to adjust the FC set-

Criterion3
value, i.e., Th, which is discussed in the following section. 0.1
0.05
0
Evaluating the proposed index
(d)
1500
To evaluate the proposed index performance under the worst
1000

FC
scenarios and to properly adjust the FC threshold for discriminat- 500
ing large inrush currents, this section reports a set of test cases Th
based on digital time-domain simulation studies. 220 240 260 280 300 320 340 360 380 400 420 440 460 480 500
(e)
Studied system Second harmonic criterion
1
Fig. 14 shows the test system under study. The system is com-
Ratio

posed of a 230/63 kV, YNd transformer and the corresponding CTs, 0.5
which is connected to a voltage source through a Thevenine impe-
dance. A new power transformer model is developed in the PSCAD/ 0
EMTDC software environment to study both inrush currents con- Gap detection criterion
Dwell time (ms)

sidering the remanent flux and also internal faults [21]. Not only 5
4
internal faults and inrush currents can be individually studied 3
using the developed transformer model, but also the model can 2
1
be used to study their combination, i.e., an internal fault subse- 0
quent to an inrush current caused by the transformer energization. 220 240 260 280 300 320 340 360 380 400 420 440 460 480 500
Time (ms)

Simulation results
(f)
Fig. 15. (a) Inrush current, (b)–(d) obtained criteria 1–3, (e) obtained final criterion,
Various simulation studies were performed on the system and (f) conventional approaches performance, for a large inrush current.
under study to investigate the proposed approach performance
and specially to adjust the Th parameter. As an example, Fig. 15
illustrates transformer phase-A current including 92% remanent Table 1
Proposed criteria amount and relay maloperation blocking time interval for conven-
flux energized at 1 ms after the voltage zero crossing instant. The
tional approaches, large inrush currents.
proposed three criteria are evaluated for the phase-A differential
current which are depicted in Fig. 15b–d. When the simulation Case Minimum value of criteria Maloperation time interval
no. (ms)
time elapses, the second and third criteria gradually decrease;
however, the first criterion increases at the same time. Conse- Criter. Criter. Criter. FC Second Gap
1 2 3 harmonic detection
quently, these criteria supplement each other, where they are com-
bined to produce the final criterion. The FC is shown in Fig. 15e 1 0.594 11.98 0.228 8480 366 142
which is greater than 1070 for the whole simulation time. 2 0.354 5.47 0.13 1200 276 243
3 0.181 3.38 0.09 276 165 243
Performance of the proposed approach for various inrush cur- 4 0.472 6.4 0.154 2330 164 0
rent conditions and CT saturations is summarized in Table 1. Since 5 0.537 12 0.21 6950 335 222
a combined scheme would be utilized for decision making, the pro- 6 0.306 5.8 0.121 1076 352 303
posed index should recognize inrush currents for which the con- 7 0.156 3 0.074 175 225 321
8 0.41 5.8 0.126 1496 52 241
9 0.93 18.1 0.29 26,200 0 NO

VTh 230 kV YNd11 63 kV


ZTh
Y Δ ventional approaches fail to correctly operate. Obtained amount
0.553 + j 11.05 Ω
230 kV of the final criterion for transformer energizing at 0, 1, 2, and
Sn = 160 MVA
3 ms after the voltage zero crossing instant of phase-A are given
%UK=%16.5
in cases 1–4, respectively. It should be noted that the transformer
Fig. 14. Single-line diagram of the system under study. includes 84% remanent flux for all of these studied cases.
H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82 81

Table 2 When a faulty transformer is energized, short circuit currents


Proposed criteria amount and relay blocking maloperation time interval for conven- are contaminated with inrush current, and therefore the differen-
tional approaches, internal faults.
tial relay may be blocked due to maloperation of the inrush current
Case Maximum value of criteria Maloperation time interval detection unit. Table 2 illustrates the performance of the proposed
no. (ms) criterion as compared with that of the conventional approaches
Criter. Criter. Criter. FC Second Gap when a faulty transformer with zero remanent flux is energized
1 2 3 harmonic detection at the voltage zero crossing. Cases 1–4 represent internal phase
1 0.044 1.435 0.008 2.78 NO 0 to ground faults at 1–4% of the HV winding, respectively. Further-
2 0.045 1.581 0.009 3.27 NO 0 more, cases 5 and 6 show turn-to-turn faults at 4% and 10% of the
3 0.048 1.9044 0.0109 5.067 NO 0
HV winding, respectively. As given in Table 2, gap detection
4 0.0644 2.56 0.017 13.993 220 0
5 0.047 1.81 0.0103 4.43 NO 0 method does not operate for such cases. However, the differential
6 0.055 2.27 0.014 8.85 NO 0 relay blocking condition cannot be released due to operation of the
7 0.1309 1.4835 0.0386 37.443 NO 0 second harmonic method. Maximum obtained FC is about 14 for
these cases, which is much lower than that of the previous cases,
i.e. large inrush currents.
Tables 1 and 2 suggest that by using a simple fixed threshold,
4 the defined final criterion can reliably identify large inrush cur-
rents. In order to compensate for probable effects of noise in fault
3
currents, a relatively large security margin is considered to deter-
Current (pu)

2 mine the final threshold, which is selected equal to Th = 100.

1
Real disturbance
0
To more precisely investigate the proposed strategy perfor-
-1 mance, energizing a real YNd11, 230/63 kV, 180 MVA transformer
0 20 40 60 80 100 120 140
is practically examined. The inrush current signals are recorded
(a) by the relay fault recorder at the sampling frequency of 1600 Hz.
400 The inrush current of phase-C is illustrated in Fig. 16a. Since the
inrush current contains a large decaying DC component, CT satu-
rates four cycles after transformer energization, which distorts
300
the inrush current. Fig. 16b illustrates the obtained FC whose least
amount is 148. Since Th is set at 100, the proposed approach can
FC

200 well recognize such an inrush current, especially after the CT


saturation.
Th In this paper a real 160 MVA transformer (230/63 kV) was mod-
eled, because various parameters of this transformer, especially its
dimensional parameters were available in detail. Next, simulation
0 20 40 60 80 100 120 140
results of this transformer are exploited to propose an appropriate
Sample
index for discrimination between inrush and internal fault currents
(b) and also to adjust the corresponding settable parameters of the
index. All the performed studies show excellent performance,
Fig. 16. A real case study of large inrush current caused by energizing a 180 MVA
power transformer. (a) Differential current of phase-C, and (b) calculated FC. and the proposed index can reliably discriminate between different
currents. In addition, we tried to verify performance of the
proposed approach for another case study, preferably a different
transformer. To do this, inrush current of a 180 MVA transformer
To investigate the proposed approach performance under the
which was available to us was investigated. This transformer has
worst condition, the source equivalent Thevenine impedance is
a different manufacturer and technical specifications and its
considered equal to zero. The FC is also evaluated for such an
current signals are registered with higher sampling frequency than
inrush current condition, when the transformer is energized at 0,
that of the simulation studies. This study clearly reveals that the
1, 2, and 3 ms after the phase-A voltage zero crossing instant. Cases
proposed approach is equally applicable even to the case where a
5–8 illustrate the results for such conditions. The CT saturation
different transformer with the simulated one is investigated.
causes higher asymmetry in the inrush current and consequently
results in a greater FC. For example, if CT saturates for the case 1
inrush current, the calculated FC increases to about three times, Discussion
as shown in case 9. It should be noted that the CT knee-point volt-
age for this case is determined based on the requirement of an The proposed final criteria with the predefined Th can reliably
industrial relay [5]. recognize large inrush currents, while its performance subjected
As given in Table 1, both of the conventional approaches are not to mild inrush currents is not always guaranteed. On the contrary,
able to detect large inrush currents for several cycles after energiz- both the second harmonic and gap detection criteria reliably
ing the transformer. For example for case 1, the second harmonic operate for mild inrush currents while fail to operate for large
and the gap detection methods block the differential relay 366 inrush currents. Accordingly, supplement of the conventional
and 142 ms after transformer energization, respectively. Thus, methods with the proposed approach can be used to effectively
the differential relay maloperates for such conditions. It should detect inrush currents and prevent the differential relay maloper-
be noted that for a small inrush current, e.g., case 4, at least one ation due to inrush currents and also to avoid its misoperation
of the conventional approaches can reliably operate and block for internal faults. Thus, the proposed strategy has the potential
the differential relay. for industrial implementation as depicted in Fig. 17.
82 H. Dashti et al. / Electrical Power and Energy Systems 75 (2016) 74–82

Gap References
detectin
OR [1] Sidhu TS, Sachdev MS. Online identification of magnetizing inrush and internal
Second faults in three-phase transformers. Power Deliv, IEEE Trans 1992;7:1885–91.
harmonic Differential [2] Horowitz SH, Phadke AG. Power system relaying. 3nd ed. England: John Wiley
OR relay
blocking
& Sons, Ltd; 2008.
[3] Pei L, Malik OP, Deshu C, Hope GS, Yong G. Improved operation of differential
protection of power transformers for internal faults. Power Deliv, IEEE Trans
FC ≥ Th
AND 1992;7:1912–9.
[4] Myong-Chul S, Chul-Won P, Jong-Hyung K. Fuzzy logic-based relaying for large
power transformer protection. Power Deliv, IEEE Trans 2003;18:718–24.
[5] Technical reference manual of RET 521 ⁄ 2.3 (Transformer protection
Fig. 17. Block diagram of the proposed inrush current blocking strategy.
terminal). ABB relay catalogue-1MRK 504 016-UEN. <www.abb.com>.
[6] Technical manual of P631, P632, P633, P634 (Transformer differential
protection). AREVA relay catalogue-P63x/UK M/A54.
Conclusion [7] Hooshyar A, Afsharnia S, Sanaye-Pasand M, Ebrahimi BM. A new algorithm to
identify magnetizing inrush conditions based on instantaneous frequency of
differential power signal. Power Deliv, IEEE Trans 2010;25:2223–33.
Conventional inrush current blocking methods including the [8] Pihler J, Grcar B, Dolinar D. Improved operation of power transformer
second harmonic and gap detection criteria are prone to maloper- protection using artificial neural network. Power Deliv, IEEE Trans 1997;
12:1128–36.
ate for large inrush currents. To overcome such a problem, a novel
[9] Wiszniewski A, Kasztenny B. A multi-criteria differential transformer relay
approach is presented in this paper. This approach exploits large based on fuzzy logic. Power Deliv, IEEE Trans 1995;10:1786–92.
inrush currents asymmetries caused by the helping effect, which [10] Bejmert D, Rebizant W, Schiel L. Transformer differential protection with fuzzy
can be observed for three-limb YNd transformers. A set of experi- logic based inrush stabilization. Int J Electr Power Energy Syst December 2014;
63:51–63.
mental and comprehensive time-domain simulation cases demon- [11] Rasoulpoor M, Banejad M. A correlation based method for discrimination
strate that the proposed approach has the following features: between inrush and short circuit currents in differential protection of power
transformer using discrete wavelet transform: theory, simulation and
experimental validation. Int J Electr Power Energy Syst October 2013;
(1) This method can reliably discriminate large inrush currents 51:168–77.
from fault currents. [12] Vahidi B, Ghaffarzadeh N, Hosseinian SH. A wavelet-based method to
(2) CT saturation due to the DC component of an inrush current discriminate internal faults from inrush currents using correlation
coefficient. Int J Electr Power Energy Syst September 2010;32(7):788–93.
may cause the gap detection method to maloperate whereas [13] Mao PL, Aggarwal RK. A wavelet transform based decision making logic
the proposed approach has reliable performance for such a method for discrimination between internal faults and inrush currents in
case. power transformer. Int J Electr Power Energy Syst August 2000;22(6):389–95.
[14] Ge B, de Almeida AT, Zheng Q, Wang X. An equivalent instantaneous
(3) The correct performance of this method is not affected by
inductance-based technique for discrimination between inrush current and
the fault current asymmetry caused by the decaying DC internal faults in power transformers. Power Deliv, IEEE Trans 2005;
component and thus the approach has appropriate perfor- 20:2473–82.
[15] Technical manual of T60 (Transformer protection system). GE relay catalogue.
mance for internal faults.
[16] Mekic F, Girgis R, Gajic Z, teNyenhuis E. Power transformer characteristics and
(4) This method can well recognize an internal fault current their effect on protective relays. Presented at the 33rd Western protective
contaminated with an inrush current and therefore releases relay conference, October 17–19, 2006.
the inrush current blocking of the differential relay. How- [17] Manual of SIPROTEC 7UT612 (Differential Protection). Siemens relay
catalogue, 4.0 ed; 2002.
ever, conventional approaches block the differential relay [18] Kulkarni SV, Khaparde SA. Transformer engineering: design and practice. CRC
for a long duration that may cause subsequent severe Press; 2004.
damages to the faulty transformer. [19] Heathcote MJ. The J & P transformer book, a practical technology of the power
transformer. 13th ed. Elsevier Ltd.; 2007.
(5) Supplementary of the conventional methods with the [20] Sonnemann WK, Wagner CL, Rockefeller GD. Magnetizing inrush phenomena
proposed approach can be used to effectively detect both in transformer banks. Power Apparatus Syst, Part III. Trans Am Instit Electr Eng
mild and large inrush currents. 1958;77:884–92.
[21] Davarpanah M, Sanaye-Pasand M, Iravani R. Performance enhancement of the
transformer restricted earth fault relay. Power Deliv, IEEE Trans 2013;
Accordingly, the proposed combined strategy has the potential 28:467–74.
for industrially implementations.

You might also like