IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO.

3, JULY 2014

747

Three-Phase Fault Direction Identication for

Distribution Systems With DFIG-Based Wind DG
Ali Hooshyar, Student Member, IEEE, Maher Abdelkhalek Azzouz, Student Member, IEEE, and
Ehab F. El-Saadany, Senior Member, IEEE

AbstractDistributed generation (DG) integration necessitates

upgrading some distribution system overcurrent relays to directional
ones to offer selective protection. The directional feature is conventionally achieved by phase angle comparison between phasors of
the fault current and a polarizing quantity, normally a voltage signal.
Doubly fed induction generator (DFIG)-based wind turbines constitute an appreciable portion of todays DG power. This paper unveils
that conventional directional elements malfunction during threephase short-circuits when a distribution system incorporates
DFIG-based wind DG. The maloperation is due to the exclusive fault
behavior of DFIGs, which affects the existing relaying practices. The
paper also proposes a fault current classication technique that
replaces the conventional directional element during problematic
conditions and provides accurate fault direction quickly based on
waveshape properties of the current. An extensive performance
evaluation using PSCAD/EMTDC simulation of the IEEE 34 bus
system corroborates the effectiveness of the proposed method.
Results are exceptionally encouraging in the case of resistive crowbar
circuits for DFIGs, which is the typical scenario in practice.
Index TermsDirectional relaying, distribution system, doubly
fed induction generator (DFIG), three-phase fault current, windbased distributed generation (DG).

I. INTRODUCTION
NSTALLATION of relatively small power sources embedded inside distribution networks, generally referred to as
distributed generation (DG), has extensively expanded in the last
decade and is expected to play a major role in future power
systems, as it enhances system reliability, reduces losses, and
shaves peak power, among other benets [1]. The spread of DG
also facilitates easier integration of renewable energy sources
such as wind power. Induction generators (IGs) are one of the
main technologies deployed in wind-based DGs, particularly
doubly fed IG (DFIG), which offer variable speed operation
accompanied by moderate converter sizes and hold the largest
market share for the currently installed wind capacity [2].
Despite various merits of DGs, their growth presents several
challenges to distribution system operation. A substantial portion
of these challenges appears as more rigorous requirements or
further complications for the existing distribution system protective relaying. For instance, DGs can cause mis-coordination
between overcurrent relays by increasing the fault current level,

Manuscript received September 30, 2013; revised December 02, 2013;

accepted January 02, 2014. Date of publication February 19, 2014; date of
current version June 17, 2014.
The authors are with the Department of Electrical and Computer Engineering,
University of Waterloo, Waterloo, ON N2L 3G1, Canada (e-mail: alihooshyar@
gmail.com; mazzouz@uwaterloo.ca; ehab@uwaterloo.ca).
Color versions of one or more of the gures in this paper are available online at
http://ieeexplore.ieee.org.
Digital Object Identier 10.1109/TSTE.2014.2298466

or affect the sensitivity of the protection system at certain locations

by reducing the fault current drawn from the substation or result in
fuse nuisance blowing [3]. Research into adverse impacts of DGs
on relaying systems and possible solutions is ongoing. For
example, several publications have addressed such issues by
means of communication networks, which have been reinforced
in todays power systems due to the smart grid initiative [4], [5].
Distribution systems have normally been composed of radial
feeders supplied from only one end, as a result of which the fault
currents have been unidirectional. However, DG introduction has
led to fault currents that ow bidirectionally in modern distribution
grids, thus requiring directional overcurrent relays to ensure
selectivity of the protection system [6]. Directional overcurrent
relays commonly operate according to the angle difference between phasors of the fault current and a polarizing quantity, which
is the zero-sequence, a line or a memorized voltage signal [7].
The above-mentioned protection issues can arise with almost all
DG types. Meanwhile, some relaying problems are exclusively
associated with specic DGs, such as wind-based ones. The
relaying difculties caused by wind-based DGs have been exacerbated by the new grid codes that require the wind facilities to remain
connected to the system during disturbances [8]. As a result of
these grid codes, particular fault properties of wind facilities must
be taken into consideration by modern protective relays. Therefore,
some relay manufacturers have recently initiated developing relays
that are specically intended for the protection of systems with
wind-based sources [9]. Further studies and investigations in this
area by the research community are necessary to ensure these early
steps taken by the industry side reach a mature level.
This paper reveals a serious drawback that results in spurious
operation of directional relays during three-phase faults when a
DFIG-based DG is connected to the distribution system. It is
shown that the conventional approach of nding the phase angle
between the fault current and a voltage signal fails to provide the
correct fault direction. The paper also proposes a solution based on
discrimination between DFIG and substation short-circuit currents
according to their waveshape properties. An extensive performance evaluation is carried out using the IEEE 34 bus system to
verify the efcacy of the proposed method. Successful operation
of the proposed method is actually signicantly boosted by the
crowbar resistance, normally used for DFIG units in practice.
II. PROBLEM DESCRIPTION
A. DFIG Fault Behavior
The fault behavior of a squirrel cage IG (SCIG) has been
analyzed in various publications, such as [10][12]. It is shown

1949-3029 2014 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.

748

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 3, JULY 2014

that the SCIG current after a bolted three-phase short-circuits at

the machine terminal is expressed by (1) [12]



Vm
1
1
0

if t

et=T cos1  s!t 
1  s X 0 Xs Xms
Vm
1

 0 coset=Ts
(1)
1s X
where Vm is the pre-fault voltage amplitude, s is the machine slip,
X 0 is the machine transient reactance, Xs is the stator leakage
reactance, Xms is the stator magnetizing reactance, T 0 is the
short-circuit transient time constant, which is inversely related to
the rotor resistance, ! 260 rad/s is the fundamental angular
frequency,  is the fault inception angle, and Ts is the stator time
constant.
Although (1) is for a bolted fault at the SCIG terminal, it
describes the essential pattern of actual power system fault
currents fed by SCIGs as well. The only difference comes from
the fact that (1) is derived based on a 100% voltage dip
assumption, whereas actual power system faults always include
some levels of fault resistance, which leads to nonzero fault
voltage and keeps the IG from complete demagnetization. As a
result, unlike the current of (1), which eventually declines to zero,
SCIGs can sustain a small fault current in practice.
In the event of a voltage dip, large overcurrents are generated
in the rotor winding of an IG. In a DFIG, these overcurrents can
damage the rotor-side converter. Several methods, commonly
known as fault ride-through (FRT) techniques, have been proposed in the literature to protect the converters during such
conditions and enable the wind facility to meet the previously
discussed new grid code requirements [13]. Meanwhile, the
actual FRT practice in industrial applications remains the traditional crowbar circuit method, which diverts the rotor currents
through a short-circuit path, or a simple resistor [14]. Threephase short-circuits, the focus of this study, are among the most
severe faults that cause signicant voltage drops and subsequent
crowbar activation. When the rotor circuit is shorted and the
converters are eliminated from the gure, a DFIG is in essence an
SCIG and exhibits similar fault behavior.
Although the SCIG fault current expression of (1) ts DFIG
currents after three-phase short-circuits fairly well, a number of
its parameters may have different values for DFIGs. The most
critical of these parameters with respect to directional relaying is
the machine slip. The SCIG rotor speed is conned to a narrow
range close to the synchronous speed. Hence, the slip becomes
very small and the (1  s) term, which is multiplied with the
frequency of the sinusoidal term in (1), can be neglected. A
DFIG, however, accommodates different wind speeds by allowing signicant rotor speed deviations around the synchronous
speed, resulting in a 30% band for the slip. Consequently,
(1  s) is no longer a negligible term in (1) and the fault current
fundamental frequency ranges from 42 to 78 Hz for a 60-Hz
system.
To illustrate this issue, the simple test system in Fig. 1, which
includes a DFIG-based DG, was simulated with the PSCAD/
EMTDC program. The crowbar resistance of the DFIG is zero at
this stage. The effect of nonzero crowbar resistance is studied
later on. At t 7 s, a three-phase fault is placed on bus 4, i.e., the

Fig. 1. Layout of a simple distribution system with a DFIG-based DG.

Fig. 2. Three-phase DFIG fault current for two sub- and super-synchronous rotor
speeds.

feeder without the DG. The phase A current recorded by the relay
at the beginning of the feeder with the DG is plotted in Fig. 2 for a
sub- and a super-synchronous rotor speed, resulting in positive
and negative slips, respectively. The depicted waveshapes agree
with the general fault current pattern described in (1). In addition,
the dependence of the current frequency on the slip is evident.
B. Directional Relaying in Presence of DFIGs
The relay installed at the beginning of the feeder with the DG
in Fig. 1, denoted by DiR12, has to be directional in order to avoid
unnecessary tripping as a result of the DG contribution to the
faults on the other feeder. Following the common directional
relaying practice for three-phase short-circuits implemented in
commercial digital directional elements, such as in [15] and [16],
DiR12 is memory-polarized, i.e., the angle of each phase current
is compared with that of the pre-fault voltage in the same phase.
The results of this study are independent of the directional relay
characteristic angle (RCA); so, for the sake of simplicity, the
RCA is chosen to be zero degrees. DiR12 measures the voltage
phasor using discrete Fourier transform (DFT). Due to the unique
properties of a three-phase DFIG fault current, the performance
of DiR12 is inspected by three different current phasor measurement techniques (PMTs) developed so as to suppress the impact
of the fault currents decaying dc offset on the measured phasor:
1) the 1.25 cycles cosine ltering (CF) technique [17];
2) the one-cycle least error square (LES)-based technique
developed in [18], referred to as conventional LES (CLES)
in this paper;
3) the one-cycle-modied LES (MLES) method, discussed in
the Appendix.
The rst two techniques are termed common PMTs in this
paper, as they are widely used in modern commercial relays [17].

HOOSHYAR et al.: THREE-PHASE FAULT DIRECTION IDENTIFICATION FOR DISTRIBUTION SYSTEMS WITH DFIG-BASED WIND DG

749

Fig. 3. Phase difference between the memorized voltage and fault current for the
fault in Fig. 2, considering voltage frequency for all phasor measurements.
(a) s 24% and (b) s 20%.

Fig. 4. Phase difference between the memorized voltage and fault current for the
fault in Fig. 2, considering voltage and current frequencies for their respective
phasor measurements. (a) s 24% and (b) s 20%.

The MLES method is a modied version of the CLES technique

and assumes exponentially decaying amplitude for the fundamental component, which is the case for DFIG short-circuit
currents.
Apart from the sampling rate, which is 3840 Hz in this paper,
the frequency of the signal under study is also required to build up
the digital lters associated with the above PMTs. The main
frequency tracking approach used in modern industrial relays,
such as in [19] and [20], is to nd the voltage frequency and
update all of the relay subroutines, including the digital lters that
compute the current phasor.
The same frequency tracking approach is adopted to analyze
the response of the common directional relaying practice based
on the current and voltage phase angle difference. Fig. 3 plots
the angle difference between the phase A-memorized voltage
and the current recorded by DiR12 during the fault depicted in
Fig. 2 for both the sub- and super-synchronous rotor speeds
using the three current PMTs discussed above. The positive
direction for the current recorded by DiR12 is assumed to be
from bus 1 to bus 2. Prior to the fault, the power ows from the
substation to bus 2 and the load connected to bus 3. Thus, the
pre-fault phase difference between the voltage and the current
of DiR12 is correctly around zero, inside the [90, 90] degree
range. After the fault, however, the current measured at the
DiR12 location is the DFIG contribution to the fault on bus 4. In
other words, the current direction is reversed and ows toward
bus 1, so the phase difference has to normally exit and remain
outside the [90, 90] degree interval following the initial
post-fault transient period of the PMTs in order to indicate the
backward direction of the fault. However, it is shown that the
measured phase difference starts oscillating inside and outside
the [90, 90] degree range for both rotor speeds, regardless of

the PMT employed, as a result of which DiR12 operation is

utterly unreliable.
These phase oscillations occur because the current frequency
is quite different from the 60-Hz frequency of voltage, which is
used to build the digital lters that compute the current phasor.
Hence, there exists an apparent discrepancy between the actual
fundamental frequency component of current and the measured
one, which is oscillatory both in magnitude and phase angle.
The above problem cannot be addressed simply by using the
actual current frequency to develop the digital lters of the
current PMTs. To demonstrate this issue, the phase differences
shown in Fig. 4 for the above fault conditions are obtained by
considering the current and voltage frequencies for calculating
the current and voltage phasors, respectively. Again, the phase
differences for both sub- and super-synchronous cases oscillate
inside and outside the [90, 90] degree interval, irrespective of
the PMT utilized, periodically indicating backward and forward
directions for the fault location. The oscillating phase differences
stem from the fact that the measured voltage and current phasors
correspond to different frequencies, thereby rotating at different
speeds and making it theoretically impossible to nd a meaningful and xed phase difference between them.

III. PROPOSED SOLUTION

A. Proposed Directional Relaying Scheme
It was demonstrated that the common directional relaying
based on comparing the phase difference between voltage and
current signals fails to detect the fault direction during threephase short-circuits when a distribution system includes a DFIGbased DG. This section proposes a new approach to detect the

750

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 3, JULY 2014

Fig. 5. Flowchart of the proposed scheme.

fault direction for such conditions according to the waveshape

properties of fault current. The waveshapes of DFIG three-phase
fault currents include unique features that can distinguish them
from the fault currents owing from the distribution substation.
Fig. 5 illustrates the owchart of the proposed scheme. A
typical disturbance detector (DD) unit that operates based on
the cycle-to-cycle comparison of current samples is similar to
the one discussed in [21] triggers this scheme. If the fault type is
three phase, the waveshape recognition technique, discussed in
Section III-B, will determine whether the current possesses the
distinctive waveshape attributes of DFIG fault currents. The
ow of a DFIG fault current at the relay location signies that
the fault is between the distribution substation and the relay
location, and the fault direction is determined accordingly. In
contrast, the absence of waveshape properties of DFIG fault
currents demonstrates that the current recorded by the relay is a
substation fault current and the fault location is between the
relay location and the DFIG-based DG. The specied direction
is then exported to the overcurrent element to make the nal
tripping decision.
B. Proposed Waveshape Recognition Technique
The aggregate transmission systems connected to the distribution substation have been conventionally modeled by a
Thevenin equivalent circuit for fault studies. Such modeling,
whose effectiveness is backed by several decades of successful
industrial practice [22], results in fault currents composed of a
constant sinusoidal component, a decaying dc offset with an
equal or smaller magnitude, harmonics, and noise. Neglecting
harmonics and noise, whose effects will be taken into account
later on, the fault current can be expressed as
if t Im sin!t   Im sinet=

(2)

where  is the decaying dc time constant and the fault angle 

determines the dc offsets initial magnitude. The previously
discussed PMTs effectively suppress the decaying dc offset and

Fig. 6. Comparison of the phasor measured for substation and DFIG fault
currents. (a) Current of (2) for  90 ,  40 ms, and Im 1. (b) Current
of (1) for identical ac and dc components, with 40 and 80 ms time constants,
respectively.

provide the accurate fundamental frequency phasor. For example, the current described by (2) is plotted in Fig. 6(a) for
 90 ,  40 ms, and Im 1. Fig. 6 also displays the
magnitude of the phasors obtained by the three current PMTs
mentioned earlier. For all three techniques, the calculated
phasor remains virtually xed at a very narrow margin of the
fundamental components amplitude following their initial
response time.
A DFIG fault current and its measured phasor, however,
exhibit a different pattern. As a representative case, Fig. 6(b)
shows the current described by (1) for identical unity ac and dc
component amplitudes and for dc and ac time constants equal to
40 and 80 ms, respectively. The ac component frequency is
60 Hz. The above PMTs are again employed to nd the current
phasor. The decaying nature of the ac component is directly
reected in the measured phasors. The declining phasors in
Fig. 6(b) are in clear contrast with the uniform pattern observed
for the phasors plotted in Fig. 6(a).
As dened by (3), 1 calculates the relative percentage
decline in the magnitude of the measured phasor for the fault
current in one phase:


jIf1 j  jIf2 j if -ext 
 100
(3)

1
jIf2 j
If2 
where If1 is the maximum of the measured fault current phasor
within the rst half-cycle after the initial response time of the
employed PMT. If2 is the minimum magnitude of the measured
fault current phasor in the second half-cycle after the response
time of the PMT applied to the current. if -max is the largest
magnitude for the current samples prior to If2 location and 4 ms
after the DD activation. if -max is supposed to be the extremum

HOOSHYAR et al.: THREE-PHASE FAULT DIRECTION IDENTIFICATION FOR DISTRIBUTION SYSTEMS WITH DFIG-BASED WIND DG

PARAMETERS

OF

751

TABLE I
(3) FOR THE CURRENTS SHOWN IN FIG. 6

associated with the fundamental component. The 4-ms delay

avoids mistaking the initial current extremums that might be
generated due to the resonance for the extremum associated with
the fundamental component.
If1 and If2 obtained by different PMTs along with if -max and
the resultant 1 are tabulated in Table I for the two currents
shown in Fig. 6. The almost uniform phasor measured for the
current in Fig. 6(a) has led to very small 1 s, whereas the indices
obtained for Fig. 6(b) are large due to the decaying trend of DFIG
fault current. In other words, 1 is capable of differentiating
between the fault currents owing from the distribution substation and a DFIG-based DG. The gap between 1 for substation
and DFIG fault currents is widened by the second ratio on the
right side of (3), as the dc offset of a substation fault current never
exceeds its fundamental components amplitude, which is not the
case for DFIG fault currents.
The choice of PMT for nding the parameters required by (3)
affects the measured 1 . The MLES technique tailors the
CLES technique to cope with a decaying ac component and
provides excellent results for both of the synthetic currents
depicted in Fig. 6. However, the gain responses of the sine and
cosine digital lters associated with the MLES technique,
displayed in Fig. 7(a), indicate that this technique is extremely
prone to fault current high-frequency contaminations. Harmonics are not considered for developing the lters whose frequency responses are shown in Fig. 7(a), i.e., Nh , discussed in
the Appendix, is one. Considering harmonics reduces the lter
gain for the harmonics to zero, but further deteriorates the
frequency responses for the nonharmonic components, which
are quite likely to be present in the fault current due to noise,
resonance, etc. The frequency responses of the CLES technique
sine and cosine lters, plotted in Fig. 7(b), illustrate that
common PMTs do not suffer from this deciency.
On the other hand, some oscillations are superimposed on the
phasors measured by the common PMTs when the fundamental
magnitude of current is variable. Such currents include nonharmonic components in their spectrum. However, as observed
in Fig. 8, which displays the normalized spectrum of the current
in Fig. 6(b), the resultant components are mainly situated in the
low-frequency region and are thereby damped by the sine and
cosine lters associated with the common PMTs. As shown
in Fig. 7(b), even for the higher frequency components, except
for the cosine lters response inside a limited interval, the
CLES lters gains are lower than unity. As a result, the
mentioned oscillations in the phasors provided by the common
PMTs are relatively small and do not obscure the phasors
overall decaying pattern for DFIG fault currents. Among the

Fig. 7. Frequency response of the sine and cosine digital lters. (a) MLES
method. (b) CLES method.

Fig. 8. Normalized spectrum of the current shown in Fig. 6(b).

common PMTs, the CLES technique is chosen for its faster

speed than the CF method.
3 is dened as the sum of 1 for the three-phase currents.
Because of the quite large value of 1 , and subsequently of 3 ,
for DFIG fault currents, a xed thresholding approach is applied
to classify fault currents. The threshold setting criterion is
founded upon the maximum indices obtained for the fault
currents described by (3). The set of three-phase currents expressed by (3) is obtained by adding 120 to . In (3),  and 
were varied in the ranges of [180, 180] degree and [1, 100] ms,
respectively, and the phasors for each pair of  and  were
inspected. The results, plotted in Fig. 9, show that 1 of phase
A and 3 reach 6.17% and 9.81% at their maximum, respectively.
Considering a 100% plus safety factor to address any effect of
noise and other transients, 3 20% is selected as the threshold for fault current classication.
IV. SIMULATION RESULTS
The performance of the proposed method was comprehensively studied for the IEEE 34 bus system, simulated in the

752

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 3, JULY 2014

PERFORMANCE

OF THE

TABLE II
PROPOSED METHOD

FOR

DIFFERENT FAULTS

a
b

Toward the substation.

Toward the DG.

Fig. 9. Indices obtained for the currents described by (2) with respect to the fault
inception angle and dc offset time constant. (a) 1 and (b) 3 .

Fig. 10. IEEE 34 bus system with a DFIG-based DG.

PSCAD/EMTDC program, as a benchmark distribution system

model, whose details can be found in [23]. The simulated
currents were exported to MATLAB, where the proposed method was implemented. The test system as well as the specications
of a DFIG-based DG connected to bus 848 on the far right end of
the system are displayed in Fig. 10. The control mechanism for
the DFIG converters along with the wind turbine specication
and the pitch angle control system are based on the industrial
models of [24]. After the DG is connected, the feeders shown by a
black solid line must be protected by directional relays in order to
ensure selectivity of the protection system.
Fault direction was determined using the proposed method for
a variety of conditions. The tests included different fault locations, wind speeds, fault inception angles, crowbar resistances,
etc. and were conducted for different relay locations. The results
were promising and corroborated the efcacy of the proposed
method. A part of these results that corresponds to zero crowbar
resistance is reported in Table II. Some of the performed tests are
graphically presented in this section. First, two case studies that

Fig. 11. Currents recorded at bus 844 for a fault on bus 842. (a) Phase A.
(b) Phase B. (c) Phase C.

include different fault locations and wind speeds are reviewed.

Then, the effect of crowbar resistance is analyzed, and nally, the
operation of the proposed method for a fault current owing from
the substation is demonstrated.

HOOSHYAR et al.: THREE-PHASE FAULT DIRECTION IDENTIFICATION FOR DISTRIBUTION SYSTEMS WITH DFIG-BASED WIND DG

753

Fig. 12. Currents recorded at bus 832 for a fault on bus 802. (a) Phase A.
(b) Phase B. (c) Phase C.

Fig. 13. Currents recorded at bus 842 for a fault on bus 860 considering nonzero
resistance for the crowbar circuit. (a) Phase A. (b) Phase B. (c) Phase C.

A. Case Studies

Therefore, the decline rate of the phasors in Fig. 12 has reduced,

compared to those of the previously discussed fault currents.
Nevertheless, 1 is still as large as 17.48%, 14.78%, and 9.6%
for phases A, B, and C, respectively, adding up to 3 41:86%,
which is more than twice the 20% threshold, and can reliably
identify the DFIG type of the fault current.

1) Fault on Bus 842: A three-phase fault is placed on bus 842 at

t 9 s when the DFIG slip is 6:7% and the crowbar resistance
is zero. The three-phase currents recorded at bus 844 are depicted
in Fig. 11. A relay at bus 844 must classify the current as a DFIG
fault current and, in turn, identify the fault direction toward the
distribution substation, i.e., bus 800. The decaying pattern of the
ac component and the measured phasor is obvious for all phases,
and 1 equals 39.27%, 22.58%, and 18.69% for phases A, B,
and C, respectively, which is fairly above the maximum 1
found in Fig. 9(a). As a result, 3 becomes 80.54%, i.e., more
than four times larger than its threshold, and the current is
correctly labeled as a DFIG one.
2) Fault on Bus 802: The current recorded by the relay located
at bus 832 for a fault on bus 802 at t 9 s is shown in Fig. 12.
Due to increased wind speed, the DFIG slip at the fault instant is
20%. Here again, the crowbar resistance of the DFIG is zero.
Unlike the last-studied fault, which was relatively close to the
DG, this fault is located farthest from the DFIG, and therefore the
short-circuit is not severe, leaving higher voltage at the DG
terminal and a lower level of demagnetization for the DFIG.

B. Effect of Crowbar Resistance

For all of the faults studied so far, the crowbar resistance was
zero. However, in practice, some resistance is added to the
crowbar circuit, as it improves the DFIG operation during
disturbances by decreasing the short-circuit current level,
dampening the generators torsional oscillations, etc. [12].
Fig. 13 shows the currents recorded at bus 842 for a fault on
bus 860, while the machine slip is 18% and the crowbar
resistance is 0.1 p.u. The decaying rate of the sinusoidal
component in Fig. 13 is signicantly larger than that of the
previously shown fault currents, as a result of which 1 has
grown to 372.11%, 198.21%, and 188.37% for phases A, B, and
C, respectively, and 3 equals 758.69%, which is considerably
in excess of the assigned threshold.

754

Fig. 14. Currents recorded at bus 852 for a fault on bus 860. (a) Phase A.
(b) Phase B. (c) Phase C.

The improved performance of the proposed method when

resistance is inserted into the crowbar path was expected. The
crowbar resistance raises the effective resistance of the rotor
circuit. Thus, the short-circuit transient time constant, i.e., T 0 in
(1), is decreased and the amplitude of the ac component declines
faster. Consequently, 3 grows excessively and the proposed
method performs exceptionally well for the more commonly
observed shape of DFIG fault currents.
C. Substation Fault Currents
The almost constant ac magnitude for the substation shortcircuit currents leads to very small values for 3 . For the fault
studied in Section IV-B, the current owing through bus 852,
which originates from the substation, is shown in Fig. 14. The
measured phasors exhibit constant magnitudes for all three
phases, and 1 is as small as 0.3241%, 4.9101%, and
0.1337% in phases A, B, and C, respectively, resulting in
3 5:37%, which is safely below the set threshold. Hence,
the proposed method identies the fault location to be between
the relay location and the DG. The proposed method was tested
for several other substation fault currents, and 3 did not reach
even half of its threshold.

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 3, JULY 2014

Fig. 15. Currents recorded by the relay at bus 854 for a fault next to the bus in the
downstream direction when an SG-based DG is behind the relay. (a) Phase A.
(b) Phase B. (c) Phase C.

D. Effect of Synchronous-Based DGs

Apart from DFIGs, the ac component of synchronous generator (SG) fault currents may also exhibit a slightly decaying nature
if the three-phase fault is electrically close to the SG terminal. As
a result, the fault current recorded by a relay located between the
fault and the distribution substation may have a decaying ac
component if an SG-based DG is located between that relay and
the substation. However, the decline in the ac component of such
fault currents will not be as high as those of the DFIG fault
currents. An SG three-phase fault current is composed of a
decaying dc, a steady-state, a transient, and a sub-transient ac
component. The rst two components are either dc or are
constant ac signals, and thereby do not affect 1 . The time
constant of the transient component is in the order of 1 s. Hence,
its variation inside the short window used by the proposed
method is negligible. Not only the sub-transient component is
one of the four components of an SG fault current, its relative
magnitude diminishes when the fault is not electrically close to
the SG terminal [12]. Even if the fault is next to the SG-based DG,
the impedance of the DG transformer increases the electrical
distance between the SG and the fault. Furthermore, the current

HOOSHYAR et al.: THREE-PHASE FAULT DIRECTION IDENTIFICATION FOR DISTRIBUTION SYSTEMS WITH DFIG-BASED WIND DG

measured at the relay location for the above-mentioned scenario

is the sum of the SG-based DG and the substation fault current,
whose magnitude is almost constant.
To test the performance of the proposed method for such
conditions, a 1.7-MW SG-based DG was connected to bus 854
through a transformer with a leakage impedance of 0.1 p.u. This
SG is large enough to cause about 60% reverse power ow at the
substation, which is the maximum allowed reverse power ow
for distribution systems in Ontario, Canada. A bold three-phase
fault is placed next to bus 854 in the downstream direction, which
is the worst-case scenario, as discussed above. The currents
recorded by a relay located at the same bus are shown in Fig. 15.
The ac component of the current does decrease, but very
gradually, because the SG fault current is mixed with the
substation fault current, and also the main decaying ac component of the SG fault current is the transient component, which has
a long time constant. 1 for phases A, B, and C is 7.19%, 3.69%,
and 2.88%, respectively, which add up to 3 13:76%. Therefore, the proposed method correctly determines that the fault is
between the relay location and the DFIG.
V. CONCLUSION
The frequency of memorized voltage that polarizes directional
elements during three-phase short-circuits is closely tied with the
nominal frequency. The fault current frequency for DFIG-based
wind DGs, however, may considerably deviate from the nominal
frequency depending on wind speed. On this basis, this paper
demonstrated the impossibility of nding fault direction through
the common method of phase angle comparison between the
current and voltage phasors. In addition, a novel method was
proposed to address this problem by classifying fault currents
according to their waveshape properties. The decaying pattern of
the ac component for DFIG fault currents proved to be a key
feature in discriminating such currents from the short-circuit
currents that originate from the distribution substation. The index
1 measured the percentage decline in the current phasor within
the two cycles after the fault inception. The sum of 1 for the
three phases was able to reliably classify fault currents. Operation
of the proposed method was tested for the IEEE 34 bus system.
Successful results were obtained regardless of the relay and fault
location. Particularly, promising performance was observed
when a resistance is placed along the DFIGs crowbar path,
which is the most likely situation in practice.
APPENDIX
The MLES PMT is based on the same approach as the CLES
technique [17], [18], with the only difference being that the
amplitude considered for the fundamental component of current
is not constant, but exponentially decaying, i.e., the nth sample of
the discretized current is described as
in I1 enT =ac sin!nT 1 Io enT =dc

Nh
X
h2

Ih sinh!nT h

(4)

755

where T is the sampling period, I1 , ac , !, and 1 are the

amplitude, time constant, angular frequency, and phase angle of
the fundamental component, respectively, Io and dc are the
initial magnitude and time constant of the decaying dc offset,
respectively, Ih and h are the amplitude and phase angle of the
hth harmonic respectively, and Nh is the number of harmonics
taken into consideration. The more generic relation of (4) ts the
pattern of DFIG fault currents as well. Rearranging (4) after
expressing the two exponentials by the rst two terms of their
Taylor series expansion and expanding the sinusoidal terms
using trigonometric identities yields

in

Nh h
X
Ih cosh sinh!nT
h1

Ih sinh cosh!nT

I1 cos1 =ac nT sin!nT

I1 sin1 =ac nT cos!nT
Io 1 Io =dc nT

(5)

where the underlined terms are the unknowns. The next steps use
(2) for the samples inside a window to form a matrix equation
whose LES-based solution provides the above unknowns and, in
turn, the current phasor. These steps are similar to the CLES
technique [17], [18].

REFERENCES
[1] R. Dugan and T. McDermott, Distributed generation, IEEE Ind. Appl.
Mag., vol. 8, no. 2, pp. 1925, Mar./Apr. 2002.
[2] A. D. Hansen, Generators and power electronics for wind turbines,
in Wind Power in Power Systems, 2nd ed. Hoboken, NJ, USA: WileyBlackwell, 2012, pp. 73104.
[3] T. Seegers, K. Birt, R. Beazer, M. Begovic, K. Behrendt, S. Boutilier et al.
(Aug. 2004). Impact of distributed resources on distribution relay protection. Report of working group D3 of the line protection subcommittee, IEEE
Power System Relaying Committee [Online]. Available: www.pes-psrc.
org/Reports/wgD3ImpactDR.pdf.
[4] H. J. Laaksonen, Protection principles for future microgrids, IEEE Trans.
Power Electron., vol. 25, no. 12, pp. 29102918, Dec. 2010.
[5] E. Sortomme, S. S. Venkata, and J. Mitra, Microgrid protection using
communication-assisted digital relays, IEEE Trans. Power Del., vol. 25,
no. 4, pp. 27892796, Oct. 2010.
[6] T. K. Abdel-Galil, A. E. B. Abu-Elanien, E. F. El-Saadany, A. Girgis,
Y. A.-R. I. Mohamed, M. M. A. Salama, and H. H. M. Zeineldin.
(Jun. 2007). Protection coordination planning with distributed generation.
National Resources Canada [Online]. Available: http://canmetenergy.nrcan.
gc.ca/sites/canmetenergy.nrcan.gc.ca/les/les/pubs/2007-149e.pdf.
[7] S. H. Horowitz and A. G. Phadke, Nonpilot overcurrent protection of
transmission lines, in Power System Relaying, 3rd ed. Hoboken, NJ, USA:
Wiley, 2008, pp. 75100.
[8] I. Erlich and U. Bachmann, Grid code requirements concerning connection
and operation of wind turbines in Germany, in Proc. IEEE Power Eng. Soc.
General Meeting, San Francisco, CA, USA, Jun. 2005, vol. 2,
pp. 12531257.
[9] Schweitzer Engineering Laboratories (SEL). (Sep. 2012). Instruction
manual of SEL-700GW wind generator and intertie protection relay. SEL,
Pullman, WA, USA [Online]. Available: www.selinc.com/SEL-700GW.
[10] L. Y. Yu, I. Minceff, D. W. Hamilton, and G. W. Bottrell, Motor
contribution during three-phase short circuit fault, IEEE Trans. Ind. Appl.,
vol. IA-18, no. 6, pp. 593599, Nov./Dec. 1982.
[11] F. Sulla, Island operation with induction generatorsFault analysis
and protection, Ph.D. dissertation, Dept. Measurement Technology and
Industrial Electrical Engineering, Lund Univ., Lund, Sweden, 2009.

756

[12] N. D. Tleis, Modelling of ac rotating machines, in Power Systems

Modelling and Fault Analysis, 1st ed. Oxford, U.K.: Newnes, 2008,
pp. 301396.
[13] Z. Chen, J. M. Guerrero, and F. Blaabjerg, A review of the state of the art
of power electronics for wind turbines, IEEE Trans. Power Electron.,
vol. 24, no. 8, pp. 18591875, Aug. 2009.
[14] G. Abad, J. Lopez, M. A. Rodriguez, L. Marroyo, and G. Iwanski,
Hardware solutions for LVRT, in Doubly Fed Induction Machine:
Modeling and Control for Wind Energy Generation Applications, 1st ed.
Hoboken, NJ, USA: Wiley-IEEE Press, 2011, pp. 479500.
[15] Toshiba. (2012). Instruction manual for GRE140 directional overcurrent
protection relay. Toshiba [Online]. Available: www.toshiba-tds.com/
tandd/pdf/pcsystems/Manual/6F2T0177_r2.1_GRE140_manual.pdf.
[16] Siemens. (Apr. 2013). Complete technical manual for 7SG11 argus
1-6 overcurrent protection relay. Siemens. Erlangen, Germany [Online].
Available: www.energy.siemens.com/ecc_pool/REYROLLE_Protection/
41555e0b-0bb4-41c8-8018-db7ee181ef52/7SG11%20-%20Argus%201-6
%20Complete%20Technical%20Manual.pdf.
[17] M. S. Sachdev, R. Das, A. Apostolov, S. M. Brahma, A. Darlington,
F. Lopez et al. (Jan. 2009). Understanding microprocessor-based technology applied to relaying. Report of working group I-01 of the relaying
practices subcommittee, IEEE Power System Relaying Committee
[Online]. Available: www.pes-psrc.org/Reports/UNTAR-Ed2.pdf.
[18] M. S. Sachdev and M. A. Baribeau, A new algorithm for digital impedance
relays, IEEE Trans. Power App. Syst., vol. PAS-98, no. 6, pp. 22322240,
Nov. 1979.
[19] Schweitzer Engineering Laboratories (SEL). (Jan. 2012). Instruction
manual for SEL-321-3,-4: Phase and ground distance relay directional
overcurrent relay fault locator. SEL. Pullman, WA, USA [Online]. Available: www.selinc.com/SEL-321/.
[20] ABB. (Dec. 2003). Application manual for transformer protection terminal
RET 521*2.5, version 2.5. ABB. Vasteras, Sweden [Online]. Available:
www05.abb.com/global/scot/scot354.nsf/veritydisplay/72557c65acb0b97
dc12578570041dc1b/$le/1MRK504037-UEN_en_Application_manual_
Transformer_protection_terminal_RET_521_2.5.pdf.
[21] Schweitzer Engineering Laboratories (SEL). (Jun. 2013). Instruction manual for SEL-411L relay: Protection and automation system. SEL, Pullman,
WA, USA [Online]. Available: www.selinc.com/SEL-411L.
[22] P. M. Anderson, Sequence impedance of machines, in Analysis of
Faulted Power Systems. Hoboken, NJ, USA: Wiley-IEEE Press, 1995,
pp. 183230.
[23] IEEE. IEEE PES distribution test feeders [Online]. Available: http://ewh.
ieee.org/soc/pes/dsacom/testfeeders/feeder34.zip.
[24] K. Clark, N. W. Miller, and J. J. Sanchez-Gasca. (Jun. 2008). Modeling of
GE wind turbine-generators for grid studies. GE, Schenectady, NY, USA
[Online]. Available: www.pes-psrc.org/c/C17/GE%20WTG%20Modeling-v4%202.pdf.

IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 5, NO. 3, JULY 2014

Ali Hooshyar (S12) received the B.Sc. and M.Sc.

degrees in electrical engineering from Isfahan
University of Technology, Isfahan, Iran and the
University of Tehran, Tehran, Iran, in 2006 and 2009,
respectively. Currently, he is pursuing the Ph.D.
degree in electrical engineering at the University of
Waterloo, Waterloo, Canada.
His research interests include protection and control of renewable energy systems and smart grids.

Maher Abdelkhalek Azzouz (S10) was born in

Egypt in 1986. He received the B.Sc. and M.Sc.
degrees in electrical engineering from Cairo University, Giza, Egypt, in 2008 and 2011, respectively, and
is currently pursuing the Ph.D. degree with the Department of Electrical and Computer Engineering,
University of Waterloo, Waterloo, ON, Canada.
His research interests include dynamics and control
of power converters, distributed and renewable generation, and control of smart distribution systems.

Ehab F. El-Saadany (SM05) was born in Cairo,

Egypt, in 1964. He received the B.Sc. and M.Sc.
degrees in electrical engineering from Ain Shams
University, Cairo, Egypt, in 1986 and 1990, respectively, and the Ph.D. degree in electrical engineering
from the University of Waterloo, Waterloo, ON,
Canada, in 1998.
Currently, he is working as a Professor with the
Department of Electrical and Computer Engineering,
University of Waterloo. His research interests include
smart grids operation and control, power quality,
distributed generation, power electronics, digital signal processing applications
to power systems, and mechatronics.