Vous êtes sur la page 1sur 12

ISLANDING DETECTION

METHOD
a) N.S.S.S. CHANDRA,
III/IV B.TECH E.E.E,
JNTU KAKINADA UNIVERSITY
COLLEGE OF ENGINEERING,
VIZIANAGARAM.
ABSTRACT
This paper proposes a new islanding
detection method for use in a small-scale,
grid-interconnected distributed generator
system. The proposed islanding detection
method is based on voltage fluctuation
injection, which can be obtained through
high-impedance load switching on the grid
periodically. The correlation factor between
the periodic switching signal and the
perturbed voltage is then used as an
islanding detection index in the proposed
islanding detection method. Experimental
results demonstrate the principles of the
proposed technique and show the new
proposed method is reliable, economical,
and easy to implement.
1 !T"#$%&T#!'
$istributed generator ($)* is
defined as the generator of power inside the
distribution system. $) is driven by prime
movers such as a wind turbine, water
turbine, micro-turbine, etc. to generate
electricity. $ue to the technological
innovations related to the energy conversion
in the last decade, it is now possible to have
competitive electricity generation with $)
units. The main merits of $) can be listed
as follows' reduction of power loss, voltage
profile improvement, power quality
improvement , possibility to exploit &+,
(&ombined +eat and ,ower* generation, less
polluting emissions . -ince $) is inside the
distribution system, it changes the
characteristics of the distribution system,
causing an impact in the voltage regulation
and protection scheme

.n essential requirement of the grid
interconnected $) system is the capability
of islanding detection slanding occurs when
a part of the distribution system is
electrically isolated from the main source of
supply, yet continues to be energi/ed by $).
The islanding operation of $) may cause
potential ha/ards to line-maintenance
personnel, and ris0 the $) in being
damaged by out of phase reconnection to the
grid. The majority of utilities require that
$) should be disconnected from the grid as
soon as the islanding occurs. IEEE
standard 1547 stipulates a maximum
delay of 2seconds for detection of an
islandin
The islanding detection methods can
be generally categori/ed into two groups,
passive methods and active methods.
,assive methods detect the islanding
operation of $) by monitoring selected
power system parameters, such as voltage
magnitude, the change rate of frequency,
phase displacement, and power output. The
passive methods include the change of
voltage magnitude relay the rate of change
of frequency relay the vector surge relay the
voltage unbalance and total harmonic
distortion of current relay the change of
output power relay the ratio of the frequency
change to the output power change relay the
rate of change of voltage and power factors
relay and the logical rule-based detection
technique The principles of these methods
were developed based on the fact that an
islanding will cause variations in system
parameters. +owever, when the amount of
power mismatch between the $) and local
load is not significant enough during
islanding, the methods mentioned above
may fail to signal the abnormality. 1esides,
another drawbac0 to the passive methods is
that they cannot effectively differentiate
between the islanding and other non-
islanding transients, li0e voltage flic0er or
sag.
.ctive methods detect the islanding
by directly interacting with the system under
consideration. The three main methods are
the reacti!e error export detection
"REE#$ t%e positi!e feed&ac' for po(er
loop met%od and t%e !oltae fluctuation
correlated met%od The "EE$ controls the
excitation current of $) so that it generates
a 0nown value of reactive current, which
cannot be supported unless the generator is
connected to the grid The positive feedbac0
for power loop method will result in an
unstable frequency or voltage, once the $)
is islanded. Eventually, the unstable
frequency or voltage will trip the frequency
or voltage relay to protect islanding The
small-scale $) has simple excitation,
perhaps using permanent magnets.
+ence, islanding of small-scale $) cannot
be detected effectively by controlling the
reactive power export, as in "EE$ or the
positive feedbac0 for power loop method.
.nother method is called the voltage
fluctuation correlated method %sing power
transistor switching high-impedance load
periodically near the voltage /ero crossing
point, it measures the voltage fluctuation
through the utility-interconnected point,
enabling evaluation of system source
impedance and detection of islanding. t
provides a very effective means of detection,
with the disadvantage of introducing a small
voltage perturbation at the /ero crossing
point. .ctive methods are more effective
and robust than the passive ones, but most
existing active schemes have the
disadvantages of high cost and degradation
of power quality to a certain extent.

To overcome the disadvantages of the
existing islanding detection methods, the
aim of this paper is to propose a new
correlation factor islanding detection method
for small-scale, typically less than 1 02,
grid-interconnected $). The proposed
method is more effective and economical
than conventional active methods, and has
very little impact on the power quality.
2 Basic )rinciples of t%e )roposed
*et%od+
The proposed islanding detection
method is based on the feature that the
!ariation at t%e terminal !oltae of #,
%as a stron correlation (it% its !oltae
pertur&ation source (%en #, is
operatin in islandin state. #n the
contrary, the variation at the terminal
voltage of $) has a wea0 correlation with
its voltage perturbation source when $) is
operating in parallel with the grid.
Therefore, measuring the correlation
between variation at terminal voltage and its
voltage perturbation source would show
whether the $) is operating in parallel with
the grid or functioning independently of the
grid. n the proposed islanding detection
applications, a periodically switching high-
impedance load is used, so that variation of
load voltage is restricted to the level, which
would not influence the supply.
The equivalent circuit of a $) parallel
with the grid in normal operation is shown
in 3ig. 1, where Eu and Eg are the open
circuit voltages of the utility and the $)4 Zu
is the source impedance of the utility grid4
Zg is the internal impedance of $)4 ZL is
the impedance of the local load4 Zh is the
impedance of the high-impedance load4 - is
the periodical switch of Zh. The terminal
voltage of grid-interconnected point when -
is turn-off can be expressed as'
-ince Zg is significantly greater than
Zu for distribution system, even for the long
radial system or the wea0 grid system
term is significantly greater than term
in (1*. f the term is ignored, VL1
can be approximately expressed as'

The terminal voltage of grid-
interconnected point when S is turn-on can
be expressed as'
-i. 1 E/ui!alent circuits of a #, parallel
(it% t%e rid
f the term is ignored, VL1
can be approximately expressed as'


The variation at terminal voltage due to the
- switching can be approximately expressed
as'

The equivalent circuit of a $)
subject to islanding operation is shown in
3ig. 5. The terminal voltage of grid-
interconnected point when - is turnoff can
be expressed as'
The terminal voltage of grid-interconnected
point when S is turn-on can be expressed as'
-i. 2 E/ui!alent circuits of a #, durin
islandin operation
The variation at terminal voltage due
to the - switching can be approximately
expressed as'


-ince the internal impedance of $)
(Zg* is significantly greater than the source
impedance of utility (Zu*, comparing (6*
with (7*, we have VL2 is larger than
8VL1. 9ariation at the terminal voltage due
to the switching of the high-impedance load
during islanding operation is thus larger than
that in normal operation.
The experimental system was
performed, and the results are shown in
3igs. : to ;. The $) employed in the tests
consisted of a grid interconnected, three-
phase, 55<9, :<<2 synchronous generator
and a 5<<<=high-impedance inductive load.
.s an example of the typical test for the $)
in normal operation, 3ig. : exhibits that the
variation at the terminal voltage due to the
switching of the high-impedance load is
very small. The frequency in the terminal
voltage is ><+/, and the frequency of the
switching signal is 1<+/.
.s an example of the typical test for
the $) in islanding operation, 3ig. ;
exhibits that the variation at the terminal
voltage due to the switching of the high-
impedance load in islanding operation is
larger than that in normal operation. n 3igs.
: to ;, &hannel 1 denotes the waveform of
the terminal voltage (1<<9?div*, &hannel 5
indicates the waveform of the switching
signal of the high-impedance load
(1<9?div*.
-i. 0 1a!eform of terminal !oltae and
s(itc%in sinal durin normal operation
-i. 4 1a!eform of terminal !oltae and
s(itc%in sinal durin islandin
operation
@easuring the periodical perturbation
of terminal voltage at the grid-
interconnected point, due to switching a
0nown high-impedance load, allows one to
estimate indirectly the $) operating state.
2hen the variation at the terminal voltage
changed, the islanding operation can be
easily detected accordingly.

+owever, variations in the terminal
voltage may result from some load
switching, other than the switching of the
high-impedance load. &onsequently, to
avoid false alarms, the measured variation of
terminal voltage should be closely related
with the given high-impedance load
switching signal as in the islanding
operation. !evertheless, in the case of some
load changes occurring coincidentally with
the intentional high-impedance load
switching, the measured voltage fluctuation
may not represent the supply impedance
change. $istinction between intentional and
coincidental load changes should be made
by observing a number of more switching
instances and terminal voltage changes. To
effectively distinguish variations of the
terminal voltage due to switching of the
given high impedance load from the others
should thus be based on correlation of the
measured terminal voltage changes with the
given load switching.
0 T%e )roposed #etection System+
The architecture of the proposed
correlation factor detection system is
illustrated in 3ig. 7. The insulated gate
bipolar transistor based switching circuit
performs the high-impedance load
switching. . voltage detecting interface
measures the magnitude of terminal voltage
at the grid interconnected point. The digital
signal processor calculates the correlation
factor between the periodical switching
signal and the perturbed voltage at the grid-
interconnected point and decides whether
the trip conditions are met.
.t the /ero crossing point of the
terminal voltage, the periodical electronic
switching circuit turns on?off every three
cycles. The voltage fluctuation due to the
periodical switching in an islanding
operation would be significantly greater than
that in the normal operation. -ince the
periodical electronic switch turns on?off
every three cycles, the switching signal S(j*
has a period of six cycles. S(j* has two
values, -1 for turn-on status and
A1 for turn-off status. The differential signal
BS(j* of S(j* with :-cycle time lag is
described as follows'
The corresponding time series of differential
terminal voltage with time lag of : cycles is
expressed as'
2here the VL(j* is the average terminal
voltage value at the jth cycle.
-ince the average terminal voltage
progressively increases during the switch
turn-off period and decreases progressively
during the switch turn-on period, a
proportional function P(j* is used to express
this feature. P(j* is experimentally set to be
1 for the first cycle after switching, 5 for the
second cycle, and : for the third cycle for
enhancement of the voltage progressive
varying trends after switching on and off.
-i.5 System confiuration of t%e
correlation factor islandin detector
The correlation factor between S(j* and VL(j*
is expressed as follows'
2here Fk is the proposed correlation factor, as
an islanding detection index, N is the number of
cycles of the observing window, and N is set as >
in this paper.

.s described previously2 in normal
operation t%e correlation &et(een 34"5$ and
S"5$ is (ea' and -' is muc% lo(er t%an a
t%res%old !alue. n contrast, as islanding
occurs, VL(j* and S(j* have a strong
correlation and Fk becomes significantly larger
than that in normal operation . Through the
proposed scheme, the correlation factor can be
used as an islanding detection index and serves
as a useful reference to activate the protective
relays.
-i. 6 )rocedure of t%e correlation factor
met%od
3ig. > depicts the procedure of the
correlation factor method. t consists of /ero
crossing detection, periodical switching
command generation, terminal voltage
detection, correlation factor calculation, and
decision process. The $-, measures the
value of terminal voltage over > cycles, so
that it avoids the impacts of various load
variations and real power or reactive power
disturbances. .s shown in 3ig. >, a threshold
for the islanding detection correlation factor
is defined4 the threshold is set to be :> in
this paper. &omputer simulations were
performed, and the simulation results of the
variation of the correlation factor Fk before
and after the islanding operation are shown
in 3ig. C. .s shown in 3ig.C, if the islanding
fault is occurring at the >th cycle, then the
correlation factor Fk increases during the
next > cycles. .t the 15th cycle the
correlation factor Fk larger than the
threshold Top Fk, and the islanding has been
detected.
4 Experimental Results+
The experimental tests were carried
out on two 0inds of generators' synchronous
generator and induction generator. The
procedures of the tests are to verify that the
$) systems cease to energi/e the utility grid
as specified in EEE -tandard 17;C when an
unintentional island condition is present.
-i.7 T%e simulation of t%e !ariation of
Fk &efore and after t%e islandin
operation
4.1 Islandin test for sync%ronous
enerators
The generation system employed in
the islanding tests for synchronous generator
consisted of a grid interconnected, three-
phase, 55<9, :<<2 synchronous generator
driven by a $& motor with ; types of loads,
including (a* maximum real load at unity
power factor, (b* maximum real load at rated
power factor lagging, (c* maximum real load
at rated power factor leading, and (d*
minimum load at unity power factor. The
test circuit, specified in EEE -tandard
17;C, is configured as shown in 3ig.6. The
$) was started, synchroni/ed to the utility
grid, and then the tie-switch -5 was closed
to interconnect the $) to the grid. #pen
switch -1 and record the time between the
openings of switch -1 and when the $)
ceases to energi/e the load. "epeat test to ;
types of loads for a total of 7 times. The test
is successful when the $) ceases to
energi/e the test load within the timing
requirements of EEE -tandard 17;C after
switch -1 is opened.

The effectiveness of the correlation
factor method for synchronous generator has
been validated in the experiments. The test
results for ; types of loads are shown in
Table 1. The testing results show that the
correlation factor, used as an index of
islanding detection, can detect the islanding
operation easily and accurately. The
verification results also reveal that the
proposed correlation factor method detected
the islanding event with a maximum delay
time of <.1D seconds in the 5< tests for ;
types of loads. The average detection time of
the 5< tests is <.1;7 seconds. The detection
time needed is much less than the maximal 5
seconds as specified by EEE standard 17;C.
n the typical test for the type (a*
load, as shown in 3ig. D of the typical test
for the resistive load as an instance, the
detection signal for islanding was issued in
<.176 seconds (totally D.7 cycles for
estimating the differential voltage
magnitudes were needed* after the islanding
operation started.
n the typical test for the type (b*
load, as shown in 3ig. 1<, the detection
signal for islanding was given in <.17
seconds (totally D cycles needed* after the
islanding operation started.
n the typical test for the type (c* load,
as shown in 3ig. 11, the detection signal for
islanding was li0ewise announced
successfully in <.16: seconds (totally 11
cycles needed* after the islanding operation
started.
n the typical test for the type (d*
load, as shown in 3ig. 15, the detection
signal for islanding was li0ewise announced
successfully in <.176 seconds (totally D.7
cycles needed* after the islanding operation
started.
n 3igs. D to 15, &hannel 1 denotes
the waveform of grid voltage (;<<9?div*,
&hannel 5 indicates the waveform of local
load terminal voltage (;<<9?div*, &hannel :
depicts the switching signal of periodical
electronic switch (79?div* and &hannel ;
shows the tripping signal (79?div*.
The capability of the proposed system
to avoid false alarms was verified through
the experiments of randomly switching the
loads. The random load switching tests were
ta0en 5<< times for each type of load4 results
depict that no false alarm occurred out of the
6<< switching tests. .s an example of the
typical test for the load at unity power
factor, 3ig. 1: exhibits that the detection
system does not have false alarms due to the
load switching.
n 3ig. 1:, &hannel 1 denotes the waveform
of grid voltage (;<<9?div*, &hannel 5
indicates the waveform of load current
(1.?div*, &hannel : depicts the switching
signal of periodical electronic switch
(79?div* and &hannel ; shows the tripping
signal (79?div*.
To further evaluate the impact on the
power quality due to the periodical voltage
fluctuation injection by using the correlation
factor method, three power quality indices
were measured through a power quality
analy/er. The three power quality indices
evaluated were total harmonic distortion
(T+$*, voltage fluctuation (,-T* and three
phase unbalance. &omparison results
between with and without the voltage
fluctuation injection are given in Table 5.
The table shows that in normal operation of
synchronous generator interconnected with
the utility grids, though the high-impedance
load switching occurred at the grid-
interconnected point, the terminal voltage
almost was not influenced. The differences
between with and without the voltage
fluctuation injection as shown in Table 5 are
supposed to result from errors or noise from
the measurement instrument.
4.2 Islandin test for induction
enerators+

The generation system employed in
the islanding tests for induction generator
consisted of a grid interconnected, three-
phase, 55<9, :<<2 induction generator
driven by a $& motor with : types of loads,
including (a* ::E rated load at unity power
factor, (b* 1<<E rated load at unity power
factor, (c* 15<E rated load at unity power
factor. The test circuit, specified in EEE
-tandard 17;C, is configured as shown in
3ig. 1;. The $) was started, synchroni/ed
to the utility grid, and then the tie-switch -5
was closed to interconnect the $) to the
utility grid. .djust the islanding "F& load
circuit in 3ig. 1; to provide a quality factor
of 1.< G <.<7. The reactive load is balanced
so that the resonant frequency of the island
circuit is within the under-frequency
(7D.7+/* and over-frequency (><.7+/* trip
settings of the $) and as close to nominal
frequency (><+/* as possible. #pen switch
-1 and record the time between the openings
of switch -1 and when the $) ceases to
energi/e the load. "epeat test for : types of
loads for a total of 7 times. The test is
successful when the $) ceases to energi/e
the test load within the timing requirements
of EEE -tandard 17;C after switch -1 is
opened.
The effectiveness of the correlation
factor method for induction generator has
been validated in the experiments. The test
results for : types of loads are shown in
Table :. The testing results show that the
correlation factor, used as an index of
islanding detection, can detect the islanding
operation easily and accurately. The
verification results also reveal that the
proposed correlation factor method detected
the islanding event with a maximum delay
time of <.51> seconds in the 17 tests for :
types of load combinations. The average
detection time of the 17 tests is <.11
seconds. The detection time needed is much
less than the maximal 5 seconds as specified
by EEE standard 17;C.
n the typical test for the type (a*
load, as shown in 3ig. 17, the detection
signal for islanding was issued in <.11C
seconds (totally C cycles for estimating the
differential voltage magnitudes were
needed* after the islanding operation started.

n the typical test for the type (b* load,
as shown in 3ig. 1>, the detection signal for
islanding was li0ewise announced
successfully in <.1;5 seconds (totally 6.7
cycles needed* after the islanding operation
started.
n the typical test for the type (c* load,
as shown in 3ig. 1C, the detection signal for
islanding was li0ewise announced
successfully in <.17 seconds (totally D cycles
needed* after the islanding operation started
n 3igs. 17 to 1C, &hannel 1 denotes
the waveform of grid voltage (;<<9?div*,
&hannel 5 indicates the waveform of local
load terminal voltage (;<<9?div*, &hannel :
depicts the switching signal of periodical
electronic switch (79?div* and &hannel ;
shows the tripping signal (79?div*.
To further evaluate the impact on the
power quality due to the periodical voltage
fluctuation injection by using the correlation
factor method, three power quality indices
were measured through a power quality
analy/er. &omparison results between with
and without the voltage fluctuation injection
are given in Table ;. The table shows that in
normal operation of induction generator
interconnected with the utility grids, though
the high-impedance load switching occurred
at the grid-interconnected point, the terminal
voltage almost was not influenced. The
differences between with and without the
voltage fluctuation injection as shown in
Table ; are supposed to result from errors or
noises of the measurement instrument.
5 Conclusions+
1ased on a correlated voltage fluctuation
scheme, this paper has proposed a new
method to quic0ly and reliably detect
islanding operation of a $) system. The
voltage fluctuation is injected on to the grid
interconnected point by switching a high-
impedance load periodically. #bserving the
correlation factor of the proposed scheme
through a digital signal processor,
discrimination between islanding and other
non-islanding disturbances can thus be made
accurately.
To verify the effectiveness of the
proposed technique, results obtained from
experiments were used in this paper. The
experimental results show that the proposed
index of the islanding detection correlation
factor can detect the islanding operation
satisfactorily for different types of loads
within <.51> seconds. The detection
performance is shown to be less dependent
on load quality factor and power level.
1esides, the test results also reveal that the
new proposed method is easier and more
economical for implementation as compared
to the existing active detection approaches.
The directions for future research of
the islanding detection method can be
described as follow'
To further improve the detection
performance of the proposed active
islanding detection method, the passive
islanding detection methods that detect the
islanding operation of $) by monitoring the
selected power system parameters will be
investigated and integrated in the proposed
active method. 1esides, for the passive
islanding detection methods, there are many
power system parameters to be monitored,
such as voltage magnitude, the change rate
of frequency, phase displacement, and
power output. The using of optimi/ation
search methods, such as genetic algorithm or
neural networ0s, for the best combination
selection of the selected power system
parameters will be investigated in the
passive islanding detection method.
Refee!"e#$
H1I EEE -td. 17;C, Standard for
Interconnecting
Ditri!uted "eource #ith E$ectric Po#er
S%te&, 5<<:.
H5I 2.J. &hang, +.T. Jang, .n .ctive slanding
,rotection @ethod for $istributed
-ynchronous )enerators, Proceeding of the
'th IET Internationa$ (onference on )d*ance
Po#er S%te& (ontro$+ ,peration and
-anage&ent, +ong Kong, &hina, 5<<>, ,aper
!o. .,-&#@5<<>-156.
H:I -.. Lang, K.+. Kim, .n slanding $etection
@ethod for $istributed )enerations %sing
9oltage %nbalance and Total +armonic
$istortion of &urrent, IEEE Tranaction on
Po#er De$i*er%, 9ol. 1D, !o. 5, 5<<;, pp.
C;7MC75.