2 ICPS2017Paper109
2 ICPS2017Paper109
net/publication/325939189
Fault detection & classification using innovative MFCDFT based slope tracking
method for digital distance protection
CITATIONS READS
4 207
3 authors:
Praghnesh Bhatt
Pandit Deen Dayal Petroleum University
65 PUBLICATIONS 1,059 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Emulation of Hidden Inertia of Variable Speed Wind Turbine Generators for Frequency Regulation View project
All content following this page was uploaded by Ujjaval Patel on 03 January 2019.
Abstract - The critical requirements for numerical distance Lines (ETPTL). However, the error in fault location
protection used for transmission line protection are reliable and estimation up to achieved up to 5%. Jamil et al. [12] has
fast phasor estimation of input signals to determine correct presented Generalized Neural Network (GNN) and wavelet
impedance reach. Factors influencing accurate phasor estimation based approach for fault location estimation of a transmission
are Fault Resistance (RF), Fault Type (FT), Fault Inception Angle
line during wide variation of power system disturbances. The
(FIA), Fault Location (FL) and Power Flow Angle (PFA). In this
paper, an innovative slope tracking method is presented which is performance is compared with ANN based scheme but, the
based on Modified Full Cycle Discrete Fourier Transform same is not validated for high resistance fault conditions.
(MFCDFT) and its application has been validated for distance A.Yadav et al. [13], [14] presented a transmission line relaying
relaying. Fault signals are simulated in PSCAD for all possible scheme for fault detection and classification using wavelet
types of faults and validation is performed in MATLAB®. For transform and linear discriminant analysis. However, the
transmission line protection, impedance trajectory is adaptively effect of decaying DC component and about to reach operation
calculated from Ground Distance Function (GDF) using of the relay is not considered. Compensation of GDF and
estimated values of voltages and currents. Moreover, the resistive reach assessment in quadrilateral characteristics is
proposed technique calculates operating time, fault location and
proposed by Sorrentino [15].
fault type, which are the key attributes of multi functional
numerical relay. The obtained results are compared with existing In the proposed work, phasor estimation of analog input
methods and found precise and faster. Error in fault location signals have been performed by Modified Full Cycle Discrete
estimation is achieved within 1.23% with fault classification Fourier Transform (MFCDFT) [16] with signal conditioning
accuracy of 100%. to improve accuracy. Phasor estimated voltage and current
phasor values are used to measure the fault phase and line
Keywords - Distance Protection, Fault location, Ground impedances. During high resistance fault, fault impedance
Distance Function, Modified Full Cycle Discrete Fourier trajectory can fall outside the relay characteristics and hence it
Transform, Numerical relay, Slope tracking method may result in under reaching of distance relays. In the
I. INTRODUCTION proposed work, impedance trajectory is adaptively set to
improve response time by mitigating the adverse effect of fault
In the event of a major power system disturbance,
resistance, DC component and pre-fault power flow conditions
protection and control action resumes power system
by proposed slope tracking method. System Modeling for
restoration, minimize impacts and power system degradation
Indian Power System Network in represented in next section
[1]. However, today’s protection and control actions were not
followed by proposed methodology and result discussions.
well coordinated which are the main causes of cascade
tripping and widespread blackouts [2], [3]. Recently, many II. SYSTEM MODELING
filter algorithms are available for accurate and fast phasor The single line diagram of a 220 kV power system is
estimation using wavelet analysis [4]-[6]. The main issues
shown in Figure 1. There are two generators (G1 and G2)
with these algorithms are delayed operation due to influence connected through a common transmission line. Bus voltages
of decaying DC component. Some algorithms implements and line currents measured at the generator G1 using line CTs
filters like Kalman [7], least-square-fitting [8] and mimic filter and bus PTs applied to the numerical distance relay R with
[9] but their slower convergence speed is the critical issue. sampling frequency of 4 kHz. Bergeron line model with
An application of distance relaying scheme to distributed parameters is used for modeling of transmission
compensate fault location errors due to fault resistance is
line. Generator is modelled with one damper winding in q-axis
presented in [10] based on Discrete Fast Fourier Transform and an IEEE Type 1(AC1A) solid-state exciter. The system
(DFFT). It is found that the error in estimation varies up to and line parameters are described in Appendix. The
6.5%, which is a major indicator of inappropriate phasor performance of numerical distance relay is tested for zone – 1
estimation. Wen et al. [11] presented fast distance relaying protection (100 km) of transmission line.
scheme based on Equal Transfer Process of Transmission
Figure 1: Single line diagram of power system Voltage & current phasor estimation using MFCDFT for
The proposed technique is validated for different faults all phases simultaneously
like L-G, LL, LL-G, LLL-G (10 types) at different locations
across the transmission line ranging from 0.1 km to 110 km to Signal conditioning for accurate phasor estimation
cover the effect of close-in fault, in-zone fault and out-of zone
fault. Moreover, faults are simulated with different fault Calculation of sequence components from voltage and
inception angle (0o-315o), power flow angle (5o, 10o, 15o, 25o, current phasors
35o, 45o) and fault resistance (0.01Ω to 100Ω). The load in
power system is varied by adjusting the variable load angle Adaptive estimation of Ground Distance Function for calculating
between the two generators. The voltage and current signals fault phase impedances
are captured for total 28,800 simulation cases using multi-run
facility in PSCAD/EMTDC and successfully validated with Calculate line impedances using sequence components and
the proposed slope tracking method for fault detection. GDF
ܰ
ʹ ο ο ʹߨ݇
െ
ܺ ͳݎሺܰሻ ൌ ܺ ݎሺܰሻ െ ቆ ݁ܦെ ߬ ͳ ߬ ݁ ͳܦቇ ܿ ݏ൬ ൰ሺͳሻ
ܰ ܰ
݇ൌͳ Estimate Trip signal Fault Classifier Fault location
Fault Generation module estimation
ܰ
ʹ ݇οܶ ݇οܶ ʹߨ݇ instance
െ
ܺ݅ͳ ሺܰሻ ൌ ܺ݅ ሺܰሻ െ ቆ ݁ܦെ ߬ ͳ ߬ ݁ ͳܦቇ ܿ ݏ൬ ൰ሺʹሻ
ܰ ܰ Fault Type
݇ൌͳ
182
In the proposed work, in order to estimate all the voltages as a constant value for finding the value of , it can be
and currents simultaneously at any instance, the phasors for calculated by,
any jth signal at any Nth instance can be obtained by Ͳ െ
Ͳ ൌ ሺͳͳሻ
͵
ܰ
ο
ܺ ͳݎሺ݆ǡ ܰሻ ൌ ܺ ݎሺ݆ǡ ܰሻ െ
ʹ െ ο
ቆܦሺ݆ሻ݁ ߬ሺ݆ ሻ ͳܦሺ݆ሻ݁ െ ߬ͳ ቇ ܿ ݏ൬
ʹߨ݇
൰ ሺ͵ሻ Where, ZL+ = Positive sequence impedance = ͳ and ZL0 =
ͳ
ܰ ܰ Ͳ
݇ൌͳ
Zero sequence impedance =
Ͳ
ܰ
ʹ ݇οܶ ݇οܶ ʹߨ݇
െ
ܺ݅ͳ ሺ݆ǡ ܰሻ ൌ ܺ݅ ሺ݆ǡ ܰሻ െ ቆܦሺ݆ሻ݁ ߬ሺ݆ ሻ ͳܦሺ݆ሻ݁ െ ߬ͳ ቇ ܿ ݏ൬ ൰ ሺͶሻ Similarly, the value of line impedances can be calculated
ܰ ܰ
݇ൌͳ
using following equations:
െ െ െ
The magnitude and phase angle can be calculated by using, ൌ Ǣ ൌ Ǣ ൌ ሺͳʹሻ
ʹ ʹ െ െ െ
ȁሺǡ ሻȁ ൌ ට൫ͳ ሺǡ ሻ൯ ൫ͳ ሺǡ ሻ൯ ሺͷሻ If any one of the complex phasor of phase impedances or
ͳ ሺǡ ሻ line impedances falls within quadrilateral characteristic, then
ס൫ሺǡ ሻ൯ ൌ െͳ ቆ ቇሺሻ trip signal will be generated. From the knowledge of phase or
ͳ ሺǡ ሻ
line impedance locus entering into the trip region, type of fault
Above equation gives magnitude and phase angle of jth signal generated in the power system can be determined. The
at Nth instance. Also equations (5) and (6) give maximum operating time can be calculated by considering the sample
magnitude and phase angle including the effect of three more number as time stamp using the equation,
samples and hence it can result in in-accurate phasor ሺ כοሻ
ൌ ሺͳ͵ሻ
estimation. To correct it, RMS value of magnitude can be
given by,
ȁሺǡ ሻȁ Where, K=Sample number, ο ൌ and ൌ
ൌ ሺǡ ሻ ൌ ሺሻ
ξʹ . In the next sections slope tracking
method has been explained in order to extract fault resistance
Here, cosine fourier transform has been performed for from the total fault impedance to detect high resistance faults.
estimating the phasor magnitude, hence there will be phase
displacement of 90o and delay of three samples is also to be B. Slope Tracking Method
taken into account. To compensate for the change in signal As shown in Figure 3, Z1’ indicates fault impedance for
parameters, the actual phase angle of the fundamental complex solid line-to-ground fault. Due to absence of resistance in
phasor can be given by, faulty path, the quadrilateral relay will easily detect the fault.
͵Ͳ° Whereas during high resistance fault, the fault impedance
Ʌ ൌ ס൫ሺǡ ሻ൯ ൌ ס൫ሺǡ ሻ൯ െ ቆͻͲ° ൬ ͵ כ൰ቇሺͺሻ
locus will move out of the quadrilateral characteristic and
relay will fail to detect such type of abnormality resulting in
Equations (7) and (8) give the RMS value and phase angle under reaching of the relay. In order to prevent such
of fundamental complex phasor after removal of all the breakdown, line impedance locus can be traced by considering
harmonics and decaying DC component, hence it can be used zero sequence component of the fault current and X/R ratio of
in relaying applications for the further digital signal transmission line as outlined in the algorithm of Figure 2.
processing of protective relaying. After incorporating signal
conditioning in MFCDFT, accurate phasors for bus
voltages , , and line currents , , are obtained.
Sequence components of voltages and currents can be
calculated from the phasor voltages and currents using the
equation,
ܸͳ ͳ ͳ ܽ ܽʹ ܸܽ ͳܫ ͳ ͳ ܽ ܽʹ ܽܫ
ܸʹ ൩ ൌ ͳ ܽʹ ܽ ൩ܸܾ ൩ ܽ݊݀ ʹܫ൩ ൌ ͳ ܽʹ ܽ ൩ ܾܫ൩ሺͻሻ
ܸͲ ͵ ܸܿ Ͳܫ ͵
ͳ ͳ ͳ ͳ ͳ ͳ ܿܫ
183
it rises abruptly. Here, in this investigation as shown in Fig. 2, based on conventional DFT, phase comparison principle [4],
a threshold value of 0.1pu of normal load current is wavelet transform & LDA [13] and ANFIS [17].
considered. If zero sequence current is greater than threshold All existing methods fail to operate if the fault resistance
value, it indicates the ground fault has occurred. During such is higher than the impedance setting whereas; the proposed
situation, the actual phase impedance without involving fault method operates well due to implementation of slope tracking
resistance can be found using following method: Total fault method as outlined in Table - II.
impedance including fault resistance can be given by Z T = RT Table-II: Comparison of proposed and existing methods
+ j XT. Where, RT = Total fault resistance between location of Operating time in ms using different
relay and point of fault. It consists of line and fault resistance. Fault methods
During high resistance fault, reactance seen by the relay Fault Locati RF
FIA (◦) Propose Phase
Type on (Ω) Wavelet ANFIS
will be line impedance only. Actual line reactance can be (km)
DFT d
[13]
compari
[17]
method son[4]
traced to total impedance during the fault, i.e. XL؆ XT. Actual
line resistance can be traced to RL = XLtanӨ. Where, tanӨ = L-G 0 10 15 4 6.5 8 7.4 9.23
ܴܮ ͳ L-G 0 60 35 NOP 17.6 18 21.2 25.85
= . Actual line impedance can be given by ZL = RL
ܺܮ Ȁ
L-G 0 50 0.01 12.8 12.5 15 14 15.47
+ j XL and fault resistance by RF = RT - RL. Thus, fault
resistance is estimated adaptively by tracking the slope of fault L-G 0 60 20 NOP 17.8 18 18.2 20.15
impedance trajectory and comparing it with line impedance L-G 0 70 40 NOP 21.5 20 21.2 25.89
trajectory using X/R ratio of transmission line. If obtained line L-G 90 1 0.01 7 7.5 18 7.8 11.88
impedance falls in the quadrilateral relay characteristic then
relay will issue trip signal else it will block the trip signal. L-G 90 5 5 4.5 6.8 13 7.8 12.01
L-G 180 40 0.01 18.2 12.5 17 13.8 17.22
IV. RESULT DISCUSSIONS
LL 0 65 0.01 18.3 13.2 15 14.8 21.58
A. Results of Phasor Estimation
LL 90 30 1 17 13.3 14 11.6 21.03
LL 90 80 0.01 8 14.8 16 20.6 19.64
LL-G 0 10 0.01 9.8 6.2 20 - 17.46
LL-G 0 80 60 NOP 12.3 12 NOP NOP
LL-G 0 99 100 NOP 17.1 17 NOP NOP
LL-G 90 30 1 5.8 6 4 - 21.03
LL-G 90 50 5 18.5 14.3 13 - 21.5
LLLG 90 99 0.01 17.5 17.7 18 - 22.41
LLLG 90 100 0.01 6.5 6.9 12 - 12
Figure 4: Phasor estimation for fault current using DFT and L-G 0 105 0.01 NOP NOP NOP NOP 23.12
MFCDFT during L-G fault at 0.1s (400 sample) with FIA = 0o NOP: No operation of the relay
184
The waveforms of phase voltages, line currents, fault
applied and trip signals during fault resistance of 0.01Ω and
20Ω are shown in Figure 5 and Figure 7. Figure 6 and Figure
8 indicate impedance trajectory during the same faults. It can
be noted that during the high resistance L-G fault applied at
0.1s, the phase impedance and line impedance locus remain
outside the trip region of quadrilateral relay characteristic for
all existing algorithms. On the other hand, in case of proposed
technique, the impedance trajectory of phase impedance (Za)
falls inside the trip region of quadrilateral characteristics
quickly as pointed out in Table-II.
Figure 8: Phase Impedance locus for L-G fault at 50 km at
0.1s with RF = 20Ω and FIA= 0o
C. Fault Classification
In fault classifier, six separate impedance modules are
designed, each for individual phase to ground fault (A-G, B-G
& C-G) and for line to line fault (AB, BC & CA) with output
signals TripA, TripB & TripC and TripAB, TripBC & TripCA
for line modules respectively. Each fault classifier module is
designed in such a way that if impedance locus is falling in to
the trip region of quadrilateral characteristic, then its output
will change to 1(high) otherwise it remains 0 (low).The results
of the fault classifier modules are shown in Figure 9.
During AB-G fault applied at 0.1s, output TripA, TripB and
TripAB of the fault classifier module goes high after a time
Figure 6: Phase Impedance locus for L-G fault at 50 km at delay of 12.5ms as shown in Figure 9 which is less than one
0.1s with RF = 0.01Ω and FIA= 0o cycle time. The obtained results are verified for all the
simulation cases as outlined in Section II. The results reveal
that the proposed technique classifies the fault with 100 %
accuracy at very high speed as outlined in Table II.
The fault classifier modules operate properly for all close-in
faults and in zone fault. For the out of zone faults, the fault
classifier module operates perfectly after providing suitable
delay depending on zone of operation thus maintaining
stability and selectivity of protection systems.
Figure 7: (a) Bus voltages, (b) Line Currents (c) Fault Signal
and (d) Trip Signal during L-G fault applied at 50 km at 0.1s
with RF = 20Ω and FIA = 0o
185
D. Fault Location Estimation V. CONCLUSION
The outcome of variation in fault location error with This paper presents novel decision logic to improve the
respect to fault distance and fault resistance is shown in Table- performance of the numerical distance relaying during high
III for L-G fault with FIA = 0o. The fault locator scheme resistance faults in transmission line protection. System
calculates the location of the fault by taking a ratio of modeling and simulation is performed in PSCAD software
estimated line impedance at the time of fault (without package using multi-run facility and algorithm is validated in
involving fault resistance) to the unit impedance of line). MATLAB. High resistance fault is successfully detected by
݊݅ݐܽܿܮ ݀݁ݐܽ݉݅ݐݏܧെ݊݅ݐܽܿܮ ݈ܽݑݐܿܣ implementing adaptive GDF calculation and slope tracking
Ψ ݎݎݎܧൌ *100 (14) method with quadrilateral relay characteristic. The proposed
݊݅ݐܽܿܮ ݈ܽݑݐܿܣ
technique is found to be highly precise and faster than existing
It can be narrated from the result analysis illustrated in methods for close-in fault, high resistance fault, CT saturation
Figure 10 that during close in faults up to 10km, the fault and fault with higher DC component. Fault classifier modules
location error is within 1% and towards far end the error in are designed for each phase and line modules to identify the
estimation goes up to 1.23%. Thus compared to existing type of fault occur on transmission line. In order to estimate
schemes [10]-[11] in which optimum error in estimation is accurate fault location, fault locator module is also designed
achieved up to 3%, the fault location error estimated by and error in fault location estimation is found within 1% for
proposed method is improved for different fault location, fault close in faults and up to 1.23% for boundary location faults.
types and variation in fault resistance. Moreover, the proposed fault identification algorithm can
accurately estimate fault instance and classify the faults which
Table-III: Fault Location Estimation are the key attributes for multifunctional numerical relay.
Fault Measured Fault Location % Error in estimation of
loca- (Km) fault location Appendix
tion RF = RF = RF = RF = RF= RF= RF= RF=
(Km)
Generator Data
0.01 Ω 10 Ω 30 Ω 80 Ω 0.01 Ω 10 Ω 30 Ω 80 Ω
Generator G1: Power rating: 100 MVA; Voltage rating: 13.8 kV;
2 2.009 2.011 2.012 2.014 0.45 0.55 0.60 0.70 Phase angle: 0o; Frequency = 50 Hz; Positive-sequence impedance =
5 5.021 5.028 5.037 5.038 0.42 0.56 0.74 0.76 0.871 + j9.96Ω.
Generator G2: Power rating: 100 MVA; Voltage rating: 13.8 kV;
10 10.06 10.06 10.06 10.07 0.65 0.67 0.69 0.74
Phase angle: Variable; Frequency = 50 Hz; Positive-sequence
20 20.13 20.14 20.15 20.20 0.67 0.72 0.79 1.04 impedance = 0.871 + j9.96Ω.
40 40.28 40.29 40.30 40.38 0.72 0.72 0.75 0.97 Transmission Line Data
Total Length = 110 km; System Voltage = 230 kV; Positive-sequence
50 50.35 50.38 50.41 50.49 0.70 0.77 0.82 0.98 impedance = 0.0297 + j0.332 Ω/km; Zero-sequence impedance =
60 60.43 60.49 60.51 60.60 0.73 0.82 0.85 1.01 0.162 + j1.24 Ω /km.
70 70.55 70.57 70.58 70.69 0.79 0.81 0.84 0.99 REFERENCES
80 80.68 80.72 80.74 80.88 0.85 0.90 0.93 1.10
[1] Technical Report on Power System Transmission Challenges and
85 85.76 85.82 85.84 85.99 0.89 0.97 0.99 1.17 Research Needs by Electrical Power Research Institute (EPRI),
California, USA, November 2004.
90 90.85 90.92 90.94 91.08 0.95 1.03 1.04 1.21 [2] Final Report on U.S.-Canada Power System Outage: Causes &
Recommendations, Task Force, USA – Canada, April 2004.
95 95.91 95.97 95.99 96.16 0.96 1.02 1.05 1.22
[3] Report on Grid Disturbance on 30th & 31st July, 2012, New Delhi,
India, August, 2012.
[4] S. Zubic, M. Djuric,” A distance relay algorithm based on phase
comparison principle”, Journal of Electric Power System Research,
2012, 92 (1), pp. 20 – 28.
[5] A. Rahmati and R. Adhami, “A Fault Detection and Classification
Technique Based on Sequential Components”, IEEE Transactions on
Industry Applications, Vol. 50, Issue:6, pp. 4202 – 4209, March 2014.
[6] M. Garcia, M. Montanes, N. Halabi, , “High resistive zero-crossing
instant fault detection and location scheme based on wavelet”, Journal of
Electric Power Systems Research, 2012, 92, (1), pp. 138 – 144.
[7] A. Zamora, J. Ramirez, M. Paternina and E. Martinez, “Digital filter for
phasor estimation applied to distance relays”, IET Generation,
Transmission & Distribution, 2015, 9 (14), pp. 1954 – 1963.
[8] P. Jafarian, M. Pasand,” Weighted least error squares based variable
window phasor estimator for distance relaying application”, IET
Generation, Transmission & Distribution, 2011, 5 (3), pp. 298 – 306.
[9] Chi-Shan Yu, “A Discrete Fourier Transform-Based Adaptive Mimic
Figure 10: % Error in Estimation for L-G fault, FIA = 0o Phasor Estimator for Distance Relaying Applications”, IEEE
Transactions On Power Delivery, 2006, 21 (4), pp. 1836 - 1846.
[10] N. Halabi, M. Garcia, A. Alonso, “Application of a distance relaying
scheme to compensate fault location errors due to fault resistance”,
Journal of Elect. Power System Research, 2011, 81(1), pp. 1681 – 1687.
186
[11] M. Wen, D Chen, X. Yin, “A novel fast distance relay for long
transmission line”, Journal of Electric Power and Energy Systems, 2014,
63, (1), pp. 681 – 686.
[12] M. Jamil, A. Kalam, A. Ansari, M. Rizwan, “Generalized neural
network and wavelet transform based approach for fault location
estimation of a transmission line”, Journal of Applied soft computing,
2014, 19, (1), pp. 322 – 332.
[13] A. Yadav, A. Swetapadma, “A novel transmission line relaying scheme
for fault detection & classification using wavelet transform & LDA”,
Ain Shams Engg. Journal, 2015, 6(1), pp. 199 – 209.
[14] A. Yadav, A. Swetapadma, “Improved fault location algorithm for
multi-location faults, transforming faults and shunt faults in TCSC
compensated transmission line”, IET Generation, Transmission &
Distribution, 2015, 9, (13), pp. 1597 – 1607.
[15] Sorrentino E., “Comparison of five methods of compensation for ground
distance function and assessment of their effect on the resistive reach in
quadrilateral characteristics”, Electric Power & Energy Systems, 2014,
61, (1), pp. 440 – 445.
[16] Sun-Li Y., Jyh-Cherng G., “Removal of Decaying DC in Current and
Voltage Signals using a Modified Fourier Filter Algorithm”, IEEE
Transactions on power delivery, 2001, 16, (3), pp. 372 – 378.
[17] M. Pasand, P. Jafarian, “An Adaptive Decision Logic to Enhance
Distance Protection of Transmission Lines”,IEEE Transaction on power
delivery, 2011,26,(4), pp. 2134 – 2144.
187