Aalborg Universitet

A Hybrid Islanding Detection Technique Using Average Rate of Voltage Change and
Real Power Shift

Mahat, Pukar; Chen, Zhe; Bak-Jensen, Birgitte

Published in:
IEEE Transactions on Power Delivery

DOI (link to publication from Publisher):

10.1109/TPWRD.2009.2013376

Publication date:
2009

Document Version
Publisher's PDF, also known as Version of record

Link to publication from Aalborg University

Citation for published version (APA):

Mahat, P., Chen, Z., & Bak-Jensen, B. (2009). A Hybrid Islanding Detection Technique Using Average Rate of
Voltage Change and Real Power Shift. IEEE Transactions on Power Delivery, 24(2), 764-771.
https://doi.org/10.1109/TPWRD.2009.2013376

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

- Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
- You may not further distribute the material or use it for any profit-making activity or commercial gain
- You may freely distribute the URL identifying the publication in the public portal -
Take down policy
If you believe that this document breaches copyright please contact us at [email protected] providing details, and we will remove access to
the work immediately and investigate your claim.
 1

A Hybrid Islanding Detection Technique Using

Average Rate of Voltage Change and Real
Power Shift
Pukar Mahat, Zhe Chen and Birgitte Bak-Jensen

the islanding detection techniques remain the same that some

Abstract— The mainly used islanding detection techniques of the system parameters (like voltage, frequency, etc.) change
may be classified as active and passive techniques. Passive greatly with islanding but not much when the distribution
techniques don’t perturb the system but they have larger non- system is grid connected. Islanding detection techniques can
detection zones, whereas active techniques have smaller non-
detection zones but they perturb the system. In this paper, a new broadly be divided into remote and local techniques.
hybrid technique is proposed to solve this problem. An average Remote islanding detection techniques are based on the
rate of voltage change (passive technique) has been used to communication between utilities and DGs. Supervisory
initiate a real power shift (active technique), which changes the Control and Data Acquisition (SCADA) [3] or power line
real power of distributed generation (DG), when the passive signaling scheme [4]-[6] can be used to determine when the
technique cannot have a clear discrimination between islanding distribution system is islanded. These techniques have better
and other events in the system. Simulation results show that the
proposed method is effective in detecting islanding of distribution
reliability but they are expensive to implement especially for
systems with DGs. small systems. Therefore, local techniques are widely used to
detect islanding and they can further be divided into passive
Index Terms-- Distributed generation, islanding detection, and active techniques.
real power shift, voltage change. Passive methods continuously monitor the system
parameters such as voltage, frequency, harmonic distortion,
I. INTRODUCTION etc. Based on the system characteristics, one or more of these

T here is a renewed interest in the distributed generation

mainly due to the environmental concern and electricity
market liberalization. Many utilities around the world already
parameters may vary greatly when the system is islanded.
Setting a proper threshold can help to differentiate between an
islanding and a grid connected condition. Rate of change of
have a significant penetration of DG in their systems. But output power of DG [3]-[7], rate of change of frequency [8],
there are many issues to be resolved before DG becomes an rate of change of frequency over power [9], change of source
integral part of the utilities around the world. One of the main impedance [10],[11], voltage unbalance [12],[13], harmonic
issues with DG is islanding. Islanding is a situation in which a distortion [12],[14],[15] and frequency monitoring with
distribution system becomes electrically isolated from the reconfiguration of frequency relay [16] are a few examples of
remainder of the power system, due to a fault upstream or any passive islanding detection techniques. A detection technique
other disturbance, and yet continues to be energized by the that looks into a database created by extensive off-line
DG connected to it. Current practice is that almost all utilities calculations is presented in [17] to overcome some of the
require the DG to be disconnected from the grid as soon as limitations of existing passive techniques. The main problem
possible in case of islanding. IEEE 929-1988 standard [1] with the passive detection techniques is that, it is difficult to
requires the disconnection of DG once it is islanded and IEEE detect islanding when the load and generation in the islanded
1547-2003 standard [2] stipulates a maximum delay of 2 system closely match. Furthermore, special care has to be
seconds for detection of an unintentional island and all DGs taken while setting the thresholds for these parameters. If the
ceasing to energize the distribution system. Hence, it is threshold is too low, then it could result in nuisance tripping
essential to detect the islanding both quickly and accurately. of DG and if the threshold is set too high, islanding may not
Many techniques have been proposed for detecting be detected. The limitation of the passive detection techniques
islanding [3]-[21]. Recent developments in islanding detection can be overcome by active techniques, which can detect
are reviewed in details in [22]. The core concept of most of islanding even under a perfect match of generation and load in
the islanded system.
Active methods directly interact with the power system
P. Mahat, Z. Chen and B. Bak-Jensen are with the Institute of Energy
Technology, Aalborg University, Pontoppidanstræde 101, Aalborg, DK-9220, operation by introducing perturbations. These small
Denmark (e-mails: [email protected], [email protected], [email protected]). perturbations will result in a significant change in system
The work is funded by PSO program, project 2006-1-6316, “Operation parameters when the DG is islanded, whereas the change will
and Control of Modern Distribution Systems”.
 2

be negligible when the DG is connected to the grid. Reactive whether the system has been islanded or not. If Av5 is larger
power export error detection method [8], impedance than a minimum set point, VSMin , islanding is suspected. If it
measurement method [10], slip-mode frequency shift is larger than maximum set point, VSMax , as a result of a large
algorithm (SMS) [18], active frequency drift (AFD) [19],
mismatch of generation and demand, it is clear that the
active frequency drift with positive feedback (AFDPF) [19], distribution system is islanded. However, if Av5 is between
automatic phase-shift (APS) [20] and adaptive logic phase
VSMin and VSMax , then the change in voltage could be the
shift (ALPS) [21] are a few examples of active islanding
detection techniques. The problems with these techniques are result of an islanding or any other event in the system like
switching, load change, etc. Then real power shift (RPS) is
that they introduce perturbations in the system and detection
used.
time is slow as a result of extra time needed to analyze the
The RPS will increase or decrease the real power
system response of the perturbations. Furthermore, the
generation of one of the DGs with increasing or decreasing
perturbations are injected at predefined intervals even though voltage at the corresponding bus, respectively. The change of
it is unnecessary during most operating conditions. Also, if the real power generation of DG also satisfies the condition of
islanding occurs during an interval, then it has to wait for next DG operating at the unity power factor. Now, the magnitude
perturbation to be applied before it can be detected, which of average rate of voltage change for 20 periods ( Av 20 ), after
further elongates the detection time. Applications of active initiation of RPS, is used to differentiate islanding from any
techniques are limited to the DG type and/or load, i.e. reactive other event in the system. If Av 20 is larger than VSMaxU (set
power export error detection method cannot be used when the
point to detect islanding with RPS), then it is from an
DG has to operate at the unity power factor and methods islanding condition.
based on phase shift are mostly useful for inverter based DGs.
Also, AFD is very effective for purely resistive loads but it
may fail for other loads [18]. Active methods based on
impedance measurement introduce high frequency signals,
AFD injects a distorted current waveform, and SMS, AFDPF,
APS and ALPS shifts the phase of output current. This will
often lower the quality of power. Therefore, there is a need to
develop an efficient methodology to detect islanding of the
distribution system with DG, without adverse effects to the
system.
Most of the DGs around the world are required to operate at Av5 > VSMin
unity power factor. Hence, it is likely that there will be
deficiency of reactive power once the distribution system is
islanded. Capacitor banks may be the sole source of reactive
power in the islanded distribution system with DG operating Av5 > VSMax

at unity power factor. The amount of reactive power they

produce is a function of the voltage and once the voltage
changes, as a result of islanding, the reactive power generated
by the capacitor bank will also change, which will further
change the voltage. This paper presents an integration of
average rate of voltage change (passive method) and real
power shift (active method), as a hybrid islanding detection
technique to overcome the short comings of both active and
passive techniques, to detect the islanding of a distribution Av 20 > VSMaxU

system with DGs.

The proposed methodology is explained in detail in Section
II and it is tested in a radial distribution system, which is
presented in Section III. The methodology is simulated in
DigSILENT PowerFactory 13.2.334. Different events have
been simulated and the results are presented in Section IV.
Section V concludes the paper. Fig. 1. Flow chart of the proposed methodology

II. PROPOSED METHODOLOGY

1 5 ⎛ dVDG ⎞
In figure 1, Av5 = ∑ (1)
The flow chart of the proposed methodology is shown in 5 i =1 ⎜⎝ dt ⎟⎠ i
Figure 1. Voltage is measured every period at one of the DG
buses. After dV / dt ≠ 0 is detected, the magnitude of average 1 20 ⎛ dVDG ⎞
Av 20 = ∑ (2)
rate of voltage change for 5 periods, Av5 , is used to determine 20 i =1 ⎜⎝ dt ⎟⎠ i
 3

Fig. 2. Test system

Voltage measurement can be instantaneous and is done P = P0 (1 + K f p Δf + K v p ΔV ) ⎫

every cycle. Av5 is computed based upon the voltage ⎪
⎬ (3)
measurement for the last 5 cycles that can be very fast with Q = Q0 (1 + K f q Δf + K vq ΔV ) ⎪
⎭
micro-controllers currently available in market. Hence, in case Where,
of large mismatch, islanding can be detected in almost 5
P and P0 are active power at new voltage and
voltage cycles. In case of VSMin < Av5 < VSMax , the micro-
frequency, and base voltage and frequency,
controller sends a signal to initiate RPS and computes Av 20 respectively
based on the voltage measurement of the next 20 cycles. Q and Q0 are reactive power at new voltage and
Hence, islanding can be detected in around 25 voltage cycles
even when the generation and load closely matches. The frequency, and base voltage and frequency,
respectively
values for RPS, VSMin , VSMaxU and VSMax are system specific
K f p and K v p are the coefficients of active load
and can be set accordingly. If high sensitivity is required, then
VSMin should be set smaller. RPS should be set at ramp rate dependency on frequency and voltage, respectively
rating of the DG. It is recommended that VSMax should be set K fq and K vq are the coefficients of reactive load
such that it is higher than Av5 achieved by 0.5 MW and 0.5 dependency on frequency and voltage, respectively
Mvar mismatch. VSMaxU should be set corresponding to RPS Δf and ΔV are the deviations on frequency and
such that it is more than Av 20 achieved by islanding and RPS voltage, respectively.
for a match of generation and demand. A look up table with Three different load models are simulated with
Av5 for different power mismatch is given in Table AI. The K f (= K f p = K f q ) = K v (= K v p = K vq ) = 0 (load type 1),
table may be referred to set the values for VSMin . K f = K v = 1 (load type 2) and K f = K v = 2 (load type 3).

III. TEST SYSTEM

Figure 2 shows the test system in which the proposed IV. SIMULATION RESULTS AND DISCUSSIONS
methodology is tested. The test system is a part of a Various events (islanding, load change, capacitor switching,
distribution network, owned by Himmerlands Elforsyning, in induction motor starting, short circuit and WTG switching)
Aalborg, Denmark. The line data for the test system is given have been simulated to show the effectiveness of the proposed
in Table AII. The test system consists of 6 loads, a capacitor methodology. All these events are simulated at time = 0 s. The
bank, 3 fixed speed wind turbine generators (WTGs) and a value for VSMin is set at 11 V/s, as it is observed from the
CHP. Generator data is given in Table AIII. The distribution Table AI that islanding with the mismatch of as minimum as 1
system is connected to the transmission network at Bus 05. kW and 1 kvar would result in Av5 =11 V/s. VSMax is set as
The data for the transmission grid, represented as Tran Grid in
Figure 2, is given in Table AIV. Islanding is simulated by 5000 V/s to be absolutely certain that this value is reached
opening the circuit breaker (CB). WTGs operate at the unity only for islanding and VSMaxU is set at 100 V/s. CHP is
power factor and so does the CHP. The data for the base case chosen for RPS. RPS will change the CHP’s real power at 4%
load and generation is given in Table AV. Loads are modeled per second (ramp rate rating of CHP generators). RPS is
as in (3). initiated at time = 0.1 s, if necessary. If initiated, at the ramp
rate of 4% per second, RPS will gradually increase or
 4

decrease CHP’s real power by 1.6% by 0.5 s, with increasing islanding and RPS. Av5 for islanding are 92.8 V/s, 82.9 V/s
or decreasing voltage at CHP bus, respectively. Voltage and 75.7 V/s for load types 1, 2 and 3, respectively. Similarly,
measurement from 0.1 s to 0.5 s is used to calculate Av 20 Av 20 for islanding, after RPS is initiated, are 192.96 V/s,
and RPS is deactivated if Av 20 < VSMaxU . Four different 138.95 V/s and 101.25 V/s for load type 1, 2 and 3,
scenarios, with four different loading conditions, are respectively. This means that islanding is detected correctly as
considered. Av 20 > VSMaxU .
For the event with Load 5 decreasing by 50%, Av5 for all
A. Scenario I
load types is 22.7 V/s, which is larger than VSMin but smaller
Load 5, in this case, is 4.86 MW and 1.84 Mvar resulting in
power deficiency of 0.7 MW and 0.7 Mvar in the islanded than VSMax . So, RPS is initiated. Now, Av 20 for all load types
system. Figure 3 shows the CHP bus voltage for islanding. is 1.94 V/s, which is smaller than VSMaxU . Hence, as the case
The magnitude of average rate of voltage change for the first should be, no islanding is detected by the routine.
five voltage cycles, Av5 , = 8708.6 V/s, 7733.5 V/s and 1.008
6984.9 V/s, for load types 1, 2 and 3, respectively, which are Islanding and RPS for Load Type 1
all larger than VSMax . Hence, islanding is detected without Capacitor Switching on and RPS for Load Type 1
Islanding and RPS for Load Type 2
RPS as Av5 > VSMin . 1.006

CHP Bus Voltage (p.u.)

Capacitor Switching on and RPS for Load Type 2
1.1 Islanding and RPS for Load Type 3
Islanding for Load Type 1 Capacitor Switching on and RPS for Load Type 3
1.004
Islanding for Load Type 2
1
Islanding for Load Type 3
CHP Bus Voltage (p.u.)

0.9 1.002

0.8
1
-0.1 0.1 0.3 0.5
Time (s)
0.7
Fig. 5. CHP bus voltage for capacitor switching on and RPS compared to
islanding and RPS
0.6
-0.1 0.2 0.5 0.8 1.1 1.4
Time (s) Av5 , for capacitor switching on event, is 54.5 V/s for all
load types. Again, values for Av5 is larger than VSMin but
Fig. 3. CHP bus voltage for islanding
smaller than VSMax . RPS is initiated as a result of that. Av 20
for all load types is 1.74 V/s, which is smaller than VSMaxU .
B. Scenario II
Hence, the algorithm ignores the event as any other event
1.008
Islanding and RPS for Load Type 1
other than islanding.
Load Decrease and RPS for Load Type 1 C. Scenario III
1.006 Islanding and RPS for Load Type 2
CHP Bus Voltage (p.u.)

Load Decrease and RPS for Load Type 2 1.003

Islanding and RPS for Load Type 3
Load Decrease and RPS for Load Type 3
1.004
1.001
CHP Bus Voltage (p.u.)

1.002
0.999

Islanding and RPS for Load Type 1

1 Load Increase and RPS for Load Type 1
-0.1 0.1 0.3 0.5 0.997 Islanding and RPS for Load Type 2
Time (s)
Load Increase and RPS for Load Type 2
Islanding and RPS for Load Type 3
Fig. 4. CHP bus voltage for Load 5 decreasing by 50% and RPS compared to Load Increase and RPS for Load Type 3
islanding and RPS 0.995
-0.1 0.1 0.3 0.5
Time (s)
In this case, the real power demand at Bus 05 is 3.87 MW
and the reactive power demand is 0.85 Mvar resulting in Fig. 6. CHP bus voltage for Load 5 increasing and RPS compared to islanding
deficiency of 0.01 MW of real power and surplus of 0.01 and RPS
Mvar of reactive power in the distribution system. Figures 4
and 5 show the change in CHP bus voltage for Load 5 In this case, the real power demand at Bus 05 is 3.87 MW
decreasing by 50% and RPS, and 1.5 Mvar capacitor and the reactive power demand is 0.87 Mvar, resulting in real
switching on at Bus 09 and RPS, respectively compared to and reactive power deficiency of 0.01 MW and 0.01 Mvar in
 5

the distribution system, respectively. Figures 6-11 show the calculated by measuring the voltage at the faulty phase.
change in voltage at the CHP bus for Load 5 increasing by Hence, the algorithm doesn’t detect islanding.
50% and RPS, switching off the capacitor at Bus 10 and RPS, 1.003
induction motor load of 0.5 MW starting at Bus 08 and RPS, a
single phase to ground fault at Bus 12 and RPS, a three phase
short circuit at Bus 12 and RPS, and wind turbine generator at 1.001

CHP Bus Voltage (p.u.)

Bus 12 (WTG1) switching off, respectively compared to
islanding and RPS. Av5 for islanding are 118.8 V/s, 102.3
V/s and 90.4 V/s for load types 1, 2 and 3, respectively. 0.999
Hence, RPS is initiated and Av 20 for islanding are 233.83
Islanding and RPS for Load Type 1
V/s, 150.67 V/s and 103.82 V/s for load types 1, 2 and 3, Induction Motor Starting and RPS for Load Type 1
respectively. Islanding is detected correctly as the values for 0.997 Islanding and RPS for Load Type 2
Av 20 for all load types are larger than VSMaxU . Induction Motor Starting and RPS for Load Type 2
Islanding and RPS for Load Type 3
In case of Load 5 increasing by 50% event, Av5 for all Induction Motor Starting and RPS for Load Type 3
0.995
load types is 23.0 V/s, which is larger than VSMin but smaller
-0.1 0.1 0.3 0.5
Time (s)
than VSMax . So, RPS is initiated. Av 20 for all load types is
1.94 V/s, which is smaller than VSMaxU . Hence, the algorithm Fig. 8. CHP bus voltage for induction motor starting and RPS compared to
ignores the event again as any other event other than islanding and RPS
islanding.
1.003
Av5 , for capacitor switching off event, is 54.3V/s for all
load types. RPS, initiated as a result of VSMin < Av5 < VSMax ,
results in Av 20 = 1.77 V/s for all load types. Av 20 is smaller CHP Bus Phase A L-L Voltage (p.u.)
1.001
than VSMaxU . Hence, as should be the case, no islanding is
detected again.
1.003 0.999

Islanding and RPS for Load Type 1

1-Φ Short Circuit and RPS for Load Type 1
1.001 0.997 Islanding and RPS for Load Type 2
CHP Bus Voltage (p.u.)

1-Φ Short Circuit and RPS for Load Type 2

Islanding and RPS for Load Type 3
1-Φ Short Circuit and RPS for Load Type 3
0.999 0.995
Islanding and RPS for Load Type 1 -0.1 0.1 0.3 0.5
Time (s)
Capacitor Switching off and RPS for Load Type 1
0.997 Islanding and RPS for Load Type 2 Fig. 9. CHP bus faulted phase line-line voltage for 1 phase to ground fault and
Capacitor Switching off and RPS for Load Type 2 RPS compared to islanding and RPS
Islanding and RPS for Load Type 3
Capacitor Switching off and RPS for Load Type 3
1.005
0.995
-0.1 0.1 0.3 0.5
Time (s)

Fig. 7. CHP bus voltage for capacitor switching off and RPS compared to 1
CHP Bus Voltage (p.u.)

islanding and RPS

In case of a 0.5 MW induction motor starting at Bus 08 0.995

event, Av5 is 166.6 V/s for all load types. Again, values for Islanding and RPS for Load Type 1
Av5 is larger than VSMin but smaller than VSMax . RPS 3-Φ Short Circuit and RPS for Load Type 1
Islanding and RPS for Load Type 2
initiated as a result of that would result in Av 20 of 1.97 V/s 0.99
3-Φ Short Circuit and RPS for Load Type 2
for all load types. Av 20 is smaller than VSMaxU . Hence, the Islanding and RPS for Load Type 3
3-Φ Short Circuit and RPS for Load Type 3
algorithm correctly identifies the induction motor starting
0.985
event as not islanding. -0.1 0.1 0.3 0.5
Distribution systems in Denmark use compensated Time (s)

grounding technique that permits sustained operation at single

Fig. 10. CHP bus voltage for 3 phase short circuit and RPS compared to
phase to ground faults. As the earth faults don’t produce a islanding and RPS
significant fault current, clearing the fault takes longer time. In
case of a single phase to ground fault at phase A of Bus 12, Fault current during the three phase short circuit is high and
Av5 are 164.11 V/s, 163.91 V/s and 163.72V/s, respectively results in recloser operation. Reclosers operate in a sequence
for type 1, 2 and 3 loads. RPS is initiated as a result of that. of two “fast” and two “time delayed” trip operations before
Av 20 is 2.44 V/s for all load types. Av5 and Av 20 are locking out. The fast operation allows a recloser to trip for a
 6

temporary fault downstream. The typical recloser opening Av5 for WTG1 switching on event is 115 V/s for all load
time used in Danish distribution network during the fast types. Av5 is larger than VSMin but smaller than VSMax . So,
operation is 500 ms. This means that during RPS, no recloser RPS is initiated. Now, Av 20 for all load types is 15.62 V/s,
trip action occur which could give a large Av 20 . In figure 10,
which is smaller than VSMaxU . Hence, islanding is not
for a 3 phase short circuit at Bus 12, the recloser opens at 0.05
s to clear the fault. Av5 is 102.53 V/s, 101.85 V/s and 101.21 detected. Av5 for islanding, are 146.4 V/s, 126.6 V/s and
V/s for type 1, 2 and 3 loads, respectively. RPS is initiated as 112.5 V/s for load types 1, 2 and 3, respectively. Similarly,
a result of this. Av 20 of 0.68 V/s is achieved for all load Av 20 for islanding are 241.9 V/s, 155.03 V/s and 106.45 V/s
types. Av 20 is smaller than VSMaxU for all load types. Hence, for load types 1, 2 and 3, respectively. Here, Av 20 is larger
than VSMaxU resulting in islanding being detected. All the
islanding is not detected.
1.003
results are summarized in Table I.

TABLE I
RATE OF VOLTAGE CHANGE FOR DIFFERENT EVENTS
1.001
CHP Bus Voltage (p.u.)

AV 5 (V/s) AV 20 (V/s)
0.999
Events Load Type Load Type
Islanding and RPS for Load Type 1 1 2 3 1 2 3
WTG1 Switching off for Load Type 1 Scenario I
Islanding and RPS for Load Type 2
0.997 Islanding 8709 7734 6985 - - -
WTG1 Switching off for Load Type 2
Islanding and RPS for Load Type 3 Scenario II
WTG1 Switching off for Load Type 3
Islanding 92.8 82.9 75.7 192.9 138.9 101.2
0.995
-0.1 0.1 0.3 0.5 Load Decrease 22.7 22.7 22.7 1.94 1.94 1.94
Time (s)
Capacitor Switch On 54.5 54.5 54.5 1.74 1.74 1.74
Fig. 11. CHP bus voltage for WTG1 switching off compared to islanding and Scenario III
RPS Islanding 118.8 102.3 90.4 233.8 150.7 103.8
Load Increase 23.0 23.0 23.0 1.94 1.94 1.94
Av5 for WTG1 switching off event is 1.1 V/s. It is
Capacitor Switch Off 54.3 54.3 54.3 1.77 1.77 1.77
interesting to note that Av5 for this event is even smaller than
Induction Motor starting 166.6 166.6 166.6 197 1.97 1.97
VSMin . RPS is not initiated ignoring the event as any other 1 phase to ground fault 164.1 163.9 163.1 15.1 15.1 15.1
event other than islanding. 3 phase short circuit 102.5 101.9 101.2 0.68 0.68 0.68
D. Scenario IV WTG Switching Off 1.1 1.1 1.1 - - -
1.01 Scenario IV
Islanding 146.4 126.6 112.5 241.9 155.0 106.5
WTG Switching On 115.0 115.0 115.0 15.62 15.62 15.62
CHP Bus Voltage (p.u.)

1 Results show that when there is a large mismatch in

generation and demand, islanding can easily be detected by
observing rate of voltage change. It is difficult to detect
Islanding and RPS for Load Type 1 islanding when load and generation in the islanded
0.99 WTG1 switching on and RPS for Load Type 1
Islanding and RPS for Load Type 2
distribution system closely match. Events like starting
WTG1 Switching on and RPS for Load Type 2 induction motor loads, short circuits or connecting wind
Islanding and RPS for Load Type 3 turbines to grid can have average rate of voltage change which
WTG1 Switching on and RPS for Load Type 3 can be larger than the average rate of voltage change for
0.98
-0.1 0.1 0.3 0.5
islanding. Results also show that the RPS initiated after the
Time (s)
suspicion of islanding results in higher Av 20 when the
Fig. 12. CHP bus voltage for WTG1 switching on and RPS compared to system is actually islanded compared to other events in the
islanding and RPS system. In other words, RPS doesn’t change the CHP bus
voltage much when the distribution system is connected to the
In this case, the real power demand at Bus 05 is 3.57 MW transmission grid, which supplies the deficient power or
and the reactive power demand is 0.87 Mvar, resulting in real absorb the excess power in the distribution system. However,
and reactive power deficiency of 0.01 Mw and 0.01 Mvar in RPS initiated after the suspicion of islanding changes the CHP
the islanded system, respectively. The wind turbine generator bus voltage significantly when the distribution system is
at Bus 12 (WTG1) is disconnected. Figure 12 shows the CHP islanded. Hence, the RPS can have absolute discrimination
bus voltage for WTG1 switching on event and RPS, compared between islanding and other events even when the load and
to islanding and RPS. generation of islanded system closely match.
 7

V. CONCLUSIONS TABLE AIII

GENERATORS DATA
A hybrid technique has been proposed to detect islanding of
a distribution system with multiple DG units operating at the Parameters CHP WTG
unity power factor. It combines the real power shift and Asynchronou
average rate of voltage change to efficiently detect islanding Type of generator Synchronous
s
under various loading conditions. Number of Parallel Machine 3 1
The proposed technique uses RPS only when the passive 3.3 MVA 630 kVA
technique (average rate of voltage change) cannot have a clear Transformer to connect to grid
20/6.3 kV 20/0.4 kV
discrimination between grid connected and islanding Individual generator’s rating
condition. This eliminates the necessity of injecting Rated Power 3.3 MW 630 kW
disturbance from time to time to detect islanding like other
Rated Voltage 6.3 kV 0.4 kV
active techniques. RPS only changes the real power of DG,
Stator resistance 0.0504 p.u. 0.018 p.u
which satisfies the condition of DG operating at the unity
power factor. In the proposed technique, only one DG changes Stator reactance 0.1 p.u. 0.015 p.u
the real power in contrast to positive feedback techniques, Synchronous reactance d-axis 1.5 p.u.
where all the DGs work together to inject perturbation in the Synchronous reactance q-axis 0.75 p.u.
system. Also the proposed technique is able to efficiently Transient reactance d-axis 0.256 p.u.
discriminate islanding from various other events in the Sub-transient reactance d-axis 0.168 p.u.
distribution system. Furthermore, islanding can be detected
Sub-transient reactance q-axis 0.184 p.u.
even when the load and the generation closely match. Even
though, the method may fail to detect islanding for a perfect Transient time constant d-axis 0.53 s
match of demand and generation in the islanded system, any Sub-tran. time constant d-axis 0.03 s
subsequent change in load or generation in the islanded Sub-tran. time constant q-axis 0.03 s
system will lead to change in voltage and the islanding being Mag. Reactance 4.42 p.u.
detected. Rotor Resistance 0.0108 p.u.
The proposed hybrid islanding detection technique, which
Rotor Reactance 0.128 p.u.
combines the active and passive techniques, can efficiently
detect islanding. Inertia Time Constant 0.54 s 0.38 s

TABLE AIV
APPENDIX TRANSMISSION SYSTEM DATA
TABLE AI
LOOK UP TABLE
Parameters Value
Av 5 Av 5 Maximum short circuit
Power Deficiency Power Deficiency 10000 MVA
(V/s) (V/s) power
0.1 kW + 0.1 kvar 1.54 5 kW + 5 kvar 55.09 Minimum short circuit power 8000 MVA
0.5 kW + 0.5 kvar 5.89 10 kW + 10 kvar 118.80 Maximum R/X ratio 0.1
1 kW + 1 kvar 11.00 20 kW + 20 kvar 219.75 Maximum Z2/Z1 ratio 1
2 kW + 2kvar 22.22 30 kW + 30 kvar 329.92 Maximum X0/X1 ratio 1
3 kW + 3 kvar 33.05 40 kW + 40 kvar 440.49 Maximum R0/X0 ratio 0.1
4 kW + 4 kvar 44.27 50 kW + 50 kvar 550.65
TABLE AV
LOAD AND GENERATION DATA FOR THE TEST SYSTEM
TABLE AII
LINE DATA FOR THE TEST SYSTEM
Bus PG (MW) QG (Mvar) PL (MW) QL (Mvar)
Reactance
From Bus To Bus Resistance (Ω)
(Ω) 05 0 0 3.87 0.85
5 6 0.1256 0.1404 06 6 0 0.0 0.0
5 7 0.1344 0.0632 07 0 0 0.56 0.11
7 8 0.1912 0.0897 08 0 0 0.56 0.11
8 9 0.4874 0.2284 09 0 0 0.55 0.10
9 10 0.1346 0.0906 10 0 1.5 0.85 0.20
10 11 1.4555 1.1130 11 0 0 0.51 0.13
11 12 0.6545 0.1634 12 0.31 0 0.0 0.0
12 13 0.0724 0.0181 13 0.31 0 0.0 0.0
13 14 0.7312 0.3114 14 0.31 0 0.0 0.0
 8

ACKNOWLEDGMENT [18] G. A. Smith, P. A. Onions, and D. G. Infield, “Predicting islanding

operation of grid connected PV inverters,” IEE Proc. Electric Power
Authors would like to thank Himmerlands Elforsyning Applications, vol. 147, pp. 1-6, Jan. 2000.
A.m.b.a, SEAV-NVE and DEFU, Denmark for their support [19] M. E. Ropp, M. Begovic, and A. Rohatgi, “Analysis and performance
for the work. assessment of the active frequency drift method of islanding
prevention,” IEEE Tran. Energy Conversion, vol. 14, no 3, pp. 810-816,
Sep. 1999.
REFERENCES [20] G. Hung, C. Chang, and C. Chen. “Automatic phase shift method for
[1] Recommended Practice for Utility Interconnected Photovoltaic (PV) islanding detection of grid connected photovoltaic inverter,” IEEE
Systems, IEEE Standard 929-2000, 2000. Trans. Energy Conversion, vol. 18, no. 1, pp. 169-173, Mar. 2003.
[2] IEEE Standard for Interconnecting Distributed Resources into Electric [21] J. Yin, L. Chang, and C. Diduch, “A new adaptive logic phase-shift
Power Systems, IEEE Standard 1547TM, June 2003. algorithm for anti-islanding protections in inverter-based DG systems,”
[3] M. A. Refern, O. Usta, and G. Fielding, “Protection against loss of 2005 IEEE Power Electronics Specialists Conference, pp. 2482-2486,
utility grid supply for a dispersed storage and generation unit,” IEEE 2005.
Tran. Power Delivery, vol. 8, no. 3, pp. 948-954, July 1993. [22] P. Mahat, Z. Chen, and B. Bak-Jensen, “Review of islanding detection
[4] M. Ropp, K. Aaker, J. Haigh, and N. Sabhah, “Using Power Line Carrier methods for distributed generation,” in Proc. 3rd International
Communications to Prevent Islanding,” in Proc. 28th IEEE Photovoltaic Conference on Electric Utility Deregulation and Restructuring and
Specialist Conference, pp. 1675-1678, 2000. Power Technologies, pp. 2743 – 2748, 6-9 April 2008.
[5] W. Xu, G. Zhang, C. Li, W. Wang, G. Wang, and J. Kliber, “A power
line signaling based technique for anti-islanding protection of distributed
generators—part i: scheme and analysis,” IEEE Tran. Power Delivery, Pukar Mahat received B.Eng and M.Eng degree
vol. 22, no. 3, pp. 1758-1766, July 2007. from Kathmandu University, Nepal and Asian Institute
[6] G. Wang, J. Kliber, G. Zhang, W. Xu, B. Howell, and T. Palladino, “A of Technology, Thailand, respectively.
power line signaling based technique for anti-islanding protection of He was Research Associate at Asian Institute of
distributed generators—part ii: field test results,” IEEE Tran. Power Technology, Thailand. Since 2007, he has been a PhD
Delivery, vol. 22, no. 3, pp. 1767-1772, July 2007. candidate at Institute of Energy Technology, Aalborg
[7] M. A. Redfern, J. I. Barren, and O. Usta, “A new microprocessor based University, Denmark. Mr. Mahat's research interest
islanding protection algorithm for dispersed storage and generation units,” includes Distributed Generation, Wind Power
IEEE Trans. Power Delivery, vol. 10, no. 3, pp. 1249-1254, July 1995. Generation and Distribution System Planning. He is a
[8] J. Warin, and W. H. Allen, “Loss of mains protection,” in Proc. 1990 Certified Engineer in Nepal..
ERA Conference on Circuit Protection for industrial and Commercial
Installation, London, UK, pp. 4.3.1-12. Zhe Chen (M’95, SM’98) received the B.Eng. and
[9] F. Pai, and S. Huang, “A detection algorithm for islanding-prevention of M.Sc. degrees from Northeast China Institute of
dispersed consumer-owned storage and generating units,” IEEE Trans. Electric Power Engineering, Jilin City, China, and the
Energy Conversion, vol. 16, no. 4, pp. 346-351, 2001. Ph.D. degree from University of Durham, U.K.
[10] P. O’Kane, and B. Fox, “Loss of mains detection for embedded He was a Lecturer and then a Senior Lecturer with
generation by system impedance monitoring,” in Proc. Sixth De Montfort University, U.K. Since 2002, Dr. Chen has
International Conference on Developments in Power System Protection, been a Research Professor and is now a Professor with
pp. 95-98, March 1997. the Institute of Energy Technology, Aalborg University,
[11] P. D. Hopewell, N. Jenkins, and A. D. Cross, “Loss of mains detection Denmark. He is the coordinator of Wind Power System
for small generators,” IEE Proc. Electric Power Applications, vol. 143, Research program at the Institute of Energy Technology, Aalborg University.
no. 3, pp. 225-230, May 1996. His research areas are power systems, power electronics and electric
[12] S. I. Jang, and K. H. Kim, “A new islanding detection algorithm for machines; and his main current research interests are wind energy and modern
distributed generations interconnected with utility networks,” in Proc. power systems. He has more than 170 publications in his technical field.
IEE International Conference on Developments in Power System Dr Chen is an Associate Editor (Renewable Energy) of the IEEE
Protection, vol.2, pp. 571-574, April 2004. Transactions on Power Electronics, Guest Editor of the IEEE Trans. on PE on
[13] S. I. Jang, and K. H. Kim, “An islanding detection method for Special Issue on Power Electronics for Wind Energy Conversion. He is a
distributed generations using voltage unbalance and total harmonic Member of the Institution of Engineering and Technology (London, U.K.),
distortion of current,” IEEE Tran. Power Delivery, vol. 19, no. 2, pp. and a Chartered Engineer in the U.K
745-752, April 2004. Birgitte Bak-Jensen (M’88) received her M.Sc. degree in Electrical
[14] S. Jang, and K. Kim, “Development of a logical rule-based islanding Engineering in 1986 and a Ph.D. degree in “Modeling of
detection method for distributed resources,” in Proc. IEEE Power High Voltage Components” in 1992, both degrees from
Engineering Society Winter Meeting, vol. 2, pp. 800-806, 2002. Institute of Energy Technology, Aalborg University,
[15] H. Kabayashi, K. Takigawa, and E. Hashimato, “Method for preventing Denmark.
islanding phenomenon on utility grid with a number of small scale PV From 1986-1988, she was with Electrolux Elmotor
systems,” Second IEEE Photovoltaic Specialists Conference, vol.1, pp. A/S, Aalborg, Denmark as an Electrical Design
695-700, 1991. Engineer. She is an Associate Professor in the Institute
[16] R. Belhomme, M. Plamondon, H. Nakra, G. Desrosiers, and C. Gagnon, of Energy Technology, Aalborg University, where she
“Case study on the integration of a non-utility induction generator to the has worked since August 1988. Her fields of interest are
Hydro-Quebec distribution network”, IEEE Tran. Power Delivery, vol. modeling and diagnosis of electrical components, power
10, no. 3, pp. 1677-1684, July 1995. quality and stability in power systems. During the last years, integration of
[17] K. El-Arroudi, G. Joos, I. Kamwa, and D. T. McGillis, “Intelligent- dispersed generation to the network grid has become one of her main fields,
Based Approach to Islanding Detection in Distributed Generation,” where she has participated in many projects concerning wind turbines and
IEEE Tran. Power Delivery, vol. 22, no. 2, pp. 828 – 835, April 2007 their connection to the grid.