Renewable and Sustainable Energy Reviews 35 (2014) 211–220

Contents lists available at ScienceDirect

Renewable and Sustainable Energy Reviews

journal homepage: www.elsevier.com/locate/rser

A review of islanding detection methods for microgrid

Canbing Li n, Chi Cao, Yijia Cao n, Yonghong Kuang, Long Zeng, Baling Fang
College of Electrical and Information Engineering, Hunan University, Changsha 410082, China

art ic l e i nf o a b s t r a c t

Article history: Renewable energy generation is considered as an important approach to solve energy and environment
Received 1 November 2013 problems in the future. Distributed generation technique is a primary way to utilize renewable energy.
Received in revised form Microgrid integrated with different kinds of distributed resources can improve energy efficiency and
15 March 2014
reduce the negative impact on power grid. Microgrid may operate in grid-connected or islanding mode,
Accepted 7 April 2014
Available online 24 April 2014
running on quite different strategies. Effective islanding detection methods are indispensable to realize
optimal operation of microgrid. In this paper, performance indices and critical technique problems are
Keywords: discussed. Islanding detection methods are also classified. The paper aims to discuss the improvement of
Distributed generation several performance indices, including non-detection zone (NDZ), detection time, error detection ratio
Islanding detection methods
and power quality, to evaluate different detection methods. Effectiveness in multiple-inverter cases is
Microgrid
also analyzed. According to the comparison upon the advantages and disadvantages of each method, the
Renewable and sustainable energy
applicability is discussed, which may provide a reference for researchers to select islanding detection
methods.
& 2014 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
2. IDMs performance indices and critical technical problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
2.1. Non-detection zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
2.1.1. Power mismatch space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
2.1.2. Load parameter space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
2.2. Detection time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
2.3. Error detection ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
2.4. Power quality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
2.5. Critical technical problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
3. Islanding detection methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
3.1. Passive methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
3.1.1. Voltage and current harmonics detection (HD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
3.1.2. Over/under voltage and over/under frequency (OUV/OUF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
3.1.3. Rate of change of frequency (ROCOF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
3.1.4. Rate of change of frequency over power (ROCOFOP) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
3.1.5. Rate of change of power output (ROCOP). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
3.1.6. Phase jump detection (PJD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
3.1.7. Voltage unbalance (VU) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.2. Active methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.2.1. Active frequency drift (AFD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.2.2. Frequency jump (FJ). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.2.3. Active frequency drift with positive feedback (AFDPF) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.2.4. Sandia frequency shift (SFS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
3.2.5. Sandia voltage shift (SVS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

n
Corresponding authors. Tel.: þ 86 15073116677, þ86 15973101088.
E-mail addresses: licanbing@gmail.com (C. Li), yjcao@hnu.edu.cn (Y. Cao).

http://dx.doi.org/10.1016/j.rser.2014.04.026
1364-0321/& 2014 Elsevier Ltd. All rights reserved.
212 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220

3.2.6. Sliding mode frequency shift (SMS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216

3.2.7. Variation of active and reactive power . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.2.8. Negative-sequence current injection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.2.9. Impedance measurement (IM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.2.10. Detection of impedance at specific frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.3. Remote methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.3.1. Power line carrier communication (PLCC) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
3.3.2. Signal produced by disconnect (SPD) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
3.3.3. Supervisory control and data acquisition (SCADA). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4. Analysis of IDMs performances. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.1. Non-detection zone . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.1.1. Power mismatch space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.1.2. L versus Cnorm parameter space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.1.3. Quality factor versus resonate frequency parameter space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.2. Detection time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.3. Error detection ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
4.4. Power quality. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
4.5. Effectiveness in multiple-inverter cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
4.6. Performance improvement by combination of IDMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

1. Introduction classified in this part. Section 4 compares performance among

different methods, and improvements of IDMs performance indices
In recent years, renewable energy generation techniques have are described in this part.
been intensively concerned and developed. Comparing with fossil
energy, renewable energy produces less greenhouse gases [1].
Distributed generation (DG), as a primary way to use renewable 2. IDMs performance indices and critical technical problems
energy, is effective. With microgrid, these different resources can
be integrated as a hybrid energy system, providing electric power Whether islanding can be detected accurately, timely and effec-
with optional cooling or heating [2,3]. Efficiency of renewable tively is largely determined by the performance of IDMs. These
energy is significantly developed. In addition, distributed genera- indices consist of non-detection zone, detection time, error detec-
tion techniques are promoted in microgrid. DG connected to the tion ratio and power quality.
main grid directly has a great negative impact because most of
the distributed resources are intermittent. Distributed generations 2.1. Non-detection zone
are complementary because they can support each other in the
microgrid. On one hand, microgrid can mitigate negative impact to Non-detection zone is the main reason why IDMs fail to detect
the main grid by these complementarities. On the other hand, islanding. It should be summarized in different ways according to
various kinds of distributed resources, controllable load and general different categories of IDMs. NDZ of methods based on monitoring
load can be integrated, reducing dependence on main grid. There- voltage, frequency or phase deviation is often described in power
fore, microgrid is widely favored and rapidly developed. In other mismatch space. NDZ of methods based on disturbance injection is
words, microgrid is a critical path for the sustainable energy deve- usually described in load parameter space.
lopment.
Strategies are different when microgrid operates in grid-connected
2.1.1. Power mismatch space
or islanding mode. In the case of islanding mode, to provide reliable
Variation of voltage and frequency at a point of common
power supply or supply parts of special loads under this condition is
coupling (PCC) is related with power mismatch between DG power
priority. In grid-connected mode, the objective is optimal economical
output and load consumption when microgrid operates in islanding
operation. Therefore, detecting operation condition in time is a
condition. Especially, in the condition that DG power output and
premise to optimize control in microgrid. Islanding can be divided
load are almost balanced, power mismatches ΔP and ΔQ are nearly
into planned islanding and unplanned islanding [4]. Planned islanding
equal to zero. The extent of the variation of voltage or frequency is
is that microgrids still supply electric power to local load reliably
not enough to detect islanding when microgrid disconnects from
when they are disconnected from main grid. It is a controllable
grid [7]. The range of power mismatches ΔP and ΔQ , which cannot
operation mode. Unplanned islanding is an undesired event due to
cause voltage or frequency exceeding normal limit to detect island-
line tripping, equipment failure, human errors and so on, with
ing, is NDZ.
microgrids disconnected from main grid [5,6]. Unplanned islanding
NDZ in power mismatch space is shown as [8]
is an uncontrollable operation mode which happens occasionally, and

2
2
the scope of islanding is not determined, thus affecting security of V ΔP V
microgrid. 1r r 1 ð1Þ
V max P V min
In the paper, the features to evaluate performance of islanding
detection methods (IDMs) are discussed, and critical problems to
2 !
2 !
f ΔQ f
improve performance are presented. IDMs are also classified in this Q 1 r rQ 1  ð2Þ
f min P f max
paper. In Section 2, indices of IDMs performance are discussed and
critical problems are illustrated. Section 3 discusses advantages, where V max and V min are the maximum and minimum voltages
disadvantages as well as applicability of IDMs. Main IDMs are also allowed in microgrid respectively, f max and f min are the maximum
 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220 213

and minimum frequencies respectively, V is the rated voltage, P is could accelerate detection speed, but it becomes easier for detec-
the rated active power, and Q is the quality factor. tion parameters to exceed threshold in the case of load switching
or system transient, increasing error detection ratio. On the con-
2.1.2. Load parameter space trary, if threshold is set too large, NDZ would become larger and
NDZ in parameter space is defined as detection speed might become slow despite error detection ratio
decreasing. For IDMs based on disturbance injection, reducing NDZ
F 1 ðcf ; K; Q Þ o ΔC norm o F 2 ðcf ; K; Q Þ ð3Þ
and simultaneously upgrading power quality is in dilemma. By
where cf is the chopping fraction, K is the accelerating gain, Q is injecting larger distorted disturbance or increasing accelerating gain
the quality factor and ΔC norm is the resonate capacitance in the of the feedback, the size of NDZ could be reduced, but it seriously
range of NDZ. deteriorates power quality.
Ropp et al. [7] proposed the L  C norm axis to describe NDZ in It is important to fully consider microgrid's practical operation,
load parameter space, where L is the load inductance and C norm is with all the key performance indices taken into account. In the
the normalized capacitance (C norm ¼ C=C res , C res is the capacitance case that microgrids are connected to the same PCC in parallel,
that resonates with the load inductance in grid line frequency). In measurement parameters of some IDMs might cancel out each
this method, real power between DG and load is assumed to be other because of deviating in different directions, decreasing IDMs'
matched and thus load resistance is determined. Then mapping of effect or even making it fail to detect islanding. With the devel-
NDZ could be described in the L  C norm axis. opment of microgrid, its structure and operation would become
Lopes et al. [9] utilized an alternative load parameter space more complicated, so IDMs, which can coordinate every perfor-
based on the Q  f 0 axis to describe NDZ, where f 0 is the load mance index to reach a balance, will be widely applied in the
resonate frequency. The main advantage of this method with future.
respect to L versus C norm is that it does not plot different curves
to analyze NDZ with different resistive loads. The increase in R
means the increase in Q , for a given set of L and C loads. 3. Islanding detection methods

2.2. Detection time Islanding detection methods are generally divided into local and
remote methods as shown in Fig. 1 [6,11–15]. Local methods are based
Detection time is the duration from the beginning of microgrid on measurement of some parameters or variables on the microgrid
disconnecting from main grid to the end of detecting islanding by side, including passive methods and active methods. Passive methods
IDMs, which is defined as directly monitor parameters or variables including voltage, current,
frequency and phase to detect islanding. Active methods intentionally
ΔT ¼ T IDM  T trip ð4Þ inject a disturbance to detect whether it affects voltage, frequency,
where ΔT is the run-on time, T IDM is the moment to detect islanding, power or impedance parameters. Remote methods are based on the
and T trip is the moment microgrid disconnects from the grid. communication between microgrid and main grid to monitor break-
ers immediately. Remote methods present small or even no NDZ,
2.3. Error detection ratio which have no impact on power quality. Remote methods are very
effective in multiple-inverter systems, but they need large amount of
Error detection means that IDMs detect false in islanding when investment. It is not economic in small systems.
microgrid is still connected to grid. Error detection is mainly
caused by load switching or other disturbance, leading measure- 3.1. Passive methods
ment parameters to exceed normal limit [10]. The ratio can be
defined as the ratio of error detection times to total detection 3.1.1. Voltage and current harmonics detection (HD)
times by IDMs. This method is based on the measurement of total harmonic
N error distortion (THD) at PCC to detect islanding when THD exceeds the
E¼ ð5Þ threshold. Under normal conditions, when the microgrid is con-
Nerror þ N correct
nected to main grid, PCC voltage is a standard sine wave, and thus
where E is the error detection ratio, N error is the times of error load-generated harmonics are negligible. Because the grid impe-
detection, and Ncorrect is the times of correct detection. dance is very small, the harmonics produced by the inverter are
transmitted to the grid without causing distortion at PCC. When a
2.4. Power quality microgrid is running on islanding condition, the current harmo-
nics produced by the inverter are transmitted to the load and
Methods based on injecting disturbance can significantly reduce hysteresis effect of transformer will further aggravate harmonic
NDZ when detecting islanding. However, it is inevitable for injecting distortion at PCC, which is able to detect islanding [12,13,16].
disturbance to distort power output and deteriorate power quality. This method has an advantage that its effectiveness does not
Power quality is an important index when selecting IDMs. change when multiple DGs are connected to the same PCC in
parallel, and is easy to implement. The detection time is about
2.5. Critical technical problems 45 ms under a fast detection speed and a wide range of situations
[11]. However, the threshold is difficult to select for this method
IDMs have many critical technical problems in improving perfor- because grid disturbance is easy to cause error detection. This
mance. Methods based on whether detection parameters exceed the method is prone to fail if NDZ is large for loads with a large quality
threshold caused by power mismatch are hard to eliminate NDZ. factor Q . Q is defined in [7,17], as Eq. (6). Thus the harmonic
Their detection speed is associated with power mismatch, making it detection method is difficult to apply in small single systems.
hard to predict. Methods based on disturbance injection are inevi- rffiffiffiffi
table to degrade power quality. C
Q ¼R ð6Þ
On the other hand, the improvements of different performance L
indices have many difficult problems that they might be restricted Q is equal to 2π times the ratio of the maximum stored energy
by each other. For example, reducing threshold of normal range at the resonant frequency to the energy dissipated of a cycle at
214 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220

Islanding detection
methods

Local methods Remote methods

Passive method Active method

1. Voltage and current harmonics 1. Active frequency drift

detection 1. Power Line Carrier
2. Frequency jump
2. Over/under voltage and over/under Communication
3. Active frequency drift with positive
frequency 2. Signal Produced by
feedback
3. Rate of change of frequency Disconnect
4. Sandia frequency shift
4. Rate of change of frequency over 3. Supervisory Control
5. Sandia voltage shift
power and Data Acquisition
6. Sliding mode frequency shift
5. Rate of change of power output 7. Variation of active and reactive power
6. Phase jump detection 8. Negative-sequence current injection
7. Voltage Unbalance 9. Impedance Measurement
10. Detection of Impedance at Specific
Frequency

Fig. 1. Classification of islanding detection methods.

that frequency. When the resonant frequency of the load closes to which may lead to error detection and make the selection of
the grid rated frequency, such as 50 Hz or so, the value of Q has threshold difficult. This method cannot distinguish whether the
great influences on the size of NDZ and detection accuracy. frequency change is caused by islanding or load changes [21]. Thus
ROCOF is suitable for loads with less fluctuation.
3.1.2. Over/under voltage and over/under frequency (OUV/OUF)
This technique is based on setting an allowable range for voltage 3.1.4. Rate of change of frequency over power (ROCOFOP)
and frequency. Inverters will disconnect to stop supplying power to This method is based on the measurement of the value of
the load if the voltage or frequency at PCC exceeds thresholds. The ∂f =∂P L , an index to determine whether microgrid is running on
frequency or voltage deviation after microgrid disconnection is islanding condition, where P L is the load power. This method has
mainly due to the power mismatch between distributed generation higher reliability, lower error detection ratio and its NDZ is smaller
and loads in microgrid, at PCC: than ROCOF. It can detect islanding more effectively when the
ΔP ¼ P load  P DG ð7Þ power mismatch is small between the DG and load. Detection time
of this method is about 100 ms [22]. Compared with ROCOF, this
ΔQ ¼ Q load  Q DG ð8Þ method could be applied in wider range of situations. It can detect
islanding effectively in microgrid with small power imbalance
In grid-connected operation, ΔP and ΔQ will be injected from
[20,22].
main grid to keep the balance of active and reactive power. When
islanding occurs, in order to meet the active and reactive power
balance, its voltage and frequency will drift until ΔP ¼ 0 and 3.1.5. Rate of change of power output (ROCOP)
ΔQ ¼ 0. Thus OUV/OUF can detect islanding by detecting voltage Principle of this method is the detection of the changes in the
and frequency deviations [11–13,16]. DG power output (dP=dt), because loss of grid generally produces
This method, with low cost, has no impact on power quality. load changes. For the same rate of load change, dP=dt measured
Weaknesses are those that NDZ is relatively large and detection when microgrid is islanded will be much greater than dP=dt
time is difficult to predict. The run-on time is from 4 ms to 2 s, measured before microgrid is islanded. The value of power change
even more than 2 s [18]. Detection speed is related to the power is measured over a few sample cycles. When the integrated change
mismatch between DG and loads. Therefore, this method is exceeds the threshold value, the DG will be disconnected to stop
suitable for microgrid with some power imbalance. supplying power to the load [11].
The detection time of this method is between 24 ms and 26 ms
3.1.3. Rate of change of frequency (ROCOF) [19]. Compared with OUF, its detection speed is not affected when
When the microgrid is disconnected from main grid with a the power mismatch between DG and load is small. Another
power mismatch, the frequency will change. With the value of advantage of this method is that it is able to quickly detect
df =dt measured over a few cycles, the islanding can be detected unsynchronized reclosing of the utility supply to the DG to ensure
and inverters shut down [19,20], if it exceeds a setting threshold. stable operation of power system. This method still has NDZ in the
ROCOF is more sensitive than OUV/OUF, and its detection speed condition of power balance. But with the increasing of power
is faster. If the power mismatch between DG and load is large, this imbalance, the method will become more effective to detect
method is highly reliable and timely. The detection time is up to islanding.
24 ms [19]. Even when power between DG and load is in balance,
any disturbance produced by load changes could break this balance, 3.1.6. Phase jump detection (PJD)
leading to frequency changes to detect islanding. The weaknesses of This method is based on monitoring the phase jump between
ROCOF are those that it is sensitive to load switching and fluctuation, the inverter's terminal voltage and current. When microgrid is
 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220 215

under normal operation, by using PLL to detect zero crossing of The strengths of AFD are that it is easy to implement and has a
voltage, the inverter's current will be synchronized with the small NDZ, and particularly, there is no NDZ in resistance load
voltage at PCC. In islanding operation, the inverter's output current with the detection time within 2 s [11]. The weakness is that in
remains unchanged, because phase-locked loop can only work at multiple inverters case, the method may fail to detect islanding
the zero crossings of voltage. Due to load phase angle, the voltage because of inverters in different deviations of frequency bias. With
will have a sudden “jump”. If the phase difference exceeds the the injected current distorting more heavily, the power quality of
setting threshold, the method can determine the islanding [12]. inverters output will degrade more quickly. Load parameters play
Major strengths of this method are the ease of its implementa- a great role to the effectiveness of the method. If the load is not
tion and fast detection speed, and the detection time of this method resistance, the detection time and the NDZ will increase with
is usually between 10 ms and 20 ms [23]. PJD does not affect the higher value of Q . Therefore, AFD is the best for the islanding
power quality. Effectiveness of this method is not reduced in the detection of microgrid which is just made up of resistive loads and
case of multiple inverters. Disadvantage of this method is that it is without multiple inverters.
difficult to choose thresholds. Because of the load switching,
particularly motor load, which increases the difficulty in setting
the thresholds for PJD, an error will be caused in the detection of 3.2.2. Frequency jump (FJ)
islanding. Similarly, if local loads are not producing sufficient phase FJ is a modification of AFD, which also inserts dead zones into
error, it is difficult to detect islanding by using this method. Thus PJD current waveform, but not into each cycle, for example, one dead
is suitable for microgrids in the situations where load phase angle is zone in every 3 cycles, instead. When the microgrid is connected
enough and load switching is not frequent. to main grid, the waveform of voltage in PCC, which is imposed by
the grid despite inverter's current is distorted, is not distorted.
When disconnected from main grid, islanding can be detected by a
3.1.7. Voltage unbalance (VU)
variation in voltage frequency [7,8,11].
Due to the topology changes of the networks after the microgrid
FJ is very effective in detecting microgrid without multiple
disconnected from main grid, the voltage unbalance of DG output
inverters in parallel; the disadvantage is that, like AFD, the detecting
varies. If the unbalance of three-phase output voltage of the DG
effectiveness will be reduced when multiple inverters are in parallel.
exceeds the setting threshold, it will determine an islanding operation
[20]. The voltage unbalance at the time t is defined as
NSt 3.2.3. Active frequency drift with positive feedback (AFDPF)
VU t ¼ ð9Þ
PSt To overcome the weakness of AFD in multiple inverters and
where NSt and PSt are the amplitudes of negative and positive reduce NDZ, AFDPF utilizes a positive feedback to increase chop-
sequence of voltage at the time t, respectively. ping fraction to accelerate frequency deviation, detecting islanding
The detection time is about 53 ms [24]. In the situations of more effectively.
variation caused by normal loads, this method is hard to cause cf k ¼ cf k  1 þ FðΔωk Þ ð11Þ
error detection and it is not sensitive to the system disturbance.
The disadvantage of this method is that the extraction of negative where cf k and cf k  1 are the chopping fractions of the kth and
sequence voltage component is affected by harmonics, increasing k-1th cycles, respectively. Where ωk is the frequency of the kth cycle,
the difficulty in calculating thresholds. This method is better to be Δωk ¼ ωk  1  ω0 , F is usually a linear function. The value of cf in
applied in systems with load fluctuations, such as motor starting AFDPF can be positive or negative. No matter if frequency drift is
and frequent capacitor banks switching. upward or downward, this method can reinforce the frequency
drift instead of counteracting it, overcoming the impact of the load
parameters [26]. The performance has been improved compared
3.2. Active methods
to AFD, greatly reducing the NDZ. Its disadvantages are that it
affects the power quality slightly, and still has NDZ for loads with
3.2.1. Active frequency drift (AFD)
high quality factor.
This method is based on slightly distorting the current wave-
form injected into PCC by the inverter. When connected to the
grid, the voltage and frequency of PCC will not change owing to
3.2.4. Sandia frequency shift (SFS)
the stability of main grid, and the frequency of the inverter's
As an extension of AFD, a positive feedback is applied for the
output current will not change after a phase-locked loop, either.
frequency of inverter's voltage, whose chopping fraction is [12,25,27]
When a grid disconnection occurs, because of distortion of the
injected current waveform, the zero crossing of the voltage occurs cf ¼ cf 0 þ Kðf PCC  f grid Þ ð12Þ
sooner than expected, thus giving rise to a phase error between
the voltage and the inverter's output current. It makes the inverter where cf 0 is the chopping factor with no deviation in frequency, K is
to drift frequency of output current to eliminate the phase error. the accelerating gain, f PCC is the voltage frequency in PCC, and f grid is
The voltage response of this current frequency drift causes an the frequency of the grid.
earlier zero crossing than expected again, making the inverter's When connected to main grid, the method attempts to change
output current to drift its frequency until the voltage frequency the voltage frequency of PCC but it is prevented by main grid. When
measured in PCC exceeds the threshold of OUF and then the disconnected from main grid, the chopping fraction increases with
islanding is detected [12]. the increase of f PCC . Therefore, the frequency of the inverter also
The major parameter describing the distortion of the inverter's increases, and all these processes will continue to reinforce the
injected current is the chopping fraction, which is defined in frequency shift to detect the islanding effectively. The detection
[25,26], as the following equation. time of SFS is within 0.5 s, and it even can detect islanding in
7 cycles [17]. This method, compared with another active methods,
2t z
cf ¼ ð10Þ has the smallest NDZ. In addition, SFS is very effective to compro-
T V util
mise the detecting efficiency, power quality as well as the impact on
where t z is the dead time and T Vutil is the period of the voltage. system transient response.
216 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220

3.2.5. Sandia voltage shift (SVS) current will entirely flow into the grid without impacting the
Sandia voltage shift is similar to SFS in principle. By applying a voltage of PCC. When the microgrid is islanding, injected current
positive feedback to the amplitude of voltage in PCC, the inverter will flow into the local load, contributing to an unbalance in the
changes its current output and power output. When connected to voltage of PCC and detecting islanding when the voltage unbalance
main grid, the amplitude of voltage is not affected by power change, exceeds the threshold.
whereas without the support of main grid, power output changes This method can detect islanding event with 3.5 cycles, 60 ms,
can accelerate the voltage drift to detect islanding [13,17,27]. a very short detection time compared to another active method. Its
SVS is easy to implement, and it has the same efficiency as the advantages are that it is not sensitive to load change, presents no
SFS method which is based on positive feedback. The primary NDZ and the accuracy is higher than detecting positive-sequence
weakness of SVS is that it slightly degrades power quality. Secondly, voltage [29].
because of changing the inverter's output power, it affects the
maximum power point tracking algorithm of the inverter, reducing
the inverter's operation efficiency [12,13]. 3.2.9. Impedance measurement (IM)
Impedance measurement is a method that changes the ampli-
3.2.6. Sliding mode frequency shift (SMS) tude of inverter output current in general. When disconnected
SMS utilizes positive feedback to change the voltage phase of from main grid, the voltage varies as a result of perturbation in
PCC, monitoring frequency deviation to detect islanding. In SMS, current. Variation monitored by calculating dv=di as equivalent
the current–voltage phase angle of the inverter is set as [9,20,28] impedance seen from the inverter can be used to detect islanding
! [13,30].
π f k1 f n
θ ¼ θm sin ð13Þ The detection time of the IM method is between 0.77 s and
2 f m f n
0.95 s [11], and main advantage of this method is that it has
where θm is the maximum phase angle at the frequency f m , f n is extremely small NDZ for single DG system. However, this method
k1 also has many disadvantages. For example, the detection efficiency
the rated frequency, and f is the frequency of previous cycle.
When the microgrid operates normally, its power factor operates will decrease in multiple-inverter cases, unless all inverters
with main grid. The phase angle between the inverter current and operate synchronously. The impedance threshold is also hard to
the PCC voltage is controlled to be zero or very close to it. When set because it requires exact value of grid impedance. So, this
disconnected from main grid, the phase angle of load and the method is difficult to practically apply.
frequency will vary along with the SMS curve, and thus islanding
can be detected if frequency variation exceeds the threshold.
The detection time of this method is about 0.4 s [9]. Advantages 3.2.10. Detection of impedance at specific frequency
of SMS are that it is easy to implement and has smaller NDZ than This method is a special case of harmonic detection method.
general active methods. Moreover, SMS is highly effective in multiple- Specific frequency harmonics are injected by inverters. When
inverter systems. Disadvantages of the method are that it reduces microgrid operates in grid-connected mode, the harmonic current
the grid power quality and has certain impacts on system transient will flow into grid without causing an abnormal voltage in PCC, for
stability. grid impedance is much lower than load impedance. When
microgrid disconnects from grid, the harmonic current will com-
3.2.7. Variation of active and reactive power pletely flow into local load, producing a harmonic voltage in PCC if
This method varies the output power injected by inverter and the load is linear, and islanding can be detected while harmonic
monitors the variation in voltage amplitude and frequency to voltage is large enough.
detect islanding. For example, when a microgrid is islanding, the The strength of this method is similar to harmonics detection.
active power of DG will flow into the load. To balance the active The weakness is that it is easy to cause equipments misoperation,
power between DG and the load, the voltage variation must satisfy such as transformers, unless harmonic amplitude is small. Error
[27]: islanding detection may happen if the same harmonic current is
injected by multiple inverters. So this method is not suitable for
V2 multiple-inverter cases [12].
P DG ¼ P load ¼ ð14Þ
R
Islanding can be detected when the voltage exceeds the
threshold of OUV. In a similar way, the disturbance of reactive 3.3. Remote methods
power will affect the variation in frequency, and islanding can be
detected by measuring whether the frequency exceeds the thresh- 3.3.1. Power line carrier communication (PLCC)
old or not [13]. Transmitters are set at grid side in the PLCC method, which can
The detection time of this method is between 0.3 s and 0.75 s, produce communication signal along with power line through
and its advantages are that it is easy to implement, and has a small PLCC system. And the DG side is equipped with receiver. If PLCC
NDZ with less investment [25]. The greatest weakness is that it signal is interrupted, it indicates that microgrid is islanding [31].
will lead to erroneous detect when multiple inverters are parallel The signal period of PLCC is designed with four consecutive
at the same PCC. The method continuously varies power output of cycles. Islanding can be detected if signal disappears in three
inverters, affecting the grid power quality and transient stability consecutive periods. Thus the detection time is about 200 ms [32].
greatly. Variation of active and reactive power is generally applied There is no NDZ within the range of normal loads. The method
in islanding detection for microgrid without multiple inverters. does not degrade power quality and has no impact on grid
transient response. It is proved to be highly effective in multiple-
3.2.8. Negative-sequence current injection inverter system. Furthermore, this method only uses PLCC signal
This method injects negative-sequence current into a three- which has existed in power grid to detect islanding. But the
phase voltage-sourced converter, monitoring the negative-sequence transmitter is expensive, and would not be economical in low-
voltage at PCC to detect islanding. When connected to the grid, density DG systems. So, this method is applied in the microgrid
because main grid has low impedance, injected negative-sequence with high-density DG systems.
 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220 217

3.3.2. Signal produced by disconnect (SPD) inductance. Ropp et al. [7] compared NDZ of SMS and AFD
This method is similar to PLCC, detecting islanding according to methods. The NDZ map of SMS shows that its NDZ is similar with
signal transmission between inverters of DG and external power that of AFD. Moreover, if L is above a certain threshold, SMS has no
grid. The difference between them is that signal transmission is NDZ. SFS has a smallest NDZ. Not only NDZ is narrower than that
based on microwave, telephone line and others forms. This of AFD and SMS under the condition of small L and large C values,
method also utilizes the consecutive carrier signal, to prevent but also the size of NDZ reduces faster when load inductance is
the failure caused by generator, channel or receiver [11–13]. increased. The threshold L without NDZ is smaller than the one in
SPD has no NDZ, and it allows additional control to DG by main the SMS method [7].
grid, coordinating between DG and grid power source which Load parameter space ignores effect of load resistance when
would be beneficial to black-start. The system startup character- applied to describe the map of NDZ. With the increasing of R, size
istics also can be improved by this coordination. The disadvantage of NDZ might become wider by increased load quality factor. Load
of the SPD method is that it needs large amount of investment. If parameter space cannot reflect the influence by Q values.
telephone line is used for signal transmission, the cost of commu-
nication wiring needs to be added and communication protocol
4.1.3. Quality factor versus resonate frequency parameter space
should be set up. If microwave is used to transmit signal, repeaters
Lopes et al. [9] showed the NDZ of the AFD method with different
are needed to install.
chopping fractions, concluding that increasing cf could reduce NDZ.
SMS and SFS improved performance to reduce NDZ. Furthermore,
3.3.3. Supervisory control and data acquisition (SCADA) SMS and SFS methods could eliminate NDZ for Q o 2:5 and Q o 4:8
This method measures the auxiliary contacts of grid circuit respectively.
breakers, monitoring the state of main grid. In the case of islanding
operation, SCADA system will send signals to corresponding range
4.2. Detection time
of DG, and thus the state of circuit breakers is transmitted to DG by
SCADA system [11–13].
Fast detection is a premise for microgrid to have enough time
The advantage of this method is similar to SPD that it is able to
to operate islanding strategy, assuring security and reliability.
take additional control of DG. If the system installation is proper and
Passive methods are based on monitoring transient response of
communication is available, NDZ can be eliminated and operation
parameters including voltage and frequency. Their detection speed
efficiency can be improved. The disadvantage of this method is that
is faster than most active methods generally. Run-on time of the
the detection speed is rather slow, especially under the condition of
OUF/OUV method spans widely, making it difficult to predict. In
a busy system. The investment is increased for requiring separate
the case that real power mismatch reaches 40% or reactive power
instrument and communication link in multiple-inverter case. This
mismatch exceeds 20%, the islanding could be detected in one
method requires complex installation and certification, which is not
cycle of disconnection from main grid. However, with the load
suitable for small-scale system.
matching to power generation, run-on time to detect islanding
would become longer, even within 2 s [18]. HD and VU could have
4. Analysis of IDMs performances improvements to detect islanding within 50 ms and ROCOP could
reach 24–26 ms [11,19,24]. These methods can also detect island-
4.1. Non-detection zone ing faster than OUV/OUF when power becomes matched. PJD and
ROCOF are most sensitive and they could detect islanding in 20 ms
With the characteristics of NDZ summarized, it is effective to [19,23].
present the applicability of each IDMs and factors which affect NDZ. Active methods which monitor system transient response by
Furthermore, performance of IDMs can be upgraded by comparing injecting a perturbance to detect islanding generally have a longer
NDZ of different methods. With the increasing of chopping factor or detection time. Run-on time of AFD is in 2s and it is much affected
accelerating gain, the size of NDZ can be reduced. Moreover, methods by Q values. The detection time is improved to 0.77–0.95 s by the
based on positive feedback can reinforce frequency deviation to IM method [11]. It is further improved by power variation method
reduce NDZ more effectively than frequency drift methods. to 0.3–0.75 s [25]. SMS, SFS and SVS IDMs can detect islanding in
0.5 s in most cases and their detection speed is slightly affected by
4.1.1. Power mismatch space load quality factor. Even SFS can detect islanding in 7 cycles [9,17].
OUV/OUF usually has a large NDZ. Jones et al. [18] proposed Negative-current injection method improves detection time in
that this method is more sensitive to reactive power mismatch. In 3.5 cycles with 60 ms [29].
NDZ, the range of ΔQ is smaller than that of ΔP. ROCOP directly For remote methods like SPD and SCADA which use auxiliary
monitors power output changes, reducing NDZ. ROCOF measures communication ways to transmit signal, detection speed easily
the value of df =dt. Though in the case that power is in balance, any becomes slow when the system is busy. Chowdhury et al. [33]
subsequent load change would enable this method to detect proposed that signal through direct communications is about 100–
islanding because it is very sensitive [20]. ROCOFOP reduces NDZ 150 ms for fiber-optic or microwave links and 200–300 ms for
and provides more reliable detection on the basis of ROCOF [22]. telephone lines. PLCC utilizes power line to transmit signal and its
The ΔP  ΔQ axis can only map NDZ of a specific load, because detection time is about 200 ms.
NDZ maps are meaningless when load parameters have changed.
Power mismatch method can only describe NDZ of passive 4.3. Error detection ratio
methods based on monitoring whether voltage or frequency
exceeds normal range, which cannot be utilized in active method. When setting the normal range of measurement parameters, the
scope could not be too large considering size of NDZ. On the other
4.1.2. L versus C norm parameter space side, it could not be set too narrow because of high error detection
Ropp et al. [26] discussed NDZ of AFD and AFDPF methods. ratio. IDMs which are sensitive to load fluctuation generally have
According to the discussion, NDZ would reduce with the increase higher error detection ratio, selecting a reasonable threshold
of chopping fraction. Moreover, AFDPF has a smaller NDZ than AFD difficultly. The threshold of the HD method selected must be higher
and the range of C norm in NDZ is significantly reduced with a given than THD expected in grid voltage and lower than THD produced
 218
Table 1
Summarization of IDMs.

Categories IDMs NDZ Detection time Error Power quality Effectiveness in Advantages, disadvantages and applicability
detection multiply-inverter
rate cases

Passive methods HD Large with a large 45 ms High No impact Very effective Advantages of passive methods are that they have no
value of Q impact on power quality, and detection speed is fast.
OUV/OUF Large From 4 ms to 2 s No impact Disadvantages are that NDZ is large and error detection

C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220

ROCOF Small 24 ms High No impact rate is higher than active method. Passive methods are usually
ROCOFOP Smaller than ROCOF 100 ms Low No impact applied in single DG systems with a certain power imbalance.
ROCOP Smaller than OUV/ 24–26 ms No impact
OUF
PJD 10–20 ms High No impact Not reduced
VU 53 ms Low No impact
Active methods AFD NDZ increases with Within 2 s Degrade Performance is Advantages of active methods are that they reduce the size
increasing of Q reduced of NDZ and decrease error detection ratio. But they are
FJ Small Degrade Reduced inevitable to deteriorate power quality. Active methods are
AFDPF Smaller than AFD Slightly degrade usually applied to confirm whether microgrid is islanding.
SFS Smallest In 0.5 s Low Slightly degrade Moreover,
SVS Smallest Low Slightly degrade effectiveness in multiple DG systems should be concerned.
SMS Smaller than AFD About 0.4 s Low Have impact on system Highly effective
transient stability
Power variation Small 0.3 s-0.75 s High Degrade power quality and Increasing error
affect system transient stability detection rate
Negative -sequence None 60 ms Low
current injection
IM Small NDZ for single 0.77 s–0.95 s Produce harmonics Reduced
system
Detection of impedance High Produce much harmonics Deteriorate power
at specific frequency quality heavily
Remote methods PLCC Without NDZ in 200 ms None No impact Highly effective Advantages of remote methods are that they have no NDZ,
range of normal and have no impact on power quality and system transient
loads response. Error detection also can be eliminated.
SPD None 100 ms–300 ms None No impact Highly effective Disadvantages
SCADA None Detection speed is None No impact Highly effective are that remote methods request large amount of investment.
slow if systems are Remote methods are usually applied in high-density DG
busy systems.
Hybrid methods ROCOV and power Small Low Small Hybrid methods are effective to be applied in complex systems.
variation With the combination of methods, it can improve multiple
VU and SFS, SVS Very small None Reduce negative impact Effective performance indices simultaneously by their compensation.
ROCOF and IM Small 0.216 s Low
 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220 219

during islanding. In reality, there are many occasions, such as power Summarization of IDMs different performance indices, different
electronic converters that produce harmonics resonated by power categories for advantages and disadvantages, and applicability are
system, which may cause THD increasing significantly. Another shown in Table 1. According to the table, some performance indices
transient response with a momentary increase in THD also can are not mentioned for several IDMs. The reasons may include firstly,
cause this method to detect false islanding. ROCOF is sensitive to the method to describe map of NDZ is complex, and different
load fluctuation and easy to detect false islanding. ROCOFOP, which methods have different ways to identify NDZ. We can only evaluate
monitors ∂f =∂P L to detect islanding, decreasing error detection ratio or compare those different IDMs whose NDZ are mapped in the
caused by load changing, has an improvement in performance same way. For one islanding detection method, variables in defini-
compared with ROCOF. SMS, SFS and SVS are reliable to detect tion of NDZ include chopping factor or accelerating gain, which may
islanding, which are less affected by load changing. Negative current change in reality. Thus size of NDZ will also change. Secondly,
injection method monitors current of the grid line to prevent different performance has conflict. For example, NDZ may contra-
false trips. dict with error detection ratio for some methods based on whether
the measured parameters have exceeded threshold. It makes no
4.4. Power quality sense to improve performance on the one hand while ignoring
performance request on the other hand. In further research, it is
Power variation method seriously deteriorates power quality significant to consider different performance indices comprehen-
and has much impact on system transient response, causing sively. Relationships among different indices should be summarized.
voltage flicker and instability. Detection of impedance at the We can select a suitable islanding detection method effectively
specific frequency method produces much harmonics to increase which is satisfactory to different performance requests.
amplitude of harmonic voltage especially in multiple-inverter
cases. AFD decreases power quality along with the increasing of
chopping fraction. SFS, SVS and SMS only have slight impact on 5. Conclusions
power quality.
Several islanding detection methods are presented in this
4.5. Effectiveness in multiple-inverter cases paper. Local methods have been divided into passive and active
methods, which are based on the inverter side, whereas in remote
Multiple inverters in parallel which might have impact on IDMs methods, IDMs are based on communications between main grid
performance should be considered when selecting IDMs in multi- and microgrid. Several significant indices to evaluate performance
ple inverters and DG systems. Effectiveness of IM decreases in of IDMs also have been discussed in this paper, including non-
multiple-inverter cases unless variation introduced by each inver- detection zone, detection time, error detection ratio and power
ter is synchronized. The reason is that the total current is reduced quality. Effectiveness of IDMs in multiple inverters and multiple
when more inverters are added, because they might cancel out DG systems cases is analyzed in this paper. The improvement of
each other or even make voltage variation undetectable. AFD and IDMs in each performance index is described. Their advantages,
FJ must assure that all inverters are in the same direction of disadvantages and applicability are summarized in this paper.
frequency bias in order to maintain effectiveness in multiple-
inverter cases. Power variation method easily generates false
detection of islanding when multiple inverters are connected to
Acknowledgments
same PCC of the grid. SMS and PJD are highly effective in multi-
inverter cases.
This work was supported by the National High Technology
Research and Development of China (863 Program) (Grant no.
4.6. Performance improvement by combination of IDMs
2011AA050203).
Combination of rate of change of voltage (ROCOV) and variation
of active power: this hybrid method is based on combination of References
passive (ROCOV) and active (variation of active power) methods.
When microgrids operate in grid-connected mode, only ROCOV is [1] El-Khattam W, Salama MMA. Distributed generation technologies, definitions
applied to monitor voltage change, which does not deteriorate and benefits. Electr Power Syst Res 2004;71(2):119–28.
power quality. The active method is applied only when ROCOV has [2] Huang J, Jiang C, Xu R. A review on distributed energy resources and
microgrid. Renew Sustain Energy Rev 2008;12(9):2472–83.
detected voltage change but cannot identify that the change is [3] Lidula NWA, Rajapakse AD. Microgrids research: a review of experimental
caused by islanding or another disturbance. This method can microgrids and test systems. Renew Sustain Energy Rev 2011;15(1):186–202.
efficiently reduce error detection ratio. Compared with two [4] Alvaro L, Octavian C, Jaime J, Haritza C. Survey on microgrids: unplanned
islanding and related inverter control techniques. Renew Energy 2011;36
methods applied separately, the hybrid method can improve (8):2052–61.
performance of power quality and error detection ratio simulta- [5] Chiang WJ, Jou HL, Wu JC. Active islanding detection method for inverter-
neously [34]. based distribution generation power system. Int J Electr Power 2012;42
(1):158–66.
Combination of VU and SFS, SVS: compared with SFS and SVS [6] Eltawil MA, Zhao Z. Grid-connected photovoltaic power systems: technical
methods in multiple DGs systems, this hybrid method can reduce and potential problems—a review. Renew Sustain Energy Rev 2010;14
negative impact on power system transient response. Compared (1):112–29.
[7] Ropp ME, Begovic M, Rohatgi Ajeet, Kern GA, Bonn RH, Gonzalez S. Determin-
with VU, this method can efficiently discriminate islanding and ing the relative effectiveness of islanding detection methods using phase
load switching, avoiding error detection [35]. criteria and nondetection zones. IEEE Trans Energy Convers 2000;15
Combination of ROCOF and IM: the islanding detection scheme (3):290–6.
[8] Ye Z, Kolwalkar A, Zhang Y, Du P, Walling R. Evaluation of anti-islanding
of this hybrid is based on two stages. ROCOF is used as primary
schemes based on nondetection zone concept. IEEE Trans Power Electr
protection and IM is the backup. Compared with ROCOF, this 2004;19(5):1171–6.
method can discriminate islanding and another disturbance, detect- [9] Lopes LAC, Sun H. Performance assessment of active frequency drifting
ing islanding more reliably. Compared with IM, this method can islanding detection methods. IEEE Trans Energy Convers 2006;21(1):171–80.
[10] Vieira JCM, Salles D, Freitas W. Power imbalance application region method
detect islanding satisfactorily for different types of loads in 0.216 s, for distributed synchronous generator anti-islanding protection design and
reducing detection time efficiently [36]. evaluation. Electr Power Syst Res 2011;81(10):1952–60.
220 C. Li et al. / Renewable and Sustainable Energy Reviews 35 (2014) 211–220

[11] Ahmad Ku Nurul Edhura Ku, Selvaraj J, Rahim NA. A review of the islanding [24] Jang S, Kim KH. An islanding detection method for distributed generations
detection methods in grid-connected PV inverters. Renew Sustain Energy Rev using voltage unbalance and total harmonic distortion of current. IEEE Trans
2013;21:756–66. Power Deliv 2004;19(2):745–52.
[12] PVPS IEA. Evaluation of islanding detection methods for photovoltaic utility- [25] Mango FD, Liserre M, Dell’Aquila A. Overview of anti-islanding algorithms for
interactive power systems. Report IEA PVPS T5-09. 2002. PV systems. part II: active methods. In: Proceedings of the IEEE 12th
[13] Velasco D, Trujillo CL, Garcera G, Figueres E. Review of anti-islanding international power electronics and motion control conference; 2006. p.
techniques in distributed generators. Renew Sustain Energy Rev 2010;14 1884–9.
(6):1608–14. [26] Ropp ME, Begovic M, Rohatgi A. Analysis and performance assessment of the
[14] Khamis A, Shareef H, Bizkevelci E, Khatib T. A review of islanding detection active frequency drift method of islanding prevention. IEEE Trans Energy
techniques for renewable distributed generation systems. Renew Sustain Convers 1999;14(3):810–6.
Energy Rev 2013;28:483–93. [27] Trujillo CL, Velasco D, Figueres E, Garcerá G. Analysis of active islanding
[15] Yu B, Matsui M, Yu G. A review of current anti-islanding methods for detection methods for grid-connected microinverters for renewable energy
photovoltaic power system. Sol Energy 2010;84(5):745–54. processing. Appl Energy 2010;87(11):3591–605.
[16] Mango FD, Liserre M, Dell'Aquila A, Pigazo A. Overview of anti-islanding [28] Liu F, Kang Y, Zhang Y, Duan S, Lin X. Improved SMS islanding detection
algorithms for PV systems. part I: passive methods. In: Proceedings of the IEEE method for grid-connected converters. IET Renew Power Gener 2010;4
12th international power electronics and motion control conference; 2006. p. (1):36–42.
1878–83. [29] Karimi H, Yazdani A, Iravani R. Negative-sequence current injection for fast
[17] Stevens J, Bonn R, Ginn J, Gonzalez S. Development and testing of an approach islanding detection of a distributed resource unit. IEEE Trans Power Electr
to anti-islanding in utility-interconnected photovoltaic systems. Albuquerque: 2008;23(1):298–307.
Sandia National Laboratories; 2000. [30] Kane PO, Fox B. Loss of mains detection for embedded generation by system
[18] Jones RA, Sims TR, Imece AF. Investigation of potential islanding of a self- impedance monitoring. In: Proceedings of the sixth international conference
commutated static power converter in photovoltaic systems. IEEE Trans on developments in power system protection; 1997. p. 95–8.
Energy Convers 1990;5(4):624–31. [31] Xu W, Zhang G, Li C, Wang W, Wang G, Kliber J. A power line signaling based
[19] Redfern MA, Usta O, Fielding G. Protection against loss of utility grid supply for technique for anti-islanding protection of distributed generators-part I:
a dispersed storage and generation unit. IEEE Trans Power Deliv 1993;8 scheme and analysis. IEEE Trans Power Deliv 2007;22(3):1758–66.
(3):948–54. [32] Wang W, Kliber J, Zhang G, Xu W, Howell B, Palladino T. A power line signaling
[20] Mahat P, Chen Z, Bak-Jensen B. Review of islanding detection methods for based scheme for anti-islanding protection of distributed generators—part II:
distributed generation. In: Proceedings of the IEEE third international con- field test results. IEEE Trans Power Deliv 2007;22(3):1767–72.
ference on electric utility deregulation and restructuring and power technol- [33] Chowdhurya SP, Chowdhurya S, Crossley PA. Islanding protection of active
ogies; 2008. p. 2743–8. distribution networks with renewable distributed generators: a comprehen-
[21] Dys´ ko A, Booth C, Anaya-Lara O, Burt GM. Reducing unnecessary disconnec- sive survey. Electr Power Syst Res 2009;79(6):984–92.
tion of renewable generation from the power system. IET Renew Power Gener [34] Mahat P, Chen Z, Bak-Jensen B. A hybrid islanding detection technique using
2007;1(1):41–8. average rate of voltage change and real power shift. IEEE Trans Power Deliv
[22] Pai FS, Huang SJ. A detection algorithm for islanding-prevention of dispersed 2009;24(2):764–71.
consumer-owned storage and generating units. IEEE Trans Energy Convers [35] Menon V, Nehrir MH. A hybrid islanding detection technique using voltage
2001;16(4):346–51. unbalance and frequency set point. IEEE Trans Power Syst 2007;22(1):442–8.
[23] Singam B, Hui LY. Assessing SMS and PJD schemes of anti-islanding with [36] Chang WY. A hybrid islanding detection method for distributed synchronous
varying quality factor. In: Proceedings of the IEEE international power and generators. In: Proceedings of the IEEE 2010 International power electronics
energy conference; 2006. p. 196–201. conference; 2010. p. 1326–30.