You are on page 1of 6

Power Engineering Conference (UPEC), 2014 49th International Universities

Islanding Prevention Scheme for Grid-Connected


Photovoltaic Systems in Matlab/ Simulink
Ioan Viorel Banu, Marcel Istrate
Gheorghe Asachi Technical University of Iasi, Romania
ibanu86@yahoo.com, mistrate@ee.tuiasi.ro

the islanding must be quickly detected and the PV System


Abstract-This paper describes the islanding phenomenon on must be immediately disconnected from the power grid [5].
Grid-Connected Photovoltaic (PV) Arrays for developing new After disconnection, in certain scenarios the embedded
techniques for islanding detection. A novel islanding prevention generator can maintain supply to local loads or even to a
scheme is presented. It consists in monitoring the dc-link voltage
of Voltage Source Converter (VSC). This Anti-Islanding (AI) small part of a network depending on its capacity and
method is validated by directly comparing the islanding network topology [1]. In order to reduce the cost and simplify
detection time in various scenarios with the detection time the installation of PV systems, the inverter should include
offered by the rate of change of frequency (ROCOF) technique. effective and reliable anti-islanding procedures [6]. As the PV
Index Terms--Anti-islanding; Grid-connected photovoltaic; arrays become more popular, reliable islanding detection is
Islanding detection. becoming increasingly important. New solutions are needed
as the limitations of the existing methods may eventually
I. INTRODUCTION pose a barrier to further integration of DER [1]. Anti-
Islanding (also known as loss of grid or Loss-of-Mains islanding protection involves the use of relays or controls to
LoM [1]) represents a condition in that a portion of the prevent the existence of an unintentional island [3]. The
power system that contains both load and distributed energy inverter must be able to detect an island, and take
resources (DER) remains energized while isolated from the appropriated measure in order to protect persons and
remainder of the power system [2]. Also, according to [3] a equipment [6]. IEEE standard 1547-2003 [4] requires that the
island represents a condition where a portion of an area DER cease to energize an unintentional island on the area
electric power system (EPS) is energized solely by one or EPS. If the local generator protection is unable to detect an
more local DER through the associated point of common islanding condition, a transfer trip should be installed from
coupling (PCC) while that portion of the area EPS is the upstream protective device to the generator [3].
electrically separated from the rest of the area EPS [4]. Rate of change of frequency (ROCOF) has become the
Islanding appears when some part of the utility grid loses most commonly used method of islanding protection. This is
connection with the rest of the system [1]. The islanding can a passive method based on the local monitoring of the
be intentional and unintentional [4]. An intentional island generator voltage waveform [1]. Standard protective relays
represents a planned island whereas an unintentional island is like under/over frequency protection (UFP/OFP) shield
an unplanned island [3]. An unintentional island can form consumers equipment and also serve as passive AI detection
when a portion of the area EPS is accidentally separated from methods resident in the PV inverter [7, 8]. These protection
the remainder of the area EPS and the DER continues to methods are low cost options for islanding detection. They
supply power [3]. may not be actually physical relays but procedures
Development of sensitive and reliable AI protection is very implemented in the inverter's software. All grid-connected PV
important to encourage medium scale integration of DER into inverters are required to have OFP/UFP methods to
the electrical network and avoid unnecessary tripping of DER disconnect the power if the frequency of the voltage at the
[1]. The PV inverter design will be influenced by the power PCC between the customer and the utility grid is outside the
grid requirements, including the anti-islanding (AI) prescribed limits [7]. The active AI methods are not widely
requirement which is considered the most technically utilized at this moment, due to power quality concerns [1].
challenging [5]. One of the most challenging aspects of The international standard IEEE 1547 [4] defines the over-
protection of DER is the sensitivity of islanding detection, frequency (OF) and under-frequency (UF) thresholds for
which makes the system stable to external faults and disconnection of the DER and, in particular PV systems from
frequency excursions [1]. Islanding for PV Power Systems the electrical network. For the European interconnected
appears when the power grid is tripped and the inverter does power grid (50Hz), the limits are 49.3 and 50.5Hz. According
not disconnect quickly enough and continues to operate with to Romanian power grid requirements for PV Systems the UF
the local load [5]. To perform islanding protection in some and OF trip thresholds are 49.5 and 50.5Hz [9].
specific circumstances (impossibility of forming a stable Reference [10] presents a study on various islanding
island) it may be sufficient to use a combination of scenarios of a PV System. The simulation results with
under/over frequency/voltage protection [1]. In most cases, detection time of different AI methods are depicted in Fig. 1.

978-1-4799-6557-1/14/$31.00 2014 IEEE


Date of Conference: 2-5 Sept. 2014
Date Added to IEEE Xplore: 23 October 2014
Electronic ISBN: 978-1-4799-6557-1
CD-ROM ISBN: 978-1-4799-6556-4
Conference Location: Cluj-Napoca, Romania
Fig. 2. Detailed Simulink model of the 100kW Grid-Connected PV System.

The 100kW PV Array is connected to a 20kV power grid


(Fig. 3) with the main frequency of 50Hz through a 100kVA
260V/20kV three-phase coupling transformer. The 5kHz dc-
dc step-up converter increases the dc voltage from PV
maximum natural voltage of 272V to 500V, and the 2000Hz
three-phase three-level VSC converts from 500Vdc to
Fig. 1. Performance of different AI methods for a 100kW Grid-Connected 260Vac while keeping unity power factor. The Maximum
PV System.
Power Point Tracking (MPPT) controller is implemented in
This paper presents a novel islanding prevention scheme of the step-up converter, using a StateFlow implementation of
Grid-Connected PV Systems in Matlab/ Simulink based on incremental conductance algorithms, as described in [13].
monitoring the dc-link voltage of the PV inverter. A 100kW The step-up converter controls the MPPT independently and
PV System equipped with islanding protection like frequency the VSC implements dc-link voltage control for the utility
relays, a ROCOF relay, and respectively the proposed dc-link grid connection [14]. The dc-link voltage is set at 500 V by
voltage relay are simulated under the conditions of islanding the three-phase VSC.
and the detection time for all these AI techniques are
compared. The study shows under which conditions our
proposed dc-link voltage AI relay is the most efficient.
The presented AI detection technique consists in
monitoring the dc-link voltage, which raises significantly
during transient grid faults, as shown in [10]. The raise
detection can be implemented in the Voltage Source
Converter (VSC) control of the PV inverter that monitors the
dc-link voltage. In [10] it has been determined that the Fig. 3. The 20kV utility grid model in Simulink environment.
optimal ROCOF threshold which should be exceeded under
the conditions of islanding is 12Hz/s. Using this threshold B. The Simulink Implementations of Anti-Islanding Methods
limit, the ROCOF relay and the frequency relays are
The Simulink block diagram of the AI protection relays on
implemented in Simulink according to the requirements of
the 20kV bus are depicted in Fig. 4. The trip signals of these
international standards. The novel dc-link voltage islanding
AI relays are disconnected from the three-phase circuit
detection method proposed in this paper is simulated as well
breaker CB1 of Fig. 2 on the 20kV distribution feeder at the
using Simulink. The time for islanding detection of the dc-
PCC of the PV System in order to activate all relays and
link voltage relay is compared with the detection time offered
compare the detection times of these AI techniques. During
by ROCOF and frequency relays.
simulation, the relays are activated only once.
The paper has the following structure. The first section
presents the general description of islanding of PV Solar
Power Systems. The Grid-Connected PV System and its
islanding prevention scheme are described in the second
section. The third section gives the theoretical simulation
results of islanding conditions that are detected by AI relays.
Finally, the conclusions of this study are presented.
II. THE GRID-CONNECTED PV SYSTEM AND ITS
ISLANDING PREVENTION SCHEME

A. The Simulink Model of the Grid-Connected PV System


The Simulink model of the 100kW Grid-Connected PV
System is depicted in Fig. 2 [11]. The PV Array, formed by
330 SunPower modules [12] provides in standard test
conditions (STC) a maximum power of 100.7kW and 273.5V. Fig. 4. The Simulink model of AI relays protection bus 20kV.
The Simulink implementations of under/ over frequency
relays, a ROCOF relay, and respectively the proposed dc-link
voltage relay are presented in the following paragraphs.
1) The Frequency Relays
The OF/UF relays monitors the frequency in the PV
inverter and compare this value with some preset thresholds.
The frequency relay model and its settings according to Fig. 8. Simulink implementation of the dc-link voltage AI protection.
Romanian grid requirements [9] are depicted in Fig. 5. The dc-link voltage, Vdc_m is measured by the VSC control
of the PV inverter as is presented in Fig. 9. The dc-link
voltage Vdc is the output voltage of the dc-dc boost converter.
The VSC controller set the reference dc-link voltage Vdc_ref at
the nominal dc bus voltage Vnom_dc equal to 500V.

Fig. 5. Simulink implementation of the under/over frequency protection.

The frequency is estimated by means of the phase-locked


Fig. 9. Measurement of the dc-link voltage.
loop (PLL) as depicted in the Fig. 6. The input is a vector
containing the normalized three-phase signals of primary The dc-link voltage threshold is chosen based on
voltage (Va, Vb and Vc) and the output is the measured simulations at 10% variation of normal dc-link voltage
frequency (Hz) = / (2). Vnom_dc [15]. After several experiments, it was determined
that increasing the threshold value of dc-link voltage, the
detection time of the dc-link voltage relay increases and
becomes larger than the detection time of the ROCOF relay.
The optimal threshold of the dc-link voltage relay determined
through simulations is also justified by the requirement to
operate with a dc-link voltage as low as possible, in order to
Fig. 6. Frequency estimation. keep down the switching losses of the VSC [15].
III. SIMULATION RESULTS OF ISLANDING CONDITIONS
2) The ROCOF Relay
The Simulink implementation of the ROCOF relay is The 100kW Grid-Connected PV Array from Fig. 2 is
depicted in Fig. 7. The ROCOF relay from [10] monitors the equipped with islanding protection methods like under/ over
frequency and the terminal voltage of the PV inverter. In this frequency relays, a ROCOF relay, and respectively the
paper, the ROCOF protection monitors only the frequency in proposed dc-link voltage relay. This model is simulated using
the PV inverter in order to calculate the ROCOF that is Simulink in various islanding scenarios. These scenarios are
compared with the ROCOF threshold. The relay is activated presented in the following subsections.
when this condition is satisfied and the PV array is The islanding conditions occur in the studied network from
disconnected from the power grid. As in case of frequency Fig. 3 when the three-phase circuit breaker CB2 is opened at
relays the frequency for the ROCOF Relay is estimated by time (t) = 0.1 seconds (s). The circuit breaker CB2
means of PLL. disconnects the utility grid from the rest of the network
3) The DC-Link Voltage Relay during 150ms. The dc-link voltages obtained for these
The dc-link voltage relay is implemented in Simulink as is islanding scenarios of Grid-Connected PV System are
shown in the Fig. 8. Its inputs are the dc-link voltage and the depicted in Fig. 10. It should be noted that dc-link voltage is
dc-link voltage threshold used for comparison. The operating not limited by the dc voltage protection of PV inverter.
principle of this AI method is very simple. The relay is
A. Scenario 1: Local Load Greater than Local Generation
activated when the dc-link voltage threshold is exceeded.
The values of local connected feeder loads for this scenario
are shown in the utility grid model from Fig. 3. Different
configurations of the connected feeder loads have different
effect on dc-link voltage as shown in Fig. 11.
The frequency variations and their ROCOF or df/dt, during
island conditions for the same configurations of connected
feeder loads from Fig. 11 are depicted in Fig. 12 and Fig. 13,
Fig. 7. Simulink model of the ROCOF relay.
respectively. As can be observed, for defaults loads and for
the resistive loads these two quantities are quite close. For
capacitive reactive loads, the PV inverter loses control of the
frequency.
Fig. 14 presents together the frequency and ROCOF during
island condition of PV system with default loads of utility
grid from Fig. 3. From Fig. 14 it can be observed the effect of
frequency on the islanding of PV system. During islanding
condition, the frequency decreases in a very short time. Rapid
change in frequency implies a certain variation of ROCOF.
The islanding conditions are detected when a ROCOF
threshold of 12Hz/s is exceeded. The minus sign indicates the
frequency decrease.

Fig. 13. The rate of change of frequency (ROCOF) during island conditions
of scenario 1 in different configuration of the connected feeder loads.

Fig. 10. The dc-link voltages in various islanding scenarios of PV System.

Fig. 14. The Frequency and its rate of change of frequency (ROCOF) during
island condition with default loads from Fig. 3.

The status and detection times (s) of the AI protection


methods in case of this scenario are presented in Fig. 15.
Each AI relay has two displays, one that shows the status of
the relay (0 for inactivated and 1 for trip status) and another
one indicating the trip time of the relay (time at which
islanding conditions are detected).
B. Scenario 2: Local Load Matches with Local Generation
Fig. 16 depicts the simulation results of dc-link voltage for
scenario 2. A direct comparison between frequency and dc-
Fig. 11. DC-Link Voltage during island conditions of scenario 1 in different link voltage in case of active power is depicted in Fig. 17. It
configurations of the connected feeder loads. is clearly observed that the dc-link voltage variation can be
much easier detected than variations of frequency, because it
has a larger slope and range when islanding conditions occur.

Fig. 12. Frequency variations during island conditions of scenario 1 in Fig. 15. AI Relay status and trip time (ms) in case of local connected load
different configuration of the connected feeder loads. greater than local generation.
TABLE I
DETECTION TIME OF ANTI-ISLANDING METHODS
Detection Time (s) of Anti-Islanding Relays
Scenarios DC-Link
UF OF ROCOF
Voltage
Local load greater
0.0536 0.0149 0.0135
than local generation
Local load matches
0.0632 0.0280 0.0212
with local generation
Local load less than
0.0623 0.0276 0.0222
local generation
As can be noted from Table I, the AI technique using dc-
link voltage relay has the fastest detection time. During
islanding condition, the dc-link voltage decrease with the
increase of the connected load. A small variation of inductive
power does not influence much the detection time of the AI
Fig. 16. DC-Link Voltage during island conditions of scenario 2. relays.
The frequency relays have the limits of preset thresholds
imposed by grid codes, leading to late detection of island
conditions of PV systems. The detection times of islanding
conditions of Grid-Connected PV System for the ROCOF
relay presented in [10] are comparable with the detection
times of frequency relays. The ROCOF relay implemented
that in this paper has best detection times than when this AI
relay is realized as in [10].
Fig. 19 depicts graphically the simulation results of AI
methods discussed in this paper. It can be observed that the
dc-link voltage relay offers the best detection time in all
scenarios. It can also be seen that the dc-link voltage, the
ROCOF, and the frequency relays are triggered in the same
Fig. 17. DC-Link Voltage and frequency variation in case of resistive loads.
order in all scenarios.
C. Scenario 3: Local Load Less than Local Generation IV. CONCLUSIONS
The dc-link voltage variations in different cases of this This paper presents a novel islanding prevention technique
scenario are depicted in Fig. 18. Considering the network of Grid-Connected PV Systems based on monitoring the dc-
configuration and its connected load, the results of scenarios link voltage of PV inverter. The detection time of the
in which the connected feeder load is equal or less than PV islanding conditions for the frequency relays, the ROCOF
generation were very close (Fig. 16 and Fig. 18). relay, and respectively the proposed dc-link voltage relay are
The detection times of the AI methods analyzed in this compared. This study shows under which conditions our
paper for studied scenarios discussed above are given in proposed dc-link voltage AI method is the most efficient.
Table I. The performance of proposed dc-link voltage relay is
emphasized by comparing the detection time for all the
studied relays.

Fig. 19. Performance of the AI protection techniques analyzed in this paper


Fig. 18. DC-Link Voltage during island conditions of scenario 3. for a 100kW Grid-Connected PV System.
REFERENCES [9] The National Authority for the Regulation in the Energy Sector,
Technical conditions for connection to electrical networks of public
[1] CIGRE, The Impact of Renewable Energy Sources and Distributed interest of photovoltaic power plants Romanian technical rule, 2013.
Generation on Substation Protection and Automation, Working Group [10] I. V. Banu, M. Istrate, D. Machidon and R. Pantelimon, A study on
B5.34, August 2010. anti-islanding detection algorithms for grid-tied photovoltaic systems,
[2] IEEE Recommended Practice for Utility Interface of Photovoltaic (PV) 14th International Conference on Optimization of Electrical and
Systems, IEEE standard 929, 2000. Electronic Equipment, OPTIM 2014, in press.
[3] IEEE Standard 1547.2-2008, IEEE Application Guide for IEEE [11] P. Giroux, G. Sybille, Hydro-Quebec Research Institute (IREQ), and C.
Standard 1547, IEEE Standard for Interconnecting Distributed Osorio, S. Chandrachood, The Mathworks, 100-kW Grid-Connected
Resources with Electric Power Systems. PV Array demo detailed model (PVarray_Grid_IncCondReg_det.mdl),
[4] IEEE Standard for Interconnecting Distributed Resources with Electric MathWorks File Exchange, 2012.
Power System, IEEE standard 1547, 2003. [12] https://sam.nrel.gov/ - NREL System Advisor Model.
[5] R. Teodorescu, M. Liserre and P. Rodrguez, Grid Converters for [13] I. V. Banu and M. Istrate, Modeling of maximum power point tracking
Photovoltaic and Wind Power Systems. John Wiley & Sons, Ltd., 2011. algorithm for photovoltaic systems, Electrical and Power Engineering
[6] IEA PVPS, Utility aspects of grid connected photovoltaic power (EPE), 2012 International Conference and Exposition on, 2012, pp.
systems, Report IEA PVPS T5-01:1998, 1998. 953-957.
[7] IEA PVPS, Evaluation of islanding detection methods for photovoltaic [14] K.N. Reddy and V. Agarwal, Utility-Interactive Hybrid Distributed
utility-interactive power systems, Task V, Report IEA-PVPS T5-09: Generation Scheme With Compensation Feature, Energy Conversion,
2002, March 2002. IEEE Transactions on, vol.22, no.3, pp. 666-673, 2007.
[8] Ye Zhihong, A. Kolwalkar, Y. Zhang, Du Pengwei, R. Walling, [15] M. Liserre, F. Blaabjerg and A. DellAquila, Step-by-step design
Evaluation of anti-islanding schemes based on nondetection zone procedure for a grid-connected three-phase PWM voltage source
concept, Power Electronics Specialist Conference, 2003. PESC '03. converter, International Journal of Electronics, vol. 91, no. 8, pp. 445-
2003 IEEE 34th Annual, vol.4, pp.1735-1741, 2003 460, 2004.

Citations:
I. V. Banu and M. Istrate, "Islanding prevention scheme for Grid-Connected Photovoltaic systems in Matlab/
Simulink," 2014 49th International Universities Power Engineering Conference (UPEC), Cluj-Napoca, 2014, pp. 1-6.
doi: 10.1109/UPEC.2014.6934698
URL: http://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=6934698&isnumber=6934581

You might also like