Alexandria Engineering Journal (2017) xxx, xxx–xxx

H O S T E D BY
Alexandria University

Alexandria Engineering Journal

www.elsevier.com/locate/aej
www.sciencedirect.com

ORIGINAL ARTICLE

Proposed Sandia frequency shift for anti-islanding

detection method based on artificial immune system
A.Y. Hatata *, El-H. Abd-Raboh, Bishoy E. Sedhom

Faculty of Engineering, Mansoura University, Egypt

Received 4 April 2015; revised 19 October 2016; accepted 15 December 2016

KEYWORDS
Anti-islanding detection;
Sandia frequency shift (SFS);
Non-detection zone (NDZ);
Total harmonic distortion
(THD);
Artificial immune system
(AIS);
Clonal selection algorithm
Abstract Sandia frequency shift (SFS) is one of the active anti-islanding detection methods that
depend on frequency drift to detect an islanding condition for inverter-based distributed generation.
The non-detection zone (NDZ) of the SFS method depends to a great extent on its parameters.
Improper adjusting of these parameters may result in failure of the method. This paper presents
a proposed artificial immune system (AIS)-based technique to obtain optimal parameters of SFS
anti-islanding detection method. The immune system is highly distributed, highly adaptive, and
self-organizing in nature, maintains a memory of past encounters, and has the ability to continually
learn about new encounters. The proposed method generates less total harmonic distortion (THD)
than the conventional SFS, which results in faster island detection and better non-detection zone.
The performance of the proposed method is derived analytically and simulated using Matlab/Simu-
link. Two case studies are used to verify the proposed method. The first case includes a photovoltaic
(PV) connected to grid and the second includes a wind turbine connected to grid. The deduced opti-
mized parameter setting helps to achieve the ‘‘non-islanding inverter” as well as least potential
adverse impact on power quality.
Ó 2016 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V. This is an
open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction from the remainder of the power system yet continues to be

energized by one or more DGs. An important requirement
Distributed generations (DG) are small-scale generation units to interconnect a DG to a power distributed system is the
that can be installed near to consumers with the ability of capability of the DG unit to detect islanding with the minimum
interacting with the grid importing or exporting energy. One time possible. The continued energizing of the load can lead to
of the major problems associated with such generators is the damage of equipment or injury to maintenance personnel
unwanted islanding phenomenon. Islanding occurs when a working within the islanded section without knowing the sys-
portion of the distribution system becomes electrically isolated tem is still alive. Most DG units are designed in such a way
that they are disconnecting from the grid when over/under
voltage or frequency occurs on the network. In the case that
* Corresponding author.
the grid is disconnected while the load and the source are
E-mail address: a_hatata@yahoo.com (A.Y. Hatata).
matched, the DG units will thus continue to power the line,
Peer review under responsibility of Faculty of Engineering, Alexandria
thereby leading to the formation of an island.
University.
http://dx.doi.org/10.1016/j.aej.2016.12.020
1110-0168 Ó 2016 Faculty of Engineering, Alexandria University. Production and hosting by Elsevier B.V.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
2 A.Y. Hatata et al.
DGs must detect islanding and immediately stop feeding
the utility lines with power. This is known as anti-islanding.
Anti-islanding methods assist the DG units to detect islanding
or force the islanded section out of the normal operational
specifications of the grid. This is achieved by attempting to
perturb either the voltage or the frequency of the network.
In the presence of the grid, these perturbations will have no
effect on the voltage or frequency. If the grid is disconnected,
variations in voltage or frequency can occur. These variations
can be detected by the over/under voltage or frequency protec-
tion system, and the DG is disconnected or shutdown.
Anti-islanding methods generally can be classified into four
major groups, which include passive methods, active methods,
hybrid methods and communication base methods [1]. Passive
methods monitor selected parameters, such as voltage, fre-
quency or their characteristics, and they switch off the inverter
if one of these parameters deviates outside specified boundaries
or conditions [2]. The boundary limits of these parameters
define the non-detection zone (NDZ). The passive methods
include Over/Under Voltage Protection (OVP/UVP), Over/
Under Frequency Protection (OFP/UFP), voltage phase jump
detection, and detection of voltage and current harmonics
methods [2–4]. These methods are conceptually simple and
easy to implement and do not introduce any change to the
power quality of the system. However, they have a number
of weaknesses including the inability to detect islanding
because they have a large NDZ. They tend to false trip due
to disturbances on the grid which may weaken grid stability
and security.
In order to reduce the NDZ, particularly in cases where the
local loads are close in capacity to the DG systems, active
detection methods have been proposed. Active methods per-
turb the connected circuit and then monitor the response to
determine whether islanding has occurred [2,5–8]. Active meth-
ods include Impedance measurement method [9–11], Slip
Mode Frequency Shift (SMS) [12], Active Frequency Drift
(AFD) [13–19], Sandia Frequency Shift (SFS) [13,20–22], San-
dia Voltage Shift (SVS) [5,8], Reactive Power Variation (RPV)
method [23], and Mains monitoring units with allocated all-
pole switching devices connected in series (MSD) [24]. Active
methods attempt to create a power mismatch between the load
and the DG when they are closely matched. It is possible that
some of the active methods can cancel out the mismatch in an
attempt to create one. It should also be noted that the positive
feedback in some active methods could lead to power-quality
degradation [25]. The injection signals can also induce some
voltage waveform distortion.
Among frequency drift islanding detection methods, SFS is
considered as one of the most effective methods in detecting
islanding conditions. The method can be used to improve the
NDZ and THD by using a positive feedback gain, but still
affect the power quality of the system. This paper presents a
proposed technique to modify SFS anti-islanding method
using Artificial Immune Systems (AIS). The method optimizes
the parameters of the SFS method to reduce both the NDZ
and the THD of the current waveforms.
AIS are computational paradigms that belong to the com-
putational intelligence family and are inspired by the biological
immune system. During the past decade, they have attracted a
lot of interest from researchers aiming to develop immune-
based models and techniques to solve complex computational
or engineering problems. Authors in [26,27] propose a
Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
technique that is based on AIS for fault detection. Ref. [28]
applies AIS for disturbance detection. On the other hand,
AIS was used for fault detection in the stator and rotor circuits
of induction machines in [29].
The impacts of the load parameters on the performance of
the SFS method have been discussed in many papers. However
these papers are concerned with reducing NDZ only and
ignore the impact on power quality deterioration. In [19] the
performances of SFS method and its major factors affecting
the phase-frequency characteristic of the islanding system were
analyzed. Ref. [20] proposed a technique to estimate the
parameter of SFS using fuzzy logic. Ref. [21] proposed SFS
methods that prevent islanding without any deregulation in
power quality. Authors in [22] presented a method to improve
the dynamic NDZ in load parameter space. In [30] a SFS based
PLL algorithm was proposed for islanding detection. Authors
in [31,32] discussed the impact of load frequency and DG on
the parameters of SFS method.
There are many artificial intelligence techniques used for
islanding detection [33–36]. Ref. [33] presented a passive
method for detecting islanding of DG inspired by the informa-
tion processing properties of natural immune system. Ref. [34]
proposed an islanding detection algorithm using combination
of an optimal Artificial Neural Network (ANN) based on Par-
ticle Swarm Optimization (PSO) with a simple active method.
Two-dimensional particle swarm optimization (2D-PSO)
method for optimizing the weighting in extension theory to
detect the islanding in the presence of photovoltaic (PV) power
generation systems was presented in Ref. [35]. Authors in [36]
presented an islanding detection method that is based on
neuro-fuzzy system which is trained by using four different
heuristic algorithms and finally among all of them, PSO with
the best results is elected.
This paper presents a proposed AIS-based technique to
obtain optimal parameters of SFS anti-islanding detection
method. This method improves the performance of the
conventional SFS method by reducing THD, NDZ, and
results in faster island detection. The proposed method is
derived analytically and simulated using Matlab/Simulink.
The verification of this method can be achieved by a two case
studies; the first case includes a photovoltaic (PV) connected
to grid and the second includes a wind turbine connected to
grid.
The remainder of this paper is organized as follows: Sec-
tion 2 describes the SFS anti-islanding method, Section 3
describes the system modeling and simulation, Section 4 intro-
duces the AIS method and the clonal selection algorithm, Sec-
tion 5 presents the modified SFS based AIS method and the
AIS algorithm for solving this problem, and Section 6 presents
the results and discussion. Finally, Section 7 concludes the
paper.
2. Sandia frequency shift anti-islanding method
Sandia Frequency Shift (SFS) is one of the active islanding
detection methods that rely on frequency drift to detect an
islanding condition. These methods depend on injecting a dis-
torted current waveform into the original reference current of
the inverter; therefore, in the case of islanding operation the
frequency drift up or down depending on the sign of the so
called chopping factor ‘‘Cf”. A positive feedback is utilized
Anti-islanding detection method
to prevent islanding and to decrease NDZ value. The proce-
dures of applying the SFS method can be summarized as fol-
lows [24]:
1. Inject a current harmonic signal with a limited duration
into the Point of Common Coupling (PCC) so as to comply
with the maximum THDi allowed by interconnection
standards.
2. The injected current signal distorts the inverter current by
presenting a 0 A segment for drift up operation as shown
in Fig. 1.
3. The desirable effect of the 0 A segment, is that the funda-
mental component of the inverter current leads the voltage
by a small angle hAFD , which is frequency dependent and it
creates a positive feedback.
4. When the grid is disconnected, the frequency of the voltage
of the PCC tends to drift, reaching values higher until the
frequency is out of the OFP/UFP trip window (Range)
and the inverter is disconnected.
5. A positive feedback is utilized to prevent islanding.
The NDZ of the SFS highly depends on its design parame-
ters. The design parameters include both chopping factor Cf
and feedback gain factor K. If these parameters are not prop-
erly tuned, it may result in failure of the method or deteriora-
tion of the system power quality through injection of high
amount of harmonics. However, SFS may fail to detect island-
ing considering a fact that the deviations of voltage and fre-
quency are small due to the power balance between DG
sources and local loads.
The chopping factor is a function of the error in the line fre-
quency and may be computed as follows:
Cf ¼ Cfo þ Kðfa  fline Þ
where Cfo is the chopping factor when there is no frequency
error, K is an accelerating gain that does not change direction,
fa is the line frequency measured at PCC, and fline, is the nom-
inal line frequency.
Figure 1 AFD current reference of PV inverter.
Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
3
3. System modeling and simulation
To verify the proposed method and prove its effectiveness, it is
applied to a test system shown in Fig. 2. The system consists of
a DG, connected to the grid PCC through a current controlled
VSI inverter to supply a local ac load. The load can also be
supplied by the grid through a power transformer.
The local load is represented by a parallel RLC with vari-
able resonant frequency and quality factor. The utility breaker
connects the grid to the PCC in case of normal operation. For
islanding condition the breaker is opened at a prescribed time.
The test system is modeled and simulated in MATLAB/Simu-
link environment as shown in Fig. 3. The simulation parame-
ters are presented in Table 1. The simulation is implemented in
the following steps:
 The frequency is firstly measured. If the frequency exceeds
the IEEE. Std. 929–2000 limits, then the OFP/UFP will
generate a fault signal to shut down the inverter.
 In case that load and DG output are closely mismatched,
then the frequency of the network is perturbed to create a
power mismatch between load and DG.
 A variations in frequency can be detected by OFP/UFP sys-
tem and consequently the DG is disconnected or shutdown.
 Since this perturbation signal may induce current and volt-
age waveform distortion, the load current is checked to
ensure that the THD limits are not violated.
4. Artificial immune system (AIS)
The immune system of vertebrates including human is com-
posed of cells, molecules and organs in the body which protect
ð1Þ the body against infectious diseases caused by foreign patho-
gens such as viruses and bacteria [37]. To perform these func-
tions, the immune system has to be able to distinguish between
the body’s own cells such as self-cells and foreign pathogens
such as non-self-cells or antigens. After distinguishing between
self and non-self-cells, the immune system has to perform an
immune response in order to eliminate non-self-cell or antigen
[38–40].
Clonal selection principle of AIS describes how the immune
cells capture a foreign antigen. This algorithm is simple and
efficient for achieving optimum solution [41]. The algorithm
steps involving clonal selection are as follows:
Figure 2 Single line diagram of the test system.
4 A.Y. Hatata et al.

Figure 3 Test system modeled in Matlab Simulink.

3. The immune cells that identify the antigens undergo cellular

Table 1 Simulation parameters.
reproduction. After reproducing the somatic cells reproduc-
Grid voltage (rms) 400 V tion asexual form, i.e. there are no crossovers of the genetic
Grid frequency 50 Hz material during the cell mitosis. The genetic new cells are
LC filter L = 18 mH copies (clones) of their parents as shown in antibody C in
C = 30 lF Fig. 4.
4. A portion of cloned cells endures a mutation mechanism
DG system output power Pinv = 3 kW
which is called somatic hyper mutation.
Qinv = 0 Var
5. The affinity of any cell with each other is measured by the
Parallel RLC load R = 120 O difference in similarity between them. The affinity is deter-
L = 153 mH mined by measuring the difference between two cells. The
C = 67 lF
response of antibodies found in memory has an average
affinity higher than the early first response. This phe-
1. Start with a number of antibodies (immune cells) which nomenon refers to the maturation of the immune response.
represent initial population size. After the mutation process the fitness as well as the affinity
2. When an antigen or pathogen invades the organism, a num- of the antibody to each other is changed.
ber of antibodies which identify these pathogens survive. In 6. For each iteration, among the efficient immune cells, some
Fig. 4 only the antibody C is able to identify the pathogen 3 become effecter cells (Plasma Cell) and the other cells
as its shape fits to a portion of the antigen. So fitness of become memory cells. The effecter cells secrete antibodies
antibody C is greater than others. and memory cells have a long life span so as to act faster

Figure 4 Basic of clonal selection algorithm.

Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
Anti-islanding detection method 5

and effective in the future when the organism is exposed to
the same antigen.
7. This process is continued till the termination condition is
satisfied; else the steps from 2 to 7 are repeated.
Clonal selection theory explains how the immune system
fights against an antigen. It establishes the idea that only those
cells that recognize the antigen, are selected to proliferate. The
selected cells are subjected to an affinity maturation process
which improves their affinity to the selected antigens [37,41].
5. Proposed SFS–based AIS method
The two parameters K and Cfo have different effects on the
NDZ of SFS schemes. Initial chopping factor Cfo changes
the location of NDZ on load space, but it doesn’t change the
size of NDZ. Increasing Cfo causes adverse impact on power
quality, as it perturbs the network with a distorted current.
The THD of the output current is directly proportional to
Cfo [13]. According to the impact on power quality; smaller
values are preferred in Cfo setting.
Positive feedback gain K has a positive impact on NDZ
consideration. Increasing K will decrease NDZ and push the
NDZ toward load space with higher Cfo , but it may also pro-
duce negative impact on power quality of the distribution sys-
tem. Therefore smaller values of K parameter are preferred for
efficient detection of islanding. To minimize both NDZ and
THD to acceptable limit, it is necessary to optimize the values
of Cfo and K. In this paper AIS optimization technique is used
to obtain optimal values of Cfo and K.
It is clear that selecting the parameters of islanding detec-
tion system is not a straightforward task. This paper presents
a method for optimizing the selection of these parameters so
that the best possible fast islanding detection is achieved at
the lowest cost of adverse effects on power quality. The pro-
posed method formulates SFS parameter selection as an opti-
mization problem with main objective of minimizing NDZ
while satisfying THD limit as a constraint. The optimization
method used is AIS. The following section gives detailed rep-
resentation of the optimization problem.
5.1. Problem formulation
In order to modify the SFS parameters for better islanding
detection method, AIS is used as an optimization technique
to optimize these parameters. AIS mimics these biological
principles of clone generation, proliferation and maturation.
The fundamental of AIS is inspired by the theoretical immune
system and the observed immune functions, principles, and
models [38,39]. Today, the AIS techniques are used to solve
complex problems in many areas that include engineering,
science, computing, and other research areas.
(A) Objective function
The objective function can be minimized by controlling the
design variables of the SFS. The objective function can be
expressed mathematically as
MinðDFo Þ ¼ fomax  fomin
where fomax is the maximum value of load resonant frequency
that will result in islanding operation, and fomin is the minimum
value of load resonant frequency that will result in islanding
operation, and they can be expressed as [42]:
f  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fomax ¼ max  tan hSFS ðfmax Þ þ tan2 hSFS ðfmax Þ þ 4Q2f ð3Þ
2Qf
fmin  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
fomin ¼  tan hSFS ðfmin Þ þ tan2 hSFS ðfmin Þ þ 4Q2f ð4Þ
2Qf
hSFS ðfÞ is the phase angle of the inverter and can be computed
as follows:
pCf ðfÞ
hSFS ðfÞ ¼ ð5Þ
2
where Qf is the load quality factor and is defined as follows:
pffiffiffiffiffiffiffiffiffiffi
Qf ¼ R C=L ð6Þ
where R, L and C are the load resistance, inductance and
capacitance respectively.
(B) Problem Constraints
Total harmonic distortion limits and accelerating gain limit
are the main constraints for which the objective function in (2)
is subjected:
i. Total harmonic distortion limits
Due to its negative impact on the power quality of the dis-
tribution system, THD must be limited to a preset value.
According to IEEE Std. 929–2000 limits THD must be lesser
than 5% [43].
ii. Accelerating gain limit
The performance of the SFS is affected by the positive gain
coefficient K. Better performance is achieved with larger values
of K. Meanwhile increasing K will increase the current distor-
tion and results in bad effect on system power quality. So, the
value of K factor must not violate the power quality of the sys-
tem. This constraint can be expressed mathematically as [19]
Qf
K> ð7Þ
12:5p
The problem is solved using AIS to obtain the optimal val-
ues of both Cfo and K that minimize the NDZ and THD to
acceptable values in a short time.
5.2. Proposed algorithm procedure
The following steps summarize the procedure of applying the
proposed SFS–based AIS method to solve the islanding detec-
tion problem using SFS technique.
i. Antibody representation
In applying AIS to solve a problem, the solution of the
problem is considered as a population of antibodies. In this
study both Cfo and K are considered as antibody and each
ð2Þ
parameter is represented by a gene of antibody.

Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
6 A.Y. Hatata et al.

ii. Antigen representation
The antigen can be referred to the case of islanding occur-
ring and as a consequence, the voltage, frequency, and the
THD measured from the PCC are deviated from its nominal
values. To represent this case using Matlab Simulink, the cir-
cuit breaker that connects the utility grid with the DG is
opened at a significant time forming islanding. After this, the
system values are disturbed and it is possible to detect island-
ing as fast as possible.

Start

Read all system data for the test system

(Figure 4), parameters limits

Set the initial antibodies

population for Cfo and K

For each antibody in population, simulate the

model and calculate the objective function
NDZ, F, THD, and IDT

Calculate the affinity from

equation (8)

No Yes
Reach population Select antibodies whose
size? affinity is high for cloning

Clone the selected

antibodies in set C
For each mutated antibody, simulate
the model; calculate the objective
function NDZ, F, THD, and IDT. Mutate the selected
antibodies in set T
Then calculate the affinity.

Reselect the best antibody that has the

highest affinity and remove the worst

No Yes No
Reach the Conditions are
population size? satisfied?

Yes

Print the data for K,

End Cfo, THD, NDZ, and
IDT

Figure 5 Flowchart of the modified SFS – based AIS method.

Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
Anti-islanding detection method 7

iii. Initialization vi. Cloning

The first step of applying the AIS method is the generation The selected n individuals of the population that are
of initial population. The initial population is generated cloned (reproduced) by the clonal process, give rise to a
between a pre-defined minimum and maximum values using temporary population of clones, C. The higher the affinity,
a random function. If the generated solutions do not fall into the larger the number of clones generated for each of the
the feasible range, they are ignored and the generation process n selected antibodies. Antibodies in the population must be
is repeated until the required number of solutions is generated. cloned according to their affinity and generate a temporary
The size of the initial population is determined by making a population.
trade-off between quality of a solution and computation time
to yield that solution. vii. Mutation

iv. Affinity function The cloned antibodies are mutated to create a population
T. During mutation, it assigns a lower mutation rate for higher
To introduce affinity (fitness) function, the variables should affinity antibodies than low affinity antibodies. The idea is that
be put in the model and then the difference of the estimated val- the antibodies close to a local optimum need only be fine-
ues from the actual data for each antibody is calculated and in tuned, whereas antibodies far from an optimum should move
each generation the individual with minimum difference must larger steps toward an optimum or other regions of the affinity
be returned. Individual parameters are selected randomly and landscape.
the affinity is calculated according to Euclidean distance [44]:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi viii. Reselection and diversification
Xm ðxi yi Þ2
affinityðx; yÞ ¼ i¼1
=m ð8Þ
This process reselects the improved antibodies from the
population T to update it. Finally, the diversity introduction
v. Selection
process replaces the low affinity antibodies with new ones.
A flowchart of the proposed SFS – based AIS method is
In selection step, n antibodies with highest affinity are
shown in Fig. 5.
selected to generate a new population.

(a) System Respoce (b) Voltage at PCC

100 400
System Responce (Seconds)

80
200
Vpcc (Volts)

60
0
40

-200
20
X= 0.23722
Y= 0

0 -400
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Time (Seconds) Time (Seconds)

(c) Load Current (d) System Frequency

4 150

2 100
Frequency (Hz)
Iload (Ampere)

0 50

-2 0

-4 -50
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Time (Seconds) Time (Seconds)

Figure 6 SFS anti-islanding detection method output waveform for PV (a) system response, (b) voltage at PCC, (c) load current and (d)
system frequency.

Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
8 A.Y. Hatata et al.

(a) System Respoce (b) Voltage at PCC

System Responce (Seconds) 100 500

80

Vpcc (Volts)
60
0
40

20
X= 0.2324
Y= 0
0 -500
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Time (Seconds) Time (Seconds)

(c) Load Current (d) System Frequency

8 150

6
100
4

Frequency (Hz)
Iload (Ampere)

2 50

0 0
-2
-50
-4

-6 -100
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

Time (Seconds) Time (Seconds)

Figure 7 SFS anti-islanding detection method output waveform for wind (a) system response, (b) voltage at PCC, (c) load current and
(d) system frequency.

0.8

0.6
NDZ

0.4

0.2

0
0.2
0.18
0.16 0.2
0.14 0.18
0.12 0.14 0.16
0.1 0.12
0.08 0.1
0.06 0.06 0.08
0.04 0.04
K 0.02 0.02
0 0 Cfo

Figure 8 Population and cloning of the best affinity values.

6. Results and discussion system. These two cases are used to verify the effectiveness
of the proposed method.
In this paper the grid-connected current controlled inverter fed
by a DC voltage source is disconnected after 0.228 s. Both the (A) Case 1: PV connected to grid
conventional SFS and the proposed modified SFS – based AIS
methods are applied to the test system and a comparison For the case of PV connected to grid, the DG in Fig. 2 is
between the results is presented below. represented by PV system. The simulation results show that:
the feedback gain K equals to 0.096 and Cfo equals to 0.026
with a NDZ of 0.97359 and THD of 4.43%.
6.1. Conventional SFS method Fig. 6a illustrates the system response for detecting the
islanding condition. The islanding is detected at 0.2372 s. so
The conventional SFS method is applied to the test system that the DG is disconnected 9.2 ms after the grid disconnection
shown in Fig. 2. The SFS is applied to two cases using Matlab (which occurs at 0.228 s). Fig. 6b and c shows the voltage and
Simulink. The first case is PV system and the second is wind load current waveforms during the simulation period, whereas

Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
Anti-islanding detection method 9

Fig. 6d shows the frequency change after the grid that the DG is disconnected 1.28 ms after the grid disconnec-
disconnection. tion. Fig. 9b and c shows the voltage and load current wave-
forms during the simulation period, whereas Fig. 9d shows
(B) Case 2: Wind connected to grid the frequency change after the grid disconnection.

For the case of wind connected to grid, the DG in Fig. 2 is (B) Case 2: Wind connected to grid
represented by a wind system. The simulation results show
that: the feedback gain K equals to 0.108 and Cfo equals to According to this method the optimal values of the feed-
0.027 with a NDZ of 0.97025 Hz and THD of 4.048%. back gain K and Cfo are 0.13 and 0.026 respectively, with a
The utility grid is disconnected at 0.228 s. After the discon- NDZ of 0.9641 Hz and THD of 3.35%.
nection of the grid, the system detects the islanding by apply- From Fig. 10a the system responds for detecting the island
ing SFS method and sends a trip signal to disconnect the DG condition at 0.22924 s so that the DG is disconnected after
at 0.2324 s that is after 4.4 ms from the grid disconnection as 1.24 ms from the grid disconnection. (The grid is disconnected
shown in Fig. 7a. When the system responds to detect island, at 0.228 s.) When the grid is disconnected, the voltage at PCC
the voltage at PCC and load current begins to decay until and the load current start to decay after sending a trip signal to
reaching steady state values and disconnects the DG as shown disconnect the DG, as explained in Fig. 10(b) and (c). The fre-
in Fig. 7(b) and (c). Fig. 7d shows the frequency change after quency change after the grid is disconnected is shown in
the grid is disconnected. Fig. 10d.

6.2. Proposed SFS – based AIS method

6.3. Result analysis
In this case the modified SFS-based AIS method is applied to
the studied system. Fig. 8 shows population and cloning of the A comparison between the conventional SFS islanding detec-
best affinity values. Set the parameter K as a population from tion method and the proposed SFS method is made to validate
0.01 to 1, and set the parameter Cfo as a population from 0.01 the effectiveness of the proposed method. The proposed
to 0.05. method reduced the NDZ, THD and the islanding detection
time (IDT). Table 2 illustrates a comparison between the con-
(A) Case 1: PV connected to grid ventional and the proposed SFS method for NDZ, THD and
IDT. It is observed that the modified method can reduce the
The optimal values of the feedback gain K and Cfo are 0.108 NDZ for the two case studies. There is an improvement in
and 0.036 respectively, with a NDZ of 0.970164 and THD of the THD value after applying the modified SFS method. For
3.91%. the PV system connected to grid, there is an improvement
Fig. 9a illustrates the system response for detecting the in the THD by 11.7% and for the wind system connected to
islanding condition. The islanding is detected at 0.22928 s so grid, the THD is improved by 17.2%. The IDT for the first

(a) System Respoce (b) Voltage at PCC

System Responce (Seconds)

100 400

80
200
Vpcc (Volts)

60
0
40

-200
20
X= 0.2259
Y= 0
0 -400
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (Seconds) Time (Seconds)

(c) Load Current (d) System Frequency

4 150

2 100
Frequency (Hz)
Iload (Ampere)

0 50

-2 0

-4
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Time (Seconds) Time (Seconds)

Figure 9 SFS based AIS anti-islanding detection method output waveform for PV (a) system response, (b) voltage at PCC, (c) load
current and (d) system frequency.

Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
10 A.Y. Hatata et al.

(a) System Respoce (b) Voltage at PCC

System Responce (Seconds)
100 500

80

Vpcc (Volts)
60
0
40

20
X= 0.22924
Y= 0
0 -500
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Time (Seconds) Time (Seconds)

(c) Load Current (d) System Frequency

8 150

6
100

Frequency (Hz)
Iload (Ampere)

2 50

0 0
-2
-50
-4

-6 -100
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4
Time (Seconds) Time (Seconds)

Figure 10 SFS based AIS anti-islanding detection method output waveform for wind (a) system response, (b) voltage at PCC, (c) load
current and (d) system frequency.

References
Table 2 Comparison between conventional and modified
SFS. [1] W. Teoh, C. Tan, An overview of islanding detection methods in
photovoltaic systems, World Acad. Sci., Eng. Technol. 5 (2011).
Conventional SFS Modified SFS based
[2] A. Hatata, El-H. Abd-Raboh, Bishoy E. Sedhom, A review of
method AIS
anti-islanding protection methods for renewable distributed
PV connected to THD = 4.43% THD = 3.91% generation systems, J. Electr. Eng. (JEE) 16 (2016).
grid IDT = 9.2 ms IDT = 1.28 ms [3] A. Timbus, A. Oudalov, Carl N.M. Ho, Islanding detection in
NDZ = 0.97359 Hz NDZ = 0.970164 Hz smart grids, Energy Convers. Congress Expos. (ECCE) (2010)
3631–3637, IEEE.
Wind connected to THD = 4.048% THD = 3.35%
[4] M. Hanif, M. Basu, K. Gaughan, A discussion of anti-islanding
grid IDT = 4.4 ms IDT = 1.24 ms
protection schemes incorporated in a inverter based DG, in:
NDZ = 0.97025 Hz NDZ = 0.9641 Hz
International Conference on Environment and Electrical
Engineering (EEEIC), 10th International, 8–11 May 2011.
[5] D. Velasco, C. Trujillo, G. Garcerá, E. Figueres, Review of anti-
islanding techniques in distributed generators, Renew. Sustain.
case is improved by 86.08%, and for the second case study it is Energy Rev. 14 (2010) 1608–1614.
improved by 71.8%. [6] M. Robitaille, K. Agbossou, M. Doumbia, R. Simard, Islanding
detection method for a hybrid renewable energy system, Int. J.
7. Conclusion Renew. Energy Res., IJRER 1 (1) (2011) 41–53.
[7] M. Laour, A. Mahrane, F. Akel, M. Chikh, D. Bendib,
Implementation of active anti-islanding methods protection
This paper presents a proposed technique to modify SFS anti- devices for grid connected photovoltaic systems, in: 6th
islanding method using AIS. SFS is one of an active methods International Conference on Computer and Electrical
for islanding detection and has its advantages in addition to Engineering (ICCEE 2013), October 12–13, 2013, Paris, France.
that it has a feedback gain that can be used for rapidly island- [8] B. Álvarez, O. Rivera, Survey of distributed generation islanding
ing detection. AIS is an optimization technique used to opti- detection methods, Latin America Trans., IEEE (Revista IEEE
mize the SFS parameters to achieve more efficient islanding America Latina) vol. 8 (5) (2010) 565–570.
detection. The modified SFS method is applied to two cases [9] P. Mahat, B. Bak-Jensen, Review of islanding detection methods
studies using MATLAB/SIMULINK. A comparison between for distributed generation, in: Third International Conference
on Electric Utility Deregulation and Restructuring and Power
the proposed method and conventional SFS was presented and
Technologies, 2008, DRPT 2008, 6–9 April, 2008, pp. 2743–
the results show that the proposed method generates less THD
2748.
than the conventional SFS, which results in faster islanding [10] P. Mahat, Z. Chen, B. Bak-jensen, Review on islanding
detection and better non-detection zone. The deduced opti- operation of distribution system with distributed generation,
mized parameter setting helps to achieve the ‘‘non-islanding Power Energy Soc. Gener. Meet. IEEE (2011) 1–8.
inverter” as well as least potential adverse impact on power [11] M. Roop, J. Ginn, J. Stevens, W. Bower, S. Gonzales,
quality. Simulation and experimental study of detection anti-islanding

Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
Anti-islanding detection method
method in the single inverter case, in: Conference Record of the
2006 IEEE 4th World Conference on Photovoltaic Energy
Conversion, volume 2, pp. 2379–2382, May 2006.
[12] F. Liu, Y. Kang, Y. Zhang, S. Duan, X. Lin, Improved SMS
islanding detection method for grid-connected converters,
Renew. Power Gener. IET 4 (1) (2010) 36–42.
[13] F. Liu, Y. Kang, S. Duan, Analysis and optimization of active
frequency drift islanding detection method, in: Applied Power
Electronics Conference, APEC 2007-Twenty Second Annual
IEEE, pp. 1379–1384, February 25, 2007–March 1, 2007.
[14] Z. Ye, A. Kolwalkar, Y. Zhang, P. Du, R. Walling, Evaluation
of anti-islanding schemes based on non-detection zone concept,
IEEE Trans. Power Electron. 19 (5) (September 2004) 1171–
1176.
[15] L. Phani Raghav, T. Sandhya, An active frequency drift method
for an islanding detection of grid connected micro turbine
generation system, Int. J. Innovat. Res. Sci. Eng. Technol. –
(ICETS’14) 3 (1) (2014).
[16] A. Yafaoui, B. Wu, S. Kouro, Improved active frequency drift
anti-islanding detection method for grid connected photovoltaic
systems, IEEE Trans. Power Electron. 27 (5) (2012).
[17] L. Lopes, Y. Zhang, Islanding detection assessment of multi-
inverter systems with active frequency drifting methods, IEEE
Trans. Power Deliv. 23 (1) (2008).
[18] A. Yafaoui, B. Wu, S. Kouro, Improved active frequency drift
anti-islanding method with lower total harmonic distortion, in:
IECON 2010–36th Annual Conference on IEEE Industrial
Electronics Society, pp. 3216–3221, 2010, IEEE.
[19] W. Huang, T. Zheng, F. Yuan, Z. Wang, S. Xu, Analysis of the
NDZ formulation theory of active frequency shift islanding
detection method for grid connected PV system, IEEE Power
Energy Eng. Conf. (APPEEC), PES Asia-Pacific (2013) 1–5.
[20] H. Vahedi, M. Karrari, Adaptive fuzzy sandia frequency shift
method for islanding protection of inverter based distributed
generation, IEEE Trans. Power Deliv. 28 (1) (January 2013).
[21] P. Du, Z. Ye, E. Aponte, J. Keith, L. Fan, Positive-feedback-
based active anti-islanding schemes for inverter-based
distributed generators: basic principle, design guideline and
performance analysis, IEEE Trans. Power Electron. 25 (12)
(Dec. 2010) 2941–2948.
[22] X. Wang, W. Freitas, W. Xu, Dynamic non-detection zones of
positive feedback anti-islanding methods for inverter-based
distributed generators, IEEE Trans. Power Deliv. 26 (2) (2011)
1145–1155.
[23] Y. Jung, J. So, G. Yu, J. Cho, Modeling analysis of active
islanding detection methods for photovoltaic power
conditioning systems, IEEE Electr. Comput. Eng. Canad.
Conf. (CCECE) 2 (2004).
[24] W. Bower, M. Ropp, Evaluation of islanding detection methods
for photovoltaic utility-interactive power systems, in:
International Energy Agency, Report IEA PVPST5-09, 2002.
[25] P. Ray, N. Kishor, S. Mohanty, Islanding and power quality
disturbance detection in grid-connected hybrid power system
using wavelet and S transform, IEEE Trans. Smart Grid 3 (3)
(2012).
[26] N. Bayar, S. Darmoul, S.-H. Gabouj, H. Pierreval, Fault
detection diagnosis and recovery using artificial immune systems
a review, Eng. Appl. Artif. Intell. 46 (Part A) (2015) 43–57.
[27] C.A. Laurentys, R.M. Palhares, W.M. Caminhas, A novel
artificial immune system for fault behavior detection, Exp. Syst.
Appl. 38 (2011) 6957–6966.
Please cite this article in press as: A.Y. Hatata et al., Proposed Sandia frequency shift for anti-islanding detection method based on artificial immune system, Alex-
andria Eng. J. (2017), http://dx.doi.org/10.1016/j.aej.2016.12.020
11
[28] Fernando P.A. Lima, Anna D.P. Lotufo, Carlos R. Minussi,
Disturbance detection for optimal database storage in electrical
distribution systems using artificial immune systems with
negative selection, Electr. Power Syst. Res. 109 (2014) 54–62.
[29] Z. Chilengue, J.A. Dente, P.J. Costa Branco, An artificial
immune system approach for fault detection in the stator and
rotor circuits of induction machines, Electr. Power Syst. Res. 81
(2011) 158–169.
[30] J. Lee, B. Min, T. Kim, D. Yoo, J. Yoo, Active frequency with a
positive feedback anti-islanding method based on a robust PLL
algorithm for grid-connected PV PCS, J. Power Electron. 11 (3)
(2011) 360–368.
[31] H. Zeineldin, M. Salama, Impact of load frequency dependence
on the NDZ and performance of the SFS islanding detection
method, IEEE Trans. Industr. Electron. 58 (1) (2011) 139–146.
[32] X. Wang, W. Freitas, W. Xu, V. Dinavahi, Impact of DG
interface controls on the Sandia frequency shift anti islanding
method, IEEE Trans. Energy Convers. 22 (3) (2007) 792–794.
[33] G. Yin, A distributed generation islanding detection method
based on artificial immune system, in: Transmission and
Distribution Conference and Exhibition: Asia and Pacific,
2005 IEEE/PES, pp. 1–4, 2005.
[34] H. Samet, F. Hashemi, T. Ghanbari, Minimum non detection
zone for islanding detection using an optimal artificial neural
network algorithm based on PSO, Renew. Sustain. Energy Rev.
52 (2015) 1–18.
[35] U.-D. Lai, K.-H. Chao, M.-H. Wang, Using a particle swarm
method to optimize the weighting in extension theory for the
detection of islanding in photovoltaic systems, Comput. Math.
Appl. 64 (2012) 1441–1449.
[36] H. Shayeghi, B. Sobhani, E. Shahryari, A. Akbarimajd, Optimal
neuro-fuzzy based islanding detection method for distributed
generation, Neurocomputing 177 (2016) 478–488.
[37] L. de Castro, F. Von Zuben, Artificial Immune System: Part 1 –
Basic Theory and Applications”, Technical report, December
1999.
[38] D. Dasgupta, N. Okine, Immunity-based systems: a survey, in:
Proceedings of the IEEE International Conference on Systems,
Man and Cybernetics, October 1997.
[39] L. de Castro, J. Timmis, Artificial immune systems: a novel
paradigm to pattern recognItion, in: Artificial Neural Networks
in Pattern Recognition, SOCO-2002, University of Paisley, UK,
2002.
[40] L.N. de Castro, J. Timmis, An artificial immune network for
multimodal function optimization, in: Proceedings of IEEE
Congress on Evolutionary Computation (CEC’02), vol. 1, pp.
699–674, Hawaii, May, 2002.
[41] L. de Castro, F. Von Zuben, Learning and optimization using
the clonal selection principle, IEEE Trans. Evol. Comput.,
Special Issue Artif. Immune Syst. 6 (3) (2001) 239–251.
[42] Y. Zhang, H. Sun, L. Lopes, An islanding detection enhancer
for a system with multiple photovoltaic inverters, in: 2nd
Canadian Solar Buildings Conference Calgary, June 10–14,
2007.
[43] IEEE Std. 929–2000, IEEE Recommended Practice for Utility
Interface of Photovoltaic (PV) Systems, Institute of Electrical
and Electronics Engineers Inc, New York, NY.
[44] X. Wang, J. Cheng, Z. Yin, M. Guo, A new approach of
obtaining reservoir operation rules: artificial immune
recognition system, Exp. Syst. Appl. 38 (2011) 11701–11707.