Electrical Power and Energy Systems 57 (2014) 198–205

Contents lists available at ScienceDirect

Electrical Power and Energy Systems

journal homepage: www.elsevier .com/locate / i jepes

A new technique for islanding detection using voltage phase angle ofinverter-based DGs

0142-0615/$ - see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.ijepes.2013.12.008

⇑ Corresponding author. Tel.: +98 21 73225612/939 268 7913.E-mail address: hajir_pourbabak@elec.iust.ac.ir (H. Pourbabak).

Hajir Pourbabak ⇑, Ahad KazemiCenter of Excellence for Power Systems Automation and Operation, Department of Electrical Engineering, Iran University of Science and Technology, Tehran, Iran

a r t i c l e i n f o

Article history:Received 15 July 2013Received in revised form 14 November 2013Accepted 1 December 2013

Keywords:Islanding detectionDistributed generation (DG)Active methodsVoltage anglePower mismatchInverter-based distributed generators

a b s t r a c t

In this paper, a new method is studied for the islanding detection of an inverter based distribution gen-eration. The use of parameters at PCC (such as deviation of the voltage and frequency) as a feedback hasbeen considered by researchers. The variation of voltage angle is used as a feedback in the proposedmethod. After an islanding occurrence the variation of voltage phase angle can change the referencesof active and reactive powers. The power variations cause the voltage and frequency deviations as detec-tion parameters. The performance of islanding detection of the proposed method is investigated underthe IEEE UL 1741 for the worst-case. The simulation results show that the adverse effects of this methodare negligible and non-detection zone of the proposed method is small. Also, this method has no signif-icant effects on the power quality and system normal operation mode. In addition, the results of tests formultiple DGs are suitable and there is not any significant interference between DGs in the normal oper-ation mode.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

The rate of electric energy generation by distributed genera-tions (DGs) such as wind farms, small diesel generators, tidal andwave generators and photovoltaic cells is growing in few last yearsand many distributed generators will be interconnected with theutility power systems [1,2]. The presence of DGs causes some sys-tem protection problems such as unintentional islanding. Uninten-tional islanding condition takes place when one or moredistributed generators (DG) and some loads are separated fromthe main utility system while the loads are energized by DGs [3].The voltage and frequency probably will be changed in the islan-ding situation by a power mismatch. The islanding situation causesdamages for equipment, power quality problems and has safetyhazards to personnel [4]. Thus, the island should be detectedcorrectly by islanding detection methods and then DGs should bedisconnected from the loads.

Researchers proposed various methods for the islandingdetection. These methods are divided into two major categories;communication based methods and local detection methods. Localdetection methods can be classified into two main groups: passivemethods and active methods [5].

The passive methods detect islanding condition by measuringone or more parameters such as frequency and voltage level.Therefore, these methods have no adverse impacts on the system

operation. Passive methods have significant non-detection zone.There is no enough parameters deviation for the islanding detec-tion when the power mismatch is low. Thus, only system parame-ters deviation cannot be trusted for islanding detection [6]. Someof passive methods are Over/Under Voltage Protection (OVP/UVP), Over/Under Frequency Protection (OFP/UFP) [7], rate ofchange of frequency and power [8,9] and Phase Jump Detection(PJD) [10].

Active methods use disturbance injection into the power sys-tem through a point of common coupling (PCC). Disturbance injec-tion causes significant variations in some parameters of theislanding condition versus the normal condition [11,12]. Thenon-detection zone of active methods is smaller than that of pas-sive methods. Due to the nature of active methods, they have ad-verse impacts on the power-quality [13]. Some of active methodsthat have been proposed by researchers are slip-mode frequencyshift (SMS) [14], Sandia frequency shift (SFS) [15] and negative-se-quence current injection [16]. However, active methods have ad-verse impacts on the power-quality, but are usually used morethan other methods, because active methods are cheaper thancommunication based methods and have smaller the non-detec-tion zone than that of passive methods.

DGs should operate at a power factor of more than 0.85 whenthe output is more than 10% of the rating. Inverter-based DGs oper-ate close to unity power factor to support the power system andloads in its highest active power capacity [3].

In the islanding detection studies, the RLC load is used in thesimulation, because this kind of loads has the most problems in

H. Pourbabak, A. Kazemi / Electrical Power and Energy Systems 57 (2014) 198–205 199

the islanding detection process [10]. In normal condition, the loadconsumes its required reactive power from the power system. Butin the islanding condition, DGs cannot inject the requested reactivepower, thus frequency goes toward the load resonance frequency.If the load resonance frequency is equal to the system nominal fre-quency, the frequency will not change.

This paper proposes a new active method in which the activeand reactive powers will be changed immediately after an islandis formed applying the variation of DG voltage phase angle (r).There are some active methods which use active and reactive pow-ers injection for destabilizing the parameters of system [11,17,18].However, proposed method in this paper uses voltage phase angleof inverter-based DG as a feedback signal to change powers refer-ences which is different from others methods. Since inverter-basedDG produces constant active and reactive powers in grid-con-nected mode, changes of the voltage phase angle will be insignifi-cant. Therefore, the injected disturbances of powers by thismethod are not very considerable and during non-islanding situa-tion such as load switching, voltage variations and other changes innetwork does not cause false detection. Thus, the adverse effects ofthis method on the power quality are negligible. Additionally, thismethod is simple to implement and does not cause any problemsfor the power with unbalanced loads or multiple DGs.

After islanding occurrence, the voltage and frequency will varyand the island will be detected in a sufficient time by changing inthe voltage phase angle.

The paper is organized as follows. Section 2 presents modellingof the power system and DG control scheme. In Section 3 the newmethod will be explained. Section 4 provides simulation results.

2. System modelling and control scheme

The simple model of a test system including inverter-based DGand a load that both of them are connected to the power systemnetwork is shown in Fig. 1. The power system just was modelledby a voltage source with an impedance.

The power capacity of DG is rated at 0.1 MW. The DG is tuned towork near the unity power factor. The DG contains a DC sourceconnected to an inverter as an interface with the grid [6].

The model of the three-phase RLC circuit is used as a load. Asmentioned in Section 1, RLC loads have most problems in the islan-ding detection process. According to UL1741 the load resonancefrequency is considered near the system operational frequencyabout 59.9–60.1 Hz [5], therefore the RLC load consumes the activepower at unity power factor. There is no significant power mis-match between the load and the DG output. Actually the loadshould have the active power close to the DG output, becausethe worst-case assumption should be considered in the islandingdetection test [3]. Under these conditions, the DG and the load willoperate close to the network nominal operation point, when the is-land is formed. Based on the IEEE 1547, load parameters will becalculated as

L ¼ V2

2pfQf Pð1Þ

Fig. 1. System under study.

C ¼PQ f

2pfV2 ð2Þ

R ¼ V2

Pð3Þ

The DG operates such as a constant active and reactive powersource and it has current and power controllers. The DG controlscheme is shown in Fig. 2.

To control the inverter based DG, a d–q synchronous referenceframe is provided. The DG output active power relation isexpressed in (4), as given [19]:

P ¼ vaia þ vbib þ vcic ð4Þ

By transforming to the d–q synchronous reference frame, theinstantaneous active power will be equal to

P ¼ 32ðvdid þ vqiqÞ ð5Þ

If the d-axis component coincides with the voltage vector andthe q-axis is in quadrature with that. The active power will beequal to (6), as given by [19].

P ¼ 32

vdid ð6Þ

The reactive power could be calculated in term of the aboveconditions. It is equal to

Q ¼ 32

vdiq ð7Þ

Eqs. (6) and (7) show that id and iq can control active and reactivepowers, respectively. The DG output power ((6) and (7)) will becompared with references of the active power (Pref) and reactive

Fig. 2. The DG controller.

200 H. Pourbabak, A. Kazemi / Electrical Power and Energy Systems 57 (2014) 198–205

power (Qref). In this paper, the (Pref) and (Qref) are considered to be0.1 MW and zero, respectively [6].

Currents irefd and iref

q are calculated by a PI controller of thepower regulation. The d–q components of the DG output currentswill be constant quantities [20]. The error of these componentsand related references will be applied into the PI controller of cur-rent regulation to create the d–q components of voltage references.

3. Proposed islanding detection method

The power balance between the load, DG and grid at the PCC isrepresented by

Pload ¼ PDG þ DP ð8Þ

Q load ¼ QDG þ DQ ð9Þ

DP and DQ are equal to the difference in power between the loadand DG that is exchanged with the utility grid. According to (10)and (11), after the islanding occurrence, the voltage and frequencydeviations on the PCC are affected just by PLoad and QLoad. The RLCload operates such as a purely resistive load in the normal operation[5], thus the resonance frequency in (12) will be equal to the systemnominal frequency. As mentioned before, the DG operates near theunity power factor, thus after islanding the frequency will notchange enough for islanding detection.

Pload ¼V2

PCC

Rð10Þ

Q load ¼ V2PCC

1xL�xc

� �ð11Þ

f0 ¼1

2pffiffiffiffiffiffiLCp ð12Þ

The small values of DP and DQ cause an insignificant deviationof the voltage and frequency after the disconnection of the net-work, respectively. In other words, the deviation of system param-eters depends on DP and DQ [10].

The power flow between two AC power systems is used in theproposed method. The active power that flows from the DG towardPCC is shown in (15). The reactive power that is injected into thePCC by the DG is shown in (16). The series resistance has been ne-glected for simplicity.

These equations are extracted from the phasor diagram. Thephase angle between VPCC and VDG is shown as r. According tothe phasor diagram, the voltage of PCC (VPCC) is assumed to bethe reference with a zero phase angle before and after the islandingoccurrence. Therefore r will be the angle of VDG.

The phasor diagram in Fig. 3 shows the relation of VPCC, VDG andIDG. Eqs. (15) and (16) are obtained by using (13) in (14).

I!

DG ¼V!

DG � V!

PCC

jXfð13Þ

Fig. 3. Phasor diagram.

SDG ¼ VPCC I!�

DG ð14Þ

PDG ¼VDGVPCC

Xfsin r ð15Þ

QDG ¼VDGVPCC

Xfcosr� V2

PCC

Xfð16Þ

The second part of the (16) shows the reactive power that isconsumed by Xf. Substituting (15) and (16) in (8) and (9), respec-tively and then (10) and (11) for Pload and Qload in (8) and (9), willresult to:

V2PCC

R¼ VDGVPCC

Xfsin rþ DP ð17Þ

V2PCC

1xL�xc

� �¼ VDGVPCC

Xfcos r� V2

PCC

Xfþ DQ ð18Þ

Eq. (19) is obtained by (17) and (18). The left side of (18) equalszero, because the RLC load operates such as a purely resistive loadin the normal operation. Thus, (19) can be written again as (20).

tanr ¼V2

PCCR � DP

V2PCC

1xL�xc� �

þ V2PCCXf� DQ

ð19Þ

tanr ¼V2

PCCR � DP

V2PCCXf� DQ

¼Xf V2

PCC � RDP� �

R V2PCC � Xf DQ

� � ð20Þ

As can be seen in (20), since the power system do not exchangethe required DP and DQ with the island area, after the islandingoccurrence r will vary with any changes in the active and reactivepowers. This variation is very insignificant, because the power mis-match is very small. In the proposed method, the variation of r isused like a feedback. Studies have shown that during the normaloperation mode, there is not any significant changes in r. There-fore, this feedback has no significant adverse effects on the powerquality.

After the islanding occurrence, the variation of r can change thereferences of active or reactive power with a suitable feedback. Inthis paper, the feedback is applied to both of them. The consideredway for applying r to the active power and reactive powers isbased on (21) and (22). Eq. (21) is the first part of (16) and showssending reactive power by the DG.

Q ¼ VDGVPCC

Xfcosr ¼ k cosr ð21Þ

P ¼ VDGVPCC

Xfsin r ¼ k sin r ð22Þ

Fig. 4. Additional blocks of the proposed method.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 459

59.5

60

60.5

61

61.5

62

Time (s)

Freq

uenc

y (H

z)

With Proposed Method

Without Proposed Method

Fig. 5. Frequency deviation for default parameters.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.8

0.9

1

1.1

1.2

1.3

1.4

Time (s)

Volta

ge (p

u)

With Proposed Method

Without Proposed Method

Fig. 6. Voltage deviation for default parameters.

H. Pourbabak, A. Kazemi / Electrical Power and Energy Systems 57 (2014) 198–205 201

Fig. 4 shows two additional blocks that detect variations of cosrand sinr as an error [17]. The output of these blocks is multipliedby a gain. This gain turns the variation of cosr and sinr to thepower variation and is equal to the maximum value of (21) or(22). The output signal will be added to the references of the activeand reactive power. After islanding occurrence, both of them willbe changed due to the variation of r. This causes a suitable varia-tion in the frequency and voltage. However, in the normal condi-tion, the feedback has no adverse effects on the operation of thepower system, because the DG is a constant power source and itsoutput power does not change.

4. Evaluation of the proposed method’s performance bysimulation

The system is introduced in Fig. 1 has been simulated by PSCAD/EMTDC. The output active power of the DG is rated at 0.1 MW. TheDG does not produce any reactive power. Thus, it operates close tothe unity power factor. The magnitude of the load resistance is ad-justed to 2.304 X. It is preferred to achieve the minimum powermismatch between the load and DG. The gain measurement is cal-culated by (23). The load resonance frequency is near the systemnominal frequency. Other parameters of the system are shown inTable 1 [6].

k ¼ VDGVPCC

Xfð23Þ

The feedback of the proposed method has been applied to theDG interface. The control system has been set such that the islan-ding has happened at t = 3 s. The simulation results of the proposedmethod with the mentioned conditions, as preliminary results, areshown in Figs. 5 and 6. The proposed method detects the island in asufficient time (under 2 s) for the minimum power mismatch. It isobvious in Figs. 5 and 6 that without the proposed method, it is notpossible to detect the island in a sufficient time. In the next parts,the effects of other parameters, on the proposed islanding detec-tion method are investigated.

4.1. UL 1741 testing

The test should be applied with different ratios of the load ac-tive power to the inverter output [21]. To create various scenarios,the load resistance is adjusted to set the active power at 50%, 100%,and 125% of the inverter’s output with 100% of the balanced condi-tion of reactive power. According to UL 1547.1, the test should bedone while the reactive load has been adjusted in 1% change from

Table 1Load and system parameters for UL 1741 testing.

System parametersFrequency 60 HzVoltage (line to line) 0.48 kVDG output power 0.1 MWDG input DC voltage 900 VSwitching frequency 8 kHzGrid resistance 0.02 XGrid inductance 0.307 mH

DG controller parametersPower PI controller Kp = 3 Ki = 0.07Current PI controller Kp = 2 Ki = 0.01

Load default parametersR (X) 2.304L (H) 0.003454C (lF) 20373Quality factor 1.77

95% to 105% of the balanced load value [22]. For brevity, just sometest cases will be proposed [11].

Table 2 shows different conditions of the active and reactivepowers for UL 1741 testing as a Cases 1, 2, 3, 4 and 5. Figs. 7 and8 show simulation results (the deviation of voltage and frequency)for Cases 1, 2 and 3 with the proposed method and without theproposed method. As can be seen, the island can be detected easilyby the frequency and voltage deviations in the various power mis-match scenarios with the proposed method. However, part (b) ofFig. 7 shows that the island situation cannot be detected withoutproposed method at a suitable time by OFP/UFP method. Because,for avoiding unnecessary tripping, the thresholds setting of theOFP/UFP is more than the frequency variation, which occurred inthis situation. In addition, Figs. 9 and 10 show the results relatedto the unbalanced conditions of reactive power (Cases 4 and 5)compared with the Case 1.

As can be seen, OVP/UVP method have large non-detection zoneand cannot detect the island situation. Actually, the OVP/UVPmethod can detect the island situation just in Case 2 (part (b) ofFig. 8).

Table 2Load parameters for UL 1741 testing.

No. L (H) C (lF) R (X) P (%) Q (%)

Case 1 0.003454 2037 2.304 100 100Case 2 0.003454 2037 4.605 50 100Case 3 0.003454 2037 1.843 125 100Case 4 0.003489 2037 2.304 100 99Case 5 0.003419 2037 2.304 100 101

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 459

59.560

60.561

61.562

Freq

uenc

y (H

z)

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 459

59.5

60

60.5

61

Time (s)

Freq

uenc

y (H

z)

Case 1

Case 3Case 2

Case 2

case 1 and 3

(b)

(a)

Fig. 7. Frequency deviation for different cases: (a) without the proposed methodand (b) with the proposed method.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.80.9

11.11.21.31.4

Volta

ge (p

u)

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.80.9

11.11.21.31.4

Time (s)

Volta

ge (p

u)

Case 3

Case 2 Case 1

Case 2 case 1

Case 3

(b)

(a)

Fig. 8. Voltage deviation for different cases: (a) without the proposed method and(b) with the proposed method.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 458596061626364

Freq

uenc

y (H

z)

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 459

59.5

60

60.5

61

Time (s)

Freq

uenc

y (H

z)

Case 5

Case 4Case 1

Case 5

Case 4 case 1

(b)

(a)

Fig. 9. Frequency deviation for Cases 4 and 5: (a) without the proposed method and(b) with the proposed method.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.80.9

11.11.21.31.4

Volta

ge (p

u)

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.80.9

11.11.21.31.4

Time (s)

Volta

ge (p

u)

Case 1

Case 5

Case 4

case 1,4 and 5

(b)

(a)

Fig. 10. Voltage deviation for Cases 4 and 5: (a) without the proposed method and(b) with the proposed method.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 458

59

60

61

62

63

64

Time (s)

Freq

uenc

y (H

z)

Qf=0.5

Qf=1.77

Qf=2.5

Qf=1

Qf=3

Fig. 11. Frequency deviation for different values of quality factor.

202 H. Pourbabak, A. Kazemi / Electrical Power and Energy Systems 57 (2014) 198–205

4.2. Impact of the load quality factor

The quality factor of a circuit is defined as two pi times the ratioof the maximum stored energy to the energy dissipated per cycleat the resonant frequency [3]. The UL 1741 proposes the qualityfactor is equal to 2.5 or less. Actually, it is a typical quality factor.For instance, a quality factor of 2.5 in The United States or a qualityfactor larger than 0.5 in The United Kingdom is used for testing theislanding detection methods [23].

If a RLC load resonance is near the system nominal frequency, itwill not change spontaneously. Some methods that use frequencyshifting techniques can drift the frequency after the islandingoccurrence. But, there is a strong tendency to remain at the reso-nance frequency (the system nominal frequency) with the highervalue of quality factor.

Thus, the higher quality factor causes some problems for theislanding detection methods that use the frequency shifting tech-niques. Thus, the proposed method is tested for the loads that havedifferent quality factors.

The higher values (more than 2.5) of the quality factor are unre-alistic and the islanding detection methods probably cannot detectthe island at a suitable time. Also, the lower values are not accept-able. The reasonable value of the quality factor is between 0.5 and2.5 [24].

Figs. 11 and 12 show the simulation results for different qualityfactors between 0.5 and 2.5 by adjusting the capacitance andinductance based on (1) and (2). The load parameters for differentquality factors is shown in Table 3. The load resistance is adjustedto set the active power at 100% of the inverter’s output. The test isto be done with 100% of the balanced condition of the reactivepower.

4.3. Load imbalance effects

A test procedure is presented to investigate possible effects ofunbalanced loads on the islanding detection methods in [16]. Theresistance of one phase or two must be varied to create an unbal-anced load. Three different cases of the unbalanced loads are testedand compared with Case 1.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.8

0.9

1

1.1

1.2

1.3

1.4

Time (s)

Volta

ge (p

u)

Qf=1.77

Qf=0.5

Qf=2.5

Qf=3

Qf=1

Fig. 12. Voltage deviation for different values of quality factor.

Table 3Load parameters for different values of quality factor.

Qf L (H) C (lF) R (X)

0.5 0.01222 575.6 2.3041 0.00611 1151 2.3041.77 0.003454 2037 2.3042.5 0.002445 2887 2.3043 0.0020387 3454 2.304

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.8

0.9

1

1.1

1.2

1.3

1.4

Time (s)

Volta

ge (p

u)

Case 1

Case A

Case B

Case C

Fig. 14. Voltage deviation for unbalanced load.

2 2.5 3 3.5 4 4.5 559.7

59.75

59.8

59.85

59.9

59.95

60

60.05

60.1

60.15

60.2

Time (s)

Freq

uenc

y (H

z)

Case ICase IICase III

Fig. 15. Frequency variation during load switching.

1.05

H. Pourbabak, A. Kazemi / Electrical Power and Energy Systems 57 (2014) 198–205 203

Case (A). Only the resistance of a phase is set to 95% of its ratedvalue (Phase A).Case (B). Only the resistance of a phase is set to 110% of its ratedvalue (Phase A).Case (C). The resistance of two phases is set to 95% and 110% oftheir rated value (Phase A and Phase C).

It can be seen obviously in Figs. 13 and 14, that the proposedmethod can detect the island for the unbalanced load as well asthe balanced load.

2 2.5 3 3.5 4 4.5 50.9

0.95

1

Time (s)

Volta

ge (p

u)

Case ICase IICase III

Fig. 16. Voltage variation during load switching.

4.4. Load switching effects on the proposed method

One of the problems from which some of the active islandingdetection methods probably suffer, are disturbances of loadswitchings. The load switchings may cause a false detection bythe islanding detection methods [18,25].

The performance of the proposed method has been tested forthe load switchings by adding an extra load with specific parame-ters. For brevity just three cases are proposed and their parametersare:

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 458

58.5

59

59.5

60

60.5

61

61.5

62

62.5

63

Time (s)

Freq

uenc

y (H

z) Case A

Case B

Case C

Case 1

Fig. 13. Frequency deviation for unbalanced load.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.8

0.85

0.9

0.95

1

1.05

1.1

1.15

1.2

Time (s)

Volta

ge (p

u)

P=66%

P=33%P=100%

Fig. 17. Voltage variation for different power level of DG output.

2 2.5 3 3.5 4 4.5 559.5

60

60.5

Freq

uenc

y (H

z)

2 2.5 3 3.5 4 4.5 50.9

0.95

1

1.05

1.1

Time (s)

Volta

ge (p

u)

Case 1Case 2Case 3

Case 1Case 2Case 3

(b)

(a)

Fig. 18. Frequency and voltage variations during load switching for P = 33% of ratedpower.

Case 1Case 2Case 3

Case 1Case 2Case 3

2 2.5 3 3.5 4 4.5 559.5

60

60.5

Freq

uenc

y (H

z)

2 2.5 3 3.5 4 4.5 50.9

0.95

1

1.05

1.1

Time (s)

Volta

ge (p

u)

(b)

(a)

Fig. 19. Frequency and voltage variations during load switching for P = 66% of ratedpower.

1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.659.9

59.95

60

60.05

Freq

uenc

y (H

z)

1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.60.9

0.95

1

1.05

1.1

Time (s)

Volta

ge (V

)

Fig. 20. The frequency deviation during voltage sag.

1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.60.040.060.080.1

0.120.140.16

Activ

e Po

wer

(MW

)

1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6−0.15−0.1

−0.050

0.050.1

0.15

Time (s)

Rea

ctiv

e Po

wer

(MVA

R)

V=1.03pu

V=0.97pu

V=1.05pu

V=1.05pu

V=0.97pu

V=1.03puV=0.95pu

V=0.95pu

Fig. 21. The active and reactive powers variation during voltage sag.

204 H. Pourbabak, A. Kazemi / Electrical Power and Energy Systems 57 (2014) 198–205

Case (I). The load power is 100 kVA with the unity power factor.Case (II). The load power is 80 kVA and 0.8 leading power factor.Case (III). The load power is 125 kVA and 0.8 lagging powerfactor.

These tests show how the proposed method discriminatesbetween the load switching events and the islanding situation.

All loads are switched at t = 3.1 s and switched out at t = 4.2 s inall the cases. The frequency and voltage variation of switchingevents is shown in Figs. 15 and 16, respectively. It can be seenclearly that there is no any significant interference with the oper-ation of power system. According to UL1547.1, the DG maybe oper-ate in other different operation points. The active methods cause todestabilize the power system or damage the power quality duringload switchings, especially in lower power level of DG output.Thus, the proposed methods is studied under different discretepower levels (e.g., 33%, 66% and 100% of rated power) and thepower quality is tested under different power levels.

These studies indicate that the proposed method is independentof the power level of DG output. Because, the island detection is de-tected easily by the proposed method without any significant dif-ference in detection time (Fig. 17). Also, the voltage andfrequency variations are extremely similar. Therefore, the differentpower levels have no effect on the voltage and frequency variationduring load switching (Figs. 18 and 19).

4.5. Voltage sag conditions

The voltage sag conditions has been proposed in [11,17] toinvestigate the performance of active detection method for tran-sient changes. The voltage step has been tested for 3% changes involtage level for VSAC method [17] and voltage sag has been testedwith deferent magnitude for Q–f droop curve method [11]. In thispaper, 3% and 5% voltage variation has been selected for the voltagesag magnitude. This condition is occurred at t = 2 s. As you can seein Figs. 20 and 21, the frequency deviation and powers variationare very low in comparison with Q–f droop curve and VSACmethods.

4.6. Multiple DGs effects on the islanding detection

One of the most important problems in the islanding detectionprocess is the presence of multiple DGs. Actually, the operation of asystem and DGs can be affected by the DGs interaction [5]. Also,multiple DGs interference may cause a false tripping. To evaluatethe performance of the proposed methods, another DG is con-nected to the PCC of the system under study introduced in Fig. 1.

Both DGs have the same parameters. The new system with twoDGs is shown in Fig. 22.

Each inverter has been provided by the proposed method. Thevalue of load resistance is adjusted to consume an active powerequal to the output power of multiple DGs. Some different casesare considered for the output of DGs that are shown in Table 4.The island is occurred at t = 3 s. The simulation results are shownin Figs. 23 and 24 for the frequency and voltage, respectively. Itcan be seen easily that the island was detected as well as the singleDG and there is not any significant interference between DGs bythe proposed method in the normal operation mode. Also, the

Fig. 22. Schematic diagram of the two-DG system.

Table 4Output level of DGs.

Case 1 Case 2 Case 3

DG 1 output (kW) 60 50 40DG 2 output (kW) 60 70 80

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 458

58.5

59

59.5

60

60.5

61

61.5

62

62.5

63

Time (s)

Freq

uenc

y (H

z) Case 2

Case 3

Case 1

Fig. 23. Frequency variation after the islanding occurrence for multiple DGs withthe proposed method.

2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8 40.2

0.4

0.6

0.8

1

1.2

1.4

Time (s)

Volta

ge (p

u)

Case 2

Case 3

Case 1

Fig. 24. Voltage variation after the islanding occurrence for multiple DGs with theproposed method.

H. Pourbabak, A. Kazemi / Electrical Power and Energy Systems 57 (2014) 198–205 205

different cases do not have any significant effect on the detectiontime.

5. Conclusion

In this paper, a new method based on the d–q synchronous ref-erence frame for islanding detection is presented. The methodinvestigated in this paper uses the variation of the voltage phase

angle of an inverter-based DG as a feedback. After the islandingoccurrence, the variation of this signal is applied to DGs to changethe active and reactive power references. This causes the signifi-cant variation in the system voltage and frequency. There are nosignificant adverse effects on the system normal operation mode.

This method is tested under different situations. The results ofthe simulation show that the method can detect the island situa-tion in the worst-case. Also, the proposed method detects the is-land situation with multiple DGs, as well as the single DGwithout any significant interferences between DGs in the normalcondition.

References

[1] Chiang W-J, Jou H-L, Wu J-C. Active islanding detection method for inverter-based distribution generation power system. Int J Electr Power Energy Syst2012;42:158–66.

[2] Dash PK, Padhee M, Panigrahi TK. A hybrid time–frequency approach basedfuzzy logic system for power island detection in grid connected distributedgeneration. Int J Electr Power Energy Syst 2012;42:453–64.

[3] IEEE recommended practice for utility interface of photovoltaic (PV) systems.IEEE Std 929-2000; 2000.

[4] Hashemi F, Ghadimi N, Sobhani B. Islanding detection for inverter-based DGcoupled with using an adaptive neuro-fuzzy inference system. Int J ElectrPower Energy Syst 2013;45:443–55.

[5] Bahrani B, Karimi H, Iravani R. Nondetection zone assessment of an activeislanding detection method and its experimental evaluation. IEEE Trans PowerDeliv 2011;26:517–25.

[6] Zeineldin HH, El-Saadany EF, Salama MMA. Impact of DG interface control onislanding detection and nondetection zones. IEEE Trans Power Deliv2006;21:1515–23.

[7] Ropp ME, Begovic M, Rohatgi A, Kern GA, Bonn Sr RH, Gonzalez S. Determiningthe relative effectiveness of islanding detection methods using phase criteriaand nondetection zones. IEEE Trans Energy Convers 2000;15:290–6.

[8] Vieira JCM, Freitas W, Huang Z, Xu W, Morelato A. Formulas for predicting thedynamic performance of ROCOF relays for embedded generation applications.IEE Proc Gener Transm Distrib 2006;153:399–406.

[9] Redfern MA, Usta O, Fielding G. Protection against loss of utility grid supply fora dispersed storage and generation unit. IEEE Trans Power Deliv1993;8:948–54.

[10] Bower W, Ropp M. Evaluation of islanding detection methods for utility-interactive inverters in photovoltaic systems. Sandia report SAND 3591:2002;2002.

[11] Zeineldin HH. A Q–f droop curve for facilitating islanding detection of inverter-based distributed generation. IEEE Trans Power Electron 2009;24:665–73.

[12] Hsieh C-T, Lin J-M, Huang S-J. Enhancement of islanding-detection ofdistributed generation systems via wavelet transform-based approaches. IntJ Electr Power Energy Syst 2008;30:575–80.

[13] Najy WKA, Zeineldin HH, Alaboudy AHK, Wei-Lee W. A Bayesian passiveislanding detection method for inverter-based distributed generation usingESPRIT. IEEE Trans Power Deliv 2011;26:2687–96.

[14] Yu B, Matsui M, Jung Y, Yu G. Modeling and design of phase shift anti-islandingmethod using non-detection zone. Sol Energy 2007;81:1333–9.

[15] Lopes LAC, Huili S. Performance assessment of active frequency driftingislanding detection methods. IEEE Trans Energy Convers 2006;21:171–80.

[16] Karimi H, Yazdani A, Iravani R. Negative-sequence current injection for fastislanding detection of a distributed resource unit. IEEE Trans Power Electron2008;23:298–307.

[17] Seul-Ki K, Jin-Hong J, Heung-Kwan C, Jong-Yul K. Voltage shift accelerationcontrol for anti-islanding of distributed generation inverters. IEEE Trans PowerDeliv 2011;26:2223–34.

[18] Alaboudy AHK, Zeineldin HH. Islanding detection for inverter-based DGcoupled with frequency-dependent static loads. IEEE Trans Power Deliv2011;26:1053–63.

[19] Schauder C, Mehta H. Vector analysis and control of advanced static VArcompensators. IEE Proc C Gener Transm Distrib 1993;140:299–306.

[20] Zeineldin HH, Kirtley JL. A simple technique for islanding detection withnegligible nondetection zone. IEEE Trans Power Deliv 2009;24:779–86.

[21] Inverters, converters, and controllers for use in independent power systems.UL STD 1741; 2001.

[22] IEEE standard conformance test procedures for equipment interconnectingdistributed resources with electric power systems. IEEE Std 15471-2005;2005. p. 1–62.

[23] Woyte A, Belmans R, Nijs J. Testing the islanding protection function ofphotovoltaic inverters. IEEE Trans Energy Convers 2003;18:157–62.

[24] Stevens J, Bonn R, Ginn J, Gonzalez S, Kern G. Development and testing of anapproach to anti-islanding in utility-interconnected photovoltaicsystems. Sandia National Laboratories; 2000.

[25] Menon V, Nehrir MH. A hybrid islanding detection technique using voltageunbalance and frequency set point. IEEE Trans Power Syst 2007;22:442–8.