Energy Conversion and Management 96 (2015) 228–241

Contents lists available at ScienceDirect

Energy Conversion and Management

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

A remote islanding detection and control strategy for photovoltaic-based

distributed generation systems
Gökay Bayrak ⇑
Nevsehir Haci Bektas Veli University, Faculty of Engineering and Architecture, Department of Electrical and Electronics, 50300 Nevsehir, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: This study presents a new remote islanding detection method and control system for photovoltaic (PV)
Received 19 December 2014 based Distributed Generation (DG) systems. The proposed method monitors and controls the grid, local
Accepted 1 March 2015 load and the output of the PV inverter in real time with the communication of circuit breakers. The pro-
posed remote control system detects the changes in the currents of the circuit breakers, frequency, and
the voltages by checking the defined threshold values at all electrical branches of the PV system. The pro-
Keywords: posed islanding detection algorithm was implemented by a low-cost FPGA board. The control system was
Distributed generation
also designed by considering a Very Large Scale Integration (VLSI) structure. The proposed method was
Remote islanding detection
Grid-tied PV system
verified by a developed prototype PV system constituted in the laboratory. The proposed control system
FPGA was checked in a resonance condition with an active power match, and the verified results indicated that
the developed system was also independent of the load and the inverter. Islanding detection time is
approximately 1.65 ms even in a worst-case operational scenario, and this is a significantly shorter
response time according to the existing standards. The proposed method presents a realistic solution
to islanding, is easy to implement, and is suitable for real system applications. The method also provides
a reliable islanding detection and presents a low-cost solution to the subject.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction The abnormal grid operating conditions negatively affect PV

systems [3]. Islanding is the most significant security problem in
PV technology has developed rapidly over recent years such a grid-tied PV system. Islanding operation can cause damage to
that solar energy has become the most important source of renew- both the PV system and the grid, and the grid voltage and the grid
ables [1]. Grid-tied PV systems have been coming into prominence frequency are not stable in an islanding situation [4]. These condi-
in Distributed Generation (DG) systems in parallel to this develop- tions change from the grid reference values such that the circuit
ment. There are some restrictions in connecting PV systems to uti- breaker (CB) connected between the grid and the point of common
lity grids such as the reliability of the grid, providing high power coupling (PCC) clears the fault during islanding mode. Meanwhile,
quality and safe interaction with the PV system. Islanding is maybe DG still supplies power to the local load if the CB cannot open the
the most important issue in this restriction for PV systems, provid- circuit [5]. Voltage shutdown, equipment failure, and short-circuit
ing a reliable connection and continual operation with the grid. conditions cause an unpredictable interruption of the grid, and the-
Islanding operation is defined in a DG that a situation while a se abnormal conditions cause islanding operation in a PV system
grid-tied PV system continues feeding the load, although discon- [6]. Fig. 2 shows an islanding condition in a PV-based DG system
nection of the electrical grid from the load [2]. Fig. 1 indicates a [7].
general schematic diagram of the grid-tied PV systems. Voltage There are two general types of islanding operations, intentional
and frequency of the system change from reference values in an islanding and unintentional islanding [8]. Intentional islanding cre-
islanding condition, so the grid disconnects from the grid-tied PV ates a power island when a disturbance occurs. An energy manage-
system without causing any damages to the system. Because of ment plan is essential in an intentional islanding which is
this, islanding detection methods have an important role in detect- established to supply the local load consistently by DG [9].
ing the islanding in grid-tied PV systems. Intentional islanding is a planned operation organized by the grid
operators such that it is not harmful to the power system [10].
However, unintentional islanding can damage the grid due to loss
⇑ Tel.: +90 3842281000; fax: +90 3842281123. of the synchronization of the electrical grid by causing a significant
E-mail address: gokaybayrak@gmail.com change in power system stability [11]. This situation causes the

http://dx.doi.org/10.1016/j.enconman.2015.03.004
0196-8904/Ó 2015 Elsevier Ltd. All rights reserved.
 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241 229

Fig. 1. A general schematic diagram of the grid-tied PV systems.

Pinv+jQinv P+j Q
TR
DC/DC PCC CB Electrical
Converter Inverter
Grid

Pload+jQload

Electrical
Load

Fig. 2. Islanding in a PV-based DG system.

voltage and frequency to be out of desired grid reference ranges Islanding is an important problem to solve in PV based DG sys-
which can cause damage to the electrical devices and equipment tems because it could cause serious problems with damaging
of the system in the island DG section [12]. Because people work- equipment in PV system and the danger of death for working peo-
ing on the grid-tied PV system cannot realize that DG continues to ple. There have been some standards to define the rules and
supply power to the island part of the system, this situation pre- restrictions for grid-tied PV systems in islanding mode of opera-
sents a danger. The definition of this problem in a grid-tied PV sys- tion. Mainly, IEEE-1547, IEEE 929, IEC-62116 and Japan standards
tem is an important criterion; islanding must be detected as soon are necessary for islanding.
as possible as indicated in IEEE standards [13]. This situation IEEE 929-2000 also defines frequency threshold values, voltage
always should be considered carefully by authorized DG workers threshold values and required opening time for circuit breaker (CB)
and companies. in PV based micro-grid systems. Table 1 shows these definitions
Consequently, a DG containing a PV system should be discon- and threshold values. Islanding is detected according to the nom-
nected from the local load by using a circuit breaker which is trig- inal voltage, and frequency values compared with specified values
gered by a generated control signal because of these restrictions in IEEE 929-2000. Table 1 also indicates the opening time of circuit
[14]. There have been many developments in islanding detection breaker in defining the conditions for the islanding mode of
methods and algorithms described in the literature [15]. operation.

2. Current islanding detection methods 2.1. Passive islanding detection methods

There are two main methods, referred to as local and remote Passive methods have a wide usage in PV based DG systems
detection methods [16,17]. Remote methods are related to measur- because of their smooth implementation and practical solution to
ing system parameters at a DG. In this study, a new remote method the subject, and these techniques are the primary detection meth-
between DG and the grid is used. ods for detecting islanding. In addition, passive systems do not

Table 1
IEEE 929–2000 threshold values for grid connection.

No Frequency Voltage CB opening time

1 fnom 0.5 Vnom 6 cycles
2 fnom 0.5 Vnom < V < 0.88 Vnom 2 s/120 cycles
3 fnom 0.88 Vnom 6 V 6 1.10 Vnom Normal operation
4 fnom 1.10 Vnom < V < 1.37 Vnom 2 s/120 cycles
5 fnom 1.37 Vnom 6 V 2 cycles
6 (fnom  0.7) 6 f 6 (fnom + 0.5) Hz Vnom Normal operation
7 f < (fnom  0.7) Hz Vnom 6 cycles
8 f > (fnom + 0.5) Hz Vnom 6 cycles
230 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241

produce any change in power quality. Passive methods have some researched an efficient and general detection method for the PV-
drawbacks like having a large non-detection zone (NDZ) and set- based DG systems. Distributed grid integration of the PV systems
ting threshold values with difficulty. These methods are unsuccess- is also an essential subject related to the islanding detection. A
ful for islanding detection, mainly the power balance of the load real-time dynamic model was proposed to determine the optimal
and the PV system. Active methods use passive methods in the size, location of a DG system into [22], and bidirectional power
background of the operation, so this situation has a negative flow control was also researched into [23]. The main purpose of
impact on the power quality of the system. Power quality changes the studies is to manage the DG in real-time by providing the pro-
significantly in active methods, mainly connecting more inverters tection of the DG. Energy management is a significant issue to
to the same DG. achieve a successful islanding detection, and controlling the active,
and reactive power of the DG is another approach to detecting
2.2. Active islanding detection methods islanding [3,24]. A real-time Labview-based islanding detection
system was also proposed in the last few years, and the remote
Active methods use a disturbing signal for detecting islanding control of the PV-based DG system was proposed in these studies
mode of operation [18]. The disturbance signal changes in very [26,27]. A local energy management of a hybrid PV system was also
short limitations when the grid connected; but when the grid dis- considered into [28], and it is obvious that the researchers have
connects from PV system, the interference signal has a significant been focused on real-time remote control techniques for the pro-
difference according to the standard running condition. Fig. 3 also tection of DG systems in recent years. As a consequence of this
shows the overall structure of active methods in islanding mode of approach, a newest study which is a smart device for islanding
operation. detection in distribution system operation [29] was proposed by
Di Fazioa et al.
2.3. Remote islanding detection methods There have been also some different approaches based on com-
putational intelligence techniques on this subject recently [32].
Remote methods have the best performance, according to the Islanding protection using wavelet analysis and neuro-fuzzy
passive and active methods because of these methods all circuit system in inverter based distributed generation was researched
breakers monitored by the control system. Installing sensors and into [16], and an adaptive ANN-controlled was proposed into
telecommunication devices to the system makes these methods [25] to reduce the NDZ consisting in passive islanding detection
having high system and operation cost [19,20]. The power quality methods. Variable impedance insertion [30] and droop control
does not change, and the system is stable in remote methods. In [31] methods also offer different solutions to the subject. There
addition, these techniques are used for the system cost is not nec- have been a few FPGA-based islanding detection studies in recent
essary, according to power quality of the system [21]. The overall years [32], but these studies present only particular, not general
structure of remote methods also is shown in Fig. 4. solutions to the problem.
From the literature review, almost none of detection methods
2.4. A short discussion of previous works have a complete solution to the problem. Passive methods have a
large NDZ and in matching the powers of inverter, load, and grid,
The new islanding detection methods have been proposed in they could fail to detect islanding. Active techniques have almost
the last few years by the scientists, and most of them have been no NDZ, but they have a power quality problem in the system.

Iinv Igrid=0

Idisturbance
TR
DC/DC
PCC CB Electrical
Inverter
Converter Grid
Iinv+Idistrurbance

Electrical
Load

Fig. 3. The general structure of active methods.

TR
DC/DC PCC CB Electrical
Inverter R T
Converter Grid

Electrical
Load

Fig. 4. The overall structure of remote methods.

 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241 231

Table 2
A comparison of current islanding detection methods.

Features Local methods Remote methods

Passive Active Hybrid PLL SCADA
Operation Measuring the PCC Adding a disturbing The combination of active and passive Grid impedance Using receiver and
principle parameters signal to the grid methods change in PCC transmitter sensors between
DG and grid
NDZ Large Small Small No No
Response Short Shorter than Passive Longer than Active Methods Fast Faster
time Methods
Detection Yes, when matching Yes only in high- Smaller than Active and Passive Methods No, except a few No, except a few specific
failure powers of inverter, load, quality factor specific conditions
and grid conditions
Effect on No Yes Yes, but smaller than Active Methods No No
dist. grid
System cost Minimum cost Average cost High cost Very high cost Extremely high cost
Multiple No Yes Yes No No
inverter
connect
Effect on No Decreases the power Decreases the power quality of the system, No No
power quality of the system but smaller effect, according to Active
quality Methods

experimental studies about this subject because of the mentioned

limitations. Also, there are not any practical, and generalized exist-
ing experimental methods can be widely used in the real system
applications.
Reducing the NDZ, achieving a stable power quality and reduc-
ing the system cost in a PV-based micro-grid are the main objec-
tives of islanding detection methods. While there have been
many islanding detection methods described in the literature,
there is only a real time application system in Japan [21]. Table 2
shows a comparison of current islanding detection methods.

3. The proposed remote islanding detection method

Fig. 5. Non detection zone (NDZ) for passive methods.
The passive islanding detection methods have some drawbacks
Hybrid methods are still only hypothetical. Remote methods are like having a large non-detection zone (NDZ) and setting threshold
possibly the best solution for islanding, but they have extremely values for difficulty and these methods are unsuccessful for island-
high system and operating costs. Most of the proposed methods ing detection, especially the power balance of the load and the PV
are still at the idealistic level only, and there have been a few system [13]. Fig. 5 shows the NDZ, and it is very small in active

Proposed Islanding
Detection&Control
Software

FPGA
Control Unit

EMC-1 CBs EMC-2

Control
Signals

Circuit Breakers
PV Array DC/DC CB-inv CB-grid Electrical
Inverter Grid
(1 kWp) Converter
CB-load

EMC-3

Electrical Current Sensor

Load Voltage Sensor

Fig. 6. The proposed FPGA-based remote islanding detection system.

232 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241

methods according to passive methods. There is a disturbing signal proposed PV system connected to the grid and its islanding detec-
which is used in active methods and also over/under voltage and tion block diagrams. The PV system generates the electrical power
frequency passive methods are used with active methods, so this and interacts with the grid. If grid voltage or frequency is out of the
condition changes the power quality of the system. Power quality limitations, circuit breakers are tripped, and the PV system discon-
significantly changes in active methods, especially connecting nected from the grid.
more inverters to the same DG [14]. The proposed method has been tested for when the grid voltage
Remote islanding detection methods have the best perfor- is 260 V (rms) that provides the ‘300 V < Un < 253 V’ condition.
mance, according to the passive and active methods. The proposed When the grid voltage increases up to 253 V, the control system
algorithm is independent of local load and inverter because it uses waits for five cycles to check the islanding condition, and then a
many parameters to detect islanding instead of measuring only the trip signal triggers the circuit breakers. The developed simulation
PCC voltage. Thus, the proposed method is not a passive method system also has been tested for different load conditions. Fig. 9
having NDZ changing by local load, and it is a communication shows the resistive load (200 W) condition and also shows an
based (coordination of the circuit breakers) method with no NDZ. inductive load condition (200 W and 400 VAR).
In this study, the proposed method has a new remote method
by using communication and control of the circuit breakers. The
power quality does not change (because these methods are inde- 4. Developed FPGA and VLSI-based detection and control
pendent of local load and, they do not fail in a resonance condi- method
tion), NDZ is zero, and the system is stable in communication
based methods. There is no NDZ problem with this method The proposed islanding detection method was implemented by
because it checks all of the system parameters in real time, and using an FPGA board. The proposed algorithm was designed by
it proposes a communication based method by using communica- considering Very Large Scale Integration (VLSI) structure. The
tion and control of the circuit breakers. The goal of the study also
chooses a remote detection method to prevent the NDZ and, local
load problems existing in passive and active detection methods. Start
The proposed algorithm that is shown in Fig. 7 checks the cur-
rents of the circuit breakers which are placed on three sides of the
PV system. If one of these currents goes to zero or they are under/
over defined threshold values, the circuit breakers have been Measure V and f at load, grid and PV sides
tripped by using a communication based control system. This cri- of the system
terion is the main approach to detecting islanding in the proposed
method. Thus, the proposed system does not measure only at the
PCC voltage, and it is not the main control criteria for the proposed
system. The proposed algorithm also controls the frequency, active Control Difference
power, and reactive power in addition to the currents, so it is reli- Values
able to detect islanding.
Under/Over No
Many methods focus on investigating the point of common cou-
pling (PCC) and are thus limited to, use the load and have an NDZ Threshold
problem. In this study, a new FPGA-based remote islanding detec- Values?
tion method is introduced for PV-based DG systems. A real-time
controller developed by the FPGA development unit manages the
PV-based DG system for islanding detection. In the proposed
method, grid, inverter, and load are monitored differently from Yes
other methods and loss of mains is checked by controlling the volt-
ages, currents and frequency of the whole system. If one of these
variables goes to zero or is outside of the selected threshold values,
digital signals created by the FPGA board control the circuit break-
ers. Thanks to this structure, breakers clear the fault within only in No
Count samples for
a few cycles and islanding is independent of the load. Fig. 6 indi-
checking the fault?
cates the proposed FPGA-based remote islanding detection system.
The proposed islanding detection algorithm detects the differ-
ences in threshold values of grid voltage and grid frequency.
Fig. 7 indicates this proposed detection algorithm. A low-cost
Altera DE0-nano board manages the algorithm. In this structure, Yes
voltages and frequency values of the grid, load and inverter output
are measured. The proposed method, continuously checks the
threshold values in real time. If detected values are over/under Trigger Circuit Breakers for Islanding
selected threshold values, control system counts the sample to
check the islanding condition. And then a trigger signal is generat-
ed by the FPGA board which opens circuit breakers connected to
the grid at the load and inverter output sides. Disconnect PV system from grid and load

3.1. The simulink model of the proposed method

A Simulink model of both the grid-tied PV system and the End

islanding detection system is implemented before constituting
the experimental test bench. Fig. 8 shows the overall structure of Fig. 7. The proposed islanding detection algorithm.
 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241 233

Fig. 8. A simulation model of the proposed islanding detection method.

Fig. 9. (a) The resistive load condition (P = 200 W), (b) the inductive load condition (P = 200 W and QL = 400 VAR).

islanding detection algorithm runs with a low-cost Altera DE0- board. These signals are evaluated in FPGA board thanks to devel-
Nano board. This frequency is also the maximum operating fre- oped software with Quartus. All the parameters can be determined
quency for the proposed algorithm. Fig. 10 also shows the general about the microgrid thanks to this structure. Fig. 11 shows the
schematic of the proposed control system in detail. developed Electronic Measurement Card (EMC). Fig. 12 also indi-
FPGA board controls the inverter, load and the grid sides of the cates the implemented voltage measurement circuit which is
microgrid system. Electronic Measurement Cards (EMCs) have shown in Fig. 11.
obtained the parameters just like voltage, current, and frequency.
0–10 V analog control signals are defined according to these para- 4.1. Developed real-time control and islanding detection software
meters for managing the system by giving them to the analog
inputs of the FPGA board. The current, voltage and frequency are It is started to be researching FPGA based software for real-time
obtained from microgrid with EMCs by using current and voltage monitoring of the developed microgrid. The developed software is
sensors. The measured signals are converted to 0–10 V analog sig- improved for monitoring each side of the DG system (PV array,
nals and are transmitted to the analog input channels of the FPGA inverter, load, and grid) with Quartus. The voltage, current,
234 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241

frequency, active power, reactive power, apparent power, the oscillator unit is defined as clk in Fig. 16 and Fig. 17, it can only
phase angle and the peak values are determined in the developed generate a clock signal with 50 MHz frequency.
software, and the DG system is monitored in real time. The devel- In the developed system, the frequency is 2 MHz; ADC provides
oped software evaluates the analog inputs transmitted from elec- conversion throughput rates of 125 Kbps. Eqs. (1)–(3) demonstrate
trical measurement circuits and according to these parameters, this example, in which; T (data transfer rate); sps (sample per sec-
monitors the electrical parameters of the PV-based DG system. ond); dt (detection time) and spdt (sample per detection time). By
Fig. 13 shows the general structure of the developed software. Eq. (3), the time is calculated to determine the voltage rise from
Active power generated by the microgrid system, requested 311 V to 357 V. There are 205 samples between 311 V and 357 V
active power and transferred power to the grid can be monitored when the clock frequency of ADC is 2 MHz. The instantaneous volt-
in real time in developed software. The line voltages, currents, age rise must be distinguished in this case so that the definite sam-
the phase angle of the load, and the system frequency are also ple number defines a threshold constant. Namely, if the peak
monitored by the software. Active and reactive power changes of voltage of the signal suddenly rises and the total counted sample
the microgrid, load, and the grid can also be monitored together number does not reach 205, the logical status of the control signal
in real time with developed software. Developed software moni- keeps it level. According to the user manual, chip select (CS) is held
tors all parameters of the DG system and detects islanding accord- low during the conversion process. Fig. 16 shows the output of
ing to the defined threshold values in Table 3. ADC, which is presented as input_signal is used to detect the fault
In the study, an uncomplicated, suitable VLSI-based structure condition by forcing a voltage to drift from up to down.
for voltage drifting detection is proposed. The system involves
operating at very high speed. Fig. 14 shows the VLSI-based control
Clock Freq 2  106
T¼ ¼ ¼ 125 Ksps ð1Þ
system. 16 16
Fig. 15 displays the modeling and monitoring schematic for the
FPGA-based PV system. The MUX block operates as the circuit
breaker due to logical ‘‘0’’ and logical ‘‘1’’. When the control signal
is rising, logical ‘‘1’’, the constant value is connected to the com-
parator block instead of the output of the PV_Signal_Generator.
Therefore, connection of faulty PV signal is avoided using the
MUX block.
The Altera DE0-Nano board consists of an ADC128S022 that is
an eight channel 12-bit analog to digital converter (ADC). The rate
of this ADC is adjustable from 50 Kbps to 200 Kbps. Thus, digital
clock input of the ADC (SCLK) must be set to 800 Hz and
3.2 MHz. In order to operate at these frequencies, the control block
of the ADC must consist of the clock divider unit. Because the main
Fig. 11. Developed Electronic Measurement Card (EMC).

Circuit Breakers
CB_inverter CB_load CB_grid

Trigger Signals
(0-5 V)
Defined Reference
Digital Outputs of FPGA
Threshold Values

Developed Islanding
Detection and Control FPGA Board
Software with Quartus
FPGA Board
VLSI Design
Analog to Digital Converter
(ADC128S022)
Converted Electrical Converted Electrical
Signals to 0-10 V I, V, P, f and cosφ Signals to 0-10 V
Developed Electronic Developed Electronic
Analog Inputs of FPGA
Measurement Card-1 Measurement Card-2

Converted Electrical
Iinv, Vinv Igrid, Vgrid
Signals to 0-10 V

Output of the PV Developed Electronic

Measurement Card-3 Electrical Grid
Inverter
Iload, Vload

Electrical Load

Fig. 10. The general schematic of the proposed control system.

 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241 235

Fig. 12. The implemented voltage measurement circuit.

Fig. 13. The general structure of the developed software.

Table 3 12
Selected grid threshold values in the experimental system. Small_Constant
Comparator
Selected threshold values Experimental IEEE 929–2000
Over voltage 253 V (rms) 253 V (rms)
Under voltage 195 V (rms) 195 V (rms)
Nominal voltage 220 V (rms) 220 V (rms)
Over frequency 50.2 Hz 59.3 Hz
Under frequency 49.8 Hz 60.5 Hz
Counter
&
125; 000 Control Control 12
sps ¼ ¼ 2500 ð2Þ Signal
50 Big_Constant
Comparator
311 ¼ 357  sinð2p50tÞ 12
ADC
 
1 311
20ms sin 357
 20 ms
dt ¼  ¼ 1:633 ms ð3Þ
4 360
Fig. 14. VLSI based control system.
spdt ¼ 125 Kbps  1:633 ms ¼ 204:125 ) must be 205 ð4Þ
236 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241

12
Small_Constant
Comparator

Counter
&
Control Control 12
Signal 12
Big_Constant
Comparator Constant
12 Inverter
Output
Signal
12 (Digital)

Fig. 15. Modeling of the overall system.

ADC

clk INPUT clk SCLK OUTPUT SCLK

VCC
DOUT INPUT DOUT CS OUTPUT CS
VCC
OUTPUT input_signal[11..0]
input_signal[11..0]

inst

Fig. 16. ADC control block.

Control_Block

clk INPUT OUTPUT SCLK

VCC clk SCLK
PIN_R8 DOUT INPUT DOUT CS OUTPUT CS PIN_B14
VCC
PIN_A9 OUTPUT Control_Signal PIN_A10
Control_Signal
PIN_B12
inst1

Fig. 17. FPGA based control block.

Fig. 17 illustrates the control block, consisting of ADC_block, In general, the developed grid-tied PV system consists of four
counter, comparator and clock divider. The system was performed main units; a PV array, a DC/DC converter, an inverter, and the
by using very high speed integrated circuit hardware description electrical load. Fig. 18 shows the developed PV-based DG system.
language (VHDL) and schematic algorithm. Under voltage failure, Islanding detection of the PV system was investigated by creating
Control_Signal is held; logical ‘0’ to activate circuit breaker. an islanding detection test bench in the laboratory after setting up
the grid-tied PV system.
The advanced test bench for islanding detection shown in
5. Experimental study and results Fig. 19 consists of EMCs, solid state relays as circuit breakers
(CBs) and an FPGA board for monitoring and managing the system.
This section introduces the grid-tied prototype PV system con- Analog outputs of the EMCs drive the analog inputs on the FPGA
stituted in the laboratory and the obtained results from the pro- board. Obtained parameters were determined by developing anti-
posed FPGA-based islanding detection system. islanding detection software using the FPGA board and its useful
tools. The proposed method is easily implemented and low cost
5.1. Implementation of the proposed system in the laboratory because of this structure.
Fig. 19 also shows the EMCs and circuit breakers. The developed
The selected FPGA board controls the output of the inverter, the FPGA-based application software monitors the detection para-
load and the grid parts of the PV system. Unlike other islanding meters of the PV system. There are three circuit breakers in the sys-
detection methods, the proposed method monitors all sides of tem. Voltages and frequency of the system are monitored
the PV system, thus it is independent of local load and NDZ which continuously by the PV system. These breakers are switched to
is the most important problem of islanding detection methods. control the PV system. Fig. 19 shows the developed electronic mea-
Electronic Measurement Cards (EMCs) obtain the parameters such surement cards indicated as A, B and C.
as voltage, current and frequency from the developed PV system. EMCs and circuit breakers can be easily implemented in
These parameters are converted to 0–10 V analog control signals designed system. The implemented controller system requires
suitable for a real-time control system for feeding them to the ana- minimal hardware. The proposed method is not inverter resident,
log inputs of the FPGA board. so it provides a robust and reliable solution for the subject.
 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241 237

Fig. 18. Photovoltaic power generating system.

Fig. 19. Developed islanding detection test system in the laboratory.

Fig. 20 also shows the FPGA-based control with an Altera sample number does not reach 205, the logical status of the control
DE0-Nano board. signal keeps it level.
Over/under-voltage threshold values of the PV system are also
determined by the developed FPGA-based software. When the grid
5.2. Experimental results
voltage is over the defined threshold value (253 V) or under a spe-
cified threshold value (195 V), the control system waits for check-
The proposed islanding detection algorithm detects the differ-
ing the islanding condition. Then, a trigger signal is sent to the
ences in threshold values of the grid voltage and the grid frequen-
circuit breakers by the FPGA board, and islanding is detected. If
cy. In this structure, voltages, current, and frequency values of the
threshold values are over/under selected threshold values, the con-
grid, load and inverter output are measured. The FPGA board con-
trol system waits for 205 samples (defines the over voltage thresh-
tinuously controls the threshold values in real time. Table 3 shows
old value) to check the islanding condition, and then a trigger
the selected grid threshold values for this experimental study.
signal is generated to open circuit breakers.

5.2.1. Normal operation of the developed microgrid 5.2.2.1. Under/over voltage operation of the developed microgrid. The
The developed software detects the variations in threshold val- developed islanding detection process continuously evaluates
ues of the grid voltage and the grid frequency. In this arrangement, feedback from the PV system and controls all of the systems.
first of all currents, voltages, active power and frequency values of When islanding occurs, circuit breakers are switched immediately,
the grid, load and inverter output are measured. Threshold values and the PV system is disconnected from the grid. In the application,
continuously are controlled in real-time. If the threshold values are the grid voltage increases to 253 V; the control system waits for
over/under selected threshold values, control system waits for two 205 samples to check the islanding condition, and then a trip signal
cycles to check the islanding situation. Then a trigger signal is gen- triggers the circuit breakers. Fig. 22 shows the over voltage
erated for opening circuit breakers where connected grid, load, and condition.
inverter output sides. Fig. 21 shows both the normal operation of The circuit breaker of the grid (CB_grid) and the circuit breaker
the PV system and the islanding situation for over voltage condi- of the inverter output (CB_inverter) are triggered by control signals
tion. The red line is the control signal of the circuit breaker, and from the FPGA board. These circuit breakers disconnect the PV sys-
the blue line is the voltage change of the system. The control signal tem from the grid. When the grid voltage decreases to 195 V, the
is zero in a normal operation, and it is triggired by the control sys- control system waits for 205 samples to check the islanding condi-
tem in an islanding condition indicated in Fig. 21. tion, and then a trip signal triggers the circuit breakers. Fig. 23
shows the under voltage condition.
5.2.2. Operating the microgrid in an islanding condition
By Eq. (3), the time is calculated to determine the voltage rise 5.2.2.2. Under/over-frequency operation of the developed
from 311 V to 357 V. There are 205 samples between 311 V and microgrid. Fig. 24 illustrates the clock signal, the control signal,
357 V when the clock frequency of ADC is 2 MHz. The instanta- the square signal (output of comparator) and the output of the
neous voltage rise must be distinguished in this case so that the inverter. The period decreases from 20 ms (millisecond) to
definite sample number defines a threshold constant. Namely, if 18.4 ms (over frequency condition) in the last period of the signal,
the peak voltage of the signal suddenly rises and the total counted thus activating the breaker and disconnecting the PV system from
238 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241

Fig. 20. FPGA board used in the experimental study.

Fig. 21. (a) The normal and (b) the islanding mode of operations.

Fig. 22. Islanding detection for over voltage condition.

Fig. 23. Islanding detection for under voltage condition.

 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241 239

the grid. These simulation results demonstrate that the proposed to detect islanding in this condition, and these values were selected
method presents a simple, suitable and non-complex solution to for indicating the success detecting of the proposed system.
islanding detection.
1 1
x0 ¼ pffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 50 Hz ð5Þ
LC 0; 16  2; 5 mF
5.2.2.3. Operating the microgrid in a resonance condition. In the
experimental study, electrical load specification was selected as These parameters define a resonance condition, which is the
R = 300 O, L = 1,6 mH, and C = 2,5 mF. This load defines a 50 Hz grid worst scenario for islanding detection. Active powers of the load
frequency which equals the resonance frequency, and it is the and the grid were selected equally to match the active powers.
worst scenario to detect islanding. Especially passive methods fail The active power mismatch is zero, and the system is in a

Fig. 24. The clock signal, the control signal, the square signal and the output of the inverter for over frequency condition.

Fig. 25. Clock signal, the control signal, and the output current signal of the inverter.

Fig. 26. Islanding detection time for developed FPGA-based control system.
240 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241

resonance condition. This condition is the worst operational sce- References

nario for islanding detection. Fig. 25 shows the clock signal, the
control signal, and the output current signal of the inverter in [1] Marzband M, Sumper A, Domínguez-García JL, Gumara-Ferret R. Experimental
validation of a real time energy management system for microgrids in islanded
the developed software with Quartus. The system detects islanding mode using a local day-ahead electricity market and MINLP. Energy Convers
in 1.65 ms and then waits for 205 samples for checking the fault. Manage 2013;76:314–22.
Fig. 26 shows this situation. Circuit breakers clear the fault within [2] Rezaei N, Kalantar M. Economic–environmental hierarchical frequency
management of a droop-controlled islanded microgrid. Energy Convers
two cycles under this load condition. Islanding detection time is Manage 2014;88:498–515.
only 1.65 ms in the proposed method. All detection and fault clear- [3] Laghari JA, Mokhlis H, Bakar AHA, Karimi M. A new islanding detection
ing times are considerable according to IEEE 929–2000. The PV sys- technique for multiple mini hydro based on rate of change of reactive power
and load connecting strategy. Energy Convers Manage 2013;76:215–24.
tem is not connected to the grid, and the proposed method is
[4] Tsang KM, Chan WL. Rapid islanding detection using multi-level inverter for
independent of the load and the inverter. grid-interactive PV system. Energy Convers Manage 2014;77:278–86.
As described by Eqs. (3) and (4), Fig. 26 illustrates that the [5] Hassaine L, Olías E, Quintero J, Barrado A. Power control for grid-connected
detection time of the FPGA control board and the time chart of applications based on the phase shifting of the inverter output voltage with
respect to the grid voltage. Electr Power Energy Syst 2014;57:250–60.
the control signal, which is generated by a VLSI-based modeled cir- [6] Du Y, Lu DDC, James G, Cornforth DJ. Modelling and analysis of current
cuit. The detection time (1.65 ms), which is between 53.51 ms and harmonic distortion from grid connected PVinverters under different operating
51.86 ms, is clearly shown in Fig. 26. conditions. Sol Energy 2013;94:182–94.
[7] Trujillo CL, Velasco D, Figueres E, Garcerá G. Analysis of active islanding
detection methods for grid-connected micro inverters for renewable energy
processing. Appl Energy 2010;87:3591–605.
6. Conclusions [8] Eltawil ME, Zhao Z. Grid-connected photovoltaic power systems: technical and
potential problems-a review. Renew Sustain Energy Rev 2010;14:112–29.
In this study, a new remote islanding detection method was [9] Spertino F, Leo PD, Cocina V. Which are the constraints to the photovoltaic
grid-parity in the main European markets? Sol Energy 2014;105:390–400.
developed for PV-based DG systems. The developed method is a [10] Kamel RM, Chaouachi A, Nagasaka K. Enhancement of micro-grid performance
remote method that controls the load, grid and the output of the during islanding mode using storage batteries and new fuzzy logic pitch angle
inverter with the communication of circuit breakers. A real-time controller. Energy Convers Manage 2011;52:2204–16.
[11] Bayrak G, Cebeci M. A novel anti islanding detection method for grid
controller with an FPGA detects the islanding condition.
connected fuel cell power generation systems. Int J Hydrogen Energy
Remote islanding detection methods have the best perfor- 2014;39:8872–80.
mance, according to the passive and active methods. The proposed [12] Bayrak G, Cebeci M. A communication based islanding detection method for
algorithm is independent of local load and inverter because it uses photovoltaic distributed generation systems. Int J Photoenergy
2014;2014:1–17.
many parameters to detect islanding instead of measuring only the [13] Bayrak G, Cebeci M. Grid connected fuel cell and PV hybrid power generating
PCC voltage. Thus, the proposed method is not a passive method system design with Matlab Simulink. Int J Hydrogen Energy 2014;39:8803–12.
having NDZ changing by local load, and it is a communication [14] IEEE Application Guide for IEEE Std 1547. IEEE Standard for Interconnecting
Distributed Resources with Electric Power Systems; 2009.p. 1–207.
based (coordination of the circuit breakers) method with no NDZ. [15] Llaria A, Curea O, Jiménez J, Camblong H. Survey on micro grids: unplanned
In the study, the proposed method has a new remote method by islanding and related inverter control techniques. Renew Energy
using communication and control of the circuit breakers. The pow- 2011;36:2052–61.
[16] Shayeghi H, Sobhani B. Zero NDZ assessment for anti-islanding protection
er quality does not change (because these methods are indepen- using wavelet analysis and neuro-fuzzy system in inverter based distributed
dent of local load and, they do not fail in a resonance condition), generation. Energy Convers Manage 2014;79:616–25.
NDZ is zero, and the system is stable. There is no NDZ problem [17] Hossain E, Kabalci E, Bayindir R, Perez R. Microgrid testbeds around the world:
state of art. Energy Convers Manage 2014;86:132–53.
with this method because it checks all of the system parameters [18] Indu Rani B, Srikanth M, Saravana Ilango G, Nagamani C. An active islanding
in real time, and it proposes a communication based method by detection technique for current controlled inverter. Renew Energy
using communication and control of the circuit breakers. The goal 2013;51:189–96.
[19] Gabbar HA, Abdelsalam AA. Microgrid energy management in grid-connected
of the study also chooses a remote detection method to prevent the
and islanding modes based on SVC. Energy Convers Manage 2014;86:
NDZ and, local load problems existing in passive and active detec- 964–72.
tion methods. [20] Patrao I, Figueres E, González-Espı́n F, Garcerá G. Transformerless topologies
Islanding is detected by using an FPGA-based control and detec- for grid-connected single-phase photovoltaic inverters. Renew Sustain Energy
Rev 2011;15:3423–31.
tion system when the frequency and voltages at three sides of the [21] Velasco D, Trujillo CL, Garcera G, Figueres E. Review of anti-islanding
PV system exceed the allowable threshold values. The proposed techniques in distributed generators. Renew Sustain Energy Rev
method, continuously monitors the load, inverter and grid sides 2010;14:1608–14.
[22] Masoud A, Nima A. A new evolutionary solution method for dynamic
of the DG system, unlike conventional methods. The proposed expansion planning of DG-integrated primary distribution networks. Energy
algorithm also controls the frequency, active power, and reactive Convers Manage 2014;82:61–70.
power in addition to the currents, so it is reliable to detect island- [23] Manuela S, Chao WB, Fabrice L, Antoine J. DC microgrid power flow
optimization by multi-layer supervision control. Design and experimental
ing. The islanding detection algorithm uses a low-cost Altera validation. Energy Convers Manage 2014;82:1–10.
DE0-Nano board. This frequency is important for the immediate [24] Linfeng Z, Nicolae G, Lawrence HV. Energy management in a microgrid with
detection of grid failure and the generation of fault signals faster distributed energy resources. Energy Convers Manage 2014;78:297–305.
[25] Muyeen SM, Hasanien Hany M, Ahmed AD. Transient stability enhancement of
than conventional methods. wind farms connected to a multi-machine power system by using an adaptive
The proposed method was verified by the development of a pro- ANN-controlled SMES. Energy Convers Manage 2014;78:412–20.
totype PV system constituted in the laboratory. The obtained [26] Bayrak G, Cebeci M. A novel Labview based anti-islanding detection method
for grid-connected PV systems. International Conference on Renewable Energy
experimental results show that the proposed method is reliable,
Research and Applications ICRERA, Madrid, Spain October 2013;20–
low-cost and independent of the load and the inverter. The island- 23:549–54.
ing detection time is 1.65 ms even in a worst operational scenario [27] Bayrak G, Cebeci M. Monitoring a grid connected photovoltaic power
as indicated by experimental results. The islanding detection time generation system with Labview. International Conference on Renewable
Energy Research and Applications ICRERA, Madrid, Spain October 2013;20–
has a significantly shorter response time according to existing 23:562–7.
standards. Islanding detection time and the grid opening time is [28] Choudar A, Boukhetala D, Barkat S, Brucker JM. A local energy management of
two cycles even in a resonance condition. The proposed method a hybrid PV-storage based distributed generation for microgrids. Energy
Convers Manage 2015;90:21–33.
is efficient and presents a realistic solution to islanding. This [29] Di Fazioa AR, Fuscoa G, Russoa M, Valeria S, Noceb C, Amuraca G. A smart
method is also easy to implement and is suitable for real system device for islanding detection in distribution system operation. Electr Power
applications in PV-based DG systems. Syst Res 2015;120:87–95.
 G. Bayrak / Energy Conversion and Management 96 (2015) 228–241 241

[30] Papadimitriou CN, Kleftakis VA, Hatziargyriou ND. A novel islanding detection response and droop controlled distributed generation framework. Energy
method for microgrids based on variable impedance insertion. Electr Power Convers Manage 2015;92:287–301.
Syst Res 2015;121:58–66. [32] Laghari JA, Mokhlis H, Karimi M, Bakar AHA, Mohamad H. Computational
[31] Rezaei N, Kalantar M. Smart microgrid hierarchical frequency control ancillary intelligence based techniques for islanding detection of distributed generation
service provision based on virtual inertia concept: an integrated demand in distribution network: a review. Energy Convers Manage 2014;88:139–52.