ISLANDING DETECTION METHOD FOR INVERTER BASED DISTRIBUTED GENERATION

CONSIDERING IT’S SWITCHING FREQUENCY

V.L.Prathyusha B.Anudeep
Prathyusha.eee@gmail.com prudhvianudeep@gmail.com
Department of Electrical and Electronics Engineering,
RVR & JC College Of Engineering.
Chowdawaram, Guntur, 522019, INDIA
Phone: E-mail:

ABSTRACT: Islanding detection is one of the most important issues for distributed generation (DG)
systems connected to an electric power grid. The conventional passive islanding detection methods
inherently have a non-detection zone (NDZ), and active islanding detection methods might cause problems
with safety, reliability, stability, and power quality on power system. This paper proposes a passive
islanding detection method with the zero NDZ property by considering the switching frequency of an
inverter of the DG system. Moreover, the proposed method does not cause any negative effects in the above
problems because it avoids applying the intended changes such as variations of reactive power and/or
harmonics, which are required in the active islanding detection methods. Several case studies are carried out
to verify that the proposed method can detect the islanding operation within 20 milliseconds, which is less
than 150 milliseconds required in a recloser as the shortest mechanical reclosing (delay) time. This value
also satisfies the requirement of two-seconds standard given in the IEEE Std. 1547.
Terms: Distributed generation (DG), inductance concentration bus, islanding detection, load concentration bus, PWM inverter,
switching frequency.

INTRODUCTION
According to the increased prices of fossil fuels such as oil and natural gas, it is expected that the electric
power industry will undergo considerable and rapid change with respect to its structure, operation, planning,
and regulation. Moreover, trends in power system planning and operation are being toward maximum
utilization of existing infrastructures with tight operating margins due to the new constraints placed by
economical, political, and environmental factors. The distributed generation (DG) system is one of the most
possible solutions to deal with the above problems. The DG is based on the renewable energy sources. The
number of DG systems is rapidly increasing, and most of them are connected to a distribution system by
supplying power into the network as well as local loads [1]. An islanding operation occurs when the DG
continues supplying power into the network after power from the main utility is interrupted [2]-[3]. If the
islanding operation occurs, the distribution network becomes out of utility’s control. It can therefore cause a
number of negative impacts on the network and DG itself, such as the safety hazards to utility personnel and
 the public, the power quality problems, and serious damage to the network and DG unless the main utility
power is restored correctly and quickly [1]. Two types of islanding detection methods, which are the passive
and active methods, have been developed until now. Some representative passive methods detect the
islanding operation by monitoring over/under voltage (OUV) and over/under frequency (OUF) [4]-[5]. The
changes of voltage and frequency are caused by the mismatches of active and reactive power, respectively.
However, these methods have the non-detection zone (NDZ), where the OUV/OUF based detection method
fails to detect islanding. The other passive methods using voltage unbalance (VU) and total harmonic
distortion (THD) might detect the islanding operation even in the NDZ. However, those passive methods
still neither overcome the NDZ completely nor operate fast enough, despite using the VU and THD
methods. On the other hand, the active method has negative effects on the power system even though it can
reduce the NDZ and islanding detection time by injecting harmonics, varying active and reactive powers, or
changing grid impedance [3]. In other words, it might cause malfunctions in islanding detection algorithms
applied to an inverter-based DG and degrade the voltage stability, power quality, and service reliability.
Although the other active methods using supplementary transmitter and receiver have no NDZ, they impose
additional costs. The design of the above anti-islanding algorithm for small DG system requires reducing
computing efforts. Therefore, the objective of this study is to develop an islanding detection method, which
is much faster than the shortest mechanical reclosing (delay) time without extra cost and negative effects.
Generally, the shortest reclosing time is 150 m. This paper makes the new contribution by developing the
new passive method to determine the islanding operation within 20 ms based on switching frequency of
inverters.

Fig.1 Distribution network with inter connected multiple DG systems

II. Islanding Detection Method Based On Switching Frequency Of A

PWM Inverter
A. Non-Detection Zone (NDZ) by OUV/OUF Method
 After an islanding operation occurs, there might be a power mismatch between power generations and load
consumptions in the isolated system. Then, it causes changes of voltage and frequency [5], and this
islanding operation is detected by monitoring those changes if they are large enough. However, the
conventional OUV/OUF passive method has the non detection zone (NDZ) as shown in Fig. 2, where the
islanding operation is not detected due to the small changes of voltage and frequency. In Fig. 2, V min,V max ,
f minand f max are thresholds of under and over voltages and frequencies, which are dependent to some degree
on the interface control design [4].
B. Case of Grid-Connected Single DG System:
The DG system is usually located on a distribution network, and capacitor bank is also placed in the
network to compensate voltage by supplementing reactive power. The sizes and parameters of all devices in
Fig. 1 are given in Table 2.
Component Real power (kW) Reactive power (kVAR) Switching Frequency (kHz)
DG 1 15 0 15
DG 2 15 0 16
DG 3 15 0 15
LOAD 1 15 3 --
LOAD 2 250 50 --
LOAD 3 500 100 --
LOAD 4 250 50 --
LOAD 5 500 100 --
LOAD 6 15 3 --
Capacitor bank-1 -- 103 --
Capacitor bank-2 -- 100 --
TABLE 2. Sizes and Parameters of All Devices in Fig. I

The distribution network in Fig. 1 is re-formed with the simplified system shown in Fig. 3 to evaluate the
effect of the proposed islanding detection method. The output of PWM inverter includes the high frequency
harmonics resulting from its periodic fast switching operation. Because the inductance of transformer is
much larger than that of distribution lines, the high frequency components from the inverter hardly pass
through the transformer. This means that high frequency harmonics over the switching frequency in voltage
and current waveforms measured at the point of common coupling (PCC) of Fig. 3 comes only from the
inverter, but not from the grid.

Fig. 3. Simplified distribution network with single DG system. Fig. 4. Modeling of distribution network with the estimated impedances.
The impedances referred to the DG system in Fig. 3 can be estimated from the correspondingly measured
voltages and currents [11]. Then, the simplified distribution system in Fig. 3 is modeled with the estimated
impedances as shown in Fig.4. To compare the impedances before and after the islanding operation, the
associated parameters are defined as follows.
• R1: Resistance of Load-1 •C 3: Capacitance of Load-3
• L1: Inductance of Load-1 •C bk−1 : Capacitance of Capacitor bank-1
•C 1: Capacitance of Load-1 •ω 0: Fundamental system frequency
• R3 : Resistance of Load-3 •ω sw: Switching frequency of inverter of DG-1
• L 3: Inductance of Load-3
As mentioned before, the impedance of LV/HV transformer with respect is to high frequency very large
when compared to those of Load-1, Load-2, and Capacitor bank-1. Therefore, the impedance of transformer
referred to on the side of DG-1 is negligible for high frequencies over the switching frequency of the
inverter when the parallel sum of impedances with Load-3 is considered. Then, the total impedance, Z
before and after the islanding operation by the Circuit breaker-4 in Fig. 3 is expressed by (1) and (2),
respectively.
1
Z before=
1 1 1 1
R1 R3 (
+ + j ω 0 C 1−
ω0 L1 ) (
+ j ω 0 C 3 +ω 0 C bk−1 −
ω 0 L3 )
.. (1)
1
Z after =
1
+ jω 0 C 1−
1 ….. (2)
R1 ω0 L1

1
C bk −1 = 2
−C 3 ….. (3)
ω L3
0

Fig.6. Implementation of the proposed islanding detection algorithm

A capacitor bank is mostly installed to compensate voltage as well as to improve power factor. When the
capacitance of Capacitor bank-1 is given as (3), the Load-3 might have close to unity power factor because
R3 can be ignored in high frequency mode. In this case, the overall system voltage and frequency referred to
the DG-1 do not change because the impedances, Z beforeand Z after in (1) and (2) become the same. Therefore,
the conventional OUV/OUF passive method cannot detect the islanding operation. In the meanwhile, the
switching frequency of inverter ω sw is given as,ω sw =m f ω 0, where m f is the frequency modulation. Then, the
impedance, Z sw, respect to the frequency modulation is calculated by
 1 α =ω 0 ( C1 +C3 +C bk −1 );
Z sw =
1 β
+ f (mf α − ) …. (4)
R1 mf 1 1 1
β= ( + ) ….... (5)
Where ω 0 L1 L3

Fig. 5. Variation of impedance, Z swcorresponding to mf

When L1 = 0.01 H, L3 = 0.0001 H, C1 = 0.00001 F, C3 = 0.001F, and R1 = 1 Ω. It is observed that the
impedance under the islanding condition is 262 times larger than under the pre islanding condition with the
mf of 10. The mf of 21 to 330 is considered in the PWM inverter of DG system when the usual switching
frequency is 1260 Hz to 20 kHz for the grid-connected DG systems. Therefore, the islanding operation can
be detected perfectly by this variation in impedance with respect to the switching modulation of inverter.
C. Case of Grid-Connected Multiple DG Systems
Some islanding detection methods [1], [4] use the DG system, which injects harmonic currents based on the
THD factor to cancel out the high frequency harmonics resulting from nonlinear loads. However, when
these harmonics are detected on a system with multiple DG systems, it is difficult to know exactly which
DG can deal with the undesired harmonics. The proposed method is not interrupted by the above case
because it operates at much higher frequencies than the harmonics from nonlinear sensitive loads.
Moreover, the proposed method requires the only particular frequency for each DG system. In other words,
it detects the islanding condition clearly only if multiple DG systems are connected to a distribution network
with different switching frequencies of their PWM inverters. As described previously, the harmonics
corresponding to switching frequencies are highly damped by the transformer due to its large impedance.
Therefore, the proposed method can be also applied to the DG systems with even same switching
frequencies if they operate in different areas separated by a transformer. For example, the DG-1/DG-3 and
DG-2/DG-3 in Fig. 1 confirm this case.
D. Implementation of Proposed Detection Method
When the switching frequencies of two DG systems in same area are very close each other, the signal
bandwidth is wide in frequency domain. For the successful operation of the proposed islanding detection
method, it requires to make the bandwidth of its frequency response narrow with the longer sampling time.
In this case, the improvement of the resolution in frequency domain based on the fast Fourier transforms
(FFT) can make the islanding detection time longer than the shortest mechanical reclosing (delay) time
required in a recloser. Moreover, the proposed method needs to find the impedance at only one switching
frequency. The proposed algorithm is therefore implemented directly in time-domain without the analysis in
 the frequency domain such as the FFT. This is shown in Fig. 6. It firstly generates sine and cosine
waveforms with the switching frequency,ω sw. They are multiplied individually by the real measurements of
voltage and current at the PCC. Then, their average values over one period, T, of waveforms are calculated.
The required impedance is finally estimated by the Zest of (6) without direct calculation in (4) with the
manipulations of those average values, as shown in Fig. 6.
Z est= √ ¿ ¿ ¿ ¿ ….. (6)
Note that the average of two sinusoidal waveforms products at different high frequencies is zero because
they have the orthogonal characteristic in high frequency mode. Therefore, the above implementation with
consideration of a particular switching frequency is reasonable.
III. Case Studies in A Single DG System
A. Case of Small Difference between Load Consumption and DG Output Power after Islanding
The case with small difference between single DG system output and load consumption after the islanding
operation (by the circuit breaker-4) at 0.5 s is tested using MATLAB/SIMULINK simulation on the system
in Fig. 3. The result in Fig. 7 shows that the magnitude of a-phase voltage at the terminal of DG-1 rarely
changes. And, its phase response shows small high frequency perturbations after 0.5 s; therefore, the change
of corresponding system frequency is also small. As mentioned in Section II, the traditional OUV/OUF
method fails to detect the islanding. However, the voltage response corresponding to the high switching
frequency of PWM inverter will show a sudden change after islanding because there is a large difference in
the impedance, as shown in Fig. 5. The result in Fig. 8 shows the voltage at the terminal of DG-1 in this
particular high frequency mode. It suddenly increases right after the islanding occurs and there is no change
in pre-islanding operation. Therefore, the proposed method determines the islanding operation successfully
even in the NDZ, where it cannot be detected by the OUV/OUF method.

Fig.7 Responses of a-phase voltage at the terminal of DG-1 after the Fig.8 Voltage response at high switching frequency (15 kHz)of
islanding operation at 0.5s: (a) Magnitude,(b) Phase. PWM inverter in the islanding and pre-islanding operations.
B. Case of Large Difference between Load Consumption and DG Output Power after Islanding
Assume that the Load-6 shown in Fig. 1 is connected with the Load-1 in Fig. 3 in parallel. This creates a
large difference between the DG-1 output and power consumption from the Loads 1 and 6 after islanding.
The results in Figs. 9 and 10 shows that the OUV/OUF based method can be possibly used as an islanding
detection method from its voltage response. In summary, both the OUV/OUF and proposed methods
provide the desirable islanding detection performance for this case.

Fig.9 Responses of a-phase voltage at the terminal of DG-1 after the Fig.10 Voltage response at high switching frequency (15 kHz) of
islanding operation at 0.5 s:(a) Magnitude (b) Phase. PWM inverter in the islanding and pre-islanding operations.

IV. Case Studies in Multiple DG Systems

It is important to verify whether the proposed islanding detection technique remains robust and has a
negligible NDZ on a system with multiple grid-connected DG systems. A total of six case studies are carried
out on the distribution network with three DG systems in Fig. 1.Table 3 shows that the output voltage
harmonics of DG-system in the pre-islanding condition satisfies the requirement of IEEE Std. 1547 for
maximum harmonic voltage distortion in percent of rated voltage. Also, six case studies are implemented by
the different operation (call state) of total eight circuit breakers in Fig. 1, as shown in Table 4. The change
of this state followed by the islanding detection occurs at 0.5 s in each case.
Total harmonic order THD
h<11 11≤h<17 17≤h<23 23≤h<35 35≤h
IEEE std. 4.0 2.0 1.5 0.6 0.3 5.0
DG -1 1.41 0.06 0.05 0.06 0.2 1.46
Table 3. Output voltage harmonics of DG-1 system

Circuit breaker
Case study
1 2 3 4 5 6 7 8
A O X OX X O X O X
S
B O X OX X O X O X
T
C O X O X O O X O X
A X O
D O X O O X O X
T E O X X O O X O O X
E F O X X O O X O X X
Table 4. States of circuit breaker to carry out 6 case studies o: close, x: open, : change at 0.5 s
A. Case of Large Difference between Load Consumption and DG Output Power after Islanding
The test is carried out by the operation of circuit breaker-4, which opens at 0.5 s, as shown in Table III. The
results are shown in Fig. 11. As in the case study in Section III-B, the islanding operation is easily detected
by the conventional OUV/OUF method when the voltage exceeds the threshold of NDZ in Fig. 2. Also, the
proposed method shows good islanding detection performance by its estimated impedance variation, as
shown in Fig. 11 (c). From the corresponding trip signals in Fig. 11 (d), both methods provide fast operation
with about 10 ms detection time, which is much shorter than the required shortest reclosing time of 150 ms.
B. Case of Small Difference between Load Consumption and DG Output Power after Islanding
When the difference between load consumption and DG- 1’s output power is not large enough, the results
are shown in Fig. 12. The change in magnitude of a-phase voltage at the terminal of DG-1 is very small
after 0.5 s [Fig. 12 (a)]. Therefore, the OUV/OUF method fails to detect islanding. In contrast, the results
shown in Fig. 12 (c)-(d) illustrate that the proposed method still provides successful detection operation
without degrading its performance. This proves that this method id robust and most important Zero NDZ
property.
C. Connection of another DG System with Different Switching Frequency
Interaction between multiple DG systems is the one of important and difficult problems, which must be
solved with careful analysis without regard to active and passive islanding detection methods. This situation
is simulated by the operation of circuit breaker-2 in Fig. 1, which closes at 0.5 s. Then, the islanding
detection performance of proposed method is evaluated by assigning different switching frequencies of 15
kHz and 16 kHz to the DG-1 and DG-2 systems, respectively as in Fig. 13. The proposed method does not
misjudge the connection of DG-2 as an islanding operation, which could be detected by the other islanding
detection methods because of active power mismatch.

Fig 12 Fig 12.Responses in the case study - B

Fig 11. Responses in the case study -A Fig

 Fig. 13. Responses in the case study-C.
Fig. 14.Responses in the case study-D.

D. Connection of another DG System with Same Switching Frequency through a Transformer

As mentioned before, the proposed islanding detection technique is based on the estimation of system
impedance with different switching frequencies of multiple DG systems. In practice, the high frequency
harmonic components of a signal hardly pass through the transformer. Therefore, the same switching
frequency can be used in different DG systems if they are separated from each other by a transformer. For
example, the DG-1 and DG-3 systems in Fig. 1 come to this case when both DGs operate with the switching
frequencies of 15 kHz. This test is carried out by the operation of circuit breaker-3 in Fig. 1, which closes at
0.5 s. The results are shown in Fig. 14. As with the responses in case study-C (Fig. 13), the proposed
method does not misjudge the change in system condition by connecting the DG-3 system as an islanding
operation, even though they use the same switching frequencies in their inverters.
E. Connection of another Load
When the load consuming a large amount of active and reactive power is suddenly connected to the system,
the conventional OUV/OUF method is subject to decide this sudden increase of load as an islanding. This
case is simulated by the operation of circuit breaker-6, which closes at 0.5 s; therefore, the load-3
consuming 500 kW and 100 KVAR is suddenly connected to the system. The results in Fig. 15 (a) and (b)
show that the magnitude and phase of a-phase voltage at the terminal of DG-1 decrease after 0.5 s when the
load-3 is connected. In Fig. 15 (c), there exists very small change in the variation of impedance estimated by
the proposed method after 0.5 s. Therefore, it judges this change of condition correctly[Fig. 15 (d)].
However, the OUV/OUF method erroneously detects an islanding operation.
F. Connection of Nonlinear Sensitive Load
When nonlinear sensitive loads such as an induction motor are connected to an electric power grid, the
voltage and current at the PCC connected to these nonlinear loads are distorted. The islanding detection
methods based on the THD factor can malfunction in the case with severe nonlinear sensitive loads
deteriorating power quality. The induction motor rated at 4-pole, 60 Hz, and 250 kW is now connected as
the Load-4 at 0.5 s. At the same time, the induction motor starts operate. Then, the change in output voltage
harmonics is given in Table 5. It is observed that harmonic distortions become worse for individual
harmonic orders, and the corresponding value of THD also increases. Moreover, the large value of starting
current of the induction motor might cause a dip in the voltage, resulting in misjudge of islanding operation
by the OUV/OUF method. These behaviors are shown in Fig. 16. However, the proposed method still
operates correctly.

Fig. 15. Responses in the case study-E Fig. 16.Responses in the case study-F.

Individual harmonic order THD

h<11 11≤h<17 17≤h<23 23≤h<35 35≤h
before 1.41 0.06 0.05 0.06 0.20 1.46
After 1.70 0.24 0.16 0.18 0.32 1.80
Table 5.Change of Output Voltage Harmonics In Percent Of Rated Voltage After The Induction Motor Is Connected

V. CONCLUSIONS
This paper proposed a new passive islanding detection method by the estimation of system impedance based
on high switching frequency of inverter in the distributed generation (DG) system. The proposed method
provided the zero non detection zone (NDZ) property, which is superior to the conventional passive
methods. It is also robust in that it avoids applying the intended changes such as variations of reactive power
and / or harmonics, which is required in the other active islanding detection methods. Moreover, the
proposed method has been applied successfully on a distribution network with multiple grid-connected DG
systems. This method can be used as a protection algorithm when the inverter-based DG systems using
several renewable energy resources are connected to an electric power grid.

REFERENCES:
[1] S. I. Jang and K. H. Kim, “An Islanding Detection Method for Distributed Generations Using Voltage Unbalance and Total Harmonic
Distortion of Current,” IEEE Transactions on Power Delivery, Vol. 19, No. 2, pp. 745-752, April 2004.
[2] F. De Mango, M. Liserre, A. D. Aquila, and A. Pigazo, “Overview of Anti-Islanding Algorithms for PV Systems. Part I: Passive Methods,”
in proc. of IEEE Power Electronics and Motion Control Conference, pp. 1878-1883, August 2006.
[3] F. De Mango, M. Liserre, A. D. Aquila, and A. Pigazo, “Overview of Anti-Islanding Algorithms for PV Systems. Part II: Active Methods,”
in proc. of IEEE Power Electronics and Motion Control Conference, pp. 1884-1889, August 2006.
[4] H. H. Zeineldin, E. F. El-Saadany, and M. M. A. Salama, “Islanding Detection of Inverter-Based Distributed Generation,” IEE Proceedings
on Generation, Transmission and Distribution, Vol. 153, No. 6, November 2006.
[5] Z. Ye, A. Kolwalkar, Y. Zhang, P. Du, and R. Walling, “Evaluation of Anti-Islanding Schemes Based on Nondetection Zone Concept,”
IEEE Transactions on Power Electronics, Vol. 19, No. 5, pp. 1171-1175, September 2004.
[6] Soo-Hyoung Lee, Jung-Wook Park, “New Islanding Detection Method for Inverter-Based Distributed Generation Considering Its Switching
Frequency”  Industry Applications Society Annual Meeting, 2009. IAS 2009. IEEE, Digital Object Identifier: 10.1109/IAS.2009.5324872