Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

www.ietdl.

org
Published in IET Generation, Transmission & Distribution
Received on 15th March 2013
Revised on 14th August 2013
Accepted on 15th August 2013
doi: 10.1049/iet-gtd.2013.0180

ISSN 1751-8687

Time-frequency transform-based differential scheme


for microgrid protection
Susmita Kar, Subhransu Rajan Samantaray
School of Electrical Sciences, Indian Institute of Technology, Bhubaneswar, India
E-mail: sbh_samant@yahoo.co.in

Abstract: The study presents a differential scheme for microgrid protection using time-frequency transform such as S-transform.
Initially, the current at the respective buses are retrieved and processed through S-transform to generate time-frequency contours.
Spectral energy content of the time-frequency contours of the fault current signals are calculated and differential energy is
computed to register the fault patterns in the microgrid at grid-connected and islanded mode. The proposed scheme is tested
for different shunt faults (symmetrical and unsymmetrical) and high-impedance faults in the microgrid with radial and loop
structure. It is observed that a set threshold on the differential energy can issue the tripping signal for effective protection
measure within four cycles from the fault inception. The results based on extensive study indicate that the differential energy-
based protection scheme can reliably protect the microgrid against different fault situations and thus, is a potential candidate
for wide area protection.

1 Introduction over current protection [1013] becomes in effective for


islanded mode of operation.
Installations of microgrids continue to proliferate as a viable An adaptive relay strategy has been proposed [6] for
solution to the problems of greenhouse gas, energy growth distribution systems with high DG penetration. However,
demand and the depletion of energy resources through this does not include islanded mode of operation and
enabling the use of renewable and distributed generation applicable only if the DG penetration is high. The authors
systems [1]. Microgrid implementations present an effective in [7, 8] proposed communication-based protection schemes
means of distributing high-quality power more efciently to for grid-connected microgrids, based on the assumption of
residential, urban and rural areas and to commercial large fault current contributed from utility grid. Although an
facilities. They will enable Distributed Energy Resources option for microgrids with no sensitive loads, this does not
(DER) usage worldwide at various levels of power delivery include fault detection in islanded mode of operation.
and increase the efciency with which multiple renewable Another non-communication-based protection scheme is
energy sources such as photovoltaic cells, fuel cells and proposed [9], which is effective, however, does not consider
wind power can provide aggregate power to a group of a proper microgrid structure.
loads. Such a sub-network, known as microgrid [2], should In another work [14], sequence components are derived to
be able to operate in grid-connected mode and islanded design the protection and relay coordination. However,
mode (independently), while meeting the standards of the relay coordination depends on multiple settings on the
service reliability and power quality. To meet the above derived sequence components, making the relaying scheme
requirements of microgrid operation, an effective microgrid more complex. Microgrid protection using voltage protection
protection scheme is essential, which is highly challenging. [15] uses synchronous reference frame to compare with the
The protection scheme of a microgrid must ensure safe reference of the phase voltage at the DG resource. Another
operation of microgrid against all types of faults in grid technique uses over current differential protection [16] on
connected as well as in islanded mode. The protection issue each line with back-up voltage and frequency protection at
becomes critical as the fault current is large in each DG. Moreover, the above-mentioned approaches are
grid-connected mode because of contribution from utility suitable for solid shunt (LG, LLG) faults and fail to detect
grid and may be signicantly smaller because of limited high-impedance faults (HIF).
current contribution from the DER units [35]. This is a Considering the above issues, there is a strong motivation
standard practice to employ conventional over current relays to develop the protection scheme including both solid and
[69] for protection in grid-connected mode, despite the HIF faults in grid connected as well as islanded mode of
fact that the protection may fail in some situations. In microgrid operation. Recently, microgrid protection scheme
islanded mode, the fault current magnitude may vary over a using communications assisted digital relays [17] has been
wide range depending on the power generated by the proposed. This scheme is based on differential current
distributed generations (DG) units. Thus, the conventional approach where time-synchronisation is not considered.

310 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0180
www.ietdl.org
spectral analysis, where the standard deviation is an inverse
function of the frequency, thus reducing the dimension of the
transform. The expression for S-transform of a continuous
signal x(t) is given as (time domain)
Fig. 1 Schematic diagram of the proposed technique 1
s(t, f ) = x(t)w(t t, f )e2pift dt (1)
1
However, no time-synchronisation will have serious impact as
differential current is the key indicator for issuing tripping where f is the frequency, t is the time and t is a parameter
signal. To overcome such type of problems, differential that controls the position of the Gaussian window on the t-axis.
energy-based microgrid protection scheme is proposed, The window function can be expanded and the S-transform
which is less sensitive to synchronisation error compared (time domain) is given as follows
with time-domain data difference such as differential 1  
currents. The proposed study includes all types of shunt 1 2 2

faults (LG, LL, LLG, LLLG) and HIF faults compared s(t, f ) = x(t)  e (tt) /2sf ei2pft dt (2)
1 s(f ) 2p
with only LG and HIF faults [17] and registering the fault
patterns (types of fault) for different kinds of fault The standard deviation ( f ) of the Gaussian window is related
situations, required for autoreclosure. to frequency f as
The proposed technique develops a differential energy-based
protection scheme for the microgrid using time-frequency 1
transform such as S-transform. The scheme can be viewed as s(f ) =   (3)
f
a pattern recognition approach for microgrid protection. The
S-transform [1821] is an invertible time-frequency spectral To improve the performance of the S-transform in terms of
localisation technique that combines elements of wavelet time-frequency contours for signal detection, the standard
transforms and short-time Fourier transform. The S-transform deviation of the Gaussian window is further modied as
uses an analysis window, whose width is decreasing with
frequency providing a frequency-dependent resolution. This k
transform may be seen as a continuous wavelet transform s(f ) =   (4)
with a phase correction. It produces a constant relative a + b/ f
bandwidth analysis like wavelets while it maintains a direct
link with Fourier spectrum. The S-transform has an advantage where a and b are
small positive

 constants varying between 0
in that it provides multiresolution analysis while retaining the and 1 and k , a2 + b2 . Using the modied standard
absolute phase of each frequency. S-transform has been found deviation, the Gaussian window is modied as
  2 2  2 
better compared with another time-frequency transform
technique such as wavelet transform, as it is (wavelet a+b f a+b f t /2k
transform) highly prone to noise and provides erroneous w(t, f ) =  e (5)
results even with signal-to-noise ratio (SNR) of 30 dB [22]. k 2p
In the proposed technique, fault currents at both ends of the
Thus, the continuous modied S-transform (Fourier domain)
respective feeders in a microgrid, are retrieved and processed
can now be represented as
through S-transform to nd out the spectral energy content.
The difference in spectral energy (differential energy) content 1   2
of time-frequency contours of both signals is the key 2p2 a2 k 2 / a+b |f |
S(t, f ) = X (a + f )e e2piat da (6)
indicator for registering the fault patterns in the microgrid. 1
Further, the tripping signal can be issued based on a set
threshold on the differential energy for different fault The discrete S-transform of the signal is given as
situations in microgrid. The proposed technique provides
highly promising results on extensive studies carried out
N 1  2 2 2   2
during testing for different fault conditions in microgrid S(j, n) = X (n + m) e 2p m k / a+b |f | ei(2pmj)/N ) (7)
(radial and loop network structures) at grid-connected and m=0
islanded mode. The schematic diagram of the proposed
differential energy-based technique is outlined in Fig. 1. The and j = 1 N 1, n = 0, 1 N 1. Here j and n indicate
following sections deal with S-transform, studied systems, the time samples and frequency steps (voice frequency),
computational results, discussion and conclusions. respectively. X(n + m) is obtained by shifting X(n) by m,
whereas X(n) is the Fourier transform of x(k) and given as

2 Generalised S-transform 1 N
1
X (n) = x(k) ei(2pnk/N ) (8)
N k=0
The S-transform [1821] is an extension to the Gabor transform
and wavelet transform, and is based on a moving and scalable
localising Gaussian window. The interesting phenomena in Now the discrete S-transform can also be represented in terms of
the S-transform are that it is fully convertible both forward amplitude and phase spectrum as
and inverse from time domain to frequency domain. This
property is due to the fact that the modulating sinusoids are S(j, n) = A(j, n) eiw(j,n) (9)
xed with respect to the time axis whereas the localising
scalable Gaussian window dilates and translates. The where A( j, n) is the amplitude spectrum and ( j, n) is the phase
S-transform falls within the broad range of multi-resolution spectrum of the original time series signal.

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320 311
doi: 10.1049/iet-gtd.2013.0180 & The Institution of Engineering and Technology 2014
www.ietdl.org
The spectral energy content E of the transformed signals is primary protection at R 1 (for a fault on the feeder between
obtained as follows B-1 and B-2), the fault currents are retrieved at B-1 and B-2,
and the differential energy is calculated. For back-up
E = {abs(S(j, n))}2 (10) protection at R-1(for fault on feeder between B-2 and B-3),
the fault currents at B-1 and B-3 are considered for computing
where S( j, n) is the generated S-matrix for different time and the differential energy. Similarly for primary protection at R-2
frequency steps. The various parameters selected in our (fault on feeder between B-2 and B-3), differential energy
proposed study are as a = 0.5, b = 0.5 and window width is computed considering the currents at B-2 and B-3,
( f ) = 0.4. respectively. For back-up protection at R-2, differential
In the present application, the time-domain sampled current energy is computed considering the fault currents at B-2 and
signals retrieved at both ends are processed through B-4. This approach is considered for different fault conditions
S-transform to generate time-frequency contours. The in primary and back-up zone in the micogrid to compute the
differential energy [23] of the time-frequency contours of differential energy for effective protection measure.
the fault current signals is computed to detect the fault The spectral energy content of the fault currents retrieved at
patterns in the microgrid. bus S and T are calculated as given in (11) and (12), which
are derived from (10). The corresponding differential energy
is computed as the difference between the two and given in
3 System studied (13). For primary and back-up protections, the busses are
selected as mentioned earlier
The detailed studied system is shown in Fig. 2. The base
power has been chosen as 10 MVA. The studied system
2
consists of radial distribution system with four DG units EBS = abs(SBS (j, n)) (11)
(wind farms), connected to the main supply system through
point of common coupling (PCC). The DG units are placed

2
at a distance of 20 km with distribution lines of pi-sections.
EBT = abs(SBT (j, n)) (12)
The details of the generator, DGs, transformers, distribution
lines and loads are mentioned in Appendix 1.
The model is simulated using SIMULINK (MATLAB) Differential energy = EBS EBT (13)
package. The sampling frequency is 1.2 kHz at 60 Hz base
frequency (20 samples per cycle). Fault currents at sending
and receiving ends of respective feeders are retrieved and Figs. 3a and b show the fault currents and generated
processed through S-transform to compute the differential time-frequency contours against samples (time) for the fault
energy for registering the fault patterns. Different types of currents at B-1 and B-2 for LG fault created on feeder
shunt faults such as lineground (LG), linelineground between B-1 and B-2. It is observed that the time-frequency
(LLG), lineline (LL), linelinelineground (LLLG) contours effectively localise the fault patterns immediately
and HIF are created on the distribution line at different after the fault occurrence. The spectral energy content of
locations. The faults are created in both grid-connected and frequency contours of the fault currents at B-1 and B-2, and
islanded mode. The CB_MAIN is activated (off) for the corresponding differential energy (computed by
islanded mode and CB_LOOP is activated (on) for loop subtracting the spectral energy content of the current signal
network of microgrid. at B-2 from spectral energy content of the current signal at
B-1) are shown in Fig. 3c. The differential energy and the
threshold setting for LG fault on the feeder between B-1
4 Simulation results and B-2 (primary protection at R-1 with radial network) are
shown in Fig. 3d. It is observed that with a threshold
4.1 Microgrid protection in grid-connected mode setting of 1, the differential energy-based technique takes
3.5 cycles (70 samples) from the fault inception to issue the
In the proposed technique, the fault current signals retrieved at
tripping signal. Similar observations are made for LLG
sending and receiving end of the respective feeder (considering
fault on the feeder between B-3 and B-4: primary protection
the primary and back-up protection), are processed through
at R-3 (loop network), taking 3.6 cycles after fault inception
S-transform to produce the time-frequency contours. For
to issue the tripping signal as shown in Fig. 4. The
threshold setting is conrmed after testing on various cases
of fault situations across different sections of the microgrid.
The proposed technique is also tested for registering fault
patterns in back-up zone. Fig. 5 shows the differential
energy and the threshold setting for LG fault on the feeder
between B-2 and B-3: back-up protection at R-1 (radial
network). In this case, the tripping signal can be issued
after 3.75 cycles from fault inception. In back-up
protection, the set threshold is 0.5 (conrmed after testing
on different fault situations in back-up zone) for issuing the
tripping signal. Fig. 6 shows the differential energy and the
set threshold for LLG fault on the feeder between B-6 and
B-7: back-up protection at R-5 (loop network), and the
tripping signal is issued after 3.7 cycles from fault
inception. The back-up zone is provided with a time delay
of 0.3 s. Thus, the relay will operate after a time delay of
Fig. 2 Microgrid with multiple DG interface 0.3 s plus the response time based on the set threshold of 0.5.

312 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0180
www.ietdl.org

Fig. 4 Differential energy and the threshold setting for LLG fault
on the feeder between B-3 and B-4: primary protection at R-3 (loop
network)

Fig. 5 Differential energy and the threshold setting for LG fault


on the feeder between B-2 and B-3: back-up protection at R-1 (radial
network)

One of the important events in protection measure is the


fault pattern recognition. This is the process of identifying
the faulty phase (phases) involved during the shunt fault

Fig. 3 Differential energy for fault protection


a Fault current and time-frequency contours of phase-a for LG fault on the
feeder between B-1 and B-2, retrieved at B-1
b Fault current and time-frequency contours of phase-a for LG fault on the
feeder between B-1 and B-2, retrieved at B-2
c Spectral energy of the fault current at B-1, B-2 and differential energy for the
for LG fault on the feeder between B-1 and B-2: primary protection at R-1
(radial network)
Fig. 6 Differential energy and the threshold setting for LLG fault
d Differential energy and the threshold setting for LG fault on the feeder on the feeder between B-6 and B-7: back-up protection at R-5 (loop
between B-1 and B-2: primary protection at R-1 (radial network) network)

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320 313
doi: 10.1049/iet-gtd.2013.0180 & The Institution of Engineering and Technology 2014
www.ietdl.org

Fig. 7 Differential energy of a, b and c-phase for LLG fault


Fig. 9 Differential energy and the threshold setting for LG fault
(ab-g) on feeder between B-2 and B-3: primary protection at R-2
on the feeder between B-1 and B-2: primary protection at R-1 (radial
(fault pattern recognition: radial network)
network)

process. The differential energy for a, b and c-phase for LLG


opens in case of islanded mode. The differential
fault (ab-g) on feeder between B-2 and B-3 is shown in Fig. 7
energy-based scheme has been tested for different fault
and is observed that the differential energy of a, b-phase are
situation in microgrid in islanded mode. Fig. 9 shows the
growing to higher value, whereas the differential energy of
differential energy and the threshold setting for LG fault
c-phase does not grow. While testing for LLLG fault
on the feeder between B-1 and B-2: primary protection at
(abc-g) on feeder between B-5 and B-6 (primary protection
R-1 (radial network). It is observed that with a threshold
at R-5 in loop network), the differential energy for all
setting of 0.3, the response time achieved is 3.8 cycles
phases are growing effectively to recognise the fault
(76 samples) from fault inception. Fig. 10 shows the
patterns as shown in Fig. 8. The response time for the
differential energy and the threshold setting for LG fault
above two cases are 3.55 and 3.65 cycles from fault
on the feeder between B-3 and B-4: primary protection at
inception, respectively. Thus, the fault pattern recognition is
R-3 (radial network), with a response time of 3.75 cycles
reliably achieved using the differential energy in microgrid
from fault inception. In the islanded mode, effectiveness of
protection.
the back-up protection is also taken into considerations.
Fig. 11 shows the differential energy and the threshold
setting for LLG fault on the feeder between B-2 and B-3:
4.2 Microgrid protection in islanded mode back-up protection at R-1 (radial network). Similarly,
The fault currents in microgrid with grid-connected mode and Fig. 12 shows the differential energy and the threshold
islanded mode are signicantly different and thus, protection setting for LG fault on the feeder between B-6 and B-7:
measures face difculty to provide a unique solution. In case back-up protection at R-5 (radial network). In case of
of faults in islanded mode, all loads remain in service and back-up protection, the threshold is set at 0.3 and the
the generation load balance is maintained. The fault tripping signal can be issued within four cycles after fault
transients are of different order compared with grid inception.
connected mode because of the reversal of power ow For microgrid faults in islanded mode, the fault pattern
along many of the lines at the instant the switching device recognition is also achieved effectively with the proposed
differential energy-based approach. Fig. 13 shows the fault

Fig. 8 Differential energy of a, b and c-phase for LLLG fault Fig. 10 Differential energy and the threshold setting for LG fault
(abc-g) on feeder between B-5 and B-6: primary protection at R-5 on the feeder between B-3 and B-4: primary protection at R-3 (radial
(fault pattern recognition: loop network) network)

314 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0180
www.ietdl.org

Fig. 11 Differential energy and the threshold setting for LLG Fig. 14 Differential energy of a, b and c-phase for LLG fault
fault on the feeder between B-2 and B-3: back-up protection at (ab-g) on feeder between B-3 and B-4: primary protection at R-3
R-1 (radial network) (fault pattern recognition: radial network)

as shown in Fig. 14. It is observed that the differential


energy changes substantially in faulted phase compared to
unfaulted one, and thus, identifying the phase involved in
the fault process.

4.3 HIF detection in microgrid

HIFs usually occur in distribution networks and occur when a


primary conductor makes electrical contact with a road
surface, sidewalk, sod, tree limb, or with some other
surfaces or object, restricting the ow of fault current to a
level below that to be reliably detectable by conventional
overcurrent devices. The main purpose in HIF detection, in
contrary to short-circuit faults, is not to protect the system,
but to protect the human lives and prevent re hazards
Fig. 12 Differential energy and the threshold setting for LG fault because of the arcing phenomenon [24].
on the feeder between B-6 and B-7: back-up protection at R-5 (radial HIF are difcult to model as they are non-linearly
network) deterministic in nature. The HIF model [25] is developed
using anti-parallel diodes with nonlinear resistance and DC
source connected together for each phase as shown in
pattern recognition for LG fault (a-g) on feeder between B-2 Fig. 15. Typical HIF current for the HIF created on the
and B-3. Thus, the differential energy for a-phase grows in feeder between B-1 and B-2, is shown in Fig. 16. The HIFs
negative direction, whereas of phase-b and c is steady at are created on the distribution feeders and differential
base. Similar observation is made for LLG fault (ab-g) on energy is computed to detect the fault.
the feeder between B-3 and B-4 (primary protection at R-3),

Fig. 13 Differential energy of a, b and c-phase for LG fault (a-g)


on feeder between B-2 and B-3: primary protection at R-2 (fault
pattern recognition: radial network) Fig. 15 HIF model

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320 315
doi: 10.1049/iet-gtd.2013.0180 & The Institution of Engineering and Technology 2014
www.ietdl.org
4.4 Validation on standard IEC microgrid structure
The proposed method is also validated on the standard IEC
61850-7-420 microgrid [26] structure containing both
induction (wind) and inverter (PV)-based DGs as shown in
Fig. 18. The details of the IEC microgrid are mentioned in
Table 1 and the system is operating at 25 kV. Fig. 19a)
shows the differential energy-based fault pattern recognition
with LG fault situation (fault occurs between B-3 and B-4:

Fig. 16 Typical HIF current for the HIF created on the feeder
between B-1 and B-2

The HIF current and time-frequency contours of phase-a


current for HIF on the feeder between B-1 and B-2, are
shown in Fig. 17a (retrieved at B-1). Fig. 17b shows the
differential energy and the threshold setting for HIF on the
feeder between B-1 and B-2: primary protection at R-1
(radial network). In case of HIF faults, the threshold is set
at 0.1 and the tripping signal can be issued within 3.3
cycles after fault inception.

Fig. 18 Microgrid model according to IEC 61850-7-420

Table 1 Parameters of the IEC microgrid model


Transformer Voltage Capacity

TR1 12025 kV 20 MVA


TR2 575 V25 kV 10 MVA
TR3, TR4 230 V25 kV 2 MVA
DG1 (wind) 575 V 6 MW
DG2, DG3 (PV) 230 V 1 MW
load P (MW) Q (MVAR)
L1 2.0 0.6
L2 1.5 0.3
L3 1.5 0.3
L4 1.0 0.2
L5 1.0 0.2
distribution lines line paramaters
Fig. 17 Protection against HIF fault DL1, DL2, DL3, DL4, DL5 R0 = 0.1153 /km, R1 = 0.413 /
km, L0 = 1.05e3 H/km, L1 =
a Fault current and time-frequency contours of phase-a for HIF on the feeder 3.32e3 H/km, C0 = 11.33e009
between B-1 and B-2, retrieved at B-1 F/km, X1 = 5.01e009 F/km, 20
b Differential energy and the threshold setting for HIF on the feeder between km each
B-1 and B-2: primary protection at R-1 (radial network)

316 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0180
www.ietdl.org
current-based schemes with SNR 20 dB as depicted in
Table 2. It is observed that the conventional differential
current-based scheme fails in case of faults at islanded
mode and some cases (faults with high resistance of 100 )
on grid connected mode with loop structure. Thus, even if
the magnitude of differential current is sufciently low for
some fault cases to detect the faulty pattern, however, the
differential energy derived from the spectral energy content
of the faulted signals succeeds as it includes time-frequency
information in the transformed signal. In case of islanded
mode of microgrid operation, the fault current contribution
will be low containing more frequency distortions because
of possible active and reactive power mismatch in the
microgrid. Furthermore, the proposed differential
energy-based scheme using S-transform is immune to noise
[21] and thus, found to be robust compared with
conventional differential current-based scheme. It is also
found that the proposed differential energy-based scheme
effectively detects the faulty situations on the standard IEC
microgrid with both wind (induction based) and PV
(inverter based) penetration. This establishes the potential
ability of the proposed differential scheme providing
reliable protection measure for microgrid including different
types of DGs.

5 Discussion
In the proposed technique, the differential energy of the fault
currents at respective bus ends (depending on primary and
back-up protection), are used to register the fault patterns
Fig. 19 Differential energy based fault pattern recognition for issuing the tripping signal for reliable microgrid
a Differential energy of a, b and c-phase for L-G fault (a-g) on feeder between protection. It is observed that the differential energy grows
B-3 and B-4 (on DL-3): primary protection at R-5 (fault pattern recognition:
radial network)
in positive direction for faults in grid-connected mode and
b Differential energy of a, b and c-phase for LLLG fault (abc-g) on feeder becomes negative in islanded mode. This occurs as the fault
between B-5 and B-6 (on DL-5): primary protection at R-9 (fault pattern currents is large in grid-connected mode largely contributed
recognition: loop network) from utility grid compared with islanded mode and thus,
the currents at both end buses are signicantly different.
This clearly distinguishes the faults in microgrid, indicating
on DL-3) in radial network at grid connected mode. The fault whether the fault exists in grid-connected or islanded mode.
pattern recognition at islanded mode for LLLG fault (fault In case of HIF, the fault current magnitude is small and
occurs between B-5 and B-6: on DL-5) with loop structure arcing in nature. Thus, the differential energy remains at
(both CB_LOOP1 and CB _LOOP2 are made ON) at lower value, requiring a lower threshold band compared
islanded mode is shown in Fig. 19b. The response time is with shunt fault situations.
almost same as compared with earlier cases. Further, a The proposed technique is also tested on similar fault
qualitative comparison has been made between the situations in radial and loop network providing reliable
proposed differential energy and conventional differential protection measure. The most important issue in the

Table 2 Comparison results between proposed differential energy and conventional differential current-based schemes (IEC
microgrid) with SNR 20 dB

Fault conditions Proposed differential energy scheme Existing differential current scheme

Faults Primary protection Back-up protection Primary protection Back-up protection

Faults at grid-connected mode


a-g fault on DL-1 (radial) fault detected (R-1) not valid fault detected (R-1) not valid
a-b fault on DL-2 (radial) fault detected (R-3) not valid fault detected (R-3) not valid
bc-g fault on DL-3 (loop) fault detected (R-5) fault detected (R-3) fault detected (R-5) fails (R-3)
abc-g fault on DL-5 (loop) fault detected (R-9) fault detected (R-7) fails (R-9) fails (R-7)
HIF on DL-3 (loop) fault detected (R-5) fault detected (R-3) fails (R-5) fails (R-3)
Faults at islanded mode
a-g fault on DL-1 (radial) fault detected (R-1) not valid fails (R-1) not valid
a-b fault on DL-2 (radial) fault detected (R-3) not valid fails (R-3) not valid
bc-g fault on DL-3 (loop) fault detected (R-5) fault detected (R-3) fails (R-5) fails (R-3)
abc-g fault on DL-5 (loop) fault detected (R-9) fault detected (R-7) fails (R-9) fails (R-7)
HIF on DL-3 (loop) fault detected (R-5) fault detected (R-3) fails (R-5) fails (R-3)

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320 317
doi: 10.1049/iet-gtd.2013.0180 & The Institution of Engineering and Technology 2014
www.ietdl.org
proposed technique is that it uses time-frequency contours of
the faulted currents to derive the differential energy, thus
involving both time and frequency information instead of
either one generally used in conventional relaying for
decision making process. The ow chart of the proposed
scheme is shown in Fig. 20 with experienced thresholds for
fault detections in different modes. The thresholds
generated and the results obtained are for the adopted
microgrid in the proposed study and based on the per unit
fault current signal. It is observed over the extensive study
that the threshold bands are apart from each other and thus,
not interfering for decision making process.
Normally the response time depends on the thresholds as it
decides the numbers of cycles (which intern represents the
time) from fault inception. In our study, the response time
is about 3 and 1/2 cycles (70 samples or 58.33 ms) from
the fault inception, which is highly suitable for relaying
strategies (45 cycles for high-speed relays), meeting the
grid code requirements. The response time vary marginally
as the fault cases are considered under variations in types of
faults with fault resistance, fault location, DG penetration,
fault inception loading parameters in the microgrid. It has
been observed that the response time of the proposed
scheme is almost same for grid connected as well as
islanded mode compared with existing work [27], where the
response time differs signicantly depending on the grid

Fig. 21 Differential energy with sampling time difference


a Differential energy with sampling time difference of 5 ms between two ends
(a-phase for LG fault (a-g) on feeder between B-2 and B-3)
b Differential energy (providing delayed response) with sampling time
difference of 10 ms between two ends (b-phase for LLG fault (b-g) on
Fig. 20 Flow chart of the proposed technique feeder between B-2 and B-3)

318 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0180
www.ietdl.org
connected and islanded mode of operation. The response time 2 Lasseter, R.H.: Microgrids. Proc. IEEE Power Eng. Soc. Winter
can further decrease if the threshold setting is selected at Meeting, January 2002, pp. 305308
3 Girgis, A., Brahma, S.M.: Effect of distributed generation on protective
different level. However, the selected thresholds are opted device coordination in distribution system, system. Proc. Large
for more secured and reliable decision making for effective Engineering System Conf. Power Engineering, July 2001, pp. 115119
microgrid protection. Further, the thresholds can be 4 Green, T.C., McDonald, J.D.: Modeling and analysis of fault behavior
adaptive depending on the microgrid structure, operating of inverter microgrids to aid future fault detection. IEEE Int. Conf.
conditions, fault situations etc. System of Systems Engineering, 2007, pp. 16
5 Mahat, P., Chen, Z., Bak-Jensen, B., Bak, C.L.: A simple adaptive over
It is observed that the differential energy is less sensitive to current protection of distributed systems with distributed generations,
synchronisation error compared with time-domain data IEEE Trans. Smartgrid, 2011, 2, (3), pp. 428437
difference such as differential currents. The differential 6 Brahma, S.M., Girgis, A.: Development of adaptive protection scheme
energy works ne with a time-sampling difference of 5 ms for distribution systems with high penetration of distributed generation,
IEEE Trans. Power Deliv., 2004, 19, (1), pp. 5663
(quarter cycle) as shown in Fig. 21a, with 3.7 cycle 7 Wan, H., Li, K.K., Wong, K.P.: A multi-agent approach to protection
response time as usual. As the time difference between two relay coordination with distributed generators in industrial power
end signals increases to 10 ms, and then the response time distribution system. Proc. Ind. Appl. Conf. 40th Ind. Appl. Soc.
increases to 4.5 cycles (which is delayed) from fault Annual Meeting, October 2005, pp. 830836
inception, as shown in Fig. 21b. Thus, the differential 8 Hadzi-Kostova, B., Styczynski, Z.: Network protection in distribution
systems with dispersed Generation. Proc. Transmission and
energy can suitably work for time difference up to 5 ms Distribution Conf. Exhibition, May 2006, pp. 16
between two end signals without affecting the performance. 9 Zamani, M.A., Sidhu, T.S., Yazdani, A.: A strategy for protection
This indicates that the energy variable is less sensitive to coordination in radial distribution networks with distributed
time error: for instance, the energy of one cycle sine and generators. Proc. IEEE Power Energy Society General Meeting, July
cosine is the same, even though they are time-shifted by a 2010, pp. 18
10 Nikkhajoei, H., Lasseter, R.: Microgrid protection. Proc. IEEE Power
quarter of cycle. Thus, the proposed S-transform-based Engineering Society General Meeting, June 2007, pp. 16
differential energy is highly effective in detecting faults and 11 Tumilty, R.M., Brucoli, M., Green, T.C.: Approaches to network
fault patterns in microgrids, as the tripping signal can be protection for inverter dominated electrical distribution systems. Proc.
issued within four cycles from the fault inception for the set Third IET Int. Conf. Power Electronics, Machines and Drives, April
2006, pp. 622626
threshold, considering different faulty situations and 12 Al-Nasseri, H., Redfern, M., Li, F.: A voltage based protection for
operating conditions. Furthermore, the proposed differential microgrids containing power electronic converters. Proc. IEEE Power
energy-based technique is validated for faults on the IEC Engineering Society General Meeting, June 2006, pp. 17
microgrid structure including both wind- and PV-based 13 Loix, T., Wijnhoven, T., Deconinck, G.: Protection of microgrids with a
DGs and, the performance is similar to that of the earlier high penetration of Inverter-coupled energy sources. Proc. IEEE Power
Energy Soc./CIGRE Symp., June 2009, pp. 16
studied system with wind-based DGs only. It is observed 14 Nikkhajoei, H., Lasseter, R.H.: Microgrid protection. IEEE Power
that even if fault current is further limited in inverter-based Engineering Society General Meeting, June 2007, pp. 16
DG (PV) compared with induction-based DG (wind), the 15 Redfern, M.A., Al-Nasseri, H.: Protection of microgrids dominated by
proposed differential energy detects the fault patterns distributed generation using solid state converters. IET Ninth Int. Conf.
effectively, providing a reliable protection measure. on Develop. Power Syst. Protection, 2008, pp. 670674
16 Zeineldin, H.H., El-Saadany, E.F., Salama, M.M.A.: Distributed
generation microgrid operation: control and protection. Proc. Power
Syst. Conf., March 2006, pp. 105112
6 Conclusions 17 Sortomme, E., Venkata, S.S., Mitra, J.: Microgrid protection using
communication-assisted digital relays, IEEE Trans. Power Deliv.,
A new approach for microgrid protection using differential 2012, 25, (4), pp. 27892796
energy based on time-frequency transform is presented. 18 Mansinha, L., Stockwell, R.G., Lowe, R.P.: Pattern analysis with two
Spectral energy contents of the fault current signals dimensional spectral localization: application of two dimensional
S-transforms, Physica A, 1997, 239, pp. 286295
retrieved at both ends of the feeders are found out using 19 Pinnegar, C.R., Mansinha, L.: The S-transform with windows of
time-frequency transform and differential energy is arbitrary and varying window, Geophysics, 2003, 68, pp. 381385
computed to register the fault patterns. A threshold can be 20 Stockwell, R.G., Mansinha, L., Lowe, R.P.: Localization of complex
set on the differential energy to issue the tripping signal for spectrum: the S-transform, J. Assoc. Expl. Geophys., 1996, 17, (3),
different faulty situations in the microgrid (both radial and pp. 99114
21 Dash, P.K., Samantaray, S.R., Panda, G., Panigrahi, B.K.:
loop structure) at grid-connected and islanded mode. The Time-frequency transform approach for the protection of parallel
extensive test results indicate that the proposed scheme can transmission lines, IET Gener., Transm. Distrib., 2007, 1, (1),
reliably protect the microgrid against shunt faults and HIFs pp. 3038
in different topology and thus, can be extended for large 22 Yang, H., Liao, C.: A de-noising scheme for enhancing wavelet-based
power quality monitoring system, IEEE. Trans. Power Deliv., 2001, 16,
power distribution network with multiple DG interface. (3), pp. 353360
23 Samantaray, S.R., Tripathy, L.N., Dash, P.K.: Differential energy based
relaying for Thyristor controlled series compensated line, Int. J. Electr.
7 Acknowledgment Power Energy Syst., Elsevier Sci., 2012, 43, (1), pp. 621629
24 High Impedance Fault Detection Technology: Report of IEEE PSRC
The authors acknowledge the funds received from Central Working Group D15, March 1996. Available at http://grouper.ieee.org/
Power Research Institute, Bangalore, Government of India, groups/td/dist/documents/highz.pdf
25 Samantaray, S.R., Panigrahi, B.K., Dash, P.K.: High impedance fault
through the project RSOP-IITB-028 to carry out the detection in power distribution networks using time-frequency
research work. transform and probabilistic neural network, IET Gener., Transm.
Distrib., 2008, 2, (2), pp. 261270
26 Ustun, T.S., Ozansoy, C., Zayegh, A.: Modeling of a centralized
8 References microgrid protection system and distributed energy resources
according to IEC 618507-420, IEEE Trans. Power Syst., 2012, 27,
1 Stadler, M., Siddiqui, A., Marnay, C., Aki, H., Lai, J.: Control of (3), pp. 15601567
greenhouse gas emissions by optimal DER technology investment and 27 Zamani, M.A., Sidhu, T.S., Yazdani, A.: A protection strategy and
energy management in zero-net-energy buildings, Eur. Trans. Electr. microprocessor based relay for low voltage microgrid, IEEE Trans.
Power, 2009, 2, (2), pp. 12911309 Power Deliv., 2011, 26, (3), pp. 18731883

IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320 319
doi: 10.1049/iet-gtd.2013.0180 & The Institution of Engineering and Technology 2014
www.ietdl.org
9 Appendix Distribution lines (DL): DL-1, DL-2, DL-3, DL-4, DL-5,
DL-6: PI-section, 20 km each, rated kV = 25, rated MVA =
Generator: Rated short-circuit MVA = 1000, f = 60 Hz, 20, Vbase = 25 kV, R0 = 0.1153 /km, R1 = 0.413 /km, L0
rated kV = 120, Vbase = 120 kV. = 1.05e3 H/km, L1 = 3.32e3 H/km, C0 = 11.33e009 F/
Distributed generations (DGs): Wind farm (9 MW) km, X1 = 5.01e009 F/km.
consisting of six 1.5-MW wind turbines is connected to a Normal loading data:
25-kV distribution system exports power to a 120-kV grid L-1 = 10 MW, 3.5 MVAR
through a 30-km, 25-kV feeder. The doubly fed induction L-2, L-3, L-4, L-5 = 6 MW, 2.5 MVAR.
generator (DFIG) has been considered for the proposed study. HIF model:
Transformer T1: Rated MVA = 50, f = 60 Hz, rated kV =
120/25, Vbase = 25 kV, R1 = 0.00375 pu, X1 = 0.1 pu, Rm = Diode: resitance Ron = 0.001 , forward voltage = 0.8
500 pu, Xm = 500 pu. V, snubber resistance = 500 , snubber capacitance = 250e
Transformer T2, T3, T4, T5: Rated MVA = 10, f = 60 Hz, 9 F.
rated kV = 575 V/ 25 kV, Vbase = 25 kV, R1 = 0.00375 pu, Nonlinear resistance: protection voltage 45e3 V.
X1 = 0.1 pu, Rm = 500 pu, Xm = 500 pu. DC voltage source: 15e3 V.

320 IET Gener. Transm. Distrib., 2014, Vol. 8, Iss. 2, pp. 310320
& The Institution of Engineering and Technology 2014 doi: 10.1049/iet-gtd.2013.0180

You might also like