Overview of Anti-Islanding Algorithms for PV

Systems. Part I: Passive Methods

Francesco De Mango', Marco Liserre2, Antonio Dell'Aquila2 and Alberto Pigazo3

1GRTN S.p.A, Rome, Italy2Dept. Electrical and Electronic Engineering, Politecnico di Bari, Bari, Italy3Dept. Electronics and Computers, University of Cantabria, Santander, Spain

Abstract- This paper offers an overview of the passivemethods used for islanding detection. A monitoring PLL hasbeen adopted for the estimation of the voltage amplitudeand frequency. Passive inverter resident methods discussedare the over/under voltage (OUV), over/under frequency(OUF), the voltage harmonic monitoring (with a study onthe effects of the grid impedance value and of the inverterdc voltage ripple) and the phase monitoring (different fromthe classical phase jump method). A harmonicsynchronization PLL is used to monitor the 3rd, 5th and 7thharmonics. All the proposed algorithms are validated bysimulations and experimental results obtained inaccordance with the IEEE Std. 929-2000 procedure.

I. INTRODUCTIONInadvertent islanding can be one of the main technical

issues for small Distributed Generation (DG) systems [1][2]. Islanding operations ofDG occurs when power supplyfrom the main grid is interrupted due to several reasonsbut the DG keeps supplying power into distributionnetworks. The grid disconnection is usually in response toa fault. Ideally, the fault should be detected by the DGprotection system and the DG tripped before an island canoccur. DG grid-connected islanding may occur as a resultof the following conditions:* a fault that is detected by the grid but which is not

detected by the PV inverter or protection devices;* accidental disconnection of the normal grid supply by

equipment failure;* intentional disconnection of the line for servicing;* human error or malicious mischief; or,* an act of nature.It is necessary to predict the possibility that islandingoccurs evaluating the probability of line disconnection andof the contemporary occurrence ofpower balance betweenthe PV-source and the load [3]. In fact in this case thenormal protection does not stop the PV-system to energizethe Point of Common Coupling (PCC). It is possible todefine the index of penetration of a generic photovoltaicsystem RP = Pload /PDG , where Pload is the active powerabsorbed by the load and PDG is the power injected bythe distributed generation system. The probability ofislanding is proportional to the amount of energy injectedby the PV system so that unitary RP means a highprobability of an islanding phenomenon.

Inadvertent islanding causes a number of safety,commercial, power quality, and system integrityproblems [4]-[8]. Essentially the risks for the line workersare tightly dependent by the probability of islandingduring an unplanned outage and by the time ofpersistence of the electric island [9]. The voltage andfrequency provided to other customers connected to theisland can be out of the grid control, yet the grid remainsresponsible to those customers. Moreover it is possiblethat grid breakers or circuit reclosers reconnect the islandto the grid system when out of phase with a consequentovercurrent and trip-out of the PV-system.

For these reasons islanding detection is anindispensable feature for DG-systems and manyalgorithms have been developed in the past decade. Thesemethods may be divided into four categories: passiveinverter-resident methods, active inverter-residentmethods, active methods not resident in the inverter andmethods based on the use of communications between thegrid and PV inverter. Passive inverter-resident methods,discussed in the present paper, rely on the detection of adisturbance in the voltage at the PCC. A generic systemfor anti-islanding study defined in IEEE Std. 929-2000and in IEEE Std. 1547 is depicted in Fig. 1 where thegrid, the RLC load and the PV inverter are connected inthe PCC [10], [11] . The IEEE Std. 929-2000 set thequality factor (q = R C/L ) to be 2.5 as a testingcondition.

II. PASSIVE INVERTER-RESIDENT ISLANDING DETECTIONMETHODS

A. Non detection zoneThe non detection zone (NDZ) is the range (in terms of

power mismatch between the DG inverter and the load orload parameters) in which the islanding detection schemeunder test fails to detect islanding. The NDZ can be usedas a performance index to evaluate different anti-islandingalgorithms [12]-[14].

Fig. 1. Interconnection ofPV source to the grid and the load.

1-4244-0121-6/06/$20.00 C 2006 IEEE 1 878 EPE-PEMC 2006, Portoro2, Slovenia

The aim of all islanding detection methods is to reducethe NDZ near to zero.

B. Overlunder voltage (OUV) and overlunderfrequency(OUF) methodThe grid is subject to numerous disturbances as voltage

dips, over voltage, harmonic distortion and frequencyvariations. It is necessary to create islanding protectionimmunes to these disturbances. The grid voltage andfrequency limits are given in Table I [10]: all grid-connected PV inverters are required to have an over/underfrequency and over/under voltage protections.

Fig. 1 shows the balance of the power in the system

Pload =PDG +AP(1Qload = QDG +AQ (2)

If Pload =PDG and/or if Qload = QDG there is no

active/reactive power mismatch between the PV systemand the grid.

The behaviour of the system when the grid isdisconnected will depend on AP and AQ [15], [16]. Ifthe resonant frequency of RLC load is the same as gridline frequency the linear load does not absorb reactivepower. Active power is directly proportional to thevoltage. After the disconnection of the grid, the activepower of the load is forced to be the same of the PVsystem, hence the grid voltage change into

V =K*V (3)where

K = VPDG /Pload (4)

When PDG > Pload there is an increase ofthe amplitudeof the voltage if PDG < Pload there is a decrease of theamplitude. Reactive power is tied up to the frequency andamplitude of voltage

Qload = QDG K Lc (5)co) L

In this way is possible to calculate the islandingpulsation (o )

C0=

_ QDG QDG 4V+

2 V 2 +

2A small AP results in an insufficient change in voltage

amplitude and small AQ results in a inadequate change infrequency to effectively disconnect the PV and preventislanding.

It is possible to calculate the NDZ area from themismatches of active and reactive power and setting thevalues of threshold for frequency and amplitude of thevoltage (Fig. 2).

Fig. 2. Non detection zone ofOUV and OUF.

The probably that AP and AQ fall into the NDZ ofOUV/OUF can be significant. Because of this concern, thestandards over/under voltage and frequency protectivedevices alone are generally considered to be insufficientanti-islanding protections.

C. Voltage harmonic monitoring methodThe goal of this method is to monitor the voltage

harmonic distortion to detect an islanding condition [18],[19], [20]. In normal operation the voltage at PCC iscontrolled by the grid; in islanding condition DG controlsthe PCC voltage and its harmonics. It is possible toconsider all the harmonics using the Total HarmonicDistortion (THD) of the PCC voltage or only the mainharmonics: the 3rd, 5th and 7th. It is possible to use aPhase Looked Loop (PLL) to provide the values of themonitored harmonics. It is very hard to detect islanding ifthe grid voltage harmonic distortion is not high or lowenough such as the THID changes when islanding occurs.

The maximum amplitude of the grid voltage harmonicsis reported in Table II. The grid voltage harmonicdistortion can change in the time depending on the gridimpedance; moreover the inverter system can present anharmonic distortion in DC bus with a consequentharmonic distortion generated on the grid side. Generallya PV system is connected to the grid through atransformer that can affect the harmonic distortionespecially in case of electric island. For this reason it isnot always possible to select a trip threshold that providesreliable islanding protection.

The NDZ of the harmonic-based methods is stronglytied to the load. RLC parallel resonant load present a lowpass characteristic in frequency that can filter low orderharmonics more than other and influence the detection ofislanding. In case of a huge variation of harmonicsamplitude, these methods have a small NDZ.

D. Effects ofthe grid impedance and ofthe inverter dcvoltage rippleHarmonic distortion can be influenced by the value of

grid impedance. The purpose is to determine how theamplitude of the voltage harmonics changes as the gridimpedance changes.

Grid impedance calculated for the generic voltageharmonic k is equal to

Zgrdk = (k) arctg (7)

where Rg is the grid resistance and Lg is the gridinductance. Considering an RLC load in resonantcondition (inductive and capacitive reactive powers equaleach other) and the PV inverter producing no reactivepower, there will be reactive power associated to the gridvoltage harmonic distortion.

It has been simulated the amplitude variations of the3rd harmonic (AV3) as the grid impedance varies. TheTHD is almost immune to the resistance variation while itis more influenced by the inductance variation (Fig. 3a).

TABLE I GRID VOLTAGE AND FREQUENCY LIMITS [17].Value Minimum Maximum

Frequency fmin=49 Hz fmax =51 Hz

Voltage V i =0.9 pu Vmax =11 pu

1879

NDZ

TABLE II HARMONICS VOLTAGE AMPLITUDE [17].

Harmonic order (h) 3 5 7 9 11 13Amplitude (%) 5 6 | 5 T 1.5 3.5 3

As it regards the dc voltage ripple influence, it shouldbe considered that the ac voltage produced by the inverteris

Vac(t) =JM Ivd,C(t) sen(co t) (8)where M is the modulation index, vdc (t) is the dc

voltage. The dc voltage harmonic content is00

Vdc Vdc + VdCk sen(ka + Sk) (9)k=2

From (8) and (9) it is possible to obtain:

Vac(t) = M-Vdc -sen(w.t)+ 2 M. VdCk -cos(kcowt+kS-co)t)-k=2

1 (10)-- M VdCk cos(ko- t +z}> + D -t) = M Vdc sen(o t) +

k=2

+- Vdk -[cos(o t (k-1) + Sk)-cos(- t (k +1) + )]

Therefore, dc voltage ripple involves the generation ofac voltage harmonics of k+1 and k-i orders [21]. Thepresence of a second order harmonic in dc voltageproduces a 3rd and 1st harmonic in the inverter output.Considering an index i0 as the ratio between a generic k-order voltage harmonic and the dc voltage it is possible tostudy the variation of the 3rd harmonic as a consequenceof the dc ripple

Vdc (1 1)

It is possible to evaluate the variation of AV3 infunction of the grid impedance and of the dc voltageripple as reported in Fig. 3b. The minimum value of AV3is obtained for a low grid impedance and a high dc rippleindex.

E. Phase monitoring methodThe method consists in detecting a sudden "jump" in

the phase displacement between inverter terminal voltageand its output current [12]-[14]. However if a fast PLL isimplemented this phase jump is negligible since thecurrent will be always in phase with the voltage at theinverter terminals. Hence it is proposed to modify thephase jump method as described in the following. Thisnew method can be more correctly defined phasemonitoring method.

Under normal operation (inverter producing zeroreactive power) there is no phase displacement betweenvoltage and current at PV system output terminals. Thereference current for the inverter control is synchronizedwith the fundamental voltage at the PCC.

The variation of the voltage frequency, consequent toislanding, causes a shift of the voltage vector incomparison to the axe d and a consequence change of thephase. The detected angle is stored and compared with thevalue measured after a whole multiple of the period of thefundamental, hence

ASi& Si_ / 1 1Aau-= Ut -ut-l t1 zJThe output phase is a ramp with a certain slope. A

frequency change causes an alteration of the slope that canbe detected.

It is possible to calculate the phase of the load with thefollowing equation:

= tan1LR 9c J1 (13)

where the pulsation co corresponds to the nominalfrequency of the grid.

If islanding occurs with a load resonating at the gridfrequency, the phase does not vary (iX = 0), while if theload is resonating at a difference frequency the phasechanges. It is possible to set an angle OS of the anti-islanding method. When fY| > Os is possible to effectivelydetect an islanding operation of the DG. The methodperformance is strongly dependent on the reactiveelements. The NDZ of this method is therefore the sameas the over/under frequency method.

III. MONITORING AND HARMONIC SYNCHRONIZATION PLLFOR THE PASSIVE ANTI-ISLANDING METHODS

For the grid-connected PV inverters in the range of 1-5kW, the most common control structure for the dc-ac gridconverter is a current-controlled H-bridge PWM inverterhaving a LCL low-pass output filter [22]-[24].

The control of the PV-converter plays a more importantrole for the active anti-islanding methods; hence it will betreated in more details in the paper "Overview of anti-islanding algorithms PV systems. Part II: active methods".On the contrary in case of the passive methods it is

worth focusing on the synchronization device (Fig. 4).The phase lock is obtained controlling to zero the sinus ofthe phase difference between the reference and the outputphases.

The time domain transfer function of this PLL systemcan be derived as

K,p s+Kp T

s2 +Kp .s+KPlT(14)

where Kp and T, are the parameters of the PI controllerthat can be derived as follows [25]

=9.2 T 32 4.6TS 2.3 ,weeo- (15)

7~~~~~~~~~~~~~~~~~~~~~~~~~~

(azdcac m)0 0.2 Grid -eistanc (,hmi)

(a)

1.6 0

Grid imeac(om 0 dc ripple? index (%)

Fig. 3. Variation of the amplitude of 3rd harmonic due to islanding a)versus grid inductance and resistance b) versus inverter dc voltage

ripple and grid impedance.

A. Monitoring PLLIt is possible to use a second PLL designed as

monitoring device for the frequency, the phase andamplitude of the voltage (Fig. 5).

1880

Loiop filter

\Iq p~~~~~~has

Grid _ Fundamentalvoltage x i sine

Fig. 4. PLL scheme.

This second PLL is designed with these constraints* time for islanding detection compliant with the IEEE

Std. 929-2000 [10] (< 2 s.),* good estimation of fundamental voltage with a THD

<2%,As a consequence the settling time should be 20 times

higher than that of PLL used for the controllersynchronization. Optimum results have been obtaineddesigning the PI controller with Ts =0.4 s and 6=0.707.When the voltage vector is synchronized to the dq rotatingframe it is possible to estimate frequency and amplitude ofthe voltage. It is necessary to introduce a filter to extractthe exact amplitude and frequency of the PCC voltage. Itis used a first order Butterworth filter with a cut-offfrequency of 2.5 Hz to get dc value out of the signals.

B. Harmonic synchronization PLLUsing multiple frames synchronized with the speeds of

the selected harmonics it is possible to monitor theselected harmonics present in the grid voltage.

This harmonic synchronization PLL is designed with aPI controller that provides the frequency of thefundamental. Fig. 6 shows the monitoring of the 3rd, 5thand 7th harmonics by the harmonic synchronization PLL.

Considering the 3rd harmonic synchronization, the newrotating frame called d'q' rotates at the same speed of theselected harmonic. The amplitude of each voltageharmonic is obtained by the calculation of Vqt andVd ' components. It is used a 3rd order Butterworth filterto eliminate the ripple and extract the dc value of thesingle component. The same operation is repeated for theother harmonics. This device is used in the harmonicmonitoring method. Generally for islanding detection it isnecessary monitoring only the 3rd, 5th and 7th harmonic.

IV. SIMULATION RESULTS

Simulated results are presented to illustrate theeffectiveness of the discussed anti-islanding algorithms.The DG power is controlled to be constant with an unitypower factor. A current controller has been designed tocompensate the 3rd, 5th and 7th harmonics. An anti-windup system has been adopted for the controller.

Gridvoltage

--(p Vd._oilter 'oItage

_ do aFltr| "plitude

PLL _* lb FMlter Frequency

_

,p Phase

lrO lt3g e M OnitbrR vg

_ w~~~~~~~~~~~~~~~q VdV 2 tV5

V7

Fig. 6. Scheme of the harmonic synchronization PLL with monitoringof 3rd, 5th and 7th harmonics.

A. Overlunder voltage and overlunder frequencymethodThe behaviour of the system at the time of grid

disconnection depends on AP and AQ at the instantbefore the switch opens to form the island. Fig. 7 showsthe cases when the active power mismatch between theDG system and the load is large (AP =25% and AQ =0%and when AP =0% AQ=-15%).

B. Voltage harmonic monitoring methodThis method is based on the monitoring of all the

harmonics by the total harmonic distortion (THD) of thePCC voltage or on the monitoring of only the greaterharmonics, generally the 3rd, 5th and 7th , through theharmonic synchronization PLL. Considering a grid modelin which the harmonic distortion is very low and a PVsystem that produces only the fundamental voltage at itsterminals, in the islanding condition the THD does notchange. In the case of a grid model characterised by theharmonic pollution reported in Table II in Section II,during islanding transient the THD decreases significantlybecause the inverter becomes the only source in theelectrical island. These two situations are shown in Fig. 8in the worst case for islanding detection (AP = AQ =0).

C. Phase monitoring methodThe aim of this method is the monitoring of the phase

variation after a grid disconnection.PCC voltage PCC voltage

600 ,600Ann I islanding I nn I islandingI

1.5 2 2.6 3 3.6 1.6 2 2.6Time (s) Time(s)(a) (b)

Fig. 7. Islanding in the case of a) AP =25% and AQ=0, b) AQ=15%and AP 0.

1881

Fig. 5. Scheme of the monitoring PLL.

Due to the previous analysis the voltage phase dependson the frequency variation and on the reactive powermismatch. Fig. 10 shows the phase change with differentvalues of AQ.

D. Comparison among the simulated anti-islandingmethodsTable III shows a comparison among the simulated

anti-islanding methods. A good result is obtained usingthe voltage harmonic monitoring method. In this methodthe NDZ is reduced to zero and the trip time is within thespecified limit of IEEE 929-2000.

V. EXPERIMENTAL RESULTS

Fig. 10 shows the scheme of the experimental setup(Politecnico di Bari, Italy) of a single phase PV invertergrid-connected to test the inverter-resident anti-islandingpassive methods. The inverter is current controlled by adigital signal processor (DSP) and the anti-islandingalgorithms are implemented in the DSP. Two dc powersupplies have been used to simulate PV panels (Vdc =460V).

A. Overlunder voltage and over underfrequencyIn Fig. 11 the results of two tests, made in power

unbalance conditions, have been reported. It is possible toverify that the variation of amplitude and frequency of thegrid voltage is in accordance with the theory and thesimulation results. The two main differences are in thetransient over/undershoots that are not present insimulation because the simulation model does not takeinto account the dynamic of the dc-link voltage and thepresence of oscillations due to the harmonics present inthe system. The little variations are due to a small powerleakage in the actual system due also at a non ideal RLCparallel resonant load that absorbs a small reactive power.

20isadn

THD voltage

1~~~~~~~~ ~islanding |, ,.

0 20 - d114 04 4

1.4 1.45 1.5 1.55Time (s) 10X

THD voltage15 a T d r islanding r1a 4 1(5o1pf1) 7 1.i 9

10 _ zr<=~~~~~~~~101.5 Time (s) 1.55 16 °4 1.5 1.6 1.7 1.8 19

Time (s)

(a) (b)Fig. 8. a) THD during islanding transient (top-left) in the case of a

sinusoidal grid voltage (bottom-left) in the case of a high grid voltageharmonic pollution, b) variation ofthe magnitude of the 3rd, 5th, 7thharmonics voltage during islanding transient detect by the harmonics

synchronization PLL.

0

TABLE III COMPARISON AMONG PASSIVE ANTI-ISLANDING METHODS

Method NDZ Trip time (powerMethod_______ -17 00<AP<24 00 - balance)OUV -17 %<AP<24 % Not applicableOUF -5%<AQ<5% Not applicable -

Phase monitoring -5%<AQ<5% Not applicableHarmonic Absent 0.1-:0.2 smonitoring

B. Voltage harmonic monitoring methodThe islanding condition has been tested in the case of

AP= AQ = 0 . All the harmonics amplitudes present anoscillatory behaviour during the transient with largevariation from normal operation to islanding condition.

The 5th harmonic suffers of a drastic variation whilethe 3rd and the 7th do not change a lot. This is illustratedin Fig. 12a. It is worth noting that the 3rd and 7th are alsovery small when the grid is still connected and it isdifficult to set the thresholds to realize a reliable method.The islanding detection is obtained using a set of

harmonic amplitude variation threshold specified in TableIV that enables the inverter trip as shown in Fig. 12b. Thetrip time is proportional to the set values.

C. Phase monitoring methodIt has been tested the phase monitoring method

proposed. In normal condition the frequency is stablebecause it is forced by the grid; in the worst case forislanding detection, the frequency does not change andtherefore the phase change is zero. In case of AQ=-15%and AP=0 there is a phase change proportional to thefrequency variation and islanding can be detected. Theresults are showed in Fig. 13.

Fig. 10. Scheme of anti-islanding test bench.

400 PGC voltage amplitude

350-

S 300

250

200IO 0.6 1,5 2

Time (s)51Frequency

48 L0

Fig. 9. Variation ofphase after an islanding condition with differentvalues of AQ.

(a) (b)Fig. 11. Islanding in case of a) AP =25% and AQ =0, b) AQ =-15%

and AP=0.

1882

,) G,rid

Time (s) Time (s)

14F

2 Voltage harmonic distortion 1212 - Normal operation

m Islanding 10

8 * O 8

1]~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~-~

2- 2

12 34. 6 7 9Harmonic order

(a)

0 0.125 0.25 0.3Z5 0 5 0.625Trip signal

0 0.125 0.25 0.375 0.5 0.625Tirne (s)

(b)Fig. 12. a) Harmonic distortion in normal operation and during

islanding, b) 3rd, 5th and 7th harmonic voltage amplitude estimatedby the harmonic synchronization PLL in the islanding transient in case

of AP =AQ =O.

TABLE IVHARMONIC AMPLITUDE THRESHOLDS FOR ISLANDING DETECTION

Harmonic order (h) 3 5 7Set threshold for 020.3 0.50.7 0.40.5

harmonic monitoring method (V)

Phase change60

30c c

- -30 z-

0 0.5 15 2Time (s)

0 05Time (s)

(a) (b)Fig. 13. Phase variation during islanding transient, in case of (a)power

balance condition, (b) AQ=-15% and AP =o.

VI. CONCLUSIONSThis paper offers an overview of passive anti-islanding

algorithms for PV grid-connected distribution systems. Allthe methods have been tested according to IEEE Std. 929-2000. The discussed islanding detections algorithms havebeen optimized: the harmonic monitoring method hasbeen studied taking into account the grid impedanceinfluence and the dc-link ripple, the phase-jump methodhas been modified into a phase monitoring method.

VII. REFERENCES[1] V. Task, "Grid aspects of grid-connected photovoltaic power

systems," Tech. Rep. IEA-PVPS T5-01:1998, December 1998.[2] I. Task, "Added values of photovoltaic power system," Tech. Rep.

IEA-PVPS T 1 -09:2001, March 2001.[3] V. Task, "Probability of islanding in grid networks due to grid-

connected photovoltaic power systems," Tech. Rep. IEA-PVPST5-07:2002, September 2002.

[4] A. Woyte, R. Belmans, J. Nijs, "Power flow fluctuations indistribution grids with high penetration",17th Europeanphotovoltaic solar energy conference and exhibition , Monaco,Germania, 22-26 October 2001.

[5] M. Begovic, A. Pregelj, A. Rohatgi D. Novosel, "Impact ofrenewable distributed generation on power systems," Proceedingsof the 34th Hawaii International Conference on System Sciences,IEEE, pp.1 -10,2001.

[6] B. A. Archer, J. B. Davies, "System islanding considerations forimproving power system restoration at Manitoba Hydro,"Proceedings of the 2002 IEEE Canadian Conference on Electrical& Computer Engineering, IEEE, pp. 60-65, 2002.

[7] V. Task, "Grid connected photovoltaic power systems: survey ofinverter and related protection equipments," Tech. Rep. IEA-PVPS T5-05:2002, December 2002.

[8] Walling, R.A.; Miller, N.W. "Distributed Generation Islanding-Implications on Power System Dynamic Performance."Proceedings of the IEEE/PES Summer Power Meeting, Chicago,Illinois, pp. 92-96, July 2002.

[9] V. Task, "Risk analysis of islanding of photovoltaic powersystems within low voltage distribution networks," Tech. Rep.IEA-PVPS T5-08:2002, March 2002.

[10] IEEE 929, "IEEE recommended practice for grid interface ofphotovoltaic (PV) systems." New York, NY: Institute of Electricaland Electronics Engineers, April 2000.

[11] IEEE 1547, "Standard for interconnecting distributed resourceswith electric power systems." New York, NY: Institute ofElectrical and Electronics Engineers, 2003.

[12] V. Task, "Evaluation of islanding detection methods forphotovoltaic grid-interactive power systems," Tech. Rep. IEA-PVPS T5-09:2002, December 2002.

[13] Z. Ye, A. Kolwalkar, Y. Zhang, P. Du, R. Walling, "Evaluation ofanti-islanding schemes based on nondetection zone concept,"IEEE transactions on power electronics, vol. 19, no. 5, pp. 1171-1176, September 2004.

[14] M. E. Ropp, M. Begovic, A. Rohatgi, G. A. Kern, R. H. Bonn, Sr.,and S. Gonzalez, "Determining the relative effectiveness ofislanding detection methods using phase criteria and nondetectionzones," IEEE transactions on energy conversion, vol. 15, no. 3,pp.290-296, September 2000.

[15] C. Jeraputra, P.N. Enjeti, "Development of a robust anti-islandingalgorithm for grid interconnection of distributed fuel cell poweredgeneration," IEEE transactions on power electronics, vol.19, no.5,pp. 1163 -1170, September 2004.

[16] R. A. Jones, T. R. Sims, A. F. Imece, "Investigation of potentialisland of self commutated static power converter in photovoltaicsystems," IEEE transactions on energy conversion, vol. 5,no. 4,pp. 624-631, December 1990.

[17] IEC 61727 Draft.[18] R. M. Hudson, T. Thome, F. Mekanik, M. R. Behnke, S.

Gonzalez, J. Ginn, "Implementation and testing of anti-islandingalgorithms for IEEE 929-2000 compliance of single phasephotovoltaic inverters," Photovoltaic specialists conference,conference records of the 29th IEEE, pp. 1414-1419, 19-24 May2002.

[19] S. Jang, K. Kim, "An islanding detection method for distributedgenerations using voltage unbalance and total harmonic distortionof current," IEEE transactions on power delivery, vol. 19, no. 2,pp. 745-752, April 2004.

[20] H. Zeineldin, M. I. Marei, E. F. El-Saadany, M. M. A. Salama,"Safe controlled islanding of inverter based distributedgeneration," 35th annual IEEE power electronics specialistconference, pp. 2515-2520, Aachen, 2004.

[21] L. Morhn, P. D. Ziogas, G. Joos, "Design aspects of synchronousPWM Rectifier-Inverter Systems under Unbalanced Input VoltageConditions," IEEE transactions on industry applications, vol. 25,no. 6, pp. 1286-1293, November/December 1992.

[22] M. Liserre, F. Blaabjerg, R. Teodorescu, Z. Chen, "Powerconverters and control of renewable energy systems," ICPE 2004,Busan, Korea, Invited paper.

[23] S. J. Ranade, N. R. Prasad, S. Omick, F. Kazda, "A study ofislanding in grid-connected residential photovoltaic systems Part1- Models and analytical methods," IEEE transaction on energyconversion, vol. 4, no. 3, pp. 436-445, September 1989.

[24] J. Liang, T.C. Green, G. Weiss, Q. Zhong, "Hybrid control ofmultiple inverters in an island-mode distribution system," PESC'03. IEEE 34th annual conference on power electronics specialist,2003, vol. 1, pp. 61-66, June 2003.

[25] M. Liserre, A. Timbus, R. Teodorescu, F. Blaabjerg,"Synchronization methods for three phase distributed powergeneration systems. An overview and evaluation," PESC 2005,Recife, Brasile.

1883

本文献由“学霸图书馆-文献云下载”收集自网络，仅供学习交流使用。

学霸图书馆（www.xuebalib.com）是一个“整合众多图书馆数据库资源，

提供一站式文献检索和下载服务”的24 小时在线不限IP

图书馆。

图书馆致力于便利、促进学习与科研，提供最强文献下载服务。

图书馆导航：

图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具