2013 IEEE GCC Conference and exhibition, November 17-20, Doha, Qatar

A Simulink Model of an Active Island Detection

Technique for Inverter-Based Distributed
Generation
Faisal Khan, member, IEEE, A. Massoud, senior member, IEEE, A. Gastli, senior member, IEEE

Abstract— With the increased involvement of Distributed system is to disconnect the DPGS as soon as the primary grid
Power Generation Systems (DPGSs) into the conventional power fails to energize the distribution line. It is crucial to understand
system, the structure has evolved and therefore has brought in that the anti-islanding requirements are only for the low-power
various challenges albeit improving flexibility and smartness of
DPGSs like the PV panels, as opposed to the higher power
the system. This paper addresses modeling one of these
challenges where a Simulink model for inverter-based distributed DPGSs like wind plants, that due to their communication
generation (IBDG) active islanding detection technique is benefits and supervisory control can be help improving the
introduced. Out of the various types of active islanding detection stability of the grid, and are therefore not required to be
methods, the modeling of the general electric islanding detection disconnected during faults.
method which uses the positive feedback of the voltage or As the main concern of the paper is the island detection,
frequency at the point of common coupling (PCC) for the uncontrolled islanding is the main concern therefore the anti-
detection of an island is presented. This methodology is modeled
islanding detection technique should have the following
and applied for an IBDG connected to a low voltage distribution
network. The simulation results are presented for a 20kW, three- characteristics; the detection system should be reliable and
phase IBDG showing that the system is able to detect islanding selective, and allow minimum disturbance to the power quality
and cease the current flow from the IBDG even under critical of the grid in order to ensure suitable operation of parallel
operating condition of a close matching between the power DPGSs. A reliable system would be functional in all types of
delivered by the inverter and the load demand (zero non- grids and involvement of DPGSs, and a selective system
detection zone operation). would differentiate between a true islanding condition and a
temporary simple disturbance in the grid [1][2].
Index Terms— Active Island Detection, Inverter-Based It should also be noted that the main parameters which can
Distributed Generation, Distributed Generators, Low Voltage be used to identify the missing of a grid connection (island) at
Distribution Networks.
the point of common coupling (PCC) are the variations of the
voltage amplitude and frequency. Having a grid-connected
I. INTRODUCTION
system would not allow for such variations. In case of close

W ITH the development of distributed generation, the

electrical infrastructure and connection to the grid has
evolved as some of the generation has been transferred to the
matching between the DPGS and the load powers, the
fluctuations in the PCC voltage and/or frequency are
negligible. This zone of operation is considered as a non-
distribution site. This evolution has brought new challenges to detection zone (NDZ). The NDZ of an anti-island method can
deal with such as non-uniform voltage distribution along the be used as a measure of its reliability [2][3], which is defined
feeders in which the load may experience overvoltage problem LQWKHSRZHUPLVPDWFKǻ3YHUVXVǻ4DWWKH3&&WKDWJRHV
particularly at light loading, increased fault current undetected and might cause parasitic trips.
contribution, and island detection. According to IEEE Std. The balanced power equations in a typical grid connected
LVODQGLVGH¿QHGDVWKHFRQGLWLRQLQZKLFKDSRUWLRQRI system can be represented as follows:
an electric power system is energized solely and separated Pload = PDG ¨3 (1)
from the rest of the electric power system.
The reason that island mode is hazardous can be classified 4load = 4DG ¨4 (2)
into two main parts. Firstly, it is a safety hazard [1], where if
the personnel are physically present on the grid-site and Where ܲ௟௢௔ௗ , ܲ஽ீ , ܽ݊݀ οܲ are the load, DPGS, and
performing checks, they might get electrocuted due to the fact difference of active power, respectively. Similarly regarding
that the inverter is still being fed by the DPGSs. Secondly, the reactive powers (4load , 4DG DQG¨4). There is no active
there could also be a problem of reclosing of the circuit power mismatch if the power of the DPGS equals the power of
breakers before the island has been extinguished. If this the load (Pload =PDG ), and the same applies to reactive power
happens, there is a problem as the island is desynchronized mismatch.
from the grid, and a connection might cause damage to the If the resonant frequency of the RLC load equals the grid
electrical equipment and components [2]. frequency, no reactive power would be absorbed by the linear
Monitoring of the grid condition is an important feature of loads. On the other hand, active power is directly proportional
the DPGSs, and the required objective of an anti-islanding to the voltage, and when there is a disconnection of the grid,

978-1-4799-0724-3/13/$31.00 ©2013 IEEE 315

 2013 IEEE GCC Conference and exhibition, November 17-20, Doha, Qatar

the load power stays the same using the power generated by be able to detect an island. Different methods of performing
the PV system. The grid voltage then can be illustrated by (3). active island detection include:
1. Drifting the frequency to activate over/under frequency
PDG
V' =ඨ *V (3) (OUF) protection
Pload 2. Drifting voltage amplitude to activate over/under
As seen in (3), if PDG <Pload , there will be a decrease in the voltage (OUV protection
amplitude of voltage and vice versa. 3. Detecting islanding by measuring the Grid Impedance
The key point to note here is that if the power mismatch is 4. Using Phase Locked Loop (PLL) based technique
very small, which it is hard to detect, then this leads to the 5. Negative Sequence detection.
problem of NDZ [3] and could keep the DPGS connected to
III. GENERAL ELECTRIC ANTI-ISLANDING T(&+1,48(
the PCC despite islanding situation. In the simplest of terms, it
is required to know how the anti-islanding performs when the The General Electric Anti-Islanding (GEAI) technique is
power mismatch is very close to zero (i.e. no power one of the active methods. The primary idea utilized in the
mismatch). This means that, while in grid-connected mode, GEAI scheme is to use the positive feedback to basically
the grid does not contribute to feeding the load and when there reduce the NDZ completely, which is a major drawback of the
is a fault in the grid, its removal does not affect the operation passive anti-islanding systems. The positive feedback would
and goes unnoticed. This is where the main advantage of the help identify an island by drifting the voltage and frequency
parameters out of the OUV and OUF limits.
active islanding method lies because it introduces a small
x Positive Feedback
perturbation in the DG output which helps in amplifying the
If the voltage at the inverter output increases, then the
voltage and frequency fluctuation and hence detecting the
controller instructs the inverter to increase the active power
island situation.
according to (4).
The primary purpose of this paper is to model, using
V2
Matlab/Simulink, an inverter-based distributed generation P= (4)
(IBDG) system where islanding can be detected under the R
most critical condition of a zero power mismatch [3] [4]. An
active method which can achieve a zero NDZ is selected and
simulated.
The key approach here is to positively feedback the value of
the voltage and frequency parameters at the PCC to increment
the current references that are going to be fed to the PWM
generator. Any present island would be detected as these
fluctuations would result in the voltage and frequency limits to
be crossed and therefore disconnecting the IBDG.
This paper is organized as follows. Section II presents a
short overview of different island detection methods. Section
III describes and models the selected active method, which is
based on the General Electric technique. Section IV gives and
discusses the simulation results. Finally, section V concludes
this paper.

II. ISLAND DETECTION METHODS

There are different techniques for addressing the problem of
islanding detection and they are classified into two different
classes.
Fig.1. Flow Chart of the island detection using the GEAI technique
A. Passive Methods
The passive methods are used to measure a specific In order to balance the power, the voltage will keep
parameter at the PCC, to see if there has been any alteration increasing which in turn increases the power. This cycle goes
due to the disconnection of the grid [4], and thus detect a on until the voltage limits are exceeded. Similar steps can be
possible presence of an island situation. Most of the passive carried out by sensing the frequency and increasing the
methods just compare the actual parameter value at the PCC to reactive power output and this would lead to the frequency to
a threshold value, to compare and decide if the values exceed cross the set limits indicating the presence of an island.
the normal limit. Fig. 1 demonstrates the GEAI algorithm being used for the
detection of an island. The key methodology here is to
B. Active Methods measure the voltage and frequency at the PCC, and since abc-
These methods are based on the concept of introducing a dq transformation is usually employed within the inverter
disturbance to the output of the inverter [4] to vary slightly the control algorithm, it is relatively simple to obtain these
frequency, phase, harmonics, or active and reactive power, to parameters. These measured values are compared to the grid

316
 2013 IEEE GCC Conference and exhibition, November 17-20, Doha, Qatar

voltages and frequencies to determine any variations present x Simulink Model

in the system. The algorithm shown in Fig. 2 is implemented with
These differences are passed through a gain, which is Simulink. A low-pass filter is added in order to mitigate the
present to enable the detection of any island due to any high frequency noise in the system.
fluctuations that might occur. The gain should be chosen so Since we expect this islanding detection system to either
that its effect is minimal while in grid-connected mode. These increase or decrease the frequency and voltage parameters [8],
new values are added to both the active and reactive power there needs to be another part which is responsible for tripping
references which in turn change the current references. Based the IBDG when a fault is detected.
on these changing references the inverter output continues to For the purpose of introducing a fault, a circuit breaker was
change in case the grid is not present, and this drifts the used on the grid side, and after detecting the fault, an
voltage and frequency measurements beyond the OUV and embedded function was created to force the inverter gate
OUF limits [5],[6]. A simple algorithm can also be signals to be zero to disconnect the IBDG. A zero-order hold
implemented to disconnect the inverter side from the grid to was used and this introduced some delay before tripping the
prevent the flow of any current and therefore shutting off the IBDG after the fault had occurred. Table 1 provides the
DPGS. parameters that were used in the model.

x System Block Diagram TABLE I

SIMULINK MODEL PARAMETERS
Fig. 2 shows the simple system block diagram that is used
to detect the presence of an island based on positive feedback. Parameter Value
The primary function of the control block is to detect if any Grid Voltage 339 V (P-P)
islanding has occurred due to a grid fault. The marked area on Grid Frequency 50 Hz
DC Voltage 800 V
the figure shows how a difference between the grid voltage Grid Impedance 5 Pȍ/ ȝ+
and the measured voltage at PCC are compared and multiplied Line Inductance L = 5 mH
by a specific gain. The result is added to the reference current Active Power 2000 Watt
which is feedback to the inverter through the PWM generator Reactive Power 0 VAR
Gain Voltage = 10, Frequency = 4
[7]. The same process is also applied to the frequency Zero-order hold Sample Time 2.1 ms
measurement at the PCC. It is clearly seen here that the Low Pass Filter ݂௖ = 160 Hz
positive feedback of the voltage is used to change the active
power, and the frequency feedback is responsible for the
change in the reactive power reference. After the positive feedback has been implemented and the
voltage and frequency parameters have been altered, the next
step is to disconnect or trip the IBDG in order to stop feeding
the load after the grid fault has occurred. So as a continuation
of Fig. 2, Fig.3 illustrates the logic used to trip the IBDG. Fig.
3 briefly illustrates the algorithm implemented in Simulink for
tripping the IBGD. The key is to turn of the IBDG by
multiplying a zero to the gate signals when the OUV or OUF
limits are exceeded. This figure shows only the OUV case,
and the same can be done by comparing the frequency
measurement at the PCC with the OUF limits.

Fig. 3. IBDG tripping logic

Fig. 2. System Block Diagram for Island Detection

317
 2013 IEEE GCC Conference and exhibition, November 17-20, Doha, Qatar

For comparison of the voltage and frequency parameters to during grid fault. In fact, when the fault occurs at 0.04
the OUV and OUF limits, a simple Matlab embedded function seconds, the product of voltage difference and the gain is large
was incorporated in the Simulink model. A zero-hold function enough to drift both the voltage and frequency levels outside
with a sample time of 2.1 ms was used to implement this the OUV and OUF limits.
algorithm successfully. Fig. 7 shows the current waveforms at the PCC before and
after a grid fault occurring at 0.04 seconds. Clearly notice that
IV. SIMULATION RESULTS the currents have increased alongside the frequency, due to the
added positive feedback of both voltage and frequency
A. Simple Case
deviations due to the fault.
Fig. 4 demonstrates how the active power changes when a
grid fault is introduced at time t=0.4 seconds. With the Ia
situation shown in Fig. 3, the system should normally detect 10 Ib

Current (Ampere)
that islanding has occurred and the breaker at the inverter side Ic
should be tripped in order to stop providing current to the load 0
while the grid has been disconnected. Similarly, Fig. 5, shows
how the frequency varies after the fault has occurred in the
-10
grid.
0.01 0.02 0.03 0.04 0.05 0.06 0.07
Time (s)
2800 Fig. 7. Current at the PCC after fault occurs at 0.04 seconds
Active Power (Watts)

2600
Fig. 7 illustrates also how the circuit breaker at the inverter
side is tripped after the grid fault occurs at 0.04 seconds, and
2400
the load is not fed anymore as no current is allowed to flow.
2200 B. Zero Power Mismatch Case
2000
Zero power mismatch is the critical case we are primarily
interested in, where all the load power is being fed by the
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Time (s) DPGS solely and the grid disconnection due to a fault would
Fig. 4. Active Power response when a fault occurs at 0.04 seconds not be recognized by passive methods [8][9]. It is critical to
test our control system’s performance in such a scenario. This
150 condition was simulated with a grid fault occurring at 0.04
seconds as shown in Fig. 8.
Frequency (Hz)

100
6000
Active Power (W)

50 5000
4000
0 3000
0.01 0.02 0.03 0.04 0.05 0.06 0.07
Time (s) 2000
Fig. 5 Frequency response when a fault occurs at 0.04 seconds
1000
0.01 0.02 0.03 0.04 0.05 0.06 0.07
800 Time (s)
Fig. 8. Active Power before and after fault occurring at 0.04 seconds for the
(Vgrid - Vd)*Gain

600 case of zero power mismatch

400 Comparing Fig. 8 with Fig. 4 we can notice the difference

200 in case of zero power mismatch. A gain of 10 was used in the
zero power mismatch case. After the occurrence of the fault,
0 the effect of the high gain can be seen with the higher
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Time (s) magnitudes of the parameters as shown in Fig. 8, 9 and 10.
Fig. 6 Voltage difference response when a fault occurs at 0.04 seconds In case of a zero power mismatch, it is usually difficult to
notice an apparent change or fluctuations in the voltage and
Figs. 4 and 5 demonstrate how the variables such as voltage currents at the PCC, as all the power required by the load is
and frequency respond to a grid fault. The changes in these being delivered by the IBDG. For this reason, a higher gain
variables lead to changes introduced in the active and reactive must be selected which would drift both voltage and frequency
power references that in turn lead to changing the currents outside the limits.
references that are finally used to generate the PWM signal. A low pass filter with a cut-off frequency of 160 Hz was
Fig. 6 shows the effect of having an appropriate gain value also used to reduce any high frequency noise introduced in the
that does not impact the stability of the system in grid- system.
connected mode while increasing the sensitivity of the system To track this fault and provide the detection solution, the

318
 2013 IEEE GCC Conference and exhibition, November 17-20, Doha, Qatar

same algorithm was implemented and we came up with the time of 0.021 seconds is chosen and the control signal is
following results. Fig. 8 shows how the voltage (Vd) at the multiplied by the modulation index that is fed to the PWM
PCC behaves as the fault occurs. There is a fluctuation generator. The implemented system is able to perfectly detect
detected, that would be helpful in setting the criteria for islanding and disconnect of the IDBG even during zero power
switching the DPGS off. mismatch scenario.

500 V. CONCLUSION
Voltage, Vd (V)

400 The key advantage of active methods over passive methods

300 is their capability to identify an island even if the active and
200
reactive power mismatch at the PCC is close to zero. This
implies that if all the load power demand is being delivered by
100
the inverter and the grid is not contributing to feeding the load,
0 then an island would be harder to detect with a passive
0.01 0.02 0.03 0.04 0.05 0.06 0.07
Time (s) technique. Having both voltage and frequency sensing at the
Fig. 9. Voltage Vd in the case of zero power mismatch PCC improves the sensitivity of the system as minor
fluctuations and variations can be picked up and detecting the
4000
island would be possible using the positive feedback. This
(Vgrid - Vd) * Gain

3000 paper presented the simulation of an active islanding detection

2000 technique, known as the general electric anti-islanding
1000 technique. The Simulink model was created keeping in view
the concept of positive feedback for islanding detection, as
0
done in this anti-islanding technique. Both voltage and
-1000
frequency drifts were introduced in this model and the
0.01 0.02 0.03 0.04 0.05 0.06 0.07
Time (s) switching off of the IBDG was simply done by comparing the
Fig. 10. Voltage difference at PCC in the case of zero power mismatch voltage values at the PCC to the OUV limits. It is critical to
know the importance of the gain used to amplify the voltage
Fig. 10 simply illustrates the effect of the gain before and and frequency drifts, as it should not affect the system in grid-
after the occurrence of the fault. It is clearly seen that in grid- connected mode and also play a vital role in the fast detection
connected operation, the increment to the current references is of a grid fault.
zero, and effect of the gain is quite apparent in islanding
mode. Finally, the voltage and frequency variations were used VI. REFERENCES
to create an embedded function to trip the IBDG after [1] Hua Geng; Dewei Xu; Bin Wu; Geng Yang, "Active Islanding Detection
occurrence of the fault. Fig. 11 and Fig. 12 illustrate how the for Inverter-Based Distributed Generation Systems With Power Control
currents and voltages at the PCC behave, respectively, before Interface," Energy Conversion, IEEE Transactions on , vol.26, no.4,
pp.1063,1072, Dec. 2011
and after the fault occurrence. [2] Riming Shao; Kaye, M.; Liuchen Chang, "Advanced building blocks of
power converters for renewable energy based distributed
Fault Occurs Ia generators," Power Electronics and ECCE Asia (ICPE & ECCE), 2011
Ib IEEE 8th International Conference on , vol., no., pp.2168,2174, May 30
Current (Ampere)

10
Ic 2011-June 3 2011
[3] Xiaoyu Wang; Freitas, W.; Wilsun Xu, "Dynamic Non-Detection Zones
0 of Positive Feedback Anti-Islanding Methods for Inverter-Based
Distributed Generators," Power Delivery, IEEE Transactions on , vol.26,
-10 no.2, pp.1145,1155, April 2011
IBDG Disconnected [4] Pengwei Du; Zhihong Ye; Aponte, E.E.; Nelson, J.K.; Lingling Fan,
"Positive-Feedback-Based Active Anti-Islanding Schemes for Inverter-
0.01 0.02 0.03 0.04 0.05 0.06 0.07 Based Distributed Generators: Basic Principle, Design Guideline and
Time (s)
Performance Analysis," Power Electronics, IEEE Transactions on ,
Fig. 11. Current at PCC after fault occurs vol.25, no.12, pp.2941,2948, Dec. 2010
[5] Pengwei Du; Aponte, E.E.; Nelson, J.K., "Performance analysis of
Fault Occurs Va positive-feedback-based active anti-islanding schemes for inverter-based
400 Vb distributed generators," Transmission and Distribution Conference and
Vc Exposition, 2010 IEEE PES , vol., no., pp.1,9, 19-22 April 2010
Voltage (V)

200 [6] Hassan, F.; Bollen, M., "Active islanding of a current-controlled

converter-interfaced DG," PowerTech, 2009 IEEE Bucharest , vol., no.,
0
pp.1,6, June 28 2009-July 2 2009
-200 [7] Xiaoyu Wang; Freitas, W., "Impact of Positive-Feedback Anti-Islanding
IBDG Disconnected Methods on Small-Signal Stability of Inverter-Based Distributed
-400 Generation," Energy Conversion, IEEE Transactions on , vol.23, no.3,
0.01 0.02 0.03 0.04 0.05 0.06 0.07 pp.923,931, Sept. 2008
Time (s)
[8] Z. Ye, R. Walling, L. Garces, R. Zhou, L. Li, T. Wang,”Study and
Fig. 12. Voltage at PCC after fault occurs Development of Anti-Islanding Control for Grid-Connected Inverters”
[9] J. Stevens, R. Bonn, J. Ginn, S. Gonzalez, and G. Kern, "Development
It has to be noted that in order for the IDGB to be and testing of an approach to anti-islanding in utility-interconnected
disconnected after the fault detection, a zero-order hold block photovoltaic systems," Sandia National Laboratories, Albuquerque, NM,
in Simulink is used for implementation purposes. A sample Lab Rep. SAND2000-1939, Aug. 2000.

319