Академический Документы
Профессиональный Документы
Культура Документы
I. INTRODUCTION
Manuscript received May 27, 2009; revised October 01, 2009, November 24,
2009. First published April 19, 2010; current version published June 23, 2010.
This work was supported by the National Science Council under Grant NSC
96-2628-E-027-115-MY2. Paper no. TPWRD-00401-2009.
The author is with the Department of Digital Technology Design, National
Taipei University of Education, Taipei, Taiwan, R.O.C. (e-mail: chsyu@tea.
ntue.edu.tw).
Digital Object Identifier 10.1109/TPWRD.2010.2045137
samples were used to detect CT saturation. In [9], a gradientbased criterion was used to detect the non-smooth variations.
In [10], a simpler algorithm that used an identified reference
point was proposed to discriminate saturated periods. After the
saturated periods were identified, other algorithms were used to
correct the saturated current samples. Regression or least-square
analyses on the unsaturated samples [6], [10] were widely used
for corrections. An artificial intelligence method [7] was also
used to correct the saturated samples.
CT saturation can be accurately detected using the algorithms
proposed in [3][10]. However, some problems may still arise.
For example, the performances of CT model-based algorithms
[3][5] may be degraded when the CT parameters are changed.
Moreover, mild CT saturation cannot easily be detected using
waveform-based algorithms [6][9]. Because the detection algorithms of [6], [7] were developed based on edge-trigger concepts, the saturation period has the chance to be discriminated
as an unsaturated period.
Generally, CT saturation is caused by the decaying dc component in a fault current. When a CT is saturated, the parameters of the decaying dc component will be changed. Hence,
this paper attempts to use the information contained in such a
component to detect CT saturation. To achieve this goal, the decaying dc component is estimated using a phasor-based computation, and the decaying factor in the decaying dc component is
used to define a detection index. When a CT is unsaturated, this
index will be within a small pre-known range. When a CT is
saturated, this index will oscillate within a wide range. Thus, a
level-trigger concept can be developed to detect CT saturation.
When a saturated current is detected, the current samples and
phasors in the latest unsaturated period are used to correct the
saturated current. Although the proposed detection index is designed based on the variation of a decaying factor, ac saturation
can still be detected using the proposed index. The proposed algorithm was tested using the MATLAB/SIMULINK [11] simulator to demonstrate its effectiveness. A DSP Starter Kit (DSK)
for TMS32C6416 [12] was also used to evaluate the real-time
applicability of the proposed algorithm.
II. THE PROPOSED DETECTION ALGORITHM
A. The Saturation Detection Index
containing a
To define a detection index, a fault current
fundamental component and a decaying dc component is described as follows:
(1)
1341
where
and
denote the magnitude and phase
and deangle of the fundamental component, respectively.
note the magnitude and time constant of the decaying dc component, respectively. If the fault current is measured by samples
per cycle, the th sample of the fault current can be represented
as follows:
(2)
, the sampling angle
where the sampling time
and the decaying factor
.
In an unsaturated current measurement, the decaying factor
varies within a small range, although the time constant for a decaying dc component varies within a wide range. For example,
and the posif the sampling frequency is 3840 Hz
sible time constant of a decaying dc component varies within
0.1 to 5 cycles, the corresponding decaying factor only varies
within a small range [0.8553, 0.9969]. However, when the current is saturated, the measured will be outside of that small
range. Thus, the value of a decaying factor can be used to
identify whether a CT is saturated or not.
In this paper, the decaying factor was estimated using the
phasor-based computations. To achieve this estimation as fast as
possible, the fractional cycle discrete Fourier transform (DFT)
[13] was used to compute a current phasor. The obtained current
has the following form:
phasor
(3)
denotes the accurate fundamental phasor and
where
denotes the phasor domain decaying dc component. Meanwhile,
denotes that the phasor is obtained just after
the subscript
th sample is
the th sample is measured. When the
measured, the phasor
can be represented as follows:
(4)
To reduce the computational burden, recursive DFT [13] was
. Comparing the phasors
used to update the new phasor
and
obtained by the recursive DFT, the relations
, and
are as follows:
between
(5)
. When the
th sample is meawhere
can also be obtained by the recursive
sured, the phasor
DFT computations. Using the relation (5), the division of two
and
can be
phasor differences
expressed as follows:
(6)
IF
ELSE
END
In the rule mentioned above, the ratio
is used to evaluate the magnitude of a decaying dc component. The level value
is set when the de10 is selected to ensure the constant
caying dc component is very small.
To tolerate the measurement noise, the threshold boundary of
the detection index is defined as [0.85,1.05]. Thus, the proposed
CT saturation detection algorithm is described as follows.
1342
(12)
where
and
are the mean and standard deviation
values of all of the detection indices in the latest un-satuwas selected
rated period, respectively. The gain
. To prevent this detection from being
for gaining the
too sensitive, the minimum value of the standard deviation
.
is limited to
is used for
3) Next, the new threshold boundary
CT saturation detections.
III. THE PROPOSED CORRECTION ALGORITHM
Using the proposed detection algorithm, the saturated and
the unsaturated current waveform periods can thus be discriminated. When a CT is unsaturated or before the end of the first
unsaturated period is detected, the measured current samples are
used directly. When the end of an unsaturated period is detected,
the proposed algorithm begins to correct the saturated samples.
When a CT is saturated, the current measurements and the
current phasors in the latest unsaturated period are used to obtain
the current waveform parameters. Assume that the CT saturation
was detected just after the th current sample was measured, and
that there are unsaturated current phasors obtained in the latest
unsaturated period
. Using the
unsaturated phasors, three summation variables
, and
are defined as follows:
(13)
(14)
(15)
Using the three summation variables and the relation of (7), the
averaged decaying factor in the latest unsaturated period can
be obtained as follows:
(16)
1343
Then, applying the concepts of (3)(5), the accurate fundamental phasor can be obtained as follows:
Fig. 3. CT magnetizing curve with the saturation point (9.52 A, 0.48 Vs).
TABLE I
PARAMETERS OF THE SIMULATION SYSTEM
(17)
Considering the time domain waveform,
can be obtained
using the averaged computations as follows:
(18)
Finally, the corrected current sample
follows:
can be obtained as
(19)
The proposed CT saturation detection/correction algorithm is
described using a flowchart shown in Fig. 1. At first, the proposed algorithm is in the no fault stage where the fault detection algorithm proceeds. After a fault is detected, the proposed
algorithm comes into the unsaturated stage and the CT saturation detection algorithm is activated. After the end of an unsaturated period is detected, the proposed algorithm comes into
the saturated stage. In the saturated stage, the current measurements and current phasors in the latest unsaturated period are
used to correct the saturated current measurements. The corrections continue until the beginning of a new unsaturated period
is detected. Then, the proposed algorithm goes back to the unsaturated stage and the CT saturation detection will continue.
The loop between the unsaturated and the saturated stages can
be stopped by a relay when a relay has received enough fault
measurements.
IV. SIMULATION EVALUATIONS
In this section, the proposed algorithm was evaluated by a
simple power system shown in Fig. 2. The parameters of this
(21)
1344
where
denotes the rms current on the CT primary
denotes the rms current on the CT secondary side,
side,
denotes the turns ratio of a CT.
and
A. Demonstrations of the Proposed Algorithm Using Case
Studies
In the following tests, the proposed algorithm was evaluated
using some specific cases, such as severe saturation, mild saturation, high remanence, and ac saturation cases.
Case 1: Severe Saturation Case: In this case, a three phase
shorted fault was used to demonstrate the performances of the
proposed algorithm under a severe saturation condition. The
fault occurred at 0.05 s, at a 0.3 per-unit distance away from
BUS S, and with a 0.1 fault resistance. Fig. 4(a) depicts the
current waveforms, in which the corrected, saturated, and ideal
currents are compared. All currents are referred to the CT primary side. The ideal current denotes the current which ignores
the CT measurement effects. The proposed detection index is
presented in Fig. 4(b). The fault and CT saturation detection
results are shown in Fig. 4(c). Using the EI index to evaluate
accuracy, the EI index is significantly reduced from 48.8% (saturated current) to 1.71% (corrected current).
Fig. 4(b) shows that the detection index performs a transient
high value just after occurring a fault (0.05 s) to cause a momen. This is caused by the
tary saturation detection result of
mixed-data windowing effect of a DFT computation [13]. Since
the proposed correction algorithm begins to correct the saturated
current after the end of the first unsaturated period has been detected, this windowing effect will not affect the performance of
the proposed algorithm.
Fig. 4 also shows that the windowing effect delays the
proposed detection index entering the threshold boundary
at the end of each saturated period. However, the
proposed detection index can detect the beginning of a saturated
period very fast. Indeed, the main purpose of a saturation detection algorithm is to identify the unsaturated current samples
for waveform corrections. Thus, the accuracy of identifying an
unsaturated current sample is more important than the accuracy
of identifying a saturated current sample. Although a time delay
is inevitable for the proposed phasor-based algorithm to detect
the end of a saturated period, the proposed detection algorithm
can ensure the accuracy of the identified unsaturated current
samples. Thus, the identified unsaturated current samples can
be used to accurately correct the saturated current.
Case 2: Mild Saturation Case: In this case, an A-phase
grounded fault was used to demonstrate the performances of
the proposed algorithm under a mild saturation condition. The
fault occurred at 0.05 s, at a 0.7 per-unit distance away from
fault resistance. Fig. 5(a) shows the
BUS S, and with a 2
corrected, saturated and ideal A-phase current waveforms.
Notably, the CT is not saturated in the first fault cycle, and
the saturation is very mild in the second fault cycle. The corrected, saturated, and ideal current waveforms are shown in
Fig. 5(a), respectively. The obtained detection index is shown
in Fig. 5(b), while the saturation detection result is shown in
Fig. 5(c). Although the saturation is very mild, the proposed
algorithm can still be used to correct the saturated current. It
can be noted that the corrected and the ideal current waveforms
Fig. 4. Simulation results for a severe saturation case. (a) A-phase current
waveforms. (b) Proposed detection index. (c) Saturation and fault detection
results.
Fig. 5. Simulation results for a mild saturation case. (a) A-phase current waveforms. (b) Proposed detection index. (c) Saturation and fault detection results.
1345
Fig. 7. Simulation results of case 4. (a) A-phase current waveforms. (b) Proposed detection index.
Fig. 8. Simulation results of case 5. (a) A-phase current waveforms. (b) Proposed detection index. (c) Saturation and fault detection results.
1346
TABLE II
PERCENTILE ANALYSES OF THE EI INDEX FOR STATISTICAL EVALUATIONS
13.9% (saturated current) to 0.12% (corrected current). Obviously, although the proposed detection index is designed based
on the variations of a decaying factor, ac saturation can still be
detected using the proposed index.
B. Statistical Evaluations
To analyze and demonstrate the performance of the proposed
algorithm under normal fault conditions in detail, different fault
types, inception angles (0-degree to 270-degree), fault locations
(0.1 to 0.5 p.u. location from Bus S), short-circuit capacities
(1500 to 3000 MVA), fault resistances (0.1 to 4 ), phase angles
and
(10-degree to 20-degree), CT
between the voltages
burden impedance (0.7 to 1 ), and CT burden power factor (0.5
to unity) were considered to generate over 5800 cases. In all
the simulation cases, the CT saturation occurred in about 3900
cases.
The EI index was used to analyze the A-phase current waveforms of all cases, and the percentile method [10], [16] was used
to analyze the percentage distribution of all the computed EI results. Table II shows the analysis results of the saturated and
corrected currents. The maximum EI of the saturated currents is
60.6% which occurred in a severe three-phase shorted fault case.
Using the proposed algorithm, the maximum EI of the corrected
currents is therefore reduced to 7.77%, which also occurred in
a severe three-phase shorted fault case. The averaged EI for all
cases is 0.72%. In considering the corrected current results in
detail,90% of the EI are smaller than 2.05%. Thus, the proposed
algorithm can support accurate correction currents under various fault conditions and CT conditions.
C. Evaluation With Considering Low Order Harmonic
Current Sources
Here, harmonic current sources (4% at the second and 4%
at the third harmonics) were injected from the position of the
to evaluate the performance of the proposed algovoltage
rithm under an active low-order harmonic environment with cur% which exceeds the current harmonic limirent
tation recommended in [17]. For ease of comparison, the same
three-phase shorted fault used in case 1 is also tested in this case.
The simulation results are shown in Fig. 9. As can be seen, in
the pre-fault and the unsaturated periods, the computed detection index contains an oscillation which is caused by the injected
harmonics. Notably, the oscillation is larger in a pre-fault period
as the fault current with a larger fundamental component can
Fig. 9. Simulation results with considering harmonic sources. (a) A-phase current waveforms. (b) Proposed detection index. (c) Saturation detection results.
TABLE III
EI INDEX ANALYSES FOR THE HARMONIC CURRENTS
WITH DIFFERENT MAGNITUDES
C language was used to develop the DSP program which includes the fractional cycle recursive DFT subroutine, the fault
detection subroutine, the saturation detection subroutine, and
the saturation correction subroutine. The transformation of a
phasor from a rectangle form to a polar form and the division
computation of phasors were accomplished using the CORDIC
algorithm [18]. To accelerate the program speed, the DSP program was programmed using the Q16 fixed-point arithmetic.
. Code
Thus, the precision for a number can be achieved as
Composer Studio (CCS) v3.1 [19] was used to develop the program, by which we can debug the program and analyze the program performances.
The performance of the DSP program was evaluated using the
profile analysis tool in CCS to indicate the memory usage and
the execution time. The total memory required for all DSP programs was less than 11 kbytes. The execution time per sample
under the unsaturated and saturated periods is different. The
maximum execution time per sample is almost 4200 DSP cycles which occurred at a saturated period. Since the DSP on the
DSK board operated at 1 GHz, the maximum execution time per
sample was less than 4.2 s, which is significantly smaller than
s).
the fault measurements sample period (
Thus, the proposed algorithm has the potential to be applied for
practical applications.
V. CONCLUSION
In this paper, a new algorithm was proposed to detect and correct the saturated CT measurements using the information contained in a decaying dc component. A great number of computer
simulations have been performed on the MATLAB/SIMULINK
simulator to demonstrate the performance of the proposed algorithm under various fault and system conditions. The results of
the performance studies indicate that the proposed algorithm has
the following features.
1) The threshold values of the proposed detection index are
pre-known. Thus, a level detection can be designed to detect CT saturation.
2) Even in a mild saturation case, smooth waveform distortion
can still be detected by the proposed algorithm.
3) In the case of large remanence, the proposed algorithm can
work properly.
4) Although the proposed detection algorithm is designed
based on the variations of a decaying factor, ac saturation
can still be detected using the proposed detection algorithm.
5) The proposed algorithm is effective under various fault and
system conditions.
6) The proposed algorithm has been realized on a DSP to
verify the real-time applicability.
1347
REFERENCES
[1] IEEE Guide for the Application of Current Transformers Used for Protective Relaying Purposes, C37.110-2007, 2008, IEEE, New York.
[2] S. H. Horowitz and A. G. Phadke, Power System Relaying, 3rd ed.
New York: Wiley, 2008.
[3] D. C. Yu, J. C. Cummins, Z. Wang, H.-J. Yoon, and L. A. Kojovic,
Correction of current transformer distorted secondary currents due to
saturation using artificial neural networks, IEEE Trans. Power Del.,
vol. 16, no. 2, pp. 18919194, Apr. 2001.
[4] H. Khorashadi-Zadeh and M. Sanaye-Pasand, Correction of saturated
current transformers secondary current using ANNs, IEEE Trans.
Power Del., vol. 21, no. 1, pp. 7379, Jan. 2006.
[5] Y. C. Kang, U. J. Lim, and S. H. Kang, Compensating algorithm suitable for use with measurement-type current transformers for protection, IEE Proc.-Gener. Transm. Distrib., vol. 152, no. 6, pp. 880890,
2006.
[6] F. Li, Y. Li, and R. K. Aggarwal, Combined wavelet transform and
regression technique for secondary current compensation of current
transformer, IEE Proc.-Gener. Transm. Distrib., vol. 149, no. 4, pp.
497503, 2002.
[7] Y. Y. Hong and P. C. Chang-Chian, Detection and correction of distorted current transformer current using wavelet transform and artificial
intelligence, IET Gener., Transm., Distrib., vol. 2, no. 4, pp. 566575,
2008.
[8] Y. C. Kang, S. H. Ok, S. H. Kang, and P. A. Crossley, Design and evaluation of an algorithm for detecting current transformer saturation,
IEE Proc.-Gener., Transm., Distrib., vol. 151, no. 1, pp. 2735, 2004.
[9] X. Lin, L. Zou, Q. Tian, H. Weng, and P. Liu, A series multiresolution morphological gradient-based criterion to identify CT saturation,
IEEE Trans. Power Del., vol. 21, no. 3, pp. 11691175, Jul. 2006.
[10] J. Pan, K. Vu, and Y. Hu, An efficient compensation algorithm for
current transformer saturation effects, IEEE Trans. Power Del., vol.
19, no. 4, pp. 16231628, Oct. 2004.
[11] Power System Blockset Users Guide. Natick, MA: The Math Works,
Inc., 2002.
[12] R. Chassaing and D. Reay, Digital Signal Processing and Applications
With the TMS320C6713 and TMS320C6416 DSK, 2nd ed. New York:
Wiley, 2008.
[13] A. G. Phadke and J. S. Thorp, Synchronized Phasor Measurements and
Their Applications. New York: Springer, 2008.
[14] D. Tsiouvaras et al., Mathematical models for current, voltage and
coupling capacitor voltage transformers, IEEE Trans. Power Del., vol.
15, no. 1, pp. 6272, Jan. 2000.
[15] H. K. Hoidalen, ATPDRAW 3.5-Users Manual. Budapest, Hungary,
2002.
[16] W. J. DeCoursey, Statistics and Probability for Engineering Applications. New York: Newnes, 2003.
[17] IEEE Recommended Practices and Requirements for Harmonic Control in Electrical Power Systems, 519-1992, 1993, IEEE, New York.
[18] U. Meyer-Baese, Digital Signal Processing With Field Programmable
Gate Arrays, 2nd ed. New York: Springer, 2007.
[19] Code Composer Studio Users Guide, SPRU328B 2000. Dallas, TX,
Texas Instruments Incorporated.
Chi-Shan Yu (M02) received the B.S. and M.S.
degrees in electrical engineering from National
Tsing Hua University, Taipei, Taiwan, R.O.C., in
1988 and 1990, respectively, and the Ph.D. degree
in electrical engineering from National Taiwan
University in 2001.
From 2001 to 2006, he was with the National Defense University. Currently, he is an Associate Professor of Electrical Engineering with the Department
of Electrical Engineering, National Taipei University
of Technology, Taipei. His research areas are computer relay and digital signal processing.