Вы находитесь на странице: 1из 4

This article has been acceptedfrom

Downloaded for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892600, IEEE
http://iranpaper.ir
http://www.itrans24.com/landing1.html
Transactions on Power Delivery
1

Mitigation of The DC Offset by a Sub-Cycle


Sample Method M-Class PMUs
Bandi Ravi Kumar, Student Member, IEEE, Avinash Kumar, Student Member, IEEE

Abstract—The phasor estimation of the fault current signal in suitable for the power system monitoring. The mathematical
the presence of DC offset results in the significant error which formulation of the proposed method is explained in the next
is not suitable for the power system monitoring. To reduce the section.
phasor estimation error of the fault current in the presence of
DC offset, primarily the DC offset needs to be mitigated. In
this letter, a four-sample method, which requires four sub-cycle II. I MPLEMENTATION OF F OUR -S AMPLE A LGORITHM
samples as the input and estimates the DC offset parameters
present in the signal, is proposed. The test results show that the A. Formulation of DC offset Mitigation
proposed method is simple and more accurate in mitigating the
DC offset from the fault signal. The fault current signal in the presence of DC offset is
expressed as the sum of the decaying exponential component,
Index Terms—DC offset, discrete Fourier transform, fault
fundamental and harmonic components. The discrete form of
current, four-sample method, phasor measurement unit.
the fault current is given by [4],
N
−1
2
I. I NTRODUCTION (−n.∆t/τ )
X
i[n] = M0 e + Mk sin(k.ω.n + φk ) (1)
T the instant of fault, current in the inductive circuit
A cannot change instantaneously. So, the DC decaying
component appears in AC current, when the value of load
k=1

where n is the sample number, i[n] is the discrete fault current


current before the fault, is different from the AC current at fault at n-th sample, ∆t is the sampling time, τ is the time constant
incidence. And this DC offset decays exponentially with a time of DC offset, ω = 2.π
N and N is the number of samples in a
constant which is a function of X/R ratio of the system [1]. cycle. k is the harmonic order. Mk and φk are the magnitude
If this DC decaying component or DC offset is not mitigated, and initial phase angle of the k-th harmonic component. M0 is
it results in large phasor estimation error. Since the phasors the magnitude of DC offset. Discrete fault current i[n], can be
from phasor measurement units are used for power system expanded into two components namely, DC offset component
monitoring and protection, the error in these phasors can have and AC component. Let DC offset and AC component of the
the adverse effect on the same. fault current i[n] be idco [n] and iac [n], and is represented as,
Several methods have been presented [2]–[6] in the liter- N
−1
2
ature to mitigate the DC offset. In [2], the discrete Fourier
X
n
i[n] = idco [n] + iac [n] = M0 E + Mk sin(k.ω.n + φk )
transform (DFT) of even and odd sample sets of the fault k=1
current is estimated. And the difference of the DFT of even (2)
and odd sample sets is calculated. This difference is used for
the estimation of DC offset parameters. But the procedure is where,
quite complex and slow. The method proposed in [3], uses E =e(−∆t/τ ) , (3)
the mimic filter for the removal of the DC offset. But it
amplifies the harmonics of the fault current which results in idco [n] =M0 E n ,
N
the significant error in phasor estimation. In [4], a partial sum- 2−1
X
based algorithm has been proposed. It is very sensitive to iac [n] = Mk sin(k.ω.n + φk ). (4)
noise, and different values of the DC offset time constant. The k=1
methods proposed in [5], [6], generates an auxiliary signal The DFT of the signal i[n] in (2) is defined as,
using the actual signal. The DC offset parameter estimation
dco 1
procedure involves the sum of the samples in one cycle of both Idf t = Idf t + Idf t (5)
the auxiliary signal and the actual signal. But as the method
dco 1
implicates the summation of fault current data in one cycle, where, Idf t , Idf t , Idf t are DFT of i[n], idco [n] and iac [n],
dco
the estimation is highly sensitive to noise and harmonics. respectively. Idf t and Idf t can be formulated as,
This letter proposes a four-sample based algorithm to esti- N −1
mate the DC offset parameters. Its performance is validated 2 X
Idf t = i[n]e−jωn (6)
by comparing with the algorithm in [5], which makes it N n=0
N −1
1 − EN
 
Bandi Ravi Kumar and Avinash Kumar are with the Department of dco 2 X n −jωn 2
Electrical Engineering, Indian Institute of Technology Kanpur, 208016, India. Idf t = M0 E e = M0 . (7)
N n=0 N 1 − Ee−jω
(emails: ravikumarjnv.kumar@gmail.com, kumarav@iitk.ac.in)

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been acceptedfrom
Downloaded for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892600, IEEE
http://iranpaper.ir
http://www.itrans24.com/landing1.html
Transactions on Power Delivery
2

dco
By substituting the value of Idf t from (7) in (5), the In (17), Ak is given by,
fundamental phasor can be expressed as,  2π 
N .(N/2 + 2) + 2φk
2

1 − EN
 Ak = 2 sin k. .
1
Idf = I df t − M0 . (8) 2
t 2π
N 1 − Ee−j N
And, cos k. π2 = 0, ∀k = 1, 3, 5, . . . , ( N2 − 1). So, Hsum1 =

1
Idf t = MR + j.MI (9) 0.
where, j is the complex operator, MR and MI are real and Similarly addition of (14) and (15) results as,
imaginary parts of DFT of the fundamental phasor.
i[2] + i[N/2 + 2] = M0 E 2 + E N/2+2 + Hsum2 . (18)
 
Fundamental phasor magnitude (M1 ) and phase angle (φ1 )
are given by, where,
q
N
M1 = MR2 + MI2 (10) −1
2
X  
2π.2

MI Hsum2 = Mk sin k. + φk +
φ1 = tan-1 (
). (11) N
k=1
MR  
2π.(N/2 + 2)
To obtain the fundamental phasor (Idf 1
t ) in (8), DC offset
sin k. + φk ,
N
parameters (M0 , E) have to be estimated. From (8), it can be N
2 −1
observed that the accurate estimation of M0 and E, affects the
 
X π N
accuracy of fundamental phasor estimation (Idf 1 = Mk Bk . cos k. , ∀k = 1, 3, 5, . . . , − 1.
t ). DC offset 2 2
k=1
parameter estimation by proposed method is explained in next
(19)
section.
In (19), Bk is given as,
B. Proposed Method for DC offset Parameter Estimation  2π 
N .(N/2 + 4) + 2φk
To implement the proposed algorithm, four sub-cycle sam- Bk = 2 sin k. .
2
ples of the fault current signal are utilized to mitigate the DC
And, cos k. π2 = 0, ∀k = 1, 3, 5, . . . , N2 − 1. So, Hsum2 = 0.

offset as follows:
N
2 −1   Dividing (16) by (18) results in ‘y’
1
X 2π.1
i[1] = M0 E + Mk sin k. + φk (12) i[1] + i[N/2 + 1] E 1 + E (N/2+1)
N y= (20)= 2 .
k=1
N
i[2] + i[N/2 + 2] E + E (N/2+2)
  2 −1  
N 2π.(N/2 + 1)
+φk From (20), ‘E’ can be obtained as follows,
X
i +1 = M0 E (N/2+1) + Mk sin k.
2 N
k=1 yE (N/2+2) − E (N/2+1) + yE 2 − E 1 = 0
(13)
N
2 −1   E (N/2+1) .(yE − 1) + E.(yE − 1) = 0
X 2π.2
i[2] = M0 E 2 + Mk sin k. + φk (14) E (N/2+1) + E .(yE − 1) = 0

N
k=1
(E). E N/2 + 1 .(yE − 1) = 0

N
  2 −1  
N X 2π.(N/2 + 2) 1
i +2 = M0 E (N/2+2) + Mk sin k. +φk . E = 0, −1(2/N ) , . (21)
2 N y
k=1
(15) The selection of the desired root among the three possible
By adding (12) and (13), it results in (16), roots is given as follows:
i[1] + i[N/2 + 1] = M0 E 1 + E (N/2+1) + Hsum1 . (16)
 
−∆t
Case 1 : E = 0 → e τ = 0 ; (22)
where, 
 E = −1,

N (2/N )
for N = 2 
2 −1    Case 2 : E = −1 →
X 2π.1  E = a + j.b, for N > 2 
Hsum1 = Mk sin k. + φk +
N
k=1
  (23)
2π.(N/2 + 1) 1
sin k. + φk , Case 3 : E = (24)
N y
N
2 −1   2π  The root in Case 1, is an undesired root because, it is known
X .(N/2 + 2) + 2φk
= Mk 2 sin k. N . that e−∞ = 0; and in this letter, ∆t
2 τ 6= ∞; as τ 6= 0, which
k=1 ranges from 0.5-5 cycles [3], and ∆t 6= ∞, as ∆t is a finite
 2π
N .(N/2) + φk − φk

cos k. , value. In Case 2, N = f s /f h , must satisfy the condition N ≥
2 2, to satisfy Nyquist criterion, i.e, fs ≥ 2.fh , where, fs , fh
N
2 −1   are the sampling frequency and highest frequency component
X π N
= Mk Ak . cos k. , ∀k = 1, 3, 5, . . . , − 1. of the signal. And, E, an exponential, cannot be negative or
2 2 complex value (a+j.b), where j is a complex operator. So, the
k=1
(17) root in Case 3, E = (1/y), is the desired and required root

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been acceptedfrom
Downloaded for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892600, IEEE
http://iranpaper.ir
http://www.itrans24.com/landing1.html
Transactions on Power Delivery
3

Test signal Phasor Magnitude Test signal Phasor Magnitude


1.5 1.03 2 1.03
Proposed
Proposed
1 1.02 1.02 Method [5]
Magnitude (p.u)

Magnitude (p.u)

Magnitude (p.u)

Magnitude (p.u)
Method [5]
1

0.5 1.01 1.01


0
0 1 1

-1
-0.5 0.99 0.99

-1 0.98 -2 0.98
0 0.05 0.1 0.15 0 0.05 0.1 0.15 0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15
Time (s) Time (s) Time (s) Time (s)
Phase Angle Total Vector Error Phase Angle Total Vector Error
100 3 100 3
Proposed Proposed Proposed
Proposed Method [5] 2.5 Method [5]
Method [5] 2.5
Method [5] 50
50

Angle (deg)
Angle (deg)

TVE in %
2

TVE in %
2 90
0 85 1.5 0 1.5
80 80
75 1 1
70 -50 70
-50
0.5 2 3 4 5 0.5
1.5 2 2.5 3 3.5 -3
#10
-100 #10 -3 0 -100 0
0 0.05 0.1 0.15 0 0.05 0.1 0.15 0 0.05 0.1 0.15 0 0.05 0.1 0.15
Time (s) Time (s) Time (s) Time (s)

Fig. 1: Mitigation of DC offset in the presence of harmonics Fig. 3: Mitigation of DC offset in the presence of noise (SNR
for τ = 0.5 cycle. 25 dB) for τ =5 cycles.

Test signal Phasor Magnitude


1.5 1.02 TABLE I
1
Proposed
Method [5]
C OMPARISON OF THE E FFICACY OF E STIMATION M ETHODS
Magnitude (p.u)

Magnitude (p.u)

1.01
0.5

0 1
DC Offset Fundamental Method [5] Proposed
Signal
-0.5 Parameters Parameters TVE (%) TVE (%)
0.99 Type
-1 τ E M0 M1 φ1 M ax M ean M ax M ean
-1.5
0 0.05 0.1 0.15
0.98
0 0.05 0.1 0.15
0.5 0.9394 0.5 1.0 30◦ 2.682 0.463 1.477 0.275
i+Har
Time (s) Time (s)
5.0 0.9938 0.5 1.0 30◦ 4.205 1.614 2.234 0.837
Phase Angle Total Vector Error
100
Proposed
2.5 i+Noise 0.5 0.9394 0.2 1.0 40◦ 4.098 1.032 1.495 0.576
Proposed

50
Method [5]
2
Method [5] (30 dB) 5.0 0.9938 0.2 1.0 40◦ 2.085 1.118 1.757 1.006
Angle (deg)

TVE in %

88
86
84
i+Noise 0.5 0.9394 1.0 1.0 20◦ 3.150 2.108 2.792 1.413
0 1.5
82
80 (25 dB) 5.0 0.9938 1.0 1.0 20◦ 2.955 1.075 2.198 0.862
78
-50 76 1
1.5 2 2.5 3
#10 -3
-100 0.5
0 0.05 0.1 0.15 0 0.05 0.1 0.15
Time (s) Time (s)
the real and imaginary parts of actual phasor at n-th sample
Fig. 2: Mitigation of DC offset in the presence of noise (SNR instant.
30 dB) for τ =5 cycles.
A. Fault Current Signal in the Presence of Harmonics
of the polynomial equation. Subsequently, substituting E in The harmonic test signal employed to the proposed method
either (16) or (18), DC offset amplitude M0 can be calculated. is given below.
Finally, DC offset from the fault current can be mitigated by
substituting the DC offset parameters E and M0 in (8). i[n] =0.5e(−n.∆t/τ ) + 1.0 sin(ωn + 30◦ ) + 0.1 sin(3ωn)+
0.08 sin(5ωn) + 0.05 sin(7ωn) + 0.04 sin(9ωn)
(26)
III. S IMULATION R ESULTS
Test signal consists of a DC offset, and the harmonics of
The efficacy of the proposed algorithm is assessed for the order 3, 5, 7 and 9 whose amplitude values considered
the test signals, and it is validated by correlating with the as 0.1, 0.08, 0.05, and 0.04 in p.u respectively. DC offset
algorithm [5]. The fundamental frequency (f ) and the number magnitude and fundamental magnitude are 0.5 p.u and 1.0
of samples per cycle (N ) are set at 50 Hz and 32, respectively. p.u, respectively. The estimated phasor and TVE is shown in
Since DC offset time constant τ varies from 0.5 to 5.0 cycles, Fig. 1 for τ of 0.5 cycle. Table I shows the comparison of
the test results are shown for the boundary values of τ . The the accuracy of phasor estimation by proposed method and
comparison between the proposed method and the method [5] method in [5]. It can be observed that the maximum TVE and
is shown for test signals in the presence of harmonics, and mean TVE are less in value for the proposed method. In Table
noise with the signal to noise ratio (SNR) 30 dB and 25 I, ‘i + har’ represents the fault current with harmonics.
dB. The performance of the proposed method is evaluated by
performance index total vector error (TVE), and given (in %)
B. Fault Current Signal in the Presence of Noise
as,
s 2  2 The fault current signal i[n] with the white gaussian noise
MR (n) − mr (n) + MI (n) − mi (n) w(n) is given by,
T V E(n) = ×100
mr (n)2 + mi (n)2 i[n] = M0 e(−n.∆t/τ ) + M1 sin(ωn + φ1 ) + w(n). (27)
(25)
Where, MR (n) and MI (n) are the real and imaginary parts The proposed method and method in [5] are tested in the
of DFT of the estimated signal, and mr (n) and mi (n) are presence of noise (27) with SNR of 30 dB and 25 dB. M0 , M1 ,

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been acceptedfrom
Downloaded for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/TPWRD.2019.2892600, IEEE
http://iranpaper.ir
http://www.itrans24.com/landing1.html
Transactions on Power Delivery
4

TABLE II Generally, a PMU requires atleast N samples (1 cycle data


C OMPARISON OF C OMPUTATIONAL B URDEN FOR P HASOR E STIMATION for DFT calculation) for phasor estimation. The proposed
method also requires one cycle (0.02 s for 50 Hz) data for
Estimation Method [2] [5] [6] Proposed
phasor estimation, and 4 sub-cycle (within the cycle) samples
Calculation of DFT
No. of samples required N N N N for DC offset parameter estimation.
of fault current (Idf t )
DC offset parameter No. of samples required N N N 4 V. C ONCLUSION
N
estimation (M0 , E) Window span (in samples) N N N +2
2 In this letter, a simple and more accurate algorithm is
proposed for the mitigation of the DC offset from the fault
current. The dominion of proposed method is shown by
φ1 are considered as 0.2 p.u, 1.0 p.u and 40◦ , respectively for comparing it with the existing method in-terms of accuracy,
SNR of 30 dB, and 1.0 p.u, 1.0 p.u and 20◦ , respectively for complexity and computation burden. Since it requires only 4
SNR of 25 dB, as shown in Table I. The performance of both sub-cycle samples for DC offset parameter estimation, and N
methods are shown in Fig. 2 and Fig. 3, and demonstrated in samples (1 cycle) for phasor estimation, it can be used in M-
Table I, where, ‘i + N oise0 represents fault current with noise. class PMU.
IV. D ISCUSSION
From (8), it can be observed that the computation burden R EFERENCES
1
for phasor estimation (Idf t ) can be divided into two parts, i.e., [1] K. W. Min and S. Santoso, “Dc offset removal algorithm for improving
(i) Computation burden for calculation of DFT of fault current location estimates of momentary faults,” IEEE Trans. Smart Grid, 2017.
[2] S. H. Kang and D. G. Lee, “Fourier transform-based phasor estimation
(Idf t ). (ii) Computation burden for DC offset parameter esti- method and apparatus capable of eliminating influence of exponentially
mation (M0 , E). Table II demonstrates that the computation decaying dc offsets,” Mar. 27 2012, US Patent 8,145,443.
burden for DFT calculation of fault current (Idf t ) is the same [3] G. Benmouyal, “Removal of dc-offset in current waveforms using digital
mimic filtering,” IEEE Trans. Power Del., vol. 10, no. 2, pp. 621–630,
for all methods (since all methods require N samples). But the 1995.
computation burden and complexity for DC offset parameter [4] Y. Guo, M. Kezunovic, and D. Chen, “Simplified algorithms for removal
estimation (M0 , E) is drastically reduced by the proposed of the effect of exponentially decaying dc-offset on the Fourier algorithm,”
IEEE Trans. Power Del., vol. 18, no. 3, pp. 711–717, 2003.
method, where it requires only four sub-cycle samples whose [5] B. Jafarpisheh, S. M. Madani, and S. M. Shahrtash, “A new DFT-based
window length span is ‘ N2 + 2’ samples. And also Table I phasor estimation algorithm using high-frequency modulation,” IEEE
shows the improved accuracy by the proposed method. The Trans. Power Del., vol. 32, no. 6, pp. 2416–2423, 2017.
[6] B. Jafarpisheh, S. M. Madani, and S. Jafarpisheh, “Improved DFT-based
aforementioned things make the proposed method simple, phasor estimation algorithm using down-sampling,” IEEE Trans. Power
more accurate and computationally cheap which are the main Del., 2018.
contributions of the proposed method.

0885-8977 (c) 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Вам также может понравиться