International Journal of Power and Energy Systems, Vol. 36, No.

1, 2016

AN ADAPTIVE REACTIVE POWER

PERTURBATION BASED HYBRID
ISLANDING DETECTION METHOD FOR
DISTRIBUTED GENERATION SYSTEMS
Geethi Krishnan∗ and Dattatraya N. Gaonkar∗

Abstract prominent issues which can seriously impair the safety of

the DG and loads [2]. Hence according to current protec-
The intentional islanding operation of distributed generation (DG) tion practices, the DG system should disconnect the DG
systems can bring many benefits to the DG owner, distribution from the utility grid within 2 s [3], [4]. But tripping the
network operators and customers. One of the requirements for DG during a mains failure has limited the benefits offered
intentional islanding operation is the successful detection of the grid by it, particularly when it is capable of supplying the local
events which can lead to islanding of the DG system. This paper load within the statutory voltage, frequency and power
presents the hybrid islanding detection method based on adaptive quality limits. As a result, the IEEE Std. 1547-2003 states
reactive power variation with a passive algorithm that relies on the the implementation of intentional islanding of DGs as one
transient response characteristics for facilitating islanding detection. of its tasks for future consideration [5], [6]. For intention-
The proposed method presented in this paper can overcome the ally sustaining an island, a secure, reliable islanding detec-
problems in the reactive power variation based methods. Extensive tion algorithm which takes minimum time for detection is
studies are conducted for differentiating the transient characteristic required.
features for islanded and non-islanded systems. In this study Islanding detection methods discussed in the literature
the adaptive reactive power sequence is modified to enhance the can be broadly classified as active, passive and communi-
performance under sudden load changes. The proposed method cation based methods [7], [8]. Passive methods directly
proved to detect all probable islanding situations with minimum monitor local parameters like voltage (over voltage protec-
run-on time. The performance of the method is evaluated using tion (OVP)/under voltage protection (UVP)), frequency
simulation results obtained from MATLAB/SIMPOWERSYSTEMS (over frequency protection (OFP)/under frequency protec-
software considering wind power based DG system. tion (UFP)), and rate of change of frequency and detect
islanding when any of these parameters exceed the speci-
fied limits. The deviations in above specified parameters
Key Words during an islanding occurrence are subjected to the varia-
tions in network topology difference, inverter and the local
Distributed generation, islanding, point of common coupling, non-
load and lead to non-detection zones (NDZ). On the other
detection zone
hand, active methods inject disturbances, and the result-
ing parameter variation is used to detect islanding. These
1. Introduction methods have small NDZ comparing with passive methods,
but can degrade the power quality of the system. Fre-
The grid integration of distributed generation (DG) is ex- quency drift methods like active frequency drift, slip mode
pected to play an important role in the future electricity frequency shift, Sandia frequency shift and reactive power
market due to increasing demand of power. DG including variation methods are the prominent active methods dis-
photovoltaic, wind farms, fuel cell and micro turbines can cussed in the literature [9], [10]. Communication methods
resolve the rising concerns for environmental safety while like power line carrier and transfer trip being costly are
supplying a reliable, cost-effective power [1]. Despite all rarely used for small systems [11]. In hybrid methods,
these advantages, so many critical issues need to be consid- active and passive methods are combined and disturbance
ered for effective grid integration. Islanding is one of the is injected only when islanding is suspected by the passive
∗
algorithm.
Department of Electrical Engineering, National Institute of
Technology Karnataka, Surathkal, Karnataka, India; e-mail:
In [12], a low-frequency disturbance signal of 1 Hz
{kgeethi9; dngaonkar}@gmail.com is injected with the reactive current through the q axis

1
Figure 1. Single line diagram of system under study.

controller. A frequency deviation detector is used to hence negative impact of the conventional method has
detect the deviations from rated frequency and to provide been eliminated. Active method comes into actin only
a reliable detection that is not sensitive to grid impedance when passive method suspects the islanding. The devel-
variation. Reference [13] proposed injection of a harmonic oped method is capable of detecting all probable islanding
current reference of 0.5 Hz and detects islanding using situations with minimum run-on time compared with con-
phase jump detection method. In [14], frequency is used ventional method and its performance is evaluated using
as feedback to shift Q-f curve to an unstable operating wind based DG system.
point during islanding. But the proper choice of gain
value is necessary to avoid false misinterpretations due to 2. Study System
other dynamics in grid connected mode. In [15] frequency
is monitored following the injection of a periodic reactive Figure 1 shows the single line diagram of the system.
power reference is less than 2.5% of the rated kVA (kilo The system consists of a DG represented by a DC source,
volt ampere). interfaced to the grid through a voltage source inverter,
Choe et al. [16] deduce the detailed design for ob- LC filter and a breaker. The interface control is designed
taining the reactive power value based on the character- with proportional integral (PI) controllers, with the control
istic analysis of system frequency variation caused by is- variables defined in the synchronous reference frame [25]. A
landing. In [17], it has been given that intermittent bi- phase locked loop (PLL) is used to determine the frequency
lateral reactive power variation (IBRPV) injection with and angular reference at the point of common coupling
magnitude 5% of rated power intermittently can elimi- (PCC). The grid is assumed to be relatively stiff and hence
nate the NDZ. In [18] load resonant frequency is used maintains the PCC voltage and frequency at constant
as an improvement over the IBRPV method. Compar- values. Hence, in grid connected mode, the voltage source
ing with the former method, this method produces uni- inverter is controlled to deliver constant active and reactive
lateral variation by injecting minimum reactive power to power. As the voltages are synchronized to the grid through
drift the frequency out of threshold value .Conventional three-phase PLL, Vq equals zero. Hence the inverter active
hybrid methods monitor different parameters like voltage power and reactive power can be controlled separately by
unbalance, total harmonic distortion (THD), average rate varying the d axis and q axis currents separately, as shown
of change of voltage or intelligent based techniques for in (1).
generating the initial suspicion signal [19]–[22]. Notwith- 3 3
standing its merits, the need for tedious calculation and P = Vd Id , Q = − Vd Iq (1)
2 2
the dependency of the threshold setting on the geograph-
ical and topological variation can act as a hindrance in is the maximum value of the voltage at the PCC and Id
these methods. A new hybrid islanding detection method and Iq are the components of the current in the dq axis.
employing active and passive methods for synchronously In grid connected system, the total load demand is
rotating DGs is proposed in [23]. In [24] a simulation shared between the DG and the grid. But when grid gets
and experimental validation of hybrid islanding detection disconnected, the DG system should have to meet this
scheme using pulse current injection based active method whole load demand alone. As the interface controller is
and power line communication remote technique have been designed to provide some preset constant power, the gener-
presented. ation couldn’t always match the load demand [26]. Hence
In this paper hybrid technique is proposed combin- depending on the load generation mismatch, the magni-
ing the features of adaptive reactive power based active tude and frequency of the PCC voltage vary to meet the
method with a passive algorithm. The proposed method demand. If these parameters exceed the threshold spec-
does not require continuous injection of the disturbance, ified by the standards, islanding will be detected by the
2
OVP/UVP and OFP/UFP methods. But if the load gen- 3.1 Adaptive Reactive Power Calculation
eration mismatch is quite small, there will be insufficient
change in the magnitude and frequency of the PCC volt- The resonant frequency, fres and quality factor, Qf of an
age to activate any standard protection devices. This will RLC load can be expressed as:
result in NDZ [27]–[31]. For a current controlled interface, 
the NDZ can be analytically derived as: 1 C
fres = √ Qf = R (4)
2π LC L
ΔP
[ V /Vmax ] − 1 ≤ ≤ [ V /Vmin ] − 1 (2) From (4) RLC parameters can be expressed as follows:
P
R Qf P
    L= C= (5)
2 ΔQ 2 2 π fres Qf R fres 2 π
Qf 1 − [f /fmax ] ≤ ≤ Qf 1 − [f /fmin ] (3)
P
The active and reactive power consumed by a three-phase
To avoid this NDZ, an external disturbance can be in- RLC load can be expressed as follows. VP CC represents
jected such that the frequency and voltage deviate beyond the rms value of the PCC voltage:
 
the threshold value when islanded. With the constant VP2CC 2 1
power controlled interface, voltage variation is dominated Pload = 3 Qload = 3VP CC − 2πf C (6)
R 2πf L
by active power mismatch and frequency variation by reac-
tive power mismatch. As a result if active power is varied, Combining (4) and (5) with (6), an expression for Qload
the PCC voltage magnitude varies considerably comparing can be derived as:
with the frequency. Whereas for reactive power variation,    
1 fres f
the frequency drift will dominate the voltage variation. Qload = Pload R − 2πf C = Pload Qf −
Reactive power variations are more sensitive and produce 2πf L f fres
notable deviations in frequency to trigger islanding detec- (7)
tion. Moreover they are easy to implement comparing with
the former one. After islanding, the whole load demand is supplied by
the inverter. P and Q represent the output active and
3. Analysis of Adaptive Reactive Power Injection reactive power from the inverter, respectively, so
Method
Pload = P Qload = Q (8)
 
After islanding, with reactive power reference as zero, the fres f
Q = P Qf − (9)
frequency of PCC voltage will deviate to the resonance f fres
frequency of the RLC loads. If this resonance frequency is
beyond the threshold limits, the conventional OFP/UFP where f represents the steady-state frequency after island-
protection can detect islanding successfully without any ing and it varies depending on the inverter output reactive
disturbance injection. But if the resonance frequency is power, Q. Hence we have to inject Q depending upon fres
within the limit [fmin , fmax ], then a reactive power pertur- such that f will just go beyond the threshold value. The
bation is needed to drift the frequency outside the thresh- frequency value which could perturb the f to the frequency
old values [31]. The minimum reactive power perturbation lower bound fmin and upper bound fmax can be expressed
required for different RLC loads vary, depending on their respectively by equations:
resonant frequency characteristics. The adaptive reac-  
fres fmin
tive power perturbation calculates this minimum reactive Q1 = P × Qf − (10)
fmin fres
power reference based on the PCC frequency and injects
 
them as a perturbation sequence in intermittent intervals fres fmax
as shown in Fig. 2. Q2 = P × Qf − (11)
fmax fres
The relationship of Q1 and Q2 with fres is illustrated
in Fig. 3. Quality factor is taken as 2.5. From Fig. 3,
the minimum reactive power variation can be expressed as
follows:
⎧ ⎧
⎪
⎪ ⎪
⎨ Qref = Q1 : fmin < fres < 60 Hz
⎪
⎪
⎪
⎨ 0 ≤ t ≤ T1 :
⎪
⎩Q
⎪ ref = Q2 : 60 Hz ≤ fres < fmax (12)
⎪
⎪
⎪
⎪
⎩ T1 ≤ t ≤ T 1 : Q = 0

Analysing the transient characteristics of the PLL and

Figure 2. Perturbation sequence under different grid fre- considering the time requirements for islanding detection,
quencies. the reactive power variation parameters are given in [18].
3
 Figure 4. PCC frequency variation for different Qf for a
load with resonant frequency 60.2 Hz.
Figure 3. Minimum reactive power perturbation based on
resonance frequency.

3.2 Shortcoming of the Above Method

Although the merits of the above suggested method is

incredible, it does suffer certain drawbacks.
1. The specific detection timing of the islanding depends
on the time at which islanding takes place.
The above specified method injects the reactive power
disturbance intermittently in an interval of 1.8 s. If
islanding happens, just after the injection of one pulse
of reactive power reference, detection is not possible for
Figure 5. PCC frequency variation for different Qf for a
another 1.8 s. Hence, there is a direct dependency of
load with resonant frequency 59.7 Hz.
the islanding detection timing on the time of islanding.
Table 1 provides the study results to validate the
same. The parameters used for the study is included
in the Appendix A (Tables A.1 and A.2). It can be true and the hence the method will fail for loads having
observed that except for Case 1, the time taken for Qf > 2.5. Also, for loads with Qf < 2.5, the reactive
islanding detection is different for islanding occurrence power injection will not be at the required minimum.
at t = 3.5 s, t = 4 s and t = 5.5 s. The variation of PCC frequency of the island with
2. The method fails for loads with quality factor loads having different Qf following same disturbance
(Qf ) > 2.5. injection can be observed in Figs. 4 and 5.
From (10) and (11), it can be seen that the minimum 3. The method can be subjected to mal operation when
reactive power to be injected depends on the quality a simultaneous switching of load happens with the
factor (Qf ). As the RLC load parameters of the reactive power injection.
islanded region are not known, adaptive calculation of As already explained, this method injects disturbance
Qf based on system parameters is not a feasible option. intermittently in an interval of 1.8 s. The magnitude
Hence we took Qf as 2.5 as specified by the IEEE Std. of the injected reactive power is related to the fre-
929-2000 for anti-islanding test systems. This value of quency of PCC voltage of the previous sampling in-
quality factor is assigned, assuming that most of the stant. Consider that system gets islanded at t = 0.5 s.
loads will have Qf < 2.5. But this might not always be The active and reactive power demand from the load

Table 1
Effect of Time of Islanding Occurrence on Specific Detection Time

Time Taken for Islanding Detection (s)

Time of islanding occurrence Case 1 Case 2 Case 3 Case 4 Case 5
Islanding takes place (at t = 3.5 s) 0.08 1.9 0.2 0.2 1.9
Islanding takes place (at t = 4 s) 0.08 1.45 1.45 1.45 1.45
Islanding takes place (at t = 5.5 s) 0.08 1.8 1.8 1.8 1.8

4
Figure 6. Response of an islanded system for reactive power injection without load switching: (a) frequency variation of PCC
voltage; (b) output of islanding detection algorithm; and (c) reactive power disturbance.

Figure 7. Response of an islanded system for reactive power injection with simultaneous load switching: (a) frequency
variation of PCC voltage; (b) output of islanding detection algorithm; and (c) reactive power disturbance.

was 100% matching with the generation. So, a reac- 4. Even if the possibility of islanding is very less, the
tive power reference, Qref of 450 Var (corresponding reactive power variation is to be injected continuously
to fres = 59.95 Hz) is injected at t = 1.8 s. But at the in an interval of 1.8 s. Continuous injection impairs
same time a reactive load with a resonant frequency of the power quality of the system.
60.2 Hz is switched on. As a result, the frequency drift
introduced by the disturbance injection becomes insuf- 4. Proposed Hybrid Detection Method
ficient to detect islanding. As the reactive power refer-
ence will be zero for another 1.8 s, islanding detection To overcome all the above-mentioned problems, some im-
time will go beyond 2 s. provements can be incorporated with the existing tech-
The PCC frequency variation without and with load nique. Hence considering all of the above inferences, a
switching is shown in Figs. 6 and 7, respectively. It new monitoring method is devised. The following section
can be observed that, if there is no load switching, explains the proposed algorithm. The developed method
the islanding is detected within 1.4 s, whereas when classifies all the islands under four categories.
load is switched simultaneous to the reactive power Case 1 : The variation between load and generation
injection, the detection time is 3.15 s, which exceeds is considerably high such that the frequency and voltage
the maximum time limit specified by the standards. go outside the set limits. As a result, islanding can be
5
Figure 8. Response of system parameters to islanding: (a) frequency; (b) active power output; (c) d axis component of the
voltage; and (d) reactive power.

Table 2
Performance of the Proposed Algorithm with Simulated Disturbances

Event Description Expected Result Actual Result

P mismatches(±40%) with RLC load at PCC Islanding Detected
Q mismatches(±5%) with RLC load at PCC Islanding Detected
Load switching ON (50% of inverter output power) Non-islanding Not detected
Capacitor bank of 50 kVar is switched ON Non-islanding Not detected
Load switching OFF (50% of inverter output power) Non-islanding Not detected
Capacitor bank of 50 kVar is switched OFF Non-islanding Not detected
Non-linear load (a 2 kW diode rectifier) Non-islanding Not detected
Motor load switching Non-islanding Not detected
PQ mismatch Islanding Detected
Line to ground temporary fault (fault duration:0.2 s) Non-islanding Not detected

easily detected by directly monitoring these parameters. power reference. The magnitude of the perturbation is
However to avoid false detection, a time delay t1 = 60 ms is calculated based on the resonant frequency, as indicated in
considered. Fig. 3. In case of islanding, frequency deviates beyond the
Case 2 : Frequency deviation is well beyond the thresh- limits islanding will be detected.
old value because of the mismatch in reactive power. But Case 4 : The most difficult case for detecting an is-
as the active power demand of load and generation match, land is when both the voltage and frequency at PCC
voltage variation is less than set limit. As per IEEE 1547, stays within the threshold limits. However it had been
following a variation in frequency, >60.5 Hz or <59.3 Hz, observed that the transients of the voltage, frequency, ac-
<160 ms, DG should shut down. Adhering that rule, if such tive and reactive power of the inverter show some pat-
a condition retain for 120 ms, detection of the islanding tern in an islanded system which differs from the re-
takes place. sponse due to other power system dynamics. If this
Case 3 : The active power mismatch between load and pattern is known, it can be used as a trigger for gen-
generation pulls the voltage outside the limits, whereas erating an islanding suspicion signal. Figure 8 shows
frequency is within the value. To cross-check the formation an example of the response of these parameters during
of island, a small perturbation is given to the reactive islanding.
6
Figure 9. Proposed detection algorithm.

For case 4, the proposed algorithm follows a procedure values need to be chosen such that no false operation
as detailed in the steps below (also shown in Fig. 9.) happens for dynamics of the power system.
Step 1. The monitoring unit keeps on monitoring Step 3. Check for any continuous deviation in the
whether any change is observable in active (P ) and active power and voltage.
reactive (Q) power, voltage (V ) and frequency (f ). Step 4. Generate an islanding suspicion signal, if any
Step 2. After a significant change in reactive power one of the above two statements are true. If not, start
whether frequency is deviating and settling at a new from step 1.
operating point. Step 5. As per (12) calculate the required reactive
power to be injected as disturbance corresponding to
⎧ ⎫ the resonant frequency for 100 ms.
⎪
⎪ Pold + ΔP < Pnew < Pold − ΔP ⎪
⎪
⎪
⎪ ⎪
⎪ Step 6. If frequency is less than 59.5 Hz or greater
⎪
⎪ ⎪
⎪
⎨ V + ΔV < V ⎬ than 60.5 Hz, then islanding condition is detected.
old new < Vold − ΔV
(13) The performance of the proposed algorithm is tested
⎪
⎪ fold + Δf < fnew < fold − Δf ⎪
⎪
⎪
⎪ ⎪
⎪ for the system shown in Fig. 1 considering 10 kW inverter
⎪
⎪ ⎪
⎪
⎩ Q=  0 ⎭ based DG system. The results of this study is reported in
Table 2.

Pold , Vold and fold represent the inverter active power, 5. Analysis of the Developed Method with Wind
PCC voltage and frequency of the last sampling in- Power System
stant, respectively. Pnew , Vnew and fnew represent the
same parameter values for the new sampling instant, This section discusses the performance of the algorithm
respectively. ΔP, ΔV and Δf are the set deviations during islanding and non-islanding events for a wind power
limit of the inverter active power, PCC voltage and system, with common power system events. Different
frequency from the nominal values, respectively. These cases such as load variation, voltage sag, switching of the
7
Figure 10. Synchronous generator based wind DG system.

Figure 11. Response of the wind based system for load switching: (a) voltage in pu; (b) inverter active power output in MW;
(c) frequency in Hz; and (d) inverter reactive power output in Mvar.

non-linear load has been considered to prove the effec- modelled using an RLC branch which is designed to
tiveness of the developed technique. A 10 MW wind sys- consume 10 MW at unity power factor (UPF). At time
tem with synchronous generator model from MATLAB/ t = 3 s, UPF load rated at 2 MW is switched in. As
SIMPOWERSYSTEM is used for the study. The simulated shown in Fig. 11, during the switching instant, the
system used in this study is shown in Fig. 10. frequency of the voltage shows transients from the
• Load switching nominal values. However within 100 ms, the transients
Initially the system was in grid integrated mode with died down and the voltages settle down to the steady-
constant power control. The load of the inverter is state value. Because of the 100 ms delay, programmed
8
Figure 12. Response of the wind based DG system for voltage sag: (a) voltage; (b) inverter active power output; (c) frequency;
and (d) inverter reactive power output.

Figure 13. Response for islanding: (a) voltage; (b) frequency; (c) inverter active power output; (d) inverter reactive power
output; (e) reactive power reference; and (f) output of islanding detection algorithm.

within the monitoring system, reactive power is not a reduction of 0.25 pu in the grid voltage, sustaining
injected. from 3 to 3.1 s can be observed in Fig. 12. This specific
• Voltage sag scenario is classified under Case 3 by the monitoring
This section explains the performance analysis of the algorithm. As indicated earlier, the delay time for
system during voltage sag. The system response for injecting perturbation is 100 ms. The variation in
9
 Table 3
Specific Detection Time for a Wind Power System

Cases Method Islanding detection time (s)

1 2 3 4 5
Islanding takes place (at t = 3.5 s) Conventional 0.08 1.9 0.2 0.2 1.9
Hybrid 0.19 0.65 0.47 0.4 0.7
Islanding takes place (at t = 4 s) Conventional 0.08 1.45 1.45 1.45 1.45
Hybrid 0.19 0.65 0.47 0.4 0.7
Islanding takes place (at t = 5.5 s) Conventional 0.08 1.8 1.8 1.8 1.8
Hybrid 0.19 0.65 0.47 0.4 0.7

Figure 14. Response of the system for non-linear load switching: (a) magnitude of PCC voltage in pu; (b) inverter active
power output in MW; (c) frequency of PCC voltage in Hz; and (d) inverter reactive power output in Mvar.

the grid voltage settles before 100 ms, and no reactive • Islanding occurrence on effect on detection time
power injection takes place. A comparative study regarding the dependency of
• Islanding the time of islanding and the detection time required
The system is islanded at t = 3 s as shown in Fig. 13. for conventional and proposed algorithm had been
The local RLC load resonates at 60.2 Hz. There is no conducted. For the conventional method, detection
mismatch in the active power between the load and time clearly dependent on the time of occurrence of
the generation. This situation is sorted under Case island. Whereas for the proposed algorithm; the time
4 by the monitoring algorithm. Immediately after of detection shows consistency and not depends on the
the disconnection of the grid, reactive power output time of occurrence of the islanding event. The results
of the inverter deviates from zero and slowly returns are given in Table 3.
back. Frequency varies slowly and settles at a new • Switching of a non-linear load
value depending on the resonating condition. This A non-linear load rated at 2 MW is switched in at
initiates the timing counter of the algorithm. When t = 3 s. There is an initial shift in the magnitude of
the count reaches 100 ms, a reactive power of −0.026 pu PCC voltage, which deviates beyond the lower thresh-
is injected in to the grid. As a result, frequency value old value of 0.88 pu. However the system returned
exceeds the set value of 60.5 Hz and islanding happens back and settled at the nominal value before 50 ms as
(Fig. 13). shown in Fig. 14.
10
6. Conclusion References
This paper presented a hybrid technique for detecting [1] T. Ackermann, G. Andersson, and L. Soder, Distributed
islanding condition effectively. The method uses an generation: A definition, Electric Power Systems Research,
existing reactive power injection technique with a novel 57(3), 2001, 195–204.
[2] N.D. Hatziargyriou and A.P.S. Meliopoulos, Distributed energy
transient characteristics based monitoring algorithm. sources: Technical challenges, IEEE Power Engineering Society
Extensive studies are conducted for identifying the differ- Winter Meeting, New York, USA, 2002, 1017–1022.
entiating transient characteristic features during islanding. [3] IEEE Std. 1547, Standard for interconnecting distributed re-
The performance of the proposed method has been evalu- sources with electric power systems, 2003.
ated using wind based DG system using simulation results. [4] IEEE Std. 929, IEEE recommended practice for utility interface
of photovoltaic (PV) systems, 2000.
The study presented indicates relatively accurate detection [5] I.J. Balaguer, Q. Lei, S. Yang, U. Supatti, and F.Z. Peng,
for most of the practical cases. This technique can be Control for grid-connected and intentional islanding opera-
implemented without much change in the control strategy tions of distributed power generation, IEEE Transactions on
and is well suited for intentional islanding operations. The Industrial Electronics, 58(1), 2011, 147–157.
[6] H. Zeineldin, E.F. El-Saadany, and M.M.A. Salama, Intentional
NDZ of this method is dependent on the ΔP , ΔV and Δf islanding of distributed generation, IEEE Power Engineering
magnitudes in addition to the quality factor of the load. Society General Meeting, San Francisco, CA, USA, 2005,
Wise choice of these parameters can make the method 1496–1502.
a viable approach for detecting islanding with minimum [7] W. Bower and M. Ropp, Evaluation of islanding detection
methods for photovoltaic utility-interactive power systems,
power quality impacts. International Energy Agency, Technical Report. IEA PVPS
T5-09, 2002.
[8] Z. Ye, R. Walling, L. Garces, R. Zhou, L. Li, and T. Wang,
Study and development of anti-islanding control for grid-
Appendix A connected inverters, General Electric Global Research Centre,
NREL/SR-560-36 243, 2004.
The system, DG interface, and load parameters and perfor- [9] P. Mahat, Z. Chen, and B. Bak-Jensen, Review of islanding de-
mance analysis for the different islanding studies performed tection methods for distributed generation, Proc. 3rd Interna-
tional Conf. on Electric Utility Deregulation and Restructuring
are given in Tables A.1 and A.2. and Power Technologies, Nanjing, China, 2008, 2743–2748.
[10] W. Xu, K. Mauch, and S. Martel, An Assessment of DG
Islanding Detection Methods and Issues for Canada, CETC-
Varennes 2004-074 (TR), CANMET Energy Technology Centre
– Varennes, Natural Resources, Canada, July 2004, 53.
Table A.1 [11] W. Xu, G. Zhang, C. Li, W. Wang, G. Wang, and J. Kliber, A
Parameters Used for Case Studies power line signaling based technique for anti-islanding protec-
tion of distributed generators – Part i: Scheme and analysis,
Cases R (Ω) L (mH) C (mF) IEEE Transactions on Power Delivery, 22(3), 2007, 1758–1766.
[12] G.H. González and R. Iravani, Current injection for active
1 7.2 7.7166 0.93034 islanding detection of electronically interfaced distributed re-
sources, IEEE Transactions on Power Delivery, 21(3), 2006,
2 4.33 4.6407 1.54694 1698–1705.
[13] J.B. Jeong, H.J. Kim, S.H. Back, and K.S. Ahn, An improved
3 7.2 7.639 0.921036 method for anti-islanding by reactive power control, Proceedings
of the 8th International Conference on Electronic Mechanical
4 4.33 4.6173 1.5392 Systems, 2, 2005, 965–970.
[14] H.H. Zeineldin, A Q-f droop curve for facilitating island-
ing detection of inverter-based distributed generation, IEEE
Transactions on Power Electronics, 24(3), 2009, 665–673.
[15] J.B. Jeong, H.J. Kim, K.S. Ahn, and C.H. Kang, A novel
Table A.2 method for anti-islanding using reactive power, Proc. 27th In-
System Parameters ternational Telecommunication Conf., Berlin, Germany, 2005,
101–106.
[16] G.-H. Choe, H.-S. Kim, H.-G. Kim, Y.-H. Choi, and J.-C.
Inverter Parameters Kim, The characteristic analysis of grid frequency variation
under islanding mode for utility interactive PV system with
Switching frequency 10 k Hz reactive power variation scheme for anti-islanding, Proc. 37th
IEEE Power Electronics Specialists Conf., Oralando, Florida,
Input DC voltage 400 V
USA, 2006, 1–5.
Filter inductance 4.2 mH [17] J. Zhang, D. Xu, G. Shen, Y. Zhu, N. He, and J. Ma, An
improved islanding detection method for a grid-connected
Filter capacitance 31 μF inverter with intermittent bilateral reactive power varia-
tion, IEEE Transactions on Power Electronics, 28(9), 2013,
Filter resistance 0.05 Ω 268–278.
[18] Y. Zhu, D. Xu, N. He, J. Ma, J. Zhang, Y. Zhang, G. Shen, and
Output voltage (phase to ground) 120 Vrms C.Hu, A novel RPV (reactive-power-variation) anti-islanding
method based on adapted reactive power perturbation, IEEE
Output power 20 kW Transactions on Power Electronics, 28(11), 2013, 4998–5012.
[19] P. Mahat, Z. Chen, and B.B. Jensen, A hybrid islanding
Grid parameters: Frequency 60 Hz, inductance 0.05 mH detection technique using average rate of voltage change and
Resistance: 0.02 Ω real power shift, IEEE Transactions on Power Delivery, 24(2),
2009, 764–771.
11
[20] V. Menon and M.H. Nehrir, A hybrid islanding detection Dattatraya N. Gaonkar received
technique using voltage unbalance and frequency set point, his Ph.D. from the Indian In-
IEEE Transactions on Power Systems, 22(1), 2007, 442–448. stitute of Technology, Roorkee,
[21] K. El-Arroudi, G. Joos, I. Kamwa, and D.T. McGillis, Intelli-
gent based approach to islanding detection in distributed gen- India, in 2008. He was a visit-
eration, IEEE Transactions on Power Delivery, 22(2), 2007, ing scholar at the University of
828–835. Saskatchewan, Canada. He also
[22] S.-I. Jang and K.-H. Kim, An islanding detection method
visited Center for Future Energy
for distributed generations using voltage unbalance and total
harmonic distortion of current, IEEE Transactions on Power Systems (CFES) at Rensselaer
Delivery, 19(2), 2004, 745–752. Polytechnic Institute (RPI), Troy,
[23] M. Moradzadeh, M. Rajabzadeh, and S.M.T. Bathaee, A novel New York, and Energy Reliability
hybrid islanding detection method for distributed generations,
Proc. 3rd International Conf. on Electric Utility Deregulation
& Security Laboratory at Michi-
and Restructuring and Power Technologies, Nanjing, China, gan State University (MSU) at
2008, 2290–2295. East Lansing, USA, and Energy Management and Micro-
[24] C.-C. Hou and Y.-C. Chen, A hybrid islanding detection grid Laboratory at NUS, Singapore, for research interac-
for distributed generation systems using pulse current injec-
tion, IEEE International Symposium on Industrial Electronics tion. He has edited and written a chapter in the book
(ISIE), Hangzhou, China, 2012, 1554–1559. titled “Distributed Generation”, which is published by
[25] T.N. Thacker, Control of power conversion systems for in- INTECH publication, Austria. He has published many
tentional islanding of distributed generation systems, Virginia articles in international journals and conferences and is a
University Technical Report, 2005.
[26] H. Zeineldin, E.F. El-Saadany, and M.M.A. Salama, Impact of senior member of the IEEE. His main research interests
DG interface control on islanding detection and non-detection are in the areas of power system operation and control,
zones, IEEE Trans. Power Del., 21(3), 2006, 1515–1523. distributed generation and power electronics.
[27] Z. Ye, A. Kolwalkar, Y. Zhang, P. Du, and R. Walling,
Evaluation of anti-islanding schemes based on non detection
zone concept, IEEE Transactions on Power Electronics, 19(5),
2004, 171–1176.
[28] W.K.A. Najy, H.H. Zeineldin, and A.H.K. Alaboudy, A
Bayesian passive islanding detection method for inverter based
distributed generation using ESPRIT, IEEE Transactions on
Power Delivery, 26(4), 2011, 2687–2696.
[29] K. Geethi and D.N. Gaonkar, Control of grid connected and
islanding operations of distributed generation systems with
seamless transfer between modes, Proc. IEEE Multi-Conf. on
Systems and Control(MSC), Hyderabad, India, 2013.
[30] K. Geethi and D.N. Gaonkar, Intentional islanding of dis-
tributed generation system with a load shedding algorithm,
Proc. IEEE International Conf. on Power Electronics, Drives
and Energy Systems (PEDES-2012), CPRI, Bangalore, India,
2012.
[31] K. Geethi and D.N. Gaonkar, Performance evaluation of a
new hybrid islanding detection method for a wind based DG,
Proc. IEEE Conf. on Clean Energy and Technology, Langkawi,
Malaysia, 2013.

Biographies

Geethi Krishnan received her

M.Tech. (by research) from the
National Institute of Technology
Karnataka Surathkal, India, in
the year 2014. She worked as an
Assistant Professor in the Depart-
ment of Electrical and Electronics
Engineering, Manipal Institute
of Technology, Karnataka, India.
Presently working as a senior
research fellow at IIT Mumbai,
India. Her interests are in the
areas of distributed generation, power electronics and
control systems.

12