You are on page 1of 6

Band Pass Filter basedActive Island Detection

Technique for PMSGbasedWECS


Suman Murugesan, Student Member, IEEE Venkatakirthiga Murali, Member, IEEE
Department of Electrical andElectronics Engineering Department of Electrical andElectronics Engineering
National Institute of Technology, Tiruchirappalli National Institute of Technology, Tiruchirappalli
Tiruchirappalli, India Tiruchirappalli, India
sumanmnitt@gmail.com mvkirthiga@nitt.edu

Abstract—The perpetual developments in the field of power unstabilize the system on islanded condition. Active IDTs
electronics have endorsed electric power generation at the basedonsignal injectionhavebeenproposedin [5], [12], [13],
distribution system. However, power supply to loads should which do not unstabilize the systemon islanded conditions.
be suspended during an unplanned islanding. In this work, The methodproposedin [5] is utilized andvarious analysing
an active islanding detection technique based on disturbance
injection through q-axis current controller is presented which methodologies areproposedin [14]-[16]. Theislanddetection
makes the frequencytovary significantly onpost islandingand time is > 200ms in [15], [16] for perfectly matched power
asimpleanalysingmethodisproposedbasedonthecombination condition (between the DG generation and demand) and the
of band pass filter (BPF) and Mean of absolute frequency proposed technique in [14] mal-identifies the non-islanding
variation (AFVmean) for multiple converter interfaced DGs to condition as islanded condition if the operating frequency is
discriminate between the islanded and non-islanded conditions.
The simulations are executedinMATLAB R2017b/Simulink. outside ± 0.5 Hz fromthenominal frequency.
Index Terms—Unintentional Island Formation, Active Island- In this context, q-axis current perturbation based active
ingDetection, Disturbance Injection, AFVmean IDT is adopted and an analysing method based on the com-
bination of BPF and AFVmean is proposed in this work to
I. Int r o duc t io n discriminate the islanding fromnon-islanding conditions for
Detection of island formation during unintentional is- permanent magnet synchronousgenerator (PMSG) basedwind
landing is ineluctable and is termed as islanding detection energyconversionsystem(WECS). Theadvantagesof thepro-
technique (IDT) [1]-[5]. posedmethodareasfollows: applicabletoanyconverter based
The IDTs are of two types, Remote and Local IDTs. DGs, applicable to varying demands, adaptive threshold for
Remote IDTs arebasedon communication between the DGs varyingdemandsisnot needed, doesnot destabilizethesystem
andthesubstation [6], [7] andhencenot suitablefor lowrated onpost islanding, accurateandfaster islanding detectionthan
microgrid, owing to higher investment cost [2]-[4]. earlier worksbasedonq-axis current disturbanceinjectionetc.
Local IDTs are divided into passive and active IDTs.
Passive IDTs are based on analysing the parameters at the II. Test Syst em Desc r ipt io n
DG output terminal. Voltage unbalances and total harmonic The test systemadopted for this work is shown in Fig. 1.
distortion are the parameters utilized to detect the island Thetest systemencompassespermanent magnet synchronous
formation in [8]. The rate of change of voltage phase angle generator based wind energy conversion system interfaced
(RoCoVPA) is computedfromthephaseangleobtainedfrom to the utility grid at the point of common coupling (PCC)
the phase locked loop of the inverter, and it is used as a through uncontrolled three phase diode bridge rectifier, DC-
detection index in [9]. But the major drawback of passive DCconverter, inverter andinductor filter (Lf) with resistance
techniques is thenon-detection zone [2]-[4], wherethe IDTs (Rf) and parallel RLC load connected at the same PCC.
fail to detect the island formation for less or zero power The resonant tuned load is considered as the worst loading
mismatch (between supply and load). condition [17], and hence the same is adopted in this work.
Active IDTs add feedback to the control parameters The test systemspecifications aredepictedin Table I.
of the converter or inject disturbance through the DG to
make notable change in frequency, voltage magnitude and III. Ac t iv e Is l a nding Det ec t io n
THD etc. The active IDTs based on adding feedback to It is observedthat thefrequency variation onpost islanding
the control parameters are implemented in [10], [11]. The isnot significant whenthegenerationmatchesthedemandand
detectionperformancesof activefrequencydrift, improvedslip hence most of the passive islanding detection techniques fail
modefrequency shift andimproved sandiafrequency shift are [18].
comparedfor multipleinvertersin [10]. Anewactiveislanding In order to make a significant variation in frequency of
detection methodology based on modified voltage positive the voltage at PCCon post islanding, adisturbance signal is
feedback is proposed in [11]. Nonetheless, these techniques injected through the q-axis controller, which assists to detect

978-1-5386-9316-2/18/$31.00 ©2018 IEEE


Fig. 1. Schematic diagramof thetest system

TABLE I frequency anddisturbance frequency respectively.


Test Syst emSpecif icat ions
= -tan 1( R (uiC---L ) ) = 1, 2 (3)
Parameter Value Parameter Value
1
fo 50 Hz Pwind 1MW Z0 1 (4)
Vrms 0.4 kV Qf 2.5
Vdc 0.8 kV R 160 mQ
fsw 7.9 kHz Load Component L 204 ¡dH R)d+ ( - - à f i 5
Rf 1m,Q C 49.74 mF i
Lf 410 /dH f dist 20 Hz Zi = T = 1, 2 (5)

'R y ^ - ¿ l ^ 2
21
island formation. The frequency of the voltage signal at PCC The block diagramof the q-axis current controller with
is obtainedby phase locked loop and it is analysedto detect disturbance injection is shown in Fig. 2.
and distinguish between islanding and non-islanding events.
The magnitude of the disturbance signal is considered as 1%
of ratedd-axis current at 20Hz frequency for this work [18].
The output current of the inverter in dq components
including the disturbance injection through q-axis controller
is given asfollows,
ldist
id idref (1)
iqref + idr sin(ud't) Fig. 2. Block diagram of the q-axis current controller with disturbance
. iq _ injection

The dq0current components areconvertedto abc current where, idist idr sin(udt)-
components by dq0 to abc transformation, and the instan-
taneous per phase voltage is measured and observed. The IV. Pr oposed Ana l ysing Met ho do l o g y
variation of frequency of the PCCvoltage dueto the current Due to the injected disturbance signal, the measured fre-
disturbance on the islanded systemdepends on the following quency of thePCCvoltage signal varies significantly onpost
equation: islanding. Inordertoconfirmwhetherthesystemisislandedor
not, aneffectiveanalysingtechniqueis essential. Inthis work,
va Z0idsin(u0t)+Z1idrsin(u1t+d1)+Z2idrsin(u2t+dd) ananalysingtechniqueis proposedbasedonthecombination
(2) of Band Pass Filter (second order filter) and computation
U0 2f ui =2 nfi, u = 2nf2 of Mean of Absolute Frequency Variation (AFVmean). The
corresponding equations aregiven as follows:
where, 2juds
/i andf2represent thesumanddifferenceof thefundamental H(s) s2+ 2&ds + ud (6)

2
Fig. 3. Block diagramof theproposed analysing methodology

where, s=Laplaceoperator, £=dampingratio=0.707, f d = intentional delay helps to discriminate the transients as non-
20 Hz (frequency of the disturbance signal which is injected islanding fromthe islanding condition.
through the q-axis controller). Bandwidth of the band pass
filter is given by the expression 2£fd. V. Case St udies
In this section, the effectiveness of theproposedmethodis
AFVmean(t) = 3 \AFV(t) | dt (7) for thaetete
valid d for various islanding and non-islanding conditions
st systemshown in Fig. 1.
T t-T
The deviation of frequency of the voltage signal onpost- A. IslandedCondition (Exactlypower matchedcondition)
islanding for exactly matched power condition is simulated Theproposedmethodistestedfor aconditionwheretheDG
for various load quality factors and are investigated. The isolateswithloadat exactlypower matchedconditionbetween
AFVmean is obtained by the simulations for exact power thegenerationandthedemand. Thegridisdisconnectedat 0.6
matchedconditions of DGgenerationandtheloadfor various s and the responses of various signals are monitored for the
loadqualityfactors. Thechangeinfrequencyonpost islanding exactly matchedpower condition and shown in Fig. 4.
condition is almost the same for any particular load quality
factor for any equally rated DG and load. The 95%of the
magnitude of AFVmean is chosen as threshold for safety (a)
50.2
purpose. Thethresholdcorrespondingtotheloadof Qf = 2.5
ischosenascommonfor all theloadQf valuesof < 2.5, since & 50.1
thevariation of frequency for theloads of Qf < 2.5 ishigher S'
thanthefrequencyvariationfor theloadof Qf = 2.5. It avoids g o*
50.0
the needof adaptivethreshold for varying demands for loads i 49.9
of Qf < 2.5. The AFVmean values and their thresholds for
various Qf values aretabulatedin Table II. 49.8
^ 0.08
TABLE II ^ - 0.06
Thr eshol d f or var ious qual it y fact or s s
g0.04
^> 0.02
Qf AFV Mean (Hz) Threshold (Hz)
2.5 0.0393 0.0373
1.8 0.0482 0.0458 0.00
1.2 (c)
1 0.0641 0.0609
0.5 0.0823 0.0782
f>
cz5
0.8
G
Theblock diagramof theproposedmethodis depictedin ■oS 0.4
Fig. 3. The intentional time delay of 90 ms is chosen based nu
Q 0.0
on simulations to avoid the mal-identification of transient
condition as anislanded event. 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1
The frequency of the voltage waveformon pre and post Time(s)
islanding condition varies based on eqn. (2). In eqn. (2),
only thetermZ0idsin(w0t) is effectiveduring grid connected Fig. 4. Responses of exactly matchedpower condition
operation. However, on post islanding condition the voltage
waveformvaries basedon eqn. (2), so as the frequency. The It is inferred fromthe Fig. 4, that during grid connected

3
mode of operation, the change in frequency due to the dis- (a)
turbance injection is unremarkable due to the presence of
low impedance path offered by the grid and only during
the islanded mode of operation the frequency variation is
significant. The islandformation is detectedwithin 180ms as
shown in Fig. 4 (c), which is much lesser than the detection
time mentioned in the IEEE standard 1547.4 of 2s.
B. Non-islanding Conditions
Inthis section, theproposedmethodisinvestigatedfor vari-
ousnon-islandingconditionssuchasgridfrequencyvariations,
different loadswitchingeventsandwindvelocity variationsto
verify the robustness towards variations on load, source and
grid side.
1) Gridfrequency variations: The distribution systems are
prone to abrupt frequency variations due to various reasons.
Inthis section, theefficacy of theproposedmethodis verified
for abrupt changes in grid frequency. The time instants of
the abrupt change in grid frequency of the voltage signal are
tabulatedinTableIII. Theresponsesaremonitoredandplotted
in Fig. 5.
0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6
TABLE III Time(s)
Time inst ant s of abr upt changes in gr id f r equency
Fig. 5. Responses of proposed analysing methodology for grid frequency
Grid frequency variations fluctuations
Time instant (s)
From(Hz) To (Hz)
50 49.3 0.7
49.3 50.5 1
50.5 50 1.3

Although, the computedAFVmean valueexceedsthresh-


old for the grid frequency variations at 0.7, 1.0 and 1.3
s, it falls below the threshold within 90 ms. Hence, this
is distinguished as a non-islanded event by the proposed
methodology.
2) Load switching events: Load switching events arepre-
dominant in the distributions systems and hence it becomes
inevitable to test the effectiveness of the proposedmethodol-
ogy for suchoperatingconditions. Thevarious loadswitching
events aredepictedin Table IV. Theresponses aremonitored
andplotted in Fig. 6.
TABLE IV
Var ious l oad swit ching event s
Load Type Power Rating Switched ON (s) SwitchedOFF(s)
Resistive 1MW 0.75 0.95
Inductive 1MVAR 1.15 1.35
Capacitive 1MVAR 1.55 1.75 0.7 0.8 0.9 1.0 1.1 12 1.3 1.4 1.5 1.6 1.7 1.8
Time(s)
Even though, the AFVmean is higher than threshold for Fig. 6. Responses of proposed analysing methodology for load switching
switching event at 1.55 s in Fig. 6, it doesn’t sustain for 90 events

4
ms. Hence, that is discriminated as anon-islanded condition condition. Theproposedtechniquedistinguishes theislanding
by theproposedmethod. andnon-islanding conditions accurately.
3) Uncertainties in wind velocity: The wind velocity is VI. Co nc l usio n
intermittent in nature. The change in wind velocity results in In this work, q-axis disturbance injection basedactive IDT
change in output power supplied to the grid and causes fre-
quency fluctuations at thePCC. Hence, theproposedmethod- is of
adoptedandananalysingmethodbasedonthecombination
band pass filter and computation of mean of absolute
ologyisverifiedfor suchanoperatingconditioninthissection.
The wind velocity is varied drastically from 12m/s to 7 m/s Theupero
freq ncy variation is proposed for PMSG based WECS.
at 0.7 and linearly from7 m/s to 11 m/s from 1.0 to 1.2 s detect thpeose dmethodovercomestheearlier shortcomings and
isla ndformation faster thantheearlier works based
respectively. Thechangeinpower andthefrequencyvariations
areplotted in Fig. 7. o n q-a xis cu rrent disturbanceinjection. Theproposedmethod
discriminates the islanding fromthenon-islanding conditions
accurately, does not unstablizethe systemon islanded condi-
tion, is suitablefor varyingdemandsfor theloadsof Qf < 2.5
and compatible for any converter based DGs. The island
detection time is within 180 ms for any load of Qf < 2.5.
The proposed method is extendible to multiple DGs at the
samepoint of common coupling without any modifications in
thecontrol circuitry. Anexperimental setupisbeingdeveloped
to validatetheproposedtechnique as anextension.
(b) Ref er enc es
[1] “IEEEguidefor design, operation, andintegrationof distributedresource
islandsystemswithelectricpower systems,”IEEEStd1547.4-2011, pp.
1-54, July 2011.
[2] S. Murugesan, V. Murali, andS. A. Daniel, “Hybridanalyzingtechnique
for active islanding detection based on d-axis current injection,” IEEE
Systems Journal, to be published.
[3] A. Khamis, H. Shareef, E. Bizkevelci, and T. Khatib, “A review
of islanding detection techniques for renewable distributed generation
systems,” Renewable and Sustainable Energy Reviews, vol. 28, no.
Supplement C, pp. 483 - 493, 2013.
[4] S. Murugesan andV. Murali, “Hybrid analysingtechniquebasedactive
islanding detection for multiple dgs,” IEEE Transactions on Industrial
Informatics, tobepublished.
[5] G. H. Gonzalez andR. Iravani, “Current injection for active islanding
detectionof electronically-interfaceddistributedresources,”IEEETrans-
actions on Power Delivery, vol. 21, no. 3, pp. 1698-1705, July 2006.
[6] G. Isazadeh, M. Kordi, F. Eghtedarnia, andR. Torkzadeh, “A newwide
areaintelligent multi-agent islanding detectionmethodfor implementa-
tionindesignedwampac structure,”EnergyProcedia, vol. 141, pp. 443
- 453, 2017, power andEnergy Systems Engineering.
[7] M. M. Ostojic and M. B. Djuric, “The algorithmwith synchronized
voltageinputsfor islandingdetectionof synchronous generators,”Inter-
Fig. 7. Responses of proposed analysing methodology for uncertainties in national Journal of Electrical Power andEnergySystems, vol. 103, pp.
velocity 431-439, 2018.
[8] H. Laaksonen, “Advanced islanding detection functionality for future
electricitydistributionnetworks,”IEEETransactionsonPower Delivery,
The results in Fig. 7 arethe evidences that theproposed vol. 28, no. 4, pp. 2056-2064, Oct 2013.
methodology easily discriminates thewind velocity variations [9] in H. Samet, F. Hashemi, andT. Ghanbari, “Islandingdetectionmethodfor
verter-baseddistributedgenerationwithnegligiblenon-detectionzone
fromthe islanding conditions. usingenergyof rateof changeof voltagephaseangle,”IETGeneration,
Transmission Distribution, vol. 9, no. 15, pp. 2337-2350, 2015.
C. Comparison with earlier works [10] F. Liu, Y. Zhang, M. Xue, X. Lin, andY. Kang, “Investigationandeval-
uation of active frequency drifting methods in multiple grid-connected
The technique suggested in [5] is not validated for non- inverters,”IETPowerElectronics, vol. 5, no. 4, pp. 485-492, April 2012.
islanding conditions. The authors in [14] have utilized the [11] fo A. Samui andS. R. Samantaray, “Newactiveislandingdetectionscheme
r constant power and constant current controlled inverter-based dis-
workin [5] anddevelopedananalysingtechniqueto discrimi- tributedgeneration,” IETGeneration, Transmission Distribution, vol. 7,
natetheislandingfromthenon-islandingconditions. However, no. 7, pp. 779-789, July 2013.
if thegridoperatingfrequencyexceeds±0.5 Hz, it mal-detects [12] D . Reigosa, F. Briz, C. B. Charro, P. Garcia, and J. M. Guerrero,
“Activeislandingdetectionusinghigh-frequency signal injection,”IEEE
thegridconnectedconditionasislandedcondition. Theisland Transactions on Industry Applications, vol. 48, no. 5, pp. 1588-1597,
detection time is > 200 ms for the techniques suggested in Sept 2012.
[15], [16]. In this work, the disadvantage in [5] is eliminated [13] in K. Jia, H. Wei, T. Bi, D. W. P. Thomas, and M. Sumner, “An island-
g detection method for multi-dg systems based on high-frequency
by implementing a band pass filter and the island detection impedance estimation,” IEEE Transactions on Sustainable Energy,
time is within 180 ms even for an exactly power matched vol. 8, no. 1, pp. 74-83, Jan2017.

5
[14] P. Gupta, R. S. Bhatia, and D. K. Jain, “Average absolute frequency
deviationvaluebasedactiveislandingdetectiontechnique,”IEEETrans-
actions onSmart Grid, vol. 6, no. 1, pp. 26-35, Jan2015.
[15] A. Emadi andH. Afrakhte, “A referencecurrent perturbationmethodfor
islanding detection of amulti-inverter system,” Electric Power Systems
Research, vol. 132, pp. 47 - 55, 2016.
[16] P. Gupta, R. S. Bhatia, andD. K. Jain, “Activerocof relayfor islanding
detection,” IEEE Transactions on Power Delivery, vol. 32, no. 1, pp.
420-429, Feb 2017.
[17] “An assessment of dg islanding detection methods and issues for
canada,” report # CETC-Varennes 2004-074 (TR), CANMET Energy
Technology Centre Varennes, Natural Resources Canada, 2004.
[18] S. Murugesan andV. Murali, “Q-axis current perturbationbased active
islandingdetectionfor converter interfaceddistributedgenerators,” Turk-
ishJournal oFElectrical Engineeringand Computer Sciences, vol. 26,
no. 5, pp. 2633-2647, 2018.

You might also like