Use of analytical hierarchy process for selecting and prioritizing islanding detection methods in power grids

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 14, No. 3, June 2024, pp. 2422~2435
ISSN: 2088-8708, DOI: 10.11591/ijece.v14i3.pp2422-2435  2422
Journal homepage: http://ijece.iaescore.com
Use of analytical hierarchy process for selecting and prioritizing
islanding detection methods in power grids
Mohammad Abu Sarhan, Andrzej Bien, Szymon Barczentewicz
Department of Power Electronics, Faculty of Electrical Engineering, AGH University of Krakow, Krakow, Poland
Article Info ABSTRACT
Article history:
Received Oct 2, 2023
Revised Jan 12, 2024
Accepted Feb 7, 2024
One of the problems that are associated to power systems is islanding
condition, which must be rapidly and properly detected to prevent any
negative consequences on the system's protection, stability, and security.
This paper offers a thorough overview of several islanding detection
strategies, which are divided into two categories: classic approaches,
including local and remote approaches, and modern techniques, including
techniques based on signal processing and computational intelligence.
Additionally, each approach is compared and assessed based on several
factors, including implementation costs, non-detected zones, declining
power quality, and response times using the analytical hierarchy process
(AHP). The multi-criteria decision-making analysis shows that the overall
weight of passive methods (24.7%), active methods (7.8%), hybrid methods
(5.6%), remote methods (14.5%), signal processing-based methods (26.6%),
and computational intelligent-based methods (20.8%) based on the
comparison of all criteria together. Thus, it can be seen from the total weight
that hybrid approaches are the least suitable to be chosen, while signal
processing-based methods are the most appropriate islanding detection
method to be selected and implemented in power system with respect to the
aforementioned factors. Using Expert Choice software, the proposed
hierarchy model is studied and examined.
Keywords:
Analytical hierarchy process
Expert Choice
Islanding detection
Multi-criteria decision making
Power grids
This is an open access article under the CC BY-SA license.
Corresponding Author:
Mohammad Abu Sarhan
Department of Power Electronics, Faculty of Electrical Engineering, AGH University of Krakow
Aleja Adama Mickiewicza 30, Krakow, 30-059, Poland
Email: sarhan@agh.edu.pl
1. INTRODUCTION
Recently finding alternative renewable energy sources to be used in place of conventional power
systems and developing new technologies that can be employed in electricity production are both of utmost
importance. Due to the advantages that can be provided, such as lowering the upgrade of transmission and
distribution capacity, reducing distribution system losses, and improving system power quality, the
implementation of distributed generations (DGs), including solar modules, wind turbines, and synchronous
generators in power systems is significantly increasing. On the other hand, when operating DGs; several
factors including islanding circumstances that may have a detrimental effect on the system must be taken into
account.
This islanding phenomena occurs when the DGs experience a loss of grid, or electrical connection to
the primary utility grid, yet continue to provide electricity to the rest of the system [1]. As a result, this
phenomenon has a number of negative side effects on the network, including the possibility of system
parameters outside of acceptable limits, the failure of protective devices, potential harm to maintenance
personnel due to the continued operation of DGs, and potential damage to prime movers from the mechanical

Int J Elec & Comp Eng ISSN: 2088-8708 
Use of analytical hierarchy process for selecting and prioritizing islanding … (Mohammad Abu Sarhan)
2423
torque brought on by instantaneous reclosing. Therefore, it is crucial to quickly, correctly, and effectively detect
the islanding.
Numerous islanding detection methods (IDMs) have been put forth and grouped into four
categories: local techniques (passive, active, and hybrid); remote techniques; approaches based on signal
processing; and computationally intelligent techniques [2]–[6]. When choosing the most suitable technique to
be implemented in the system, various criteria must be taken into account because each method has
advantages and disadvantages over the others. Therefore, it is crucial to develop a simplified way for
determining which islanding detection technology is the most suitable for integration into the system. Multi-
criteria decision analysis (MCDA) is a good tool that can be used to solve this problem. However, depending
on the type of DG units and their connection topologies, the choice of IDM is very flexible. The selection of
islanding-detection techniques is influenced by several criteria, including the location of distributed
generation, the lifespan of distributed generation generators, and future expandability. The short circuit
capacity at point of common coupling (PCC), energy conversion/processing methods, DG unit capacity/size,
regulatory concerns enforcing requirements, and other factors can also have a considerable impact, either
directly or indirectly, on the choice of anti-islanding strategies. The proper selection of IDMs also involves
several additional considerations. There are many IDMs available, but none of them is perfect. Consequently,
a major concern is utilizing a suitable technique to assess various IDM types to determine their applicability
and to make future projections. Uncertainty prevents deterministic values from adequately accounting for the
constraints (criteria) of various IDM selection as well as the interactions between the constraints. Decision-
makers find it challenging to handle without a great deal of experience.
When a decision needs to be made after considering numerous, opposing, and negative evaluations,
MCDA is employed. These conflicts will be brought to light, and a suitable strategy will be developed to
produce a transparent procedure. The evaluation procedure in the area of power systems has already utilized
MCDA. There are numerous MCDA techniques that can be utilized to address some issues in this area,
including but not limited to the analytical hierarchy process (AHP), elimination and choice expressing reality
(ELECTRE), fuzzy sets, and evacuation management decision support system (EMDSS). Various commonly
utilized IDMs: ratio of change of frequency (RCF), phase jump detection (PJD), harmonic detection (DH),
impedance measurement (IM), slip-mode frequency shift (SMS), and Sandia frequency shift (SFS), were
examined using AHP in [7]. Both passive and active methods can be applied to those techniques. However,
no investigation was done on the other primary islanding detection categories. Additionally, it was noted that
there was a deficiency in the research conducted to date to identify a selection methodology that could be
used to the analysis of all significant islanding detection techniques, particularly those based on signal
processing and computational intelligence. Hence, this paper examines all the primary categories for
islanding detection to show how applicable AHP is to anti-islanding selection issues. This work's outcome is
accurate and efficient in comparison to the studies that were carried out. But in this work, only the primary
four criteria were considered. More criteria in the future, such as load type, dependability, applicability in the
event of multi-inverters, and sensitivity to cyber-attack, can be taken into consideration, once there are
sufficient studies covering those criteria accurately.
Two categories of islanding detection techniques were compared; conventional techniques, which
include local and remote techniques, and modern methods, which include techniques based on signal
processing and computational intelligence. Each solution is analyzed and evaluated using the AHP based on
several factors, including implementation costs, non-detected zones, power quality, and response times. As a
result, when the implementation cost requirement is the only consideration, then passive techniques are the
best choice. Selecting methods based on computational intelligence or signal processing is the best course of
action when the non-detected zone criterion is the only consideration. If the primary consideration is the
required level of power quality, then the best options are those that are passive, remote, computationally
intelligent, or based on signal processing. If the response time criterion is the only consideration, then the
best options to choose are those that rely on passive or signal processing. Nonetheless, passive and signal
processing-based approaches might be the best options provided these aspects are considered.
There are seven sections of the work that is being presented. The primary various types of islanding
detection techniques are examined in section 2. The selection criteria are described in section 3. The design and
process study of decision analysis are explained in section 4. The simulation based on expert choice software is
covered in section 5. The results and discussion are presented in section 6. The last section states with a conclusion.
2. ISLANDING DETECTION METHODS
Local approaches (passive, active, and hybrid), remote methods, signal processing-based methods,
and computationally intelligent-based methods are the four primary groups into which islanding detection
techniques fall. The operation of passive methods relies on tracking changes in system characteristics at the
point of common coupling (PCC). Active techniques alter various network injections, and the effect of the

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 14, No. 3, June 2024: 2422-2435
2424
injection on the system parameters is then examined. Active and passive techniques are used in hybrid
methods. The foundation of remote techniques is the gathering and exchange of data between the utility and
distributed generator (DG) sides. The foundation of how signal processing-based techniques work is the
extraction of system features. Methods based on computational intelligence operate through data training and
pattern recognition. The methods used to identify islanding detection are briefly described here.
2.1. Passive methods
System variables like voltage, frequency, current, power, or impedance are measured at the PCC
when passive methods are used in the system. The values of these parameters will fall within acceptable
ranges in the case of normal operation. The values of these parameters will, however, fluctuate and go above
the allowable threshold levels when islanding occurs. The protection relays that trip the main circuit breakers
to prevent the islanding action are used to examine and detect these fluctuations. Figure 1 depicts the process
involved in passive islanding detection. The term “passive methods” refers to a variety of strategies,
including voltage imbalance (VU), over/under voltage protection (O/UV), over/under frequency protection
(O/UF), rate of change of frequency (ROCOF), rate of change of active and reactive power (ROCOP), and
rate of change of frequency (ROCOF), rate of change of active and reactive power (ROCOP), voltage
unbalance (VU), and phase jump detection (PJD) [8]–[10].
2.2. Active methods
An external, tiny disturbance signal is injected into the DG output when active methods are used in
the system. Due to this injection, the system parameters will fluctuate and go above the permitted ranges
while the system is in an islanding condition. Figure 2 depicts the steps necessary for active islanding
detection. Numerous techniques fall under the category of active methods, including the active frequency
drift method (AFD), the Sandia frequency shift method (SFS), the Sandia voltage shift method (SVS), the
impedance measurement method (IM), the slip mode frequency shift method (SMFS), and the frequency
jump method (FJ) [11]–[14].
Figure 1. Flowchart of passive islanding detection
methods
Figure 2. Flowchart of active islanding detection
methods
2.3. Hybrid methods
Passive and active methodologies are used to create hybrid approaches. Hybrid method
implementation is accomplished in two parts. A passive strategy is used in the initial step primarily to
identify the islanding. An active method is utilized to precisely detect the islanding if it is still there after the
first step has been applied. Figure 3 depicts the steps necessary for hybrid islanding detection. Numerous
techniques, including the voltage imbalance and frequency set-point method, the voltage and actual power
shift method, the voltage fluctuation injection technique, the hybrid Sandia frequency shift and Q-f
technique, are included in hybrid methods [15]–[17].

2425
Figure 3. Flowchart of hybrid islanding detection methods
2.4. Remote methods
The utility side and the DG side must communicate for remote approaches to work. The islanding is
identified based on the utility's state of the circuit breakers. The DG unit is then triggered by providing the
appropriate tripping signal. The term “remote methods” refers to a variety of techniques, including power
line carrier communication (PLCC), signal produced by disconnect (SPD), supervisory control and data
acquisition (SCADA), transfer trip scheme, impedance insertion method, and phasor measuring unit
[18], [19].
2.5. Signal processing-based methods
Signal processing approaches are applied to lower the non-detection zone (NDZ) of passive methods
in islanding detection. These techniques have the additional benefit of being able to extract the voltage,
frequency, and current hidden aspects of the recorded signals at PCC when compared to passive methods.
The acquired features can then be utilized as input to a classification approach like artificial intelligence or
machine learning to determine if the system functions in an islanding situation or not. Figure 4 depicts the
steps necessary for signal processing-based islanding detection. The Fourier transformer method, Wavelet
transformer method, S-transformer method, and time-time transformer method are only a few examples of
the numerous signal processing-based techniques [20]–[22].
Figure 4. Flowchart of signal processing-based islanding detection methods

 ISSN: 2088-8708
2426
2.6. Computational intelligent based methods
Signal processing methods can increase islanding detection accuracy, but they cannot eliminate the
NDZ when the DG system is more complex. Giving the islanding detecting relay additional intelligence in
this situation can boost performance. Computationally intelligent methods for islanding detection can handle
multiple parameters at once. Choosing threshold values is not required with those methods, although there
has been a major computational overhead. Figure 5 depicts the process used in computational intelligent
islanding detection. There are several different computational intelligence-based methodologies, including
support vector machine, fuzzy logic, decision trees, and artificial neural networks [23]–[25].
Figure 5. Flowchart of computational intelligent-based islanding detection methods
3. SELECTION CRITERIA
Several factors can be used to evaluate the applicability and efficacy of islanding detection
approaches. Depending on the variables that are taken into consideration, each scenario can be successfully
handled using the most appropriate strategy. Below are the specifics of the requirements.
3.1. Implementation cost
It is considered that the cost of implementation represents a compromise between system cost and
quality. Passive approaches cost the least compared to other techniques. The most expensive approaches to
implement are remote ones because of their complexity and need for extra components. Table 1 provides a
brief comparison of islanding detection approaches based on cost [26]–[29].
Table 1. Comparison between IDMs based on cost
IDMs Cost
Passive methods Low
Active methods Low
Hybrid methods Low
Remote methods Very high
Signal processing methods Low
Computational intelligent methods High
3.2. Non-detected zone
The non-detected zone (NDZ) is the area of power imbalance where the islanding detection
method may fail to pick up the islanding. Therefore, when the power of the DGs equals the power of the

2427
load, the deviation amount of voltage and frequency can be very small, which has a significant impact on
the efficacy of detection. Passive approaches are less successful than active methods because of their
broader NDZ. Table 2 provides a brief comparison of islanding detection approaches based on non-
detected zone [26]–[29].
Table 2. Comparison between IDMs based on non-detected zone
IDMs Non-detected zone
Passive methods Large
Active methods Small
Hybrid methods Small
Remote methods Very small
Signal processing methods Very small
Computational intelligent methods Very small
3.3. Power quality
In addition to the generation requirement, the DGs must meet power quality requirements.
Electromagnetic interference, harmonic distortion, frequency deviation, and voltage fluctuation are a few
examples of power quality issues. The system's ability to recognize islanding has a significant impact on the
power quality. For instance, passive procedures do not degrade power quality but active solutions, which are
based on injections and disruption, may. Table 3 provides a brief comparison of islanding detection
approaches based on power quality [26]–[29].
Table 3. Comparison between IDMs based on power quality
IDMs Power quality
Passive methods No effect
Active methods Slightly degraded
Hybrid methods Slightly degraded
Remote methods No effect
Signal processing methods No effect
Computational intelligent methods No effect
3.3. Response time
Due to the negative impacts of islanding on network components and utility workers, the response
time of the islanding detection method is crucial and should be as quick as possible. Especially when an
island is working continuously on its own, the response times of most islanding detection approaches range
from half a second to two seconds, which is rather long. While remote techniques are faster than passive and
active methods, the passive method's response time is longer than the active method's response time. Table 4
provides a brief comparison of islanding detection approaches based on response time [26]–[29].
Table 4. Comparison between IDMs based on response time.
IDMs Response time
Passive methods Very fast
Active methods Slightly fast
Hybrid methods Slow
Remote methods Slow
Signal processing methods Very fast
Computational intelligent methods Fast
4. MULTI-CRITERIA DECISION ANALYSIS
Multi-criteria decision analysis (MCDA) is a supervisory process that employs several
methodologies and procedures for decision-making that can be used in complex decision-making situations
involving many competing criteria. Numerous MCDA techniques have been suggested and documented in
various research. The analytical hierarchy process (AHP) is one of these techniques, and it is regarded as a
straightforward and acceptable technique that can offer a thorough resolution for islanding detection
problems involving a variety of uncertainties and criteria. AHP is a decision support tool that may be used to
rank choice alternatives on a numeric scale by establishing subjectively determined qualifications for
intangible aspects.

 ISSN: 2088-8708
2428
By analyzing operational performances under various scenarios, AHP is used to choose the best
islanding detection methods for grid-connected DG systems. The following is the proposed hierarchical
model for islanding detection technique selection based on AHP: i) The main goal of the problem is to find
out the most appropriate islanding detection method; ii) The considered criteria for the decision are
implementation cost, non-detected zone, power quality, and response time; and iii) The decision alternatives
are passive methods, active methods, hybrid methods, remote methods, signal processing-based methods, and
computational intelligent-based methods.
The process begins with organizing a problem involving decision-making as an upside-down tree
with the primary objective at the top. At the second level are sub-objectives that contribute to the primary
goal. Every set at every level satisfies the goal of the level to which it is subordinate, and every partial target
at the second level can be broken down into third-level objectives. In this article, these partial objectives are
considered as criteria. At a lower level, each objective, or criterion, from the lower level is reached by
ranking the options and comparing them pairwise. Pairwise comparisons are carried out at the fundamental
scale shown in Table 5.
The number of alternatives, n, is used to assemble a n×n matrix. Matrix A is supplemented with
values 𝑎𝑖𝑗, where j is the alternative being compared with i and i is the basis alternative for comparison,
corresponding to row i, considering a specific criterion. 𝐴𝑖𝑗 takes on the value of 5, which can be interpreted
as a dominance of i over j, if the contribution of i to the criterion under consideration is highly significant in
relation to j. Values in between the ones displayed can also be taken into consideration. The following
significant associations are shown in the matrix using the procedure. Once the matrix is completed, the
procedure looks for a vector that represents each alternative's priority for the taken into consideration
criterion. The relationship between matrix A, its higher eigenvalue λ, and the related vector x is the first step
in obtaining this vector of priority, x as (1):
𝑎𝑗𝑖 = 1
𝑎𝑖𝑗
⁄ (1)
when assessments are consistent:
𝑎𝑗𝑘 =
𝑎𝑖𝑘
𝑎𝑖𝑗
⁄ (2)
where 𝑘 and 𝑗 are two alternatives being compared to 𝑖.
𝐴𝑥 = 𝜆𝑥 (3)
Every alternative is compared to every criterion, and every criterion at a given level is compared to
the higher-level criterion with which it is related. At last, every first-level criterion is contrasted with the
goal. By building matrices using the same methodology and scale as shown in Table 5, comparisons are
made. Until the priorities of the alternatives against the overall objective have been determined, the priorities
of the criteria are utilized as weights to compute the priorities of the alternatives in each criterion. Before
calculating the priorities for each matrix with n alternatives, comparisons are made given relation 1 and the
fact that the diagonal 𝑎𝑖𝑗 =1.
The following are the steps for an AHP model:
Step 1: Establish the hierarchy which contains three levels. Level 1 is the goal to achieve, level 2 is the
criteria, and level 3 is the alternatives which are presented in Figure 6.
Step 2: Create the matrix for pair-wise comparisons. As shown in Table 5, Saaty's nine-point scale serves as
the foundation for each matrix component. The decision-makers assessment of the relative weight
given to various factors is reflected in the comparison matrix.
Step 3: Construct the input matrix as presented in Table 6. The scales in the input matrix are given based on
the decision-makers.
Step 4: Create the normalized matrix as presented in Table 7. To normalize the matrix, we divide the scale
over the sum.
Step 5: Calculate the criteria weight by adding each row of the normalization matrix divided by the number
of alternatives as presented in Table 7.
Step 6: Ranking the alternatives based on the calculated weight as presented in Table 8.
To gather adequate data to assess whether the decision makers have made consistent decisions,
consistency must be assessed. The consistency ratio as 𝐶𝑅 = 𝐶𝐼/𝑅𝐼, where 𝑅𝐼 is random inconsistency and
𝐶𝐼 is the consistency index of the comparison matrix, which are both equal to 𝐶𝐼 = (𝑛𝑚𝑎𝑥 − n)/ (n − 1) and

2429
𝑅𝐼 = 1.987(𝑛 − 2)/𝑛. For total inconsistency to be considered acceptable, the consistency ratio needs to be
10% or less. If not, judgment data quality needs to be raised. The overall consistency in this study equals 0.04
as shown in the following section.
Figure 6. Flowchart of computational intelligent-based islanding detection methods
Table 5. Pair-wise comparison matrix
Intensity of relative importance Definition
1 Equally important
3 Moderately preferred
5 Strongly preferred
7 Very strongly preferred
9 Extremely preferred
2,4,6,8 Intermediate judgment between two adjacent judgments
Table 6. Input matrix
Initial
Passive Active Hybrid Remote Signal processing Computational intelligent
Criterion 1 Implementation Cost
Passive 1 3 5 9 5 7
Active 1/3 1 3 5 3 5
Hybrid 1/5 1/3 1 5 1 7
Remote 1/9 1/5 1/5 1 1/7 1/3
Signal Processing 1/5 1/3 1 7 1 7
Computational intelligent 1/7 1/5 1/7 3 1/7 1
Sum 1.987 5.066 10.342 30 10.285 27.333
Criterion 2 Non-detected Zone
Passive 1 1/5 1/5 1/7 1/9 1/9
Active 5 1 1/3 1/9 1/9 1/9
Hybrid 7 3 1 1/9 1/9 1/9
Remote 7 9 9 1 1/3 1/3
Signal Processing 9 9 9 3 1 1
Computational intelligent 9 9 9 3 1 1
Sum 38 31.2 28.53 7.37 2.67 2.67
Criterion 3 Power Quality
Passive 1 9 7 1 1 1
Active 1/9 1 1/3 1/9 1/9 1/9
Hybrid 1/7 3 1 1/7 1/7 1/7
Remote 1 9 7 1 1 1
Computational intelligent 1 9 7 1 1 1
Sum 4.254 40.000 29.333 4.254 4.254 4.254
Criterion 4 Response Time
Passive 1 5 9 7 1 3
Active 1/5 1 5 3 1/5 1/3
Hybrid 1/9 1/5 1 1/3 1/9 1/7
Remote 1/7 1/3 3 1 1/7 1/5
Computational intelligent 1/3 3 7 5 1/3 1
Sum 2.787 14.533 34.000 23.333 2.787 7.676

 ISSN: 2088-8708
2430
Table 7. Normalized matrix
Normalization
Criterion 1 Implementation cost
Passive Active Hybrid Remote Signal
processing
Computational
intelligent
Weight W%
Passive 0.503 0.592 0.483 0.300 0.486 0.256 0.436667 43.66
Active 0.167 0.197 0.290 0.166 0.291 0.182 0.2155 21.55
Hybrid 0.100 0.065 0.096 0.166 0.097 0.256 0.13 13
Remote 0.055 0.039 0.019 0.033 0.013 0.012 0.0285 2.85
Signal processing 0.100 0.065 0.096 0.233 0.097 0.256 0.141167 14.11
Computational intelligent 0.071 0.039 0.013 0.1 0.013 0.036 0.045333 4.53
Criterion 2 Non-detected zone
Passive 0.026 0.006 0.007 0.019 0.042 0.042 0.024 2.40
Active 0.132 0.032 0.012 0.015 0.042 0.042 0.046 4.60
Hybrid 0.184 0.096 0.035 0.015 0.042 0.042 0.069 6.90
Remote 0.184 0.288 0.315 0.136 0.125 0.125 0.196 19.6
Criterion 3 Power quality
Passive 0.235 0.225 0.239 0.235 0.235 0.235 0.234 23.4
Active 0.026 0.025 0.011 0.026 0.026 0.026 0.023 2.3
Hybrid 0.034 0.075 0.034 0.034 0.034 0.034 0.041 4.1
Remote 0.235 0.225 0.239 0.235 0.235 0.235 0.234 23.4
Criterion 4 Response time
Passive 0.359 0.344 0.265 0.300 0.359 0.391 0.336 33.6
Active 0.072 0.069 0.147 0.129 0.072 0.043 0.089 8.9
Hybrid 0.040 0.014 0.029 0.014 0.040 0.019 0.026 2.6
Remote 0.051 0.023 0.088 0.043 0.051 0.026 0.047 4.7
Table 8. Alternative ranking
Criterion 1 Cost
Weight (%) Ranking
Passive 43.66 1st
Active 21.55 2nd
Hybrid 13 4th
Remote 2.85 6th
Signal processing 14.11 3rd
Computational intelligent 4.53 5th
Criterion 2 Non-detected zone
Passive 2.40 5th
Active 4.60 4th
Hybrid 6.90 3rd
Remote 19.6 2nd
Signal processing 33.3 1st
Computational intelligent 33.3 1st
Criterion 3 Power quality
Passive 23.4 1st
Active 2.3 3rd
Hybrid 4.1 2nd
Remote 23.4 1st
Computational intelligent 23.4 1st
Criterion 4 Response time
Passive 33.6 1st
Active 8.9 3rd
Hybrid 2.6 5th
Remote 4.7 4th
Computational intelligent 16.6 2nd
5. SOLUTION WITH EXPERT CHOICE
The hierarchy is organized into three parts: the goal (Islanding detection method selection), criteria
(cost, non-detected zone, power quality, and response time), and alternative (passive method, active method,
hybrid method, remote method, signal processing-based method, and computational intelligent-based
method), as shown in Figure 7. After the model is constructed, the elements are evaluated using a pair-wise

2431
comparison. Comparing the alternatives considering the criteria stated in Figure 8(a) cost, 8(b) non-detected
zone, 8(c) power quality, and 8(d) response time; is how the pair-wise comparison is conducted with respect
to each criterion. The judgements are input using Saaty's 1–9 scale, where every alternative that is compared
to itself has a “1” value will show up in all alternatives of the major diagonal of any judgment matrix.
Figure 7. Hierarchy Structure
(a) (b)
(c) (d)
Figure 8. Pair-wise comparison with respect to (a) cost, (b) non-detected zone, (c) power quality,
and (d) response time
Priorities are computed when the pair-wise comparison is completed. Cost, non-detected zone,
power quality, and response time are all given similar weights in this study regarding the main objective.
However, the proprieties are determined based on the relative preference comparison for each criterion as
shown in Figure 9(a) cost, 9(b) non-detected zone, 9(c) power quality, and 9(d) response time.
The ideal mode, which uses normalization by dividing the score of each alternative solely by the
score of the best alternative under each criterion, is used to combine the local preferences across all criteria to
determine the global priority. As seen in Figure 10, the study's overall consistency is equivalent to 0.04. By
slightly altering the input data to track the impact on the outcomes, the sensitivity analysis can be applied to
decision-making. The findings are regarded as solid if the ranking stays the same. The interactive graphical
interface depicted in Figure 11 is the ideal method for carrying out the sensitivity analysis. The sensitivity
analysis shows that hybrid techniques have the lowest alternative and objective priorities (10% and 5%,
respectively) when all criteria are given equal weight. and the highest alternative and objective priority (55%
and 27%) are seen in signal processing-based techniques.

 ISSN: 2088-8708
2432
(a) (b)
(c)
(d)
Figure 9. Priorities derived from pair-wise comparison for (a) cost, (b) non-detected zone, (c) power quality,
and (d) time response
Figure 10. Global priorities using ideal mode
Figure 11. Performance sensitivity
6. RESULTS AND DISCUSSION
As illustrated in Figure 6, the criteria and alternatives are identified and then arranged in an AHP
hierarchy. Subsequently, a pair-wise comparison matrix (PCM) or decision matrix is created based on the
alternatives for each criterion. A value 𝐶𝑖𝑗, (𝑖 = 1, 2, . . . , 𝑛; 𝑗 = 1, 2, . . . , 𝑛) defined on Saaty's nine-point scale
as presented in Table 5 is used to compare objectives i and j. Moreover, 𝐶𝑗𝑖 =1/c if 𝐶𝑖𝑗 = 𝑐. Based on a nine-
point rating system, the value of 𝐶𝑖𝑗 is determined by how much an attribute is valued more highly for objective
i than for objective j. As shown in Table 6, the diagonal element of PCM, 𝐶𝑖𝑗(𝑖 = 𝑗) (𝐶11, 𝐶22, . . . , 𝐶𝑛), denotes

2433
self-importance and its value is always 1. While building a PCM, a review of the research literature already
in existence, discussions with experts in the field, and manufacturer reports can all be helpful resources for
determining values 𝐶𝑖𝑗, (𝑖 = 1, 2, . . . , 𝑛; 𝑗 = 1, 2, . . . , 𝑛). Based on the relative assigning value for the
alternatives, Table 6 illustrates the PCMs among the alternatives (objectives) regarding each criterion
(attribute). In the PCM with respect to the first criterion (implementation cost) as presented in Table 6, the
first row and first column 𝐶11 equals 1 (self-reference of passive methods), 𝐶12 = 3 = 1/𝐶21 (passive
methods are moderately preferred than active methods or active methods are moderately less preferred
than passive methods), 𝐶13 = 5 = 1/𝐶31 (passive methods are strongly preferred than hybrid methods),
𝐶14 = 9 = 1/𝐶41 (passive methods are extremely preferred than remote methods), and so on. The elements of
PCMs are assigned in this manner. As demonstrated in Figure 9, Expert Choice software was utilized to
calculate the weight factor for each of the alternatives for each criterion, complying with the AHP procedure,
such as the weight given to passive methods (44.5%, 2.1%, 23.4%, and 34%), active methods (22.5%, 3.8%,
2.3%, and 8.5%), hybrid methods (12.7%, 5.5%, 4%, and 2.5%), remote methods (2.7%, 20.7%, 23.4%, and
4.4%), signal processing-based methods (13.5%, 34%, 23.4%, and 34%), and computational intelligent-based
methods (4%, 34%, 23.4%, and 16.6%) based on the comparison criteria cost, non-detected zone, power
quality, and response time respectively. As illustrated in Figure 10, the overall weight is calculated for each
alternative such as the overall weight given to passive methods (24.7%), active methods (7.8%), hybrid
methods (5.6%), remote methods (14.5%), signal processing-based methods (26.6%), and computational
intelligent-based methods (20.8%) based on the comparison of all criteria together. Therefore, according to
the overall weight it can be observed that signal processing-based methods are the most appropriate methods
to be selected and the least one is hybrid methods. Additionally, the performance sensitivity shown in
Figure 11 demonstrates that, when all criteria are given equal weight, hybrid methods have the lowest
alternative and objective priorities (10% and 5%, respectively) and signal processing-based methods have the
highest alternative and objective priorities (55% and 27%). The performance sensitivity analysis is dynamic,
though, so each criterion's priority will adjust in accordance with whether the criteria are weighted unequal
according to the designer's assessment of their relative importance.
7. CONCLUSION
This paper offers a comprehensive analysis of several islanding detection methods. Traditional and
modern approaches are used to detect islands. Traditional techniques include local (passive, active, and
hybrid) and remote methods, whilst modern ones include signal processing and computationally intelligent
methods. Passive methods' key tenet is to monitor changes in network parameters like voltage or frequency at
PCC. Active techniques, which are based on perturbation injection, look at how injection affects system
parameters. Active and passive strategies are used in hybrid techniques. For remote approaches to function,
the utility side and the DGs side must exchange information and interact. Techniques based on signal
processing use feature extraction as their cornerstone. Pattern recognition and data training are the core of
computational intelligence methods. By contrasting the islanding detection methods based on a few factors,
including implementation cost, non-detected zone, power quality, and response time, the AHP-based
methodology is proved and proven in this work. Passive approaches are the best option to choose if the
implementation cost criterion is the sole factor considered. Signal processing-based approaches or
computationally intelligent-based methods are the most suitable options to choose if the non-detected zone
criterion is the only factor considered. Passive, remote, signal processing-based or computationally intelligent
solutions are the best ones to choose if the power quality requirement is the only factor considered. Passive or
signal processing-based solutions are the best options to select if the response time criterion is the only factor
considered. However, if these factors are considered, signal processing-based methods and passive methods
may be the ideal ones to use.
REFERENCES
[1] F. Barkat et al., “Review, analysis, and performance evaluation of the most common four active methods for islanding detection
in grid-connected photovoltaic systems,” Electric Power Systems Research, vol. 214, Jan. 2023, doi: 10.1016/j.epsr.2022.108909.
[2] M. Abu Sarhan, “An extensive review and analysis of islanding detection techniques in DG systems connected to power grids,”
Energies, vol. 16, no. 9, Apr. 2023, doi: 10.3390/en16093678.
[3] M. Y. Worku, M. A. Hassan, L. S. Maraaba, and M. A. Abido, “Islanding detection methods for microgrids: a comprehensive
review,” Mathematics, vol. 9, no. 24, Dec. 2021, doi: 10.3390/math9243174.
[4] R. S. Kunte and W. Gao, “Comparison and review of islanding detection techniques for distributed energy resources,” in 2008
40th North American Power Symposium, Calgary, AB, Canada, 2008, pp. 1-8, doi: 10.1109/NAPS.2008.5307381.
[5] C. Li, C. Cao, Y. Cao, Y. Kuang, L. Zeng, and B. Fang, “A review of islanding detection methods for microgrid,” Renewable and
Sustainable Energy Reviews, vol. 35, pp. 211–220, Jul. 2014, doi: 10.1016/j.rser.2014.04.026.
[6] K. N. E. Ku Ahmad, J. Selvaraj, and N. A. Rahim, “A review of the islanding detection methods in grid-connected PV inverters,”
Renewable and Sustainable Energy Reviews, vol. 21, pp. 756–766, May 2013, doi: 10.1016/j.rser.2013.01.018.

 ISSN: 2088-8708
2434
[7] A. Datta, A. Ray, D. Mukherjee, and H. Saha, “Selection of islanding detection methods based on multi-criteria decision analysis
for grid-connected photovoltaic system applications,” Sustainable Energy Technologies and Assessments, vol. 7, pp. 111–122,
Sep. 2014, doi: 10.1016/j.seta.2014.04.003.
[8] I. V. Banu et al., “Passive anti-islanding protection for three-phase grid-connected photovoltaic power systems,” International
Journal of Electrical Power and Energy Systems, vol. 148, Jun. 2023, doi: 10.1016/j.ijepes.2023.108946.
[9] X. Xie, C. Huang, and D. Li, “A new passive islanding detection approach considering the dynamic behavior of load in microgrid,”
International Journal of Electrical Power and Energy Systems, vol. 117, May 2020, doi: 10.1016/j.ijepes.2019.105619.
[10] A. Abyaz et al., “An effective passive islanding detection algorithm for distributed generations,” Energies, vol. 12, no. 16, Aug.
2019, doi: 10.3390/en12163160.
[11] M. S. Kim, R. Haider, G. J. Cho, C. H. Kim, C. Y. Won, and J. S. Chai, “Comprehensive review of islanding detection methods
for distributed generation systems,” Energies, vol. 12, no. 5, Mar. 2019, doi: 10.3390/en12050837.
[12] C. R. Reddy, B. S. Goud, B. N. Reddy, M. Pratyusha, C. V. Vijay Kumar, and R. Rekha, “Review of islanding detection
parameters in smart grid,” in 8th International Conference on Smart Grid, icSmartGrid 2020, Jun. 2020, pp. 78–89, doi:
10.1109/icSmartGrid49881.2020.9144923.
[13] N. Gupta, R. Dogra, R. Garg, and P. Kumar, “Review of islanding detection schemes for utility interactive solar photovoltaic
systems,” International Journal of Green Energy, vol. 19, no. 3, pp. 242–253, Jun. 2022, doi: 10.1080/15435075.2021.1941048.
[14] J. A. Cebollero, D. Cañete, S. Martín-Arroyo, M. García-Gracia, and H. Leite, “A survey of islanding detection methods for
microgrids and assessment of non-detection zones in comparison with grid codes,” Energies, vol. 15, no. 2, Jan. 2022, doi:
10.3390/en15020460.
[15] C. Rami Reddy et al., “Hybrid ROCOF relay for islanding detection,” Journal of Electrical Engineering and Technology, vol. 17,
no. 1, pp. 51–60, Aug. 2022, doi: 10.1007/s42835-021-00856-9.
[16] B. Hariprasad., P. Bharat Kumar, P. Sujatha, and G. Sreenivasan, “Island detection in inverter based distributed generation using a
hybrid method,” in 2022 Second International Conference on Artificial Intelligence and Smart Energy (ICAIS), Feb. 2022,
pp. 1702–1707, doi: 10.1109/ICAIS53314.2022.9742815.
[17] M. A. Ajith and R. M. Shereef, “Islanding detection of inverter based distributed generations using wavelet transform and KNN,”
in 2022 IEEE International Conference on Power Electronics, Smart Grid, and Renewable Energy (PESGRE), Trivandrum, India,
2022, pp. 1-6, doi: 10.1109/PESGRE52268.2022.9715940.
[18] M. Tajdinian, H. Khosravi, H. Samet, and Z. M. Ali, “Islanding detection scheme using potential energy function based criterio,”
Electric Power Systems Research, vol. 209, Aug. 2022, doi: 10.1016/j.epsr.2022.108047.
[19] S. K. G. Manikonda and D. N. Gaonkar, “Comprehensive review of IDMs in DG systems,” IET Smart Grid, vol. 2, no. 1,
pp. 11–24, Jan. 2019, doi: 10.1049/iet-stg.2018.0096.
[20] S. C. Paiva, R. L. de A. Ribeiro, D. K. Alves, F. B. Costa, and T. de O. A. Rocha, “A wavelet-based hybrid islanding detection
system applied for distributed generators interconnected to AC microgrids,” International Journal of Electrical Power and
Energy Systems, vol. 121, Oct. 2020, doi: 10.1016/j.ijepes.2020.106032.
[21] A. Taheri Kolli and N. Ghaffarzadeh, “A novel phase let-based approach for islanding detection in inverter-based distributed
generation systems,” Electric Power Systems Research, vol. 182, May 2020, doi: 10.1016/j.epsr.2020.106226.
[22] M. Shaik, A. G. Shaik, and S. K. Yadav, “Hilbert–Huang transform and decision tree based islanding and fault recognition in
renewable energy penetrated distribution system,” Sustainable Energy, Grids and Networks, vol. 30, Jun. 2022, doi:
10.1016/j.segan.2022.100606.
[23] N. A. Larik, M. F. Tahir, Z. M. S. Elbarbary, M. Z. Yousaf, and M. A. Khan, “A comprehensive literature review of conventional
and modern islanding detection methods,” Energy Strategy Reviews, vol. 44, Nov. 2022, doi: 10.1016/j.esr.2022.101007.
[24] I. Bhushan, S. Chandra, and A. Yadav, “A comprehensive portrayal of research drifts in islanding detection methods towards soft
computing,” in 2022 2nd International Conference on Power Electronics & IoT Applications in Renewable Energy and its
Control (PARC), Mathura, India, 2022, pp. 1-5, doi: 10.1109/PARC52418.2022.9726246.
[25] S. S. Mohapatra, M. K. Maharana, A. Pradhan, P. K. Panigrahi, and R. C. Prusty, “Detection and diagnosis of islanding using
artificial intelligence in distributed generation systems,” Sustainable Energy, Grids and Networks, vol. 29, Mar. 2022, doi:
10.1016/j.segan.2021.100576.
[26] F. Mumtaz, K. Imran, A. Abusorrah, and S. B. A. Bukhari, “An extensive overview of islanding detection strategies of active
distributed generations in sustainable microgrids,” Sustainability, vol. 15, no. 5, Mar. 2023, doi: 10.3390/su15054456.
[27] A. M. M. Alyasiri and S. Kurnaz, “Passive islanding detection in microgrids using artificial neural networks,” Applied
Nanoscience, vol. 13, no. 4, pp. 2885–2900, Apr. 2023, doi: 10.1007/s13204-021-02197-5.
[28] Y. A. Elshrief, D. H. Helmi, S. Abd-Elhaleem, B. A. Abozalam, and A. D. Asham, “Fast and accurate islanding detection
technique for microgrid connected to photovoltaic system,” Journal of Radiation Research and Applied Sciences, vol. 14, no. 1,
pp. 210–221, Dec. 2021, doi: 10.1080/16878507.2021.1923913.
[29] B. K. Panigrahi, A. Bhuyan, J. Shukla, P. K. Ray, and S. Pati, “A comprehensive review on intelligent islanding detection
techniques for renewable energy integrated power system,” International Journal of Energy Research, vol. 45, no. 10,
pp. 14085–14116, Apr. 2021, doi: 10.1002/er.6641.
BIOGRAPHIES OF AUTHORS
Mohammad Abu Sarhan received a B.S degree in electrical power engineering
in 2013 from Al-Balqa Applied University, Amman, Jordan, and MSc in control science and
control engineering in 2019 from China University of Geosciences, Wuhan, China. He is
currently doing his Ph.D. in electrical engineering at AGH University of Science and
Technology, Krakow, Poland. He worked as a technical engineer on many projects,
particularly those involving the marine industry. He also lectured at the Jordan Academy for
Maritime Studies, where he taught a variety of engineering and electrical power system
courses. His research interests include renewable energy resources, smart grids, and electrical
power systems control and optimization. He can be contacted at email: sarhan@agh.edu.pl.

2435
Andrzej Bien received PhD degree in electrical engineering from AGH
University of Science and Technology in 1988. Main professional interests are related to
measurements and measurement systems using fast signal processors, in particular
applications related to electricity and its quality. Measurements and related research include
broadly understood signal analysis, building new ways of measuring and new measures. In the
last years of his professional activity, he worked on the analysis of signals related to non-
stationary disturbances. Currently, he is the head of the Department of Power Electronics and
Automation of Energy Processing Systems at AGH University of Science and Technology. He
can be contacted at email: abien@agh.edu.pl.
Szymon Barczentewicz received MSc and PhD degree in electrical engineering
from AGH University of Science and Technology in 2012 and 2017 respectively. Since June
2018 he has been employed at AGH University as an assistant professor. During 2019–2021
he was a project manager of ERANET, RELflex “Renewable energy and load flexibility in
small industry” project. Since 2023 he was working as innovation project manager at
TAURON Dystrybucja (Polish DSO). His current research is focused on measurement
systems used in power grids, with particular emphasis on issues related to synchro phasor
measurements and the use of phasor technology in power quality problems. He can be
contacted at email: barczent@agh.edu.pl.

Use of analytical hierarchy process for selecting and prioritizing islanding detection methods in power grids

More Related Content

Similar to Use of analytical hierarchy process for selecting and prioritizing islanding detection methods in power grids (20)

More from IJECEIAES (20)

Recently uploaded (20)

Use of analytical hierarchy process for selecting and prioritizing islanding detection methods in power grids