Sie sind auf Seite 1von 5

Current Transformer Saturation Segmentation Using

Morphological Gradient-Based Detectors


L. L. Zhang, Member IEEE, T. Y. Ji, Member IEEE, M. S. Li, Member IEEE, Q. H. Wu, Fellow IEEE
School of Electric Power Engineering, South China University of Technology, Guangzhou 510640, China.
Email: wuqh@scut.edu.cn

Abstract—This paper presents a novel algorithm for secondary One way to avoid malfunction of differential protection
current signal segmentation of saturated current transformer when CT saturated during external fault is to blocked the
(CT) using three morphological gradient-based detectors. The protection once CT saturation is detected. This scheme is
saturation onset detector is used to monitor the occurrence of simple, but it lower the sensitivity and response speed of the
saturation and then trigger the positive/negative segmentation protection [3]. Therefore, approaches that can accurately
detector, which are designed to locate the starting point and the compensate the saturated current and calculate the phasor of
ending point of CT saturation. The saturation segments are differential current are more preferable. Several methods [4-7]
determined by combined the output of the two segmentation for compensating the saturated secondary current of CT based
detectors. Simulation studies are conducted in PSCAD/EMTDC
on approximate CT model have been presented. In [4] and [5],
to evaluate the performance of the proposed method. The
simulation results demonstrated that the proposed method can
the secondary currents is compensated by magnetizing current
accurately determine the saturation segments. calculating form the magnetization curve. These methods and
the like can be conducted only if the parameters of CT are
Index Terms—current transformer, mathematical morphology, available. Parameter identification of current transformer are
differential protection, saturation detection. introduced in [6] and [7]. Artificial neural networks (ANNs) [8-
11] are also used to reconstruct the secondary current, however,
I. INTRODUCTION which need a large data set for training and its computational
load is heavy. Reference [12, 13] present methods to detect the
Iron core current transformers (CTs) are widely used for
saturation firstly and then estimate the parameters of the
both measurement and protection purposes in electrical power
unsaturated portions and then calculate the current phasor for
systems. The iron core may saturated during an external fault
the followed saturated portions. Many methods can detect the
due to the large fault current, and thus the secondary current
starting of the saturated portion, but the detection of the ending
will be distorted. This may lead to inaccurate current
of saturated portion is still a challenge because the change in
measurement and malfunction of the differential protection.
the current at the ending point is much smoother than that at the
Thus, in order to improve the reliability of the protection system,
starting point.
the adverse impact of the CT saturation must be relays must be
eliminated. In order to avoid the shortcomings of the existing methods,
a new method based on morphological gradient (MG) is
The probability of saturation of iron core can be greatly
proposed to detect CT saturation in this paper. The
reduced by reasonably selecting current transformer in practical
morphological gradient are performed using three different
application. However, the fault current in the power system
structuring elements (SEs). The structuring element with its
may be dozens of times greater than the current normal
origin at its center is used to detect the occurrence of saturation.
operation, and thus the core flux density is much larger than the
After that, the structuring element with its origin at its left hand
steady state. To ensure the transformer core will not saturate
is used to locate the starting point and ending point of the
under all the fault transient conditions, the size and cost of the
saturation portion of the positive current. The structuring
iron core will be very high [1]. Therefore, it is not practical way
element with its origin at its right hand is used to locate the
of avoiding iron core saturation by only design and deployment.
starting point and ending point of the saturation portion of the
In recent years, current transformers based on Rogowski negative current. The final saturation segmentation is based on
coil and optical fiber are designed and applied in some the combination of the outputs of the positive detector and
demonstration projects. However, the reliability of these negative detector. The proposed method can make accurate
electronic transformers still need to be examined and improved segmentation with five-sample delay and does not need
before they can be widely used. Before that, iron core current previous knowledge of CT.
transformer will be still the main type of current transformers.
II. MATHEMATICAL MORPHOLOGY
In this respect, many researcher are inspired to work on
detection of CT saturation and compensation of saturated Mathematical morphology is a powerful image/signal
secondary current. processing tool, which was originally developed as a tool in
geometrical analysis and description [14]. It was established by
introducing fundamental operators applied to two sets. One set, processing, which is defined as the arithmetic difference
which is the image/signal, is processed by another, known as between the dilation and erosion of a signal by an SE. The MG
the structuring element (SE), which is translated over the is usually defined as:
image/signal as a probe. Morphological operators are set
transformations which are very efficient for ameliorating the G( f ) f † g  f g (7)
quality of signals or for extracting some of their features. With In this paper, a structuring element with its origin at its
appropriate use, they can process signals and preserve their center is employed to perform the MG, as it is capable of
intrinsic characteristics[15]. effectively suppressing various kinds of noises and does not
blur the sudden changes in the signals. The MG-based
There are two basic morphological operators, dilation and saturation onset detector has large impulse in the output when
erosion, which form a pair of dual transforms and are defined the secondary current signal of current transformer change
as follows: suddenly, regardless of the polarity of the change. Therefore, it
S †G (S  g ) (1) can be used to detect the onset of saturation and thus it can be
gG
used as a trigger of the following saturation segmentation
detectors.
S G (S  g ) (2)
gG
B. Saturation Segmentation Detectors
In order to use MM in signal processing where most signals At the beginning of saturation, the change in the magnitude
are not binary, morphological operators should be extended to of the current is large while the change is much smaller at the
a grey-scale level. Instead of performing dilation and erosion end of saturation, which makes methods based on threshold
by union and intersection as in the binary case, they are comparison hard to find the ending points of saturation. The
performed by algebraic addition and subtraction in the output of the saturation onset detector only gives the
greyscale case. In gray scale signal processing, these two information of the onset of saturation, therefore, two MG-based
morphological operators defined as follows: segmentation detectors are designed to determine the ending
points of the saturation.
( f † g )( x) max{ f ( x  s)  g (s)} (3)
s For a fault current with positive DC offset, the CT
secondary current is descending due to saturation. In contrast,
(f g )( x) min{ f ( x  s)  g (s)} (4)
s for a fault current with positive DC offset, the CT secondary
current is descending during saturation. In this respect, we
where f and g are the input signal and the SE, respectively; design two detectors to examine the negative change and the
( x  s)  D f , ( x  s)  D f and s  Dg ; D f and Dg are the positive change in CT secondary current, respectively. With
domains of f ( x) and g ( s) . this purpose, two morphological gradient-based detectors with
two different SEs are defined:
As explained previously, the SE is a small set used to probe
G( f g  ) f † g  f g (8)
the signal under study. Depending on the problem to be solved,
the structuring element can have different shapes and sizes. A
simple case is that the SE has the form of g(s) = 0, s Ѯ Dg, G( f g  ) f † g  f g (9)
which is referred to as a ‘flat SE’. Hence, definitions of dilation
and erosion degrade to [16]: where

( f † g )( x) max{ f ( x  s)} (5) g {g1 , g2 , }, gl 1 , gl } (10)


s

(f g )( x) min{ f ( x  s)} (6) g {g1 , g2 , }, gl 1 , gl } (11)


s

where the dilation operator calculates the maximum value where the underlined samples represent the origin. Apparently,
within the neighborhood of f ( x) , and the erosion operator the two SEs have the same shape and the same size, and the
calculates the minimum value within the neighborhood of only difference lies in the position of the origin. The detector
f ( x) . Intuitively, dilation can be imagined as swelling or with structuring element g  , namely positive detector, is for the
detection of the positive change and the detector with
expanding, and erosion can be thought of as a shrinking
procedure. structuring element g  , namely negative detector, is for the
detection of the negative change.
All of the other morphological transformations, such as
opening, closing, and morphological filters, are based on these For the saturation of single polarity, the output of
two basic operators. positive/negative detector will have two maximum outputs in
every cycle, which correspond to the saturation starting point
III. SATURATION DETECTORS and ending point respectively. For the bipolar saturation, both
the positive and negative detector has two maximum outputs in
A. Saturation Onset Dectector
every cycle. The saturation segments can be determined by the
The morphological gradient is frequency used for edge positions of these maximums.
detection in image and sudden change detection in signal
IV. PERFORMANCE EVALUATION 0% remanence within the CT core. The distortion of the
A. Simulation Model secondary current starts at the before the faulted current
reaching its peak. The output of the positive and negative
A typical model of a 50 Hz power system with a 230 kV and
detector are similar to the results shown in Fig. 2. As the
100 km transmission line, as shown in Fig. 1 is established in
PSCAD/EMTDC for the simulation studies. saturation become more severe, the saturation segment is larger.
Fig. 3(c) clearly shows that the proposed accurately detects the
.
Vm f . distorted sections of the original secondary current waveform.
M N Vn
CT 1 CT 2
50 iref i2 f g+

Current (A)
‫ = ܯܧ‬23 0ğ0°, ‫ = ܰܧ‬23 0ğ20°, ܼ‫ = ܼܰ = ܯ‬9. 186 + j 40.192¡ 0
ܴ1 = 0. 0346¡/ km, ‫ܮ‬1 = 1. 3482 mH/ km, ‫ܥ‬1 = 0. 0086­F / km
ܴ0 = 0.3 000¡/ km, ‫ܮ‬0 = 3. 6340 mH/ km, ‫ܥ‬0 = 0. 0062­F / km
-50
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Fig. 1 Model of the sample power system. Time (ms)
(a)
The Lucas model is employed to simulate the saturation 50 iref i2 f g-

Current (A)
characteristics of CT, the setting is shown in Table I [2]. For all
cases presented here, the CT saturation is simulated under the 0
condition that a phase-to-ground fault occurs at the location 30
km away from Bus 1. The sampling rate of the system is 4000 -50
Hz. 0.18 0.2 0.22 0.24
Time (ms)
0.26 0.28 0.3
(b)
1.5
TABLE I. PARAMETERS OF CT 1

Saturation status
1
Turn ratio 300/1 0.5
Fundamental Frequency 50 Hz 0

Mean core length 0.448 m -0.5


0.18 0.2 0.22 0.24 0.26 0.28 0.3
Cross section area 0.00168 m2 Time (ms)
(c)
Winding resistance 2.1 Ω Fig. 2 Saturation detection of the case with -80% residual flux in iron core;
(a) the referred, distorted secondary current and output of positive detector
Air core inductance 0.85 mH output; (b) the referred, distorted secondary current and output of negative
Saturation fux density 1.88 T detector output; (c) the combined saturation status.

Nominal burden 8 VA
50 iref i2 f g+
Current (A)

B. The effect of residual flux


0
1) -80% residual flux: Fig. 2 shows the detecting result in
the case of a remanence of -80% flux in the CT core. The -50
0.18 0.2 0.22 0.24 0.26 0.28 0.3
polarity of the remanence is opposite to that of the DC offset, Time (ms)
(a)
which results in a slight distortion beginning at the first cycle of
50
the fault secondary current, as shown in Fig. 2. In Fig. 2(a) and iref i2 f g-
Current(A)

Fig. 2(b), the dotted line represent the primary current referred 0
to secondary side, the dashed line represent the secondary
current, the solid lines represent the detecting result of positive -50
0.18 0.2 0.22 0.24 0.26 0.28 0.3
CT saturation detector and negative CT saturation detector, Time (ms)
(b)
respectively. As the saturation occurs at the positive part of the 1.5
Saturation status

secondary current, the positive detector is launched once the 1


saturation is detected. In contrast, the negative detector is 0.5
blocked and it output is always zero, as shown in Fig. 2(b). 0
Therefore, the saturation segmentation is determined by the -0.5
output of the positive detector. In every cycle, the first 0.18 0.2 0.22 0.24
Time (ms)
0.26 0.28 0.3

maximum corresponds to the starting point and the second (c)


Fig. 3. Saturation detection of the case with 0% residual flux in iron core;
maximum corresponds to the ending point. The overall (a) the referred, distorted secondary current and output of positive detector
segmentation is shown in Fig. 2(c). For the saturation status, ‘0’ output; (b) the referred, distorted secondary current and output of negative
and ‘1’ indicate that the CT is unsaturated and saturated, detector output; (c) the combined saturation status.
respectively. 3) 80% residual flux: Fig. 4 illustrates the distortion
2) 0% residual flux: The extent of the distortion gradually associated with 80% residual flux which is the maximum flux
becomes more severe as the remanence increases from a possible from a practical CT. Under this situation, the transit
negative value to a positive value. The case shown in Fig. 3 has
flux has a very short effective excursion before saturation detectors. The results shown in Fig.6(c) indicate that the
occurs. It is apparent from the figure that the distortion of the secondary current waveform includes two saturated segments
secondary current is the severest in comparison with other and two unsaturated segments per cycle.
situations. It begins only a few milliseconds after fault occurs.
50
However, this change just have a small impact on the iref i2 f g+

Current (A)
performance of the morphological gradient-based detectors.
0
The results shown in Fig. 4 indicate that the proposed method
can accurately determine the interval of saturation even when
-50
the secondary current is severely distorted due to a high residual 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Time (ms)
flux. (a)
50
iref i2 f g-

Current (A)
50 iref i2 f g+ 0
Current (A)

0
-50
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Time (ms)
-50 (b)
0.18 0.2 0.22 0.24 0.26 0.28 0.3 1.5
Time (ms)

Saturation status
(a) 1

50 iref i2 f g- 0.5
Current (A)

0
0
-0.5
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Time (ms)
-50 (c)
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Time (ms) Fig. 5. Saturation detection of the case with negative DC-offset; (a) the
(b)
1.5 referred, distorted secondary current and output of positive detector output; (b)
the referred, distorted secondary current and output of negative detector output;
Saturation status

1
(c) the combined saturation status.
0.5

0
60
-0.5 iref i2 f g+
0.18 0.2 0.22 0.24 0.26 0.28 0.3 40
Current (A)

Time (ms)
(c) 20
0
Fig. 4. Saturation detection of the case with 80% residual flux in iron core; -20
(a) the referred, distorted secondary current and output of positive detector
0.18 0.2 0.22 0.24 0.26 0.28 0.3
output; (b) the referred (dashed line), distorted secondary current and output of Time (ms)
negative detector output; (c) the combined saturation status. (a)
60
C. The Performance on Negative DC-Offest iref i2 f g-
Current (A)

40

The polarity of DC offset is related to the fault inception 20

angle. Different from the case in Fig. 3, the fault inception angle 0

is delay by 180q , which leads to the negative DC offset shown -20

in Fig. 5. In this case, the positive detector is blocked and its 0.18 0.2 0.22 0.24
Time (ms)
0.26 0.28 0.3

output is always zero. In contrast, the negative detector is 1.5


(b)

launched once the saturation occurs and its output can be used
Saturation status

1
to determine the saturation segments. The results shown in Fig.
5(c) indicate that the proposed method accurately detect the 0.5

starting point and ending point of saturation of fault current 0


with negative DC offset. -0.5
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Time (ms)
D. The Performance on AC Saturation (c)

AC saturation of a CT is caused by a high fault current Fig. 6. AC Saturation detection of the case with 0% residual flux in iron
magnitude or a high burden impedance [2]. To investigate the core; (a) the referred, distorted secondary current and output of positive detector
output; (b) the referred, distorted secondary current and output of negative
performance of the proposed method on AC saturation, a short- detector output; (c) the combined saturation status.
circuit event without a decaying dc component is simulated, as
shown in Fig. 6. It can be seen from that the secondary current E. Comparison Studies
of the CT is distorted both in the positive half cycles and To further evaluate the performance of the proposed method,
negative half cycles. Therefore, both the positive detector and comparison studies are carried out on the difference-based
the negative detector are launched and the saturation segments detection method and the SMMG-based method [17]. The
are determined by the combination of the outputs of these two SMMG method also uses MM to analyze the waveform of a
signal, and it has been proved to have relatively good ACKNOWLEDGMENT
performance when compared with other detection methods [17]. This project was supported by Guangdong Innovative
The solid lines in Fig. 7(a) and Fig. 7(b) represent the output Research Team Program (No. 201001N0104744201) and the
signals obtained by the SMMG method performed on the case Fundamental Research Funds for the Central Universities.
as depicted in Fig. 3 at levels 1 and 2, respectively. The output
signals of the SMMG method, of both 2-order and 4-order REFERENCES
SMMG, have large impulses where saturations occur in every [1] Gangadharan P K, Sidhu T S, Finlayson G J. Current transformer
cycle while its output is almost zero at the saturation ending dimensioning for numerical protection relays [J]. Power Delivery, IEEE
Transactions on, 2007, 22(1): 108-115
point, as shown in Fig. 7(b) and Fig. 7(c). Therefore, the [2] Ajaei F B, Sanaye-Pasand M, Davarpanah M, et al. Compensation of
SMMG method cannot accurately determine the saturation the current-transformer saturation effects for digital relays [J]. Power
segments. The comparison study demonstrates that the Delivery, IEEE Transactions on, 2011, 26(4): 2531-2540.
performance of the proposed method is superior to the SMMG [3] Pan J, Vu K, Hu Y. An efficient compensation algorithm for current
method. transformer saturation effects [J]. Power Delivery, IEEE Transactions
on, 2004, 19(4): 1623-1628.
[4] Kang Y C, Park J K, Kang S H, et al. An algorithm for compensating
50 iref i2
secondary currents of current transformers [J]. Power Delivery, IEEE
Current (A)

Transactions on, 1997, 12(1): 116-124.


[5] Kang Y C, Lim U J, Kang S H, et al. Compensation of the distortion in
0
the secondary current caused by saturation and remanence in a CT [J].
Power Delivery, IEEE Transactions on, 2004, 19(4): 1642-1649.
-50 [6] Locci N, Muscas C. A digital compensation method for improving
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Time (ms) current transformer accuracy [J]. Power Delivery, IEEE Transactions
(a) on, 2000, 15(4): 1104-1109.
50 [7] Bittanti S, Cuzzola F A, Lorito F, et al. Compensation of nonlinearities
in a current transformer for the reconstruction of the primary current [J].
(A)

Control Systems Technology, IEEE Transactions on, 2001, 9(4): 565-


3
2

0
573.
SMMG

[8] Haghjoo F, Pak M. Compensation of CT Distorted Secondary Current


-50 Waveform in Online Conditions [J]. Power Delivery, IEEE
0.18 0.2 0.22 0.24 0.26 0.28 0.3
Time (ms) Transactions on, 2015, PP (99): 1-1 (IEEE Early Access Articles).
(b) [9] Yu D C, Cummins J C, Wang Z, et al. Correction of current transformer
50
distorted secondary currents due to saturation using artificial neural
networks [J]. Power Delivery, IEEE Transactions on, 2001, 16(2): 189-
(A)

194.
4
2

0 [10] Rebizant W, Bejmert D. Current transformer saturation detection with


SMMG

genetically optimized neural networks [C]. Power Tech, 2005 IEEE


-50 Russia. St. Petersburg: IEEE, 2005: 1-6.
0.18 0.2 0.22 0.24 0.26 0.28 0.3 [11] Hong Y Y, Chang-Chian P C. Detection and correction of distorted
Time (ms)
(c) current transformer current using wavelet transform and artificial
Fig. 7. Saturation detection based on SMMG. (a) the referred and distorted intelligence [J]. Generation, Transmission & Distribution, IET, 2008,
secondary current; (b) output of the third-order SMMG (c) the output of the 2(4): 566-575.
fourth-order SMMG. [12] Lu Z, Smith J S, Wu Q H. Morphological lifting scheme for current
transformer saturation detection and compensation [J]. Circuits and
V. CONCLUSION Systems I, IEEE Lin X N, Zou L, Tian Q, et al. A series multiresolution
morphological gradient-based criterion to identify CT saturation [J].
This paper proposes a novel algorithm for CT saturated Power Delivery, IEEE Transactions on, 2006, 21(3): 1169-1175.
secondary current signal segmentation, which employs three Transactions on, 2008, 55(10): 3349-3357.
MG-based detectors to locate the starting point and the ending [13] Wiszniewski A, Rebizant W, Schiel L. Correction of current
point of CT saturation and then determine the saturation transformer transient performance [J]. Power Delivery, IEEE
Transactions on, 2008, 23(2): 624-632.
segments. The effect of some parameters of CT and fault [14] J. Serra, Image Analysis and Mathematical Morphology. New York,
conditions including residual flux, polarity of DC offset, and USA: Academic, 1983.
CT burden has been studied. The simulation results have [15] M. H. Sedaaghi and Q. H. Wu, “Real-time implementation of greyscale
demonstrated that the proposed method is able to accurately morphological operators,” Electron. Letters, vol. 33, no. 21, pp. 1761–
detect CT saturation intervals. The results of comparison 1763, 1997.
[16] Q. H. Wu, Z. Lu, and T. Y. Ji, “Mathematical Morphology,” in
studies demonstrate that the proposed method outperforms the Protective Relaying of Power Systems Using Mathematical
SMMG-based method. Morphology. London, U.K.: Springer, 2009, pp. 13–40.
[17] Lin X N, Zou L, Tian Q, et al. A series multiresolution morphological
gradient-based criterion to identify CT saturation [J]. Power Delivery,
IEEE Transactions on, 2006, 21(3): 1169-1175.

Das könnte Ihnen auch gefallen