You are on page 1of 6

Proceedings of the IEEE SoutheastCon 2015, April 9 - 12, 2015 - Fort Lauderdale, Florida

A Passive Islanding Detection Approach for


 

Inverter-based
 
Distributed Generation using Rate of
 
Change of Frequency Analysis
 
Bikiran Guha, Rami J. Haddad, Youakim Kalaani
  Department of Electrical Engineering
Georgia Southern University
  Statesboro, Georgia 30458
Email: {bg02728, rhaddad, yalkalaani}@georgiasouthern.edu
 

 
Abstract—Islanding occurs when a Distributed Generation to encourage their use and installations [4]. Furthermore,
(DG) source continues to energize an isolated section of a power centralized power systems are vulnerable to outages and can
system  even after it was disconnected from the main power be prone to major blackouts as witnessed in recent years [5].
grid. Since islanding can cause hazardous conditions to people
and equipment, current utility standards require that islanding DG sources can reduce the impact of outages by sustaining
 
be quickly detected by protective relays and inverters that are part of the load.
parts of the DG system. Passive islanding detection techniques, However, there are still several challenges imposed by the
unlike  their active counterparts, monitor system parameters integration of DG systems that need to be overcome. One
without injecting any disturbance into the grid. Although widely major challenge is unintentional islanding which can present
used, passive detection techniques are not very effective in
  islanding especially in cases where there is small power
detecting serious safety hazards and technical problems. Islanding oc-
mismatch and they also may trigger false detection in some non- curs when a DG unit continues to energize part of the power
islanding
  cases. To address these drawbacks, a novel and effective system disconnected from the rest of the power grid. Prolonged
passive islanding detection technique that conforms to standard islanding can be dangerous and even fatal to field personnel
regulations has been presented in this paper. The proposed since they may not be aware that a serviced power line may
  technique is based on monitoring the oscillations in
detection
the Rate of Change of Frequency (ROCOF) measured at the still be energized by a nearby DG source [6]. Since most of the
Point  of Common Coupling (PCC) in the system. The proposed existing DG systems are independently owned, electric utilities
detection technique was developed and tested on a grid connected have limited or no control over the DG units. When islanding
photovoltaic DG system using simulation. Results indicated that occurs, the DG voltage in that isolated area will go out of
 
this technique was not only capable of detecting islanding when phase with the rest of the power system. Reclosing in such a
it occurs but also able to accurately distinguish between islanding
and non-islanding under a wide range of operating conditions. condition could introduce damaging high currents and torques
  into the power grid. Reclosing may even fail if the DG source
Keywords—Anti-Islanding, Islanding Detection, Rate of change continues to feed the fault at the islanding side resulting in
 
of frequency, Distributed Generation, Passive islanding detection, prolonged outages. Standard regulations such as IEEE 1547
Power system protection, Smart Grid. [7] series and UL 1741 [8] require islanding to be detected
  within 2 seconds and require the DG sources to shut down if
I. I NTRODUCTION the islanded load cannot be sustained.
A decentralized
  electrical generation source connected to There has been considerable research efforts to develop
the distribution level of the power grid is called a Distributed quick and robust islanding detection techniques which are
Generation
  (DG) unit [1]. DG units provide solutions to classified as active, passive, machine learning and remote
various issues associated with the current structure of the techniques. Remote techniques such as the PLCC [9] and the
power  system. Traditional power systems are not efficient since SCADA [10] require collecting information from both the util-
centralized power plants have to deliver large amount of power ity and DG sides to detect islanding. PLCC monitors signals
over long distances which result in considerable losses [2]. from the utility side and detect islanding when those signals
 
In addition, environmental regulations and prohibitive costs disappear. On the other hand, SCADA detect islanding using
associated with new power plant have exacerbated these information from the auxiliary contacts of the circuit breakers.
 
problems [3]. DG sources are located near the load points Remote techniques are very reliable but they are expensive to
which can help resolve this issue. In addition, they can help implement because of the need to install transceivers and other
reduce  demand on the national power grid by providing monitoring devices [11].
much needed power. Renewable energy sources such as solar, Active techniques continuously inject a form into the sys-
wind,   biomass, and fuel cells play an important part as green tem [12] or change the power output reference of the DG
DG sources, therefore governments are providing incentives source [13]. In a grid-connected system, the grid absorbs local
 

7300-5/15/$31.00 ©2015 IEEE


978-1-4673-7300-5/15/$31.00
978-1-4673-7300-5/15/$31.00 ©2015 IEEE
©2015 IEEE
Proceedings of the IEEE SoutheastCon 2015, April 9 - 12, 2015 - Fort Lauderdale, Florida

 
disturbance and no deviation is observed in the system voltage DG system exceeds a certain threshold for a specific amount
 
or frequency. However, in the case of islanding, the disturbance of time. The monitoring is performed in the time domain
can be significant resulting in noticeable deviation. Active which simplify its integration in DG inverters. Furthermore,
  techniques are mostly efficient but can fail to detect
detection this technique has no NDZ limitation, therefore it can be used
islanding under certain conditions and may degrade the power concurrently with existing OUV/OUF techniques in modern
quality  due to the constant injection of disturbance [14]. DG inverters. The rest of the paper is organized as follows.
Machine learning detection (MLD) techniques are based The proposed technique and its development are presented in
 
on machine learning algorithm that can be trained to de- Section II, System modeling and simulation are discussed in
tect islanding using large sets of sample data representing Section III, and conclusions are presented in Section IV.
islanding
  and non-islanding events. A combination of local II. P ROPOSED ISLANDING DETECTION TECHNIQUE
parameters such as voltage, frequency, power factor, and
others  are used for training. Some popular algorithms used are DG inverters produce harmonics due to high frequency
Artificial Neural Networks [15], Decision Tree Models [16], switching, dead time effect, and DC link voltage ripples [17].
and Support Vector Machines [17]. MLD techniques have to be During islanding, harmonics cause distortions in the voltage
  waveform at the PCC level (VP CC ) that may not be noticeable
extensively trained offline to be able to accurately distinguish
between islanding and non-islanding events which requires in a grid-connected system. However, distortions in the VP CC
 
considerable computing resources not available to small DG waveform can cause sustained variations in the frequency
systems. leading to increased ROCOF. These changes in the ROCOF
In  passive techniques, local parameters such as voltage waveform can be used to detect islanding. On the other hand,
and frequency are monitored at the PCC level and islanding non-islanding events such as faults and capacitor switching
 
is detected if there are changes beyond a certain threshold can also cause transient spikes in the ROCOF waveform but
level. Over-under voltage and over-under frequency protection usually they settle down within a short period of time. The
 
(OUV/OUF) technique is a passive technique that detects difference between the ROCOF waveform for islanding and
islanding when the voltage or the frequency of the system non-islanding events are illustrated in Figure 1.
exceeds  a predefined value [11]. Other popular passive tech-
(a) Islanding event with zero power mismatch
niques rely on monitoring more elaborate parameters such as 5

phase  jump, rate of change of frequency, and rate of change Islanding

of output power [10], [18]. Most passive techniques have been


ROCOF

0
integrated
  in modern DG systems since they are not only sim-
ple and easy to implement but also accurate and cost effective.
The detection time and accuracy of passive techniques depend
  −5
0 0.5 1 1.5
on the selected threshold for islanding detection and have been Time (s)

extensively investigated in the literature [19–22]. Thresholds (b) Three phase short circuit fault (non−islanding event)
 
are generally set to prevent nuisance tripping for non-islanding
5
Fault

events such as short-circuit faults, load switching, and loss of


  feeder. This has the effect of increasing the detection
parallel
ROCOF

time to a value which is greater than 500 ms.


 
Passive techniques have large non-detection zones (NDZ).
They can also fail to detect islanding in case of small power −5
0 0.5 1 1.5
Time (s)
 
mismatch [19]. The NDZ varies based on the technique
and the threshold used. In general, these techniques fail to Fig. 1. Variations in the ROCOF waveform
detect  islanding if the real power mismatch is less than 10%
and the reactive power mismatch is less than 3%. Other Therefore, the ripple content in the ROCOF waveform is
passive  techniques convert the time domain parameters into monitored and used to detect islanding whenever a predefined
frequency domain using Fourier or wavelet transform [23]. threshold is triggered. As shown in Figure 2, a 60 Hz low pass
Frequency
  domain techniques detect islanding when one or filter (LPF) with a cut-off frequency (fc ) is used to eliminate
more frequency components exceed a certain threshold level. the noise in the ROCOF input. The ripple content in the
These techniques usually have a much smaller NDZ than ROCOF waveform is amplified by taking the instantaneous
 
the time domain passive techniques. However, the process of derivative of the filtered signal. After elimination of high
optimizing the threshold in the frequency domain is difficult frequency noise using a second LPF at the same fc , the root
since  non-linear loads can mimic the frequency response of mean square (RMS) of the resultant signal is measured to
islanding events [14]. quantify the ripple content and eliminate any DC component.
 
A computationally inexpensive passive technique for in- The output of the RMS block is smoothed by a third LPF to
verter based DG system is proposed in this paper. The pro- obtain the desired islanding detection waveform (IDW). The
posed  technique detects islanding when the ripple content in frequency of the calculated RMS (fRM S ) and fc of the third
the rate of change of frequency (ROCOF) waveform in the LPF have been set to 9 Hz and 6 Hz respectively. These values
 

7300-5/15/$31.00978-1-4673-7300-5/15/$31.00
©2015 IEEE ©2015 IEEE
Proceedings of the IEEE SoutheastCon 2015, April 9 - 12, 2015 - Fort Lauderdale, Florida

 
are chosen to get a smooth IDW waveform and set an optimal TABLE I
  level.
threshold S PECIFICATIONS OF THE SIMULATION MODEL

Parameter Value
Input
  LPF Derivative LPF RMS LPF
PV array rated power output 100 kW
Detection Detection
fc=60Hz fc=60Hz fRMS=9Hz fc=6Hz >=
delay output Load quality factor (Qf ) 2.5
  Threshold Load resonant frequency 60 Hz
Inverter switching frequency 1980 Hz
  Fig. 2. Block diagram of proposed detection technique Filter resistance 2 mΩ
Filter inductance 0.2 mH
The  relationships between the input ROCOF and IDW Transformer nominal power 150 kVA
waveforms can be mathematically modeled as follows: Infinite bus short circuit MVA 2500
  Infinite bus voltage 120 kV
  
d ROCOF
  IDW = (1)
dt
RM S breaker on the utility side is switched open in order to simulate
Let  α be the selected threshold, td be the predefined the occurrence of islanding. The utility grid model shown in
detection delay, and Δt be the time delay during which IDW Figure 3 (b) consists of a 25 kV distribution feeder and a 120
> α, then
  the detection signal OID can be obtained from the kV equivalent transmission system. The distribution feeder is
following equation, connected to two large loads ([2 MW], [30 MW, 2 MVAR])
  and a 120 kV infinite bus via the substation transformer.

1 (islanding) if IDW ≥ α , Δt ≥ td
OID = (2)
  0 (non − islanding) if otherwise Vabc_B1 PCC 3-phase voltage

Iabc_B1 PCC 3-phase current


Vdc
DC Link voltage
  +
g
+
A A a A a a A A a
N N
B b
III. S YSTEM M ODELING AND S IMULATION -
B Bb b B B b

  -
C C c
R-L filter
C c
PCC
c
260V/25kV
C C c
Circuit
Three phase
The performance of the proposed islanding detection tech- PV array
inverter
measure Transformer breaker Utility Grid
with DC-DC converter
nique   was investigated using a grid-connected PV system

A
B
C

B
C
modeled in a MATLAB/Simulink environment as shown in
Figure 3 (a). The model is designed to conform to the islanding
  Capacitor Load
detection test system specified in IEEE 1547 and UL 1741 Bank

series. System specifications are listed in Table I. The PV (a)


model  simulated the characteristics of SunPower (SPR-305) 1
A + + a A A
modules. The open-circuit voltage and short-circuit current of 2 B
B C
each  module are 64.2 V and 5.96 A respectively. The PV 3
5-km Feeder V Infinite Bus
C
array consisted of 66 strings of 5 series-connected modules Connection to
 
in parallel to produce rated power at 1000 W/m2 irradiance DG side
A

2 MW
o Load
and 25 C ambient temperature. It is connected to a DC-DC Grounding
Resistance
  with switching duty cycle controlled by a Maximum
converter
Power Point Tracking (MPPT) controller. The purpose of the
(b)
MPPT  controller, as its name indicates, is to extract the max-
imum power from the PV as irradiance and temperature vary. Fig. 3. Simulink models of (a) DG System. (b) Utility grid
The MPPT
  controller is implemented using the Incremental
Conductance and Integral Regulator technique. The converter To verify performance, a wide range of islanding and non-
is connected
  to a 3-phase inverter to convert DC into AC islanding events were simulated to present different types of
and to synchronize with the grid. The inverter is programmed disturbance as shown in Figure 4. The threshold and detection
to output constant power at unity power factor. The load is delay were chosen based on simulation results under islanding
 
connected as a parallel RLC circuit with constant impedance. and non-islanding conditions. The IDW values for normal
Load quality factor (Qf ) is selected to be 2.5 and resonant grid-connected operation are assumed to be under 55Hz/s2 .
 
frequency is set at 60 Hz to simulate the worst case scenario For islanding events, the minimum recorded value for IDW
defined by IEEE 1547 [7] and UL 1741 [8] interconnection was 65Hz/s2 , therefore, the threshold value (α) was set
 
standards. The output voltage of the inverter is 260 V and it is at 60Hz/s2 . The maximum settling time recorded for non-
connected to the grid via a step-up transformer. The RL filter islanding events was 280 ms, therefore td was chosen to be
 
and capacitor bank are used to filter harmonics produced by the 300 ms. The waveforms and detection signal for each of the
inverter and also to provide reactive power. The 3-phase circuit simulated conditions are depicted in Figure 5 through Figure 8.
 

7300-5/15/$31.00978-1-4673-7300-5/15/$31.00
©2015 IEEE ©2015 IEEE
Proceedings of the IEEE SoutheastCon 2015, April 9 - 12, 2015 - Fort Lauderdale, Florida

 
These results are discussed in subsections A and B. IDW (a)
IDW OID

 
waveform is plotted in per unit values with α as the base
3

Amplitude (pu)
2
value. 1
Islanding

  0
Detection

Zero power mismatch 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
  3
(b)

Islanding Real power mismatch

Amplitude (pu)
2

  1
Islanding
Reactive power mismatch Detection
0
Simulation
 matrix
0 0.2 0.4 0.6 0.8 1
Time (s)
1.2 1.4 1.6 1.8 2

(c)
3

Amplitude (pu)
  Three phase fault
2
Islanding
1
Detection
0
  Non-islanding
Load switching
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
Capacitor bank switching
  Fig. 5. IDW and Odetection for a) Zero mismatch b) 25% real power deficit
Loss of parallel feeder c) 1% reactive power deficit
 
Motor starting
condition. It is observed that IDW stays well above α after
 
Fig. 4. Matrix of Islanding and Non-Islanding Simulated Events islanding resulting in successful islanding detection.
The effect of changing the generating capacity of the DG
  unit was also investigated. Figure 6 shows the IDW for a
A. Islanding Events zero mismatch islanding case with different DG capacities of
All  the simulated islanding cases were chosen to lie within 50 kW, 100 kW, and 150 kW. In the grid-connected mode,
the NDZ of the OUV/OUF techniques currently used for the IDWs were almost similar in all scenarios. In islanded
 
protection in modern PV inverters. According to IEEE 1547 mode, although the waveforms look different, their minimum
and UL 1741 regulations, OUV/OUF protection should not levels were very close to each other and significantly above the
  system when the voltage and frequency are within
trip the predefined threshold. Similar patterns were observed for other
88% ≤ V ≤ 110% and 59.3Hz ≤ f ≤ 60.5Hz respec- islanded scenarios. Therefore, the technique’s performance is
tively.  For these bounds, the percentage range of real (ΔP ) not dependent on the DG capacity.
and reactive (ΔQ) power mismatch not leading to islanding
detection
  are given in Equations 3 and 4 respectively [19]. In 7
50 kW
these equations, P is the inverter rated power output. 6 100 kW
150 kW
  ΔP
−17.36% ≤ ≤ 29.13% (3) 5
P
 
Amplitude

ΔQ
−5.94% ≤ ≤ 4.11% (4) 3
  P
Positive (+) and negative (-) values for real and reactive 2

power  indicate scenarios where there is a deficit (-) or excess 1


Islanding

(+) in the power mismatch. For verification purposes, the


  power mismatch for islanding events were −20% ≤
simulated 0
0 0.5 1 1.5 2 2.5 3
Real power mismatch ≤ 30% and −6% ≤ Reactive power Time (s)

mismatch
  ≤ 6%. The plots for IDW and OID for 3 different
Fig. 6. IDW for islanding with zero mismatch and different DG capacities
islanding events are illustrated in Figure 5. As indicated,
islanding occurred at 0.5 s. Figure 5 (a) represents the zero
  B. Non Islanding Events
power mismatch condition which is considered the hardest to
detect. Thus, the proposed technique was able to successfully In this section, the response of the proposed technique to
detect  islanding within 400 ms. Figure 5 (b) represents a non-islanding events is presented. The selected non-islanding
real power deficit of 25% which was also detected within events were known to cause nuisance trips in passive detection
  After islanding, IDW almost reaches the threshold
400 ms. techniques. Figure 7 (a) depicts the IDW responses for a
(α) for a very short period of time but never goes below load switching scheme. A large load (100 MW, 50 MVAR)
 
it. Therefore, successful detection of islanding is once again connected on the utility side was switched OFF at 0.5 s and
observed. Figure 5 (c) shows 1% positive reactive mismatch switched ON at 2 s. Both of these events are known to trigger
 

7300-5/15/$31.00978-1-4673-7300-5/15/$31.00
©2015 IEEE ©2015 IEEE
Proceedings of the IEEE SoutheastCon 2015, April 9 - 12, 2015 - Fort Lauderdale, Florida

 
transient spikes in the inverter frequency which is reflected (a) Three phase fault
IDW OID

in the  IDW waveform. However, these spikes settled down


3

Amplitude (pu)
2 Fault
within the time delay (300 ms), therefore no false islanding 1
 
was detected. 0
No false detection
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Figure 7 (b) shows the IDW waveform for switching a large Time (s)
  bank connected to the 25 kV side and rated at 3
capacitor 3
(b) Loss of parallel feeder

Amplitude (pu)
MVAR. Similar to the previous event, the capacitor bank is 2 LOPF
  OFF at 0.5 s and back ON at 2 s. In both events,
switched 1

0
the IDW exhibited sharp peaks for less than 250 ms. The No false detection
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
  technique did not falsely detect islanding since the
proposed Time (s)
(c) Motor starting
settling time was less than the allowable time delay td . 3

Amplitude (pu)
 
In addition, short-circuit faults were also simulated both on
2 Motor started

1
the utility side and on the DG side for single line-to-ground, 0 No false detection
 
line-to-line, and 3-phase faults. It was observed that among 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Time (s)
all the non-islanding events, the 3-phase faults had the largest
IDW  oscillation settling time. Fig. 8. IDW and Odetection for different non-islanding events

  TABLE II
IDW OID
(a) Load switching ROCOF - BASED DETECTION TECHNIQUE COMPARISON
3

2.5   Technique Detection NDZ False Trigger


2 Opening Closing
Amplitude (pu)

1.5 Parameter Probability


1   Existing ROCOF-based ROCOF absolute Large High
0.5
detection techniques value
0
No false detection
−0.5
  Proposed detection ROCOF ripple Zero Zero
0 0.5 1 1.5 2 2.5 3 3.5
Time (s) technique content
3   (b) Capacitor switching

2.5
Opening Closing
2
  C. Comparison with existing ROCOF techniques
Amplitude (pu)

1.5

1
ROCOF is a common input parameter used in passive
0.5
  islanding detection. Existing ROCOF-based detection tech-
0
No false detection niques set thresholds on the absolute ROCOF value [24], [25]
−0.5
0
 
0.5 1 1.5 2
Time (s)
2.5 3 3.5 which results in NDZ for islanding cases with small power
mismatch. The NDZ size and the detection time are dependent
Fig. 7. IDW and Odetection for different switching events on the selected threshold. Severe non-islanding events also
 
generate large ROCOF values and have the potential to cause
false islanding detection. The proposed technique, on the other
Figure
  8 (a) shows the IDW waveform for a catastrophic 3- hand, extracts the ripple content in the ROCOF waveform and
phase fault occurring on the utility side with a very low fault
used it to detect islanding. As observed from the simulation
resistance
  (1Ω). As shown, the transient spike falls below the
results, the islanding cases exhibit a considerable change in
threshold (α) for only 200 ms with no effect on the detection
the steady ROCOF ripple content while non-islanding events
signal. The loss of parallel feeders is another non-islanding
  exhibit transient peaks only. Therefore, the proposed technique
scenario affecting the frequency of the inverter. In this event,
does not have an NDZ nor does it falsely detect islanding.
the grid short-circuit MVA was set at 500 M V Asc . Figure 8
 
(b) shows the IDW signal for loss of parallel feeder. At 0.5 s,
Table II summarizes the differences in operating principle and
performance between the existing ROCOF techniques and the
the grid M V Asc value dropped to 250 indicating the loss of
  feeder resulting in a 260 ms transient spike with no proposed technique.
a parallel
false islanding detection. IV. C ONCLUSIONS
 
The final simulated non-islanding event was the starting of In this paper, a novel passive islanding detection technique
a 460V, 300 kVA induction motor. The motor was connected for inverter based DG systems using the rate of change of
to the  utility through a transformer and it was turned on frequency was presented. The proposed technique monitors the
at 0.5 s. Although starting such a large motor draws high time domain rate of change of the inverter output frequency
  it seems to have negligible effects on the frequency
currents, and detects islanding when the ripple content exceeds a pre-
as shown in Figure 8. The waveform did not even approach defined threshold for a certain amount of time. This technique
 
the threshold level when the motor started, therefore islanding was modeled and implemented on a grid-connected PV system
was not falsely detected. representing a typical DG source. System performance was
 

7300-5/15/$31.00978-1-4673-7300-5/15/$31.00
©2015 IEEE ©2015 IEEE
Proceedings of the IEEE SoutheastCon 2015, April 9 - 12, 2015 - Fort Lauderdale, Florida

 
tested and verified for wide range of islanding and non- [15] S. Salman, D. King, and G. Weller, “Investigation into the development
  events representing realistic scenarios. The proposed
islanding of a new ANN-based relay for detecting loss of mains of embedded
generation,” in IEEE International Conference on Developments in
technique exhibited a maximum detection time of 400 ms Power System Protection, vol. 2, April 2004, pp. 579–582 Vol.2.
which  is well within the 2 s time limit set by IEEE 1547 [16] O. Faqhruldin, E. El-Saadany, and H. Zeineldin, “A universal islanding
series and UL 1741 standards. In addition, this technique can detection technique for distributed generation using pattern recognition,”
IEEE Transactions on Smart Grid, vol. 5, no. 4, pp. 1985–1992, July
easily  be integrated into any existing inverter to eliminate the 2014.
non-detection zone exhibited by the currently used OUV/OUF [17] B. Matic-Cuka and M. Kezunovic, “Islanding detection for inverter-
based distributed generation using support vector machine method,”
 
protection technique. Moreover, the proposed technique was
IEEE Transactions on Smart Grid, vol. 5, no. 6, pp. 2676–2686, Nov
proven to be insensitive to various non-islanding events which 2014.
include  load switching, capacitor bank switching, severe three- [18] P. Mahat, Z. Chen, and B. Bak-Jensen, “Review of islanding detection
phase faults, loss of parallel feeder, and induction motor methods for distributed generation,” in Electric Utility Deregulation and
Restructuring and Power Technologies (DRPT), April 2008, pp. 2743–
starting.
  Future work includes 1) testing the proposed detection 2748.
technique in a microgrid with more DG sources and buses, 2) [19] Z. Ye, A. Kolwalkar, Y. Zhang, P. Du, and R. Walling, “Evaluation
testing the possibility of reducing the detection time of the of anti-islanding schemes based on nondetection zone concept,” Power
  Electronics, IEEE Transactions on, vol. 19, no. 5, pp. 1171–1176, Sept
technique by augmenting the technique with input parameters 2004.
like voltage or power, and 3) experimentally verifying of the [20] J. C. M. Vieira, W. Freitas, W. Xu, and A. Morelato, “An investigation
results  of the proposed technique. on the nondetection zones of synchronous distributed generation anti-
islanding protection,” IEEE Transactions on Power Delivery, vol. 23,
no. 2, pp. 593–600, April 2008.
  [21] J. Yin, L. Chang, and C. Diduch, “Recent developments in islanding de-
R EFERENCES
tection for distributed power generation,” in Large Engineering systems
Conference on Power Engineering, July 2004, pp. 124–128.
[1] A.  A. Bayod-Rjula, “Future development of the electricity systems with [22] B. Singam and L. Hui, “Assessing SMS and PJD Schemes of Anti-
distributed generation,” Energy, vol. 34, no. 3, pp. 377 – 383, 2009, 6th Islanding with Varying Quality Factor,” in IEEE International Power
World Energy System Conference Advances in Energy. and Energy Conference, Nov 2006, pp. 196–201.
[2] B.  Guha, R. Haddad, and Y. Kalaani, “A novel passive islanding [23] N. W. A. Lidula and A. Rajapakse, “A Pattern Recognition Approach
detection technique for converter-based distributed generation systems,” for Detecting Power Islands Using Transient Signals - Part I: Design
in IEEE PES Conference on Innovative Smart Grid Technologies (ISGT),
 
February 2015.
and Implementation,” IEEE Transactions on Power Delivery, vol. 25,
no. 4, pp. 3070–3077, Oct 2010.
[3] J. Momoh, S. Meliopoulos, and R. Saint, “Centralized and Distributed [24] C. Affonso, W. Freitas, W. Xu, and L. C. P. Da Silva, “Performance of
 
Generated Power Systems- A Comparison Approach,” Power Systems ROCOF relays for embedded generation applications,” IEEE Proceed-
Engineering Research Center, Tech. Rep., June 2012. ings on Generation, Transmission and Distribution, vol. 152, no. 1, pp.
[4] “Renewables 2014 global status report,” Renewable Energy 109–114, Jan 2005.
 
Policy Network for the 21st Century (REN21), 2014. [Online]. [25] J. C. M. Vieira, W. Freitas, W. Xu, and A. Morelato, “Efficient
Available: http://www.ren21.net/Portals/0/documents/Resources/GSR/ coordination of ROCOF and frequency relays for distributed generation
2014/GSR2014 fullreport lowres.pdf protection by using the application region,” IEEE Transactions on Power
[5] P.  Hines, J. Apt, and S. Talukdar, “Large blackouts in North America: Delivery, vol. 21, no. 4, pp. 1878–1884, Oct 2006.
Historical trends and policy implications,” Energy Policy, vol. 37, no. 12,
pp. 5249 – 5259, 2009.
[6] R.  Walling and N. Miller, “Distributed generation islanding-implications
on power system dynamic performance,” in IEEE Power Engineering
Society Summer Meeting, vol. 1, July 2002, pp. 92–96 vol.1.
 
[7] “IEEE application guide for IEEE std 1547(TM), IEEE standard for
interconnecting distributed resources with electric power systems,” IEEE
  1547.2-2008, pp. 1–217, April 2009.
Std
[8] “UL1741 standard for safety for static converters and charge controllers
for use in photovoltaic power systems,” Underwriters Laboratories, Jan
 
2001.
[9] 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,” Power Delivery,
IEEE Transactions on, vol. 22, no. 3, pp. 1758–1766, July 2007.
[10] A. Khamis, “A review of islanding detection techniques for renewable
 
distributed generation systems,” Renewable and Sustainable Energy
Reviews, vol. 28, no. 0, pp. 483 – 493, 2013.
[11] D.  Velasco, C. Trujillo, G. Garcer, and E. Figueres, “Review of anti-
islanding techniques in distributed generators,” Renewable and Sustain-
able Energy Reviews, vol. 14, no. 6, pp. 1608 – 1614, 2010.
[12] P.  Gupta, R. Bhatia, and D. Jain, “Average absolute frequency deviation
value based active islanding detection technique,” IEEE Transactions on
Smart Grid, vol. 6, no. 1, pp. 26–35, Jan 2015.
[13] P.  Du, J. Nelson, and Z. Ye, “Active anti-islanding schemes for
synchronous-machine-based distributed generators,” IEEE Proceedings-
Generation, Transmission and Distribution, vol. 152, no. 5, pp. 597–606,
  2005.
Sept
[14] W. Bower and M. Ropp, “Evaluation of islanding detection methods for
utility-interactive inverters in photovoltaic systems,” Report SAND2002-
 
3591, Sandia National Laboratories, 2002. [Online]. Available: http://
www.prod.sandia.gov/cgi-bin/techlib/access-control.pl/2002/023591.pdf
 

7300-5/15/$31.00978-1-4673-7300-5/15/$31.00
©2015 IEEE ©2015 IEEE

You might also like