You are on page 1of 15

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/361164908

An innovative hybrid method for islanding detection using fuzzy classifier for
different circumstances including NDZ

Article in Journal of Radiation Research and Applied Sciences · June 2022


DOI: 10.1016/j.jrras.2022.04.008

CITATIONS READS

0 116

7 authors, including:

Yasser Ahmed Elshrief Amin Danial


Egyptian Academy for Engineering and Advanced Technology Affiliated to Ministr… Egyptian Academy for Engineering and Advanced Technology
14 PUBLICATIONS 56 CITATIONS 18 PUBLICATIONS 54 CITATIONS

SEE PROFILE SEE PROFILE

Belgacem Bouallegue Dalal Helmi


King Khalid University EEHC
92 PUBLICATIONS 608 CITATIONS 18 PUBLICATIONS 37 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Automatic translation system from Arabic Sign Language to Arabic Text View project

Networked control systems View project

All content following this page was uploaded by Amin Danial on 08 June 2022.

The user has requested enhancement of the downloaded file.


Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Contents lists available at ScienceDirect

Journal of Radiation Research and Applied Sciences


journal homepage: www.journals.elsevier.com/journal-of-radiation-research-and-applied-sciences

An innovative hybrid method for islanding detection using fuzzy classifier


for different circumstances including NDZ
Yasser A. Elshrief a, b, *, Amin D. Asham a, Belgacem Bouallegue c, Abdelmoty Ahmed c,
Dalal H. Helmi d, Belal A. Abozalam b, Sameh Abd-Elhaleem b
a
Electrical Engineering Department, Egyptian Academy for Engineering and Advanced Technology (EAE&AT), Affiliated to the Ministry of Military Production, Cairo,
32511, Egypt
b
Industrial Electronics and Control Engineering Department, Menofia University, Menouf, 32511, Egypt
c
Department of Computer Engineering, King Khalid University, Abha, Saudi Arabia
d
Ministry of Electricity, Sector of Electricity Market and Cross-Border Interconnection at (EEHC), Cairo, Cairo, 11765, Egypt

A R T I C L E I N F O A B S T R A C T

Keywords: Distributed energy resources (DERs) are an alternative to conventional fossil fuel-based energy resources for a
Distributed generation pollution-free operation. But the integration of these DERs to the main grid comes with certain challenges, out of
Fuzzy classifier which unintentional islanding detection is of prior importance. A lot of techniques have been used for detecting
Grid faults
islanding, among these techniques, passive and active. Passive islanding detection methods have drawbacks in
Islanding detection
Non-detection zone
the form of the existence of large non-detection zones whereas active detection methods have devastating effects
on power quality. To overcome these drawbacks, a proposed hybrid passive-active systematic methodology based
on a smart classifier that decides to use an active method instead of a passive one; is presented. The simulation
results confirm that the proposed scheme is accurate, robust, fast, simple to implement for inverter-based DGs,
and is able enough to perform effective and rapid detection of islanding in all the circumstances for the chosen
system model without degrading power quality. The proposed scheme was simulated using the MATLAB/
Simulink platform, and the results reflect its potential with a comparative study.

1. Introduction case of maintenance services. On the other hand, unintentional or un­


planned islanding happens when the utility loses control of voltage and
In recent years, the gradual shift to the non-traditional energy gen­ frequency due to any fault in the grid or malfunctions in the load.
eration such as distributed generation (DG) sources especially wind Hence, unintentional islanding is one of the main technical problems
energy, photovoltaic and fuel cell can improve reliability, power quality, that are produced when connecting DG to a power system (Elshrief et al.,
decreases the power losses in the distribution system, avoid capacity 2021a) and is assumed a disagreeable state as it may cause a shock
upgrades in power transmission and distribution, and also improves hazard for humans working on the utility side and have negative impacts
voltage profile (Baneshi et al., 2020). on management, operation, and protection of DGs. Moreover, uninten­
Despite all these advantages, using DGs would lead to the appearance tional islanding causes a reduction in the system voltage stability, and
of the islanding phenomenon, which is defined as the state in which a bad restoration (out of phase switching of re-closers results to molder the
portion of the power line consists of distributed energy resources and DG, near loads, and utility equipment). When unintentional islanding
loads are isolated from the power system, and remain energized. happens, frequency and voltage in the islanded zone cannot be
Intentional and unintentional are the types of islanding, intentional or controlled by the main utility system, so the recommendation of
planned islanding is used to permit the DG to supply power to the load in IEEE-1547- 2003 is to cut off all DGs from the main grid as soon as

Peer review under responsibility of The Egyptian Society of Radiation Sciences and Applications.
* Corresponding author. Electrical Engineering Department, Egyptian Academy for Engineering and Advanced Technology (EAE&AT), Affiliated to the Ministry of
Military Production, Cairo, 32511, Egypt.
E-mail addresses: yasserelshrief@eaeat.edu.eg, yasserelshrief@yahoo.com (Y.A. Elshrief), amin.danial@eaeat.edu.eg (A.D. Asham), belgacem.bouallegue2015@
gmail.com (B. Bouallegue), abd2005moty@yahoo.com (A. Ahmed), dalal_helmi@yahoo.com (D.H. Helmi), drbelalabozalam@yahoo.com (B.A. Abozalam), sameh.
abdalhaleem@el-eng.menofia.edu.eg (S. Abd-Elhaleem).

https://doi.org/10.1016/j.jrras.2022.04.008
Received 17 February 2022; Received in revised form 10 April 2022; Accepted 28 April 2022
1687-8507/© 2022 The Author(s). Published by Elsevier B.V. on behalf of The Egyptian Society of Radiation Sciences and Applications. This is an open access
article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 1. Power flow at Point of Common coupling (PCC).

possible within 2 s after the occurrence of islanding (Elshrief et al., distributed generation microgrid (MG) and can catch the islanding, but
2021b). it has a drawback in the case of disconnection of a single DG. A hybrid
Generally, there are two main techniques for detecting unintentional technique between an artificial neural network and Fourier transform is
islanding phenomenon, local and remote techniques. Remote techniques presented in (Abd-Elkader et al., 2014) to detect islanding on the wind
require huge infrastructure and need more reliable communication than turbine but is more complex in construction. A lot of artificial intelli­
local schemes (Elshrief et al., 2019a, 2021c). So, they are considered gence techniques are used to catch islanding as support vector machines
non-economical as they need extremely high cost and hence, these (Matic-Cuka & Kezunovic, 2014), the decision tree (Azim et al., 2015),
techniques are rarely used on small scale (Elshrief et al., 2019b). But the adaptive ensemble classifier (Khamis et al., 2018), the modified
local techniques can be categorized into three classes: passive, active, Slantlet transform (Ahmadipour et al., 2019), machine learning,
and artificial intelligence. About passive techniques, the utility param­ Hilbert-Huang transforms techniques (Mishra & Rout, 2018), and
eters like current, voltage, harmonic distortion, and frequency are neuro-fuzzy logic (Bitaraf et al., 2012).
observed from the DG side at the point of common coupling (PCC) to From the previous work, no one uses the combination between
detect the islanding occurrence (Elshrief et al., 2021c). The change of passive and active with a fuzzy classifier to solve the problem of NDZ, in
phase deviation over time was applied in (Haider et al., 2018). The this paper a novel anti-islanding detection technique is used to get zero
variations of active power with the time and frequency were used in NDZ and accurate islanding detection without degrading the power
(Karegar & Shataee, 2008). In (Freitas et al., 2005) islanding is detected quality and the results are verified by Matlab/Simulink. The results of
using the measurement of harmonic distortion. Another way is pre­ this paper are compared with the results in (Seyedi et al., 2021). As in
sented in (Samui & Samantaray, 2011) by using the displacement of (Seyedi et al., 2021) islanding is detected using a hybrid technique be­
phase angle and the voltage surge. In (Reigosa et al., 2014), the authors tween the rate of change of voltage and the rate of change of active
used the frequency and voltage abnormalities values to identify the power.
islanding incidence. All these passive techniques can’t detect islanding The main contributions of the proposed technique can be summa­
in NDZ but they can detect islanding in a shorter time. rized as:
About active techniques, they depend on disturbing the signals by
injecting some disturbances into the system (Reigosa et al., 2014). Also, 1 It is able enough to perform effective and rapid detection of islanding
signal processing approaches and artificial intelligence techniques are in all the circumstances for the chosen system model without
used. Fast Fourier transform (FFT) and Fourier transform (FT) (Hashemi degrading power quality.
et al., 2017; Raza et al., 2015)depend only on a frequency spectrum and 2 Obtaining zero NDZ.
are unable to catch islanding due to the instability features in the cur­ 3 Is accurate, robust, fast, and simple to implement for inverter-based
rent, power, and voltage waveforms. Therefore, time-frequency was DGs.
proposed to observe non-stationary disturbances, but it also failed to 4 The proposed technique not only detects islanding but also differs
catch islanding in some cases (Kim et al., 2019) owing to its between it and grid faults.
non-adaptive window size. In (Alshareef et al., 2014), the machine 5 Overcome the problem of determining the appropriate thresholds
learning algorithm was joined with wavelet transform (WT) to notice the using the fuzzy classifier.
islanding incident and this mixture gave high adaptability and more
accuracy but it has a complexity in learning the rules. The output of WT The rest of the paper is structured in four sections. Section 2 presents
was used in (ElNozahy et al., 2011) as input for neuro-fuzzy and this the definition non-detection zone. Section 3 presents the proposed
combination had many merits as simplicity and fast but not accurate. In technique and its flowchart. The simulation results of islanding condi­
(Mohanty et al., 2015) a combination of support vector machines tions that are detected by the proposed technique using MATLAB/
(SVMs) algorithm and WT with s-transform improves the performance SIMULINK are presented in Section 4. A comparison between the pro­
and minimizes noise but it can’t detect islanding in NDZ. posed technique and the other technique in (Seyedi et al., 2021) is
WT combined with a decision tree classifier to get high applicability presented in section 5. Finally, in Section 6, the conclusions are
to multi distributed generation networks and fast detection in (Lidula & summarized.
Rajapakse, 2012) but it didn’t differ between normal islanding and grid
faults. A hybrid technique of WT with fuzzy in (Samantaray et al., 2010) 2. Non-Detection Zone (NDZ)
is accomplished by a successful performance under various conditions.
All the methods based on wavelet transform depend on a selected signal The NDZ is an area where the intended detection method cannot able
for detection which has a great challenge as if the threshold of this signal to detect the islanding event properly at a certain time (Elshrief et al.,
is very small, false tripping maybe happens and if is high, the detection 2021a). It is the main parameter for evaluating the effectiveness of any
may not occur. In (Wang et al., 2018) they solve the problem of signal islanding detection method, especially the passive detection methods.
threshold selection, by applying the neuro-fuzzy algorithm on multiple As the passive islanding detection methods have the disadvantage that

130
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

( )2 ( )2
V ΔP V
− 1≤ ≤ − 1 (3)
Vmax P Vmin
( )2 ( )2
f ΔQ f
1− ≤ ≤1− (4)
fmin P fmax
As (Vmax .fmax.Vmin.andfmin ) are the maximum (voltage, frequency)
and the minimum (voltage, frequency) respectively calculated at the
PCC, and their values are detected according to the IEEE 1547 series
standard are shown in Table 1.

The load reactive power can be illustrated by equation (5) (Seyedi


et al., 2021).
( )
1
QLOAD = 3v2 − wn C (5)
wn L
Fig. 2. NDZ of islanding (shaded region).
where wn .L. ​ C ​ and ​ v are the nominal frequency load, inductance,
capacitance, and voltage, respectively. By taking into account the unity
Table 1
power factor, so the reactive power mismatch is equal to the load
Standard IEEE 1547 for anti-islanding techniques (Elshrief et al., 2021a).
reactive power (Seyedi et al., 2021) which can be calculated using (6):
Parameters Standard ( )
3v2 ( ) 3v2 wn 2
Range of Voltage 88% ≤ V ≤ 110% ΔQ = 1 − wn 2 LC = 1− (6)
Range of Frequency 49.5 HZ ≤ f ≤ 50.5 HZ wn L wn L wr 2
The maximum time for islanding detection 2s
Total Harmonic Distortion (THD %) ≤5% where wr is the load resonance frequency.
wr = wn ± Δw (7)
they are not able to work when the DG generation matches the local load
Therefore, the reactive power mismatch for NDZ calculation can be
perfectly, or the power difference between them is small. The Power
written as follows (Seyedi et al., 2021):
flow between the utility grid and DG source is shown in Fig. 1.
( )
NDZ represented by the shaded region in Fig. 2, can also be defined 3v2 fn 2
as the area where there is a power mismatch between the utility grid and ΔQ = 1− (8)
wn L (fn ± Δf )2
DG source that is established by a coordinate system using an active
power mismatch axis and a reactive power mismatch axis (ΔP and ΔQ), where wr .Δf are the resonance frequency and permissible frequency
that are declared in equation (1)&2 (Elshrief et al., 2021a): range, respectively.
ΔP = PDG − PLOAD (1) On the other hand, ΔP is declared by equation (1) (Elshrief et al.,
2021a) and the load active power can be illustrated by equation (9)
ΔQ = QDG − QLOAD (2) (Seyedi et al., 2021).

PDG andQDG are the DG source real and reactive power, respectively. 3v2
PLOAD = (9)
Similarly,PLOAD andQLOAD are loads of real and reactive powers, their R1
threshold limits are determined by equations (3) & (4) (Elshrief et al., The active power mismatch is the difference between the active
2021a). power attracted by local load resistance as R1 which causes deviation as
ΔV in voltage (Seyedi et al., 2021).

Fig. 3. The three-level VSC Stage connected to the inverter in the DG power stage.

131
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 4. The Simulink implementation of the VSC Control.

( )
3v2 1 grid voltage, then the active power and the reactive power can be
ΔP = − 1 (10) expressed by (Khamis et al., 2018):
R1 (1 ± ΔV)2
P = 3 / 2(Vd id ) (13)
3. Concept of the proposed technique / ( )
Q = − 3 2 Vd iq (14)
The concept of the proposed technique depends mainly on the Equations (13) and (14) indicate that P is proportional to id and can
voltage source converter (VSC) controller and will be described in the therefore be controlled by it. Similarly, the reactive power Q is regulated
following paragraph: by controlling iq . So the proposed technique can control islanding by
making a variation in iq that will lead to a variation in reactive power
hence the reactive power will be the main parameter for detecting
3.1. Voltage Source Converter controller
islanding.
The three-level VSC that is declared in Fig. 3 and its Simulink
implementation is depicted in Fig. 4 (Elshrief et al., 2021a) regulates the 3.2. The main concept of the proposed technique
dc bus voltage at 600 V and keeps the unity power factor.
VSC controller presented in Fig. 4, shows the control system that uses From the whole anti-islanding detection techniques mentioned in
two control loops: an external control loop that regulates the reference section 1, we conclude that the passive techniques are the simplest and
of dc-link voltage (Vdc− ref) to 600 V and an internal control loop that cheapest but it has large NDZ, on the other hand, active techniques have
regulates Id and Iq grid currents (active and reactive current compo­ a small NDZ but it degrades power quality and insert disturbances
nents). Id current reference (idref ) is the output of the dc voltage regu­ during the grid-connected operation. The proposed method aims to
lator. Iq current reference (iqref ) is set to zero to maintain the unity power detect islanding accurately even in NDZ, reduce the injection of dis­
factor. The voltage outputs of the current controller (vd &vq ) are con­ turbances and overcome the problem of determining the appropriate
verted to three modulating signals (Vvsc ) used by the pulse-width mod­ thresholds using a fuzzy classifier system. The proposed technique
ulation (PWM) three-level pulse generator. measures the reactive power of PV (QPv ), load (Qload ), and grid (QGrid )
So the active and reactive power of the PV grid converter in a dq using a power meter (PM) as depicted in Fig. 5, and the flowchart of the
frame is given by Equations (11) and (12) (Khamis et al., 2018): proposed technique is presented in Fig. 6. These three values of reactive
/ ( ) power are the inputs to the fuzzy system to make a decision, for example;
P = 3 2 Vd id + Vq iq (11)
if the value of reactive power of the grid is small and the other two
/ ( ) values aren’t small, that means, there is no connection between the DG
Q = 3 2 Vd id − Vq iq (12)
and utility grid Then synchronization is checked, if it is active and has
The idref is the desired active output power control, while iqref is the reached steady-state (SS), so islanding is detected. But if the value of
desired reactive output power, and also used to achieve unity power reactive power of the grid isn’t small, in this case, it means that there is a
factor operation, in which the grid current vector is in phase with the connection between DG and grid, and hence no islanding is detected. In
grid voltage vector, thus iqref should be zero. Aligning the d axis with the the case of a small value of the 3 m, the order for disturb is generated and

Fig. 5. Schematic diagram of the proposed technique.

132
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 6. Flowchart of the proposed technique.

Fig. 7. Matlab implementation of the proposed technique.

applies a change to the Iq of the inverter by varying power factor to be in iq is within 5% according to standards (Elshrief et al., 2021a) and it is
0.95 at minimum to guarantee that the power quality of the system at measured by equation (15)
PCC isn’t affected. This change in Iq will affect the reactive power of PV √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∑H 2̅
by a small variation, if there isn’t any change on the reactive power of h=2 Ih
the grid, and synchronization is still active, hence islanding is detected, THD = × 100 (15)
I1
but if there is a change that means there is no islanding and repeats the
As Ih is the RMS value of the hth harmonic component of the current
measurements. THD is an indication to ensure that the effect of variation
and I1 is the RMS value of the fundamental component of the current.

133
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 8. Membership function for reactive power of PV (X1).

3.3. Fuzzy inference system 5. If (PV-Q is Small) AND (Grid-Q is Medium) AND (Load-Q is Me­
dium) THEN (Island is Small) AND (Disturb is Small)
According to the frequent parameters variation of the studied sys­ 6. If (PV-Q is Small) AND (Grid-Q is Large) AND (Load-Q is Medium)
tem, the fuzzy classifier is more effective compared with the crisp THEN (Island is Small) AND (Disturb is Small)
classifier. As shown in Fig. 7, three reactive power measurements of PV, 7. If (PV-Q is Small) AND (Grid-Q is Small) AND (Load-Q is Large)
load, and grid are entering the fuzzy system as inputs, and the outputs THEN (Island is Medium) AND (Disturb is Medium)
are the island and disturb. If the fuzzy output is an island, then it checks 8. If (PV-Q is Small) AND (Grid-Q is Medium) AND (Load-Q is Large)
the synchronization between grid and DG if yes then the islanding de­ THEN (Island is Small) AND (Disturb is Medium)
cision is energized. But if the fuzzy output is disturb, then it enters the 9. If (PV-Q is Small) AND (Grid-Q is Large) AND (Load-Q is Large)
Matlab function to make the change in Iq of the inverter by the way that THEN (Island is Small) AND (Disturb is Small)
doesn’t be less than 0.95 power factor, then the three reactive values are 10. If (PV-Q is Medium) AND (Grid-Q is Small) AND (Load-Q is
checked again to get the final decision if islanding or repeats the Small) THEN (Island is Medium) AND (Disturb is Small)
sequence of disturbance after guaranteeing the synchronization between 11. If (PV-Q is Medium) AND (Grid-Q is Small) AND (Load-Q is Me­
grid and DG. So the fuzzy rules are in the form of (if, then rule) and are dium) THEN (Island is Large) AND (Disturb is Medium)
represented as follows: 12. If (PV-Q is Medium) AND (Grid-Q is Small) AND (Load-Q is Large)
THEN (Island is Large) AND (Disturb is Small)
1. If (PV-Q is Small) AND (Grid-Q is Small) AND (Load-Q is Small) 13. If (PV-Q is Medium) AND (Grid-Q is Medium) AND (Load-Q is
THEN (Island is Small) AND (Disturb is Large) Small) THEN (Island is Small) AND (Disturb is Medium)
2. If (PV-Q is Small) AND (Grid-Q is Medium) AND (Load-Q is 14. If (PV-Q is Medium) AND (Grid-Q is Medium) AND (Load-Q is
Small) THEN (Island is Small) AND (Disturb is Small) Medium) THEN (Island is Small) AND (Disturb is Small)
3. If (PV-Q is Small) AND (Grid-Q is Large) AND (Load-Q is Small) 15. If (PV-Q is Medium) AND (Grid-Q is Medium) AND (Load-Q is
THEN (Island is Small) AND (Disturb is Small) Large) THEN (Island is Medium) AND (Disturb is Small)
4. If (PV-Q is Small) AND (Grid-Q is Small) AND (Load-Q is Me­ 16. If (PV-Q is Medium) AND (Grid-Q is Large) AND (Load-Q is Small)
dium) THEN (Island is Medium) AND (Disturb is Medium) THEN (Island is Small) AND (Disturb is Small)

Fig. 9. Membership function for reactive power of grid (X2).

134
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 10. Membership function for reactive power of load (X3).

Fig. 11. Membership function for the island (Q1).

17. If (PV-Q is Medium) AND (Grid-Q is Large) AND (Load-Q is Me­ 20. If (PV-Q is Large) AND (Grid-Q is Small) AND (Load-Q is Medium)
dium) THEN (Island is Small) AND (Disturb is Small) THEN (Island is Large) AND (Disturb is Small)
18. If (PV-Q is Medium) AND (Grid-Q is Large) AND (Load-Q is Large) 21. If (PV-Q is Large) AND (Grid-Q is Small) AND (Load-Q is Large)
THEN (Island is Small) AND (Disturb is Small) THEN (Island is Large) AND (Disturb is Small)
19. If (PV-Q is Large) AND (Grid-Q is Small) AND (Load-Q is Small) 22. If (PV-Q is Large) AND (Grid-Q is Medium) AND (Load-Q is Small)
THEN (Island is Small) AND (Disturb is Large) THEN (Island is Medium) AND (Disturb is Small)

Fig. 12. Membership function for Disturb (Q2).

135
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 13. Islanding detection in normal cases.

23. If (PV-Q is Large) AND (Grid-Q is Medium) AND (Load-Q is Me­ 4. The Matlab implementation results of a hybrid passive-active
dium) THEN (Island is Small) AND (Disturb is Small) technique with a fuzzy classifier scheme
24. If (PV-Q is Large) AND (Grid-Q is Medium) AND (Load-Q is Large)
THEN (Island is Small) AND (Disturb is Small) The proposed technique is applied to the model presented in (Elshrief
25. If (PV-Q is Large) AND (Grid-Q is Large) AND (Load-Q is Small) et al., 2021d). All figures of the results consist of three columns; the
THEN (Island is Small) AND (Disturb is Medium) three input signals of the fuzzy system (Q-PV, Q-load, Q-Grid) and total
26. If (PV-Q is Large) AND (Grid-Q is Large) AND (Load-Q is Medium) harmonic distortion of voltage (THD-V) are presented in the first col­
THEN (Island is Small) AND (Disturb is Small) umn. Active powers (P-PV, P-Load, P-Grid) and THD of current (THD-I)
27. If (PV-Q is Large) AND (Grid-Q is Large) AND (Load-Q is Large) are presented in the second column, the third column consist of the
THEN (Island is Small) AND (Disturb is Small) synchronization signal between DG and grid, the variation signal for the
inverter (iq), and the two outputs of the fuzzy system; disturb and island.
3.4. Membership function formulation The results are classified into two types: by applying varying load and
grid faults as presented below:
There are many different choices of membership functions, and after
trying, the trapezoidal membership function is the most effective
compared to the triangular and Gaussian one. Reactive powers of PV, 4.1. Simulation results of diff erent values of load for islanding
Grid, and load are taken to be the fuzzy inputs (x1, x2, and x3), but the phenomenon
island and disturb are the fuzzy outputs (Q1 and Q2). The following
Figs. 8–12 are the membership functions for the monitored parameters 4.1.1. Power mismatch between DG and Load (normal case)
inputs and outputs. Fig. 13 presents the simulation result of applying islanding at 0.3 s
after disconnection of the circuit breaker in case of a load that consumes
25 kW, 10 kVAr. So in this figure, the Q-grid starts to supply the Q-load

Fig. 14. Islanding detection in NDZ case.

136
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 15. Applying single-phase fault to the proposed technique.

until the synchronization becomes active then it continues supplying the 4.1.2. Power match between DG and Load (NDZ case)
Q-load and it is found that Q-PV is still zero as iq equals zero. This will In the case of (NDZ) in Fig. 14 the load consumes 100 kW, 0 kVAr,
last until islanding occurs, then the value of Q-Grid decrease to zero and after disconnection, it is found that the value of Q-PV equals to Q-Grid
the Q-PV start increasing to substitute the value of Q-load. In this case, which means an active technique must be used so the fuzzy generates the
the decision of fuzzy is to generate the islanding detection signal and be disturbance signal and repeats the measurements if the status of disturb
sure that the synchronization signal between grid and DG is active, and is held active for 0.01 s, Iq will be changed according to the change of
then islanding is detected. In this situation, there is no variation in Iq and power factor to be 0.95 instead of 1. THD is an indication to ensure that
hence no disturb signal, and so the THD-I and THD-V of the system don’t the effect of variation in iq is within 5% according to standards (Elshrief
reach 5% in the zone of the detection according to IEEE standards hence et al., 2021b). If this variation affects the Q-grid and the synchronization
there is no effect on the power quality of the system. Fig. 13 also presents signal is still active that means there is no islanding, but if there is no
the relation between the active powers as the P-grid starts to supply the variation in the Q-grid, islanding will be detected. If the synchronization
P-load until the synchronization becomes active then the P-PV switch to is failed for any reason islanding can’t be detected.
supply the P-load instead of the P-grid and the remaining power went to Fig. 14 also presents the relation between the active powers as the P-
the grid, which will last until islanding occurs, then the value of P-grid grid starts to supply the P-load until the synchronization becomes active,
decrease to zero and the whole P-PV went to load. then the P-PV switch to supply the P-load instead of the P-grid that will

Fig. 16. Applying double phase fault to the proposed technique.

137
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 17. Applying double phase fault with the ground to the proposed technique.

Fig. 18. Applying three-phase fault to the proposed technique.

last without any change after islanding occurrence so active power can’t and THD-V of the system don’t reach 5% and hence no effect on the
be assumed a good parameter to detect islanding in case of NDZ. power quality of the system. Fig. 15 also presents the relation between
the active powers as the P-grid starts to supply the P-load until the
4.2. Simulation results of diff erent types of faults impact synchronization becomes active then the P-PV switch to supply the P-
load instead of the P-grid and the remaining power went to the grid,
Faults are applied to clarify the accuracy of the proposed technique which will last until a fault occurs. That will last until a fault occurs, then
by observing the behavior of each fault. For example, a single-phase the value of P-PV decrease to zero, and the grid return to supply the P-
fault applied at 0.3 s for a load that nearly consumes 25 kW, 10 kVAr, load. The other faults as a double-phase fault, double-phase fault with
and its performance presented in Fig. 15, the Q-grid starts to supply the the ground, three-phase fault, and three-phase fault with the ground are
Q-load until the synchronization becomes active. Then it continues applied with the same load, and their performance are shown in
supplying the Q-load and it is found that Q-PV is still zero as Iq equals Figs. 16–19 sequentially.
zero, which will last until a fault occurs. In this case, there is a discon­
nection in the synchronization signal between grid and DG due to the
fault, so fuzzy didn’t generate the islanding detection signal. As there is
no variation in Iq hence no disturb signal, as a result of that the THD-I

138
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 19. Applying three-phase fault with the ground to the proposed technique.

thresholds detected by IEEE standards. The results of these techniques


are classified into two types: by applying varying load and grid faults as
declared below:

5.1. Simulation results of diff erent values of load for islanding


phenomenon in (Seyedi et al., 2021)

Fig. 21 presents the simulation result using the combination between


ROCOV and ROCOAP (Seyedi et al., 2021) of applying islanding at 0.3 s
after disconnection of the circuit breaker that connects the DG with the
utility grid in case of a low load value that consumes 25 kW, 10 kVAr. So
in this figure, the connected load has a power less than the PV genera­
tion. If the ROCOV is above zero value, the ROCOAP is activated. Then,
the over-voltage relay exceeds the threshold value in case of a lower load
and hence islanding is detected.
Fig. 22 presents the simulation results of applying relays in (Seyedi
et al., 2021) in the case of the non-detection zone, by using loads that
consume 103 kW, 0 kVAr. It is clear from this figure that the combina­
tion between ROCOV and ROCOAP can’t detect islanding in NDZ. The
case of applying a large load is presented in Fig. 24.

5.2. Simulation results of diff erent types of faults impact in (Seyedi et al.,
2021)

Faults are applied to clarify the accuracy of the proposed technique


and the behavior is recorded in each case. For example, a double-phase
Fig. 20. The flowchart of the ROCOV and ROCOAP technique (Seyedi fault is applied on the grid at 0.3 s in case of similarity between the load
et al., 2021). consumption power and the power generated from the DG, and its effect
on islanding detection relays mentioned in reference (Seyedi et al.,
5. Comparison between the proposed technique and the hybrid 2021) and is shown in Fig. 23. The performance of the other faults as
technique published in [ 29] single-phase fault, double-phase fault with the ground, three-phase
fault, and three-phase fault with the ground are summarized in Fig. 24
The proposed technique is compared with a hybrid technique pre­ showing the superiority of the proposed technique over the compared
sented in (Seyedi et al., 2021). one in (Seyedi et al., 2021) and that is presented in values with yellow
It consists of a combination of the rate of change of voltage (ROCOV) colors.
and the rate of change of active power (ROCOAP) presented in (Seyedi
et al., 2021)and its flowchart depicted in Fig. 20. The ROCOAP is only 6. Conclusion and future work
activated when the islanding condition is suspected by ROCOV and the
decision is taken if the value of voltage or frequency relay exceeds the This paper makes a comprehensive analysis of a new hybrid meth­
odology for detecting islanding operations in inverter-based DGs. This

139
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 21. Islanding detection relays when low power load in (Seyedi et al., 2021).

Fig. 22. Islanding detection relay of NDZ in (Seyedi et al., 2021).

proposed technique is accurate, robust, fast, simple to implement for passive detection technique if the decision of fuzzy is islanding, but if it
inverter-based DGs, and can work properly with grid faults. The pro­ can’t decide, the FIS system activates the active mode and injects
posed method combines the passive method and active method with a reactive power. The proposed hybrid technique is compared with the
fuzzy classifier system, which has three inputs measurements (reactive hybrid technique between ROCOV and ROCOAP presented in (Seyedi
power of, PV, load, and grid). According to their values, FIS decides et al., 2021). These techniques are tested by considering different cir­
whether the islanding is detectable, or a reactive power must be injected cumstances of loads including NDZ and the grid faults in the power
to be able to detect the islanding situation. Hence, the system acts as a system. From the comparison, it is clarified that the hybrid proposed

140
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Fig. 23. Double phase fault in NDZ in (Seyedi et al., 2021).

Fig. 24. Comparison between the hybrid proposed technique and the hybrid technique in (Seyedi et al., 2021) in terms of the time of detection in (sec).

technique has superiority in time detection over the hybrid technique Acknowledgments
presented in (Seyedi et al., 2021) in all cases of varying load including
the NDZ also it has accuracy in differentiating between islanding cases The authors extend their appreciation to the Deanship of Scientific
and grid faults. A suggestion for future work is to extend the study. Our Research at King Khalid University for funding this work through Large
study only focuses on single DG so it would be interesting to introduce Groups [Grant No. RGP.2/208/43).
the proposed technique for a multi DG environment due to its less
computational burden and higher accuracy with smaller detection time. References
Another extension of this study is to include testing of adaptive fuzzy
(Elshrief et al., 2016a, 2016b, 2017; Wan et al., 2021), also a combi­ Abd-Elkader, A. G., Allam, D. F., & Tageldin, E. (2014). Islanding detection method for
DFIG wind turbines using artificial neural networks, 62. https://doi.org/10.1016/j.
nation between active frequency drift (Elshrief et al., 2020) and rate of ijepes.2014.04.052
change of frequency (Elshrief et al., 2019c) with ANFIS (Karaboga &
Kaya, 2019) can be introduced.

141
Y.A. Elshrief et al. Journal of Radiation Research and Applied Sciences 15 (2022) 129–142

Ahmadipour, M., et al. (2019). A Novel Islanding Detection Technique Using Modified Hashemi, F., Mohammadi, M., & Kargarian, A. (2017). Islanding Detection Method For
Slantlet Transform in Multi-distributed Generation, 112. https://doi.org/10.1016/j. Microgrid Based On Extracted Features From Differential Transient Rate Of Change Of
ijepes.2019.05.008 Frequency, 11. https://doi.org/10.1049/iet-gtd.2016.0795
Alshareef, S., Talwar, S., & Morsi, W. G. (2014). A New Approach Based on Wavelet Design Karaboga, D., & Kaya, E. (2019). Adaptive network based fuzzy inference system (ANFIS)
and Machine Learning for Islanding Detection Of Distributed Generation, 5. https://doi. training approaches: A comprehensive survey. Artificial Intelligence Review, 52(4),
org/10.1109/TSG.2013.2296598 2263–2293.
Azim, R., et al. (2015). A Decision Tree-based Approach For Microgrid Islanding Detection. Karegar, H. K., & Shataee, A. (2008). Islanding Detection Of Wind Farms by THD. https://
https://doi.org/10.1109/ISGT.2015.7131809 doi.org/10.1109/DRPT.2008.4523885
Baneshi, et al. (2020). Coordinated Power Sharing In Islanding Microgrids For Parallel Khamis, A., et al. (2018). Faster Detection Of Microgrid Islanding Events Using An Adaptive
Distributed Generations, 9. https://doi.org/10.3390/electronics9111927, 11. Ensemble Classifier, 9. https://doi.org/10.1109/TSG.2016.2601656
Bitaraf, H., et al. (2012). Neuro-fuzzy Islanding Detection In Distributed Generation. https:// Kim, M. S., et al. (2019). Comprehensive Review Of Islanding Detection Methods For
doi.org/10.1109/ISGT-Asia.2012.6303292 Distributed Generation Systems, 12. https://doi.org/10.3390/en12050837
ElNozahy, M. S., El-Saadany, E. F., & Salama, M. M. A. (2011). A Robust Wavelet-ANN Lidula, N. W. A., & Rajapakse, A. D. (2012). A Pattern-Recognition Approach For Detecting
Based Technique For Islanding Detection. https://doi.org/10.1109/PES.2011.6039158 Power Islands Using Transient Signals-part II: Performance Evaluation, 27. https://doi.
Elshrief, Y., Asham, A., Helmi, D., & Abozalam, B. (2019a). On remote anti-islanding org/10.1109/TPWRD.2012.2187344
detection techniques. The Future of Electricity Challenges and Opportunities, 297–304. Matic-Cuka, B., & Kezunovic, M. (2014). Islanding Detection For Inverter-Based Distributed
Elshrief, Y. A., Atlam, G. A., & Abozalam, B. A. (2016a). Adaptive–fuzzy logic power filter Generation Using Support Vector Machine Method, 5. https://doi.org/10.1109/
for nonlinear systems, 11 pp. 66–73). https://doi.org/10.9790/1676-1102016673, 2. TSG.2014.2338736
Elshrief, Y. A., Atlam, G. A., & Abozalam, B. A. (2017). Adaptive Hysteresis Active Power Mishra, M., & Rout, P. K. (2018). Detection And Classification Of Micro-Grid Faults Based on
Filter Using Fuzzy-Logic Controller For Nonlinear Systems, 26 pp. 99–112). https://doi. HHT And Machine Learning Techniques, 12. https://doi.org/10.1049/iet-
org/10.21608/mjeer.2017.63429, 1. gtd.2017.0502
Elshrief, Y. A., et al. (2016b). Comparison between the fixed_band HCC and adaptive HCC Mohanty, S. R., et al. (2015). Comparative Study Of Advanced Signal Processing Techniques
used for APF control, 11 pp. 54–59). https://doi.org/10.9790/1676-1103045459, 3. for Islanding Detection in a Hybrid Distributed Generation System, 6. https://doi.org/
Elshrief, Y. A., et al. (2019b). Merits And Demerits Of The Distributed Generations Connected 10.1109/TSTE.2014.2362797
To The Utility Grid, 28 pp. 259–262). https://doi.org/10.21608/mjeer.2019.70907, Raza, S., et al. (2015). Application of Signal Processing Techniques For Islanding Detection of
1. Distributed Generation in Distribution Network: A Review, 96. https://doi.org/10.1016/
Elshrief, Y. A., et al. (2019c). ROCOF for detecting Islanding of Photovolt system, 28 pp. j.enconman.2015.03.029
225–258). https://doi.org/10.21608/mjeer.2019.70897. ICEEM2019-Special Issue. Reigosa, D., et al. (2014). Active Islanding Detection For Multiple Parallel-Connected
Elshrief, Y. A., et al. (2020). AI Protection Algorithms For PV-grid Connection System (pp. Inverter-Based Distributed Generators Using High-Frequency Signal Injection, 29. https://
334–341). https://doi.org/10.1109/ITCE48509.2020.9047774 doi.org/10.1109/TPEL.2013.2263845
Elshrief, Y. A., et al. (2021a). Methods For Protecting Network From Islanding Danger, 9. Samantaray, S. R., et al. (2010). A Fuzzy Rule-Based Approach For Islanding Detection In
https://doi.org/10.36909/jer.v9i2.9695 Distributed Generation, 25. https://doi.org/10.1109/TPWRD.2010.2042625
Elshrief, Y. A., et al. (2021b). On active anti-islanding techniques: Survey, 22. https://doi. Samui, A., & Samantaray, S. R. (2011). Assessment of ROCPAD Relay for Islanding
org/10.11591/ijeecs.v22.i2.pp609-618 Detection in Distributed Generation, 2. https://doi.org/10.1109/TSG.2011.2125804
Elshrief, Y. A., et al. (2021c). Fast And Accurate Islanding Detection Technique For Microgrid Seyedi, M., et al. (2021). A Hybrid Islanding Detection Method Based On The Rates Of
Connected To Photovoltaic System, 14. https://doi.org/10.1080/ Changes In Voltage And Active Power For The Multi-Inverter Systems, 12. https://doi.
16878507.2021.1923913 org/10.1109/TSG.2021.3061567
Elshrief, Y. A., et al. (September 2021). A New Passive Islanding Detection Technique For Wan, J., et al. (2021). Compulsory islanding transition strategy based on fuzzy logic
Different Zones In Utility Grid, 9 pp. 131–143). https://doi.org/10.36909/jer. control for a renewable microgrid system. Mathematical Problems in Engineering,
v9i3A.9699 (3B). 2021, 13, 9959222.
Freitas, W., Huang, Z., & Xu, W. (2005). A Practical Method For Assessing The Effectiveness Wang, G., et al. (2018). State Variable Technique Islanding Detection Using Time-Frequency
Of Vector Surge Relays For Distributed Generation Applications, 20. https://doi.org/ Energy Analysis for DFIG Wind Turbine In Microgrid System, 80. https://doi.org/
10.1109/TPWRD.2004.838637 10.1016/j.isatra.2018.07.017
Haider, R., et al. (2018). Passive Islanding Detection Scheme Based On Autocorrelation
Function Of Modal Current Envelope For Photovoltaic Units, 12. https://doi.org/
10.1049/iet-gtd.2017.0823

142

View publication stats

You might also like