Certified Journal | Page630
FAULT DETECTION SCHEME OF 5 BUS BY ANN AND ANFIS
Due to progress in Distributed Generation (DG) development [1], microgrids are gathering attention from industry and researchcommunity.Thankstoimprovedpowerefficiency,reliabilityandquality, thesewillbringadvantagestomodern power system control. Microgrids can be operated on either grid connected or in the Icelandic mode if the external grid suffersfromdisruptionslikedeviationsoffrequencyandvoltagefluctuations.Criticalloadsin microgridscanbesupplied without the external grid in island mode by load side DGs. In the meantime, microgrid protection is one of the main and critical tasks[1] [3]. With the progressive use in modern power systems of renewable energy sources, micro grids are commonly integrated with interface DGs, such as PVDG and battery energy storage systems. inter face DGs are also commonly integrated (BESS). Large fault currents depend on traditional protective relays for distribution system fault detection.However,onlynegligiblefaultcurrentscancontributetoIIDG'sthatprotectschemesarenotenabled[4].These relayscanthereforenotprotectmicrogrids [5]providesadetailedanalysisofthedynamicsofcurrentandvoltageinsuch microgrids.Microgridfailuredetectionusuallyhasthreegoals.Ifthesystemhasafault,thetypeoffault(e.g.single phase to ground,three phase short,etc.)shouldbedetermined,andthefaultphaseshouldbedetermined.Thefirsttwoenable isolation operations for fault subsequently, and the latter can be restored to service. According to [5], the operation of soundstagesinunbalancedshortcircuiterrorsinordertosupporttheintegrationofone stageprotectiondevicesshould continue with modern microgrids Selective phase tripping can be achieved given accurate fault type and fault phase information[6]. As a result, the reliability of sys tems is improving significantly [7], and this protection system is being gradually adopted by utilities [8]. In addition, accurate detection of the failure location can considerably reduce service restorationefforts[9],whichisincreasinglyimportantifundergroundoperationsaretoberestored.
Student, Prof Ram Meghe College of Engineering and Management, Badnera 2Asst.Prof. Prof Ram Meghe College of Engineering and Management, Badnera ***
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056
ABSTRACT: In micro grid control and operation, failure detection is essential because it makes the system quick to isolate and recover from the fault. Due to the use of inverter interfaced distributed generation in microgrids, standard fault detection systems are no longer appropriate because they rely on large fault currents. We propose an intelligent microgrid fault detection technique based on wavelet transform neural networks and ANFIS in this paper. The scheme uses a discrete wavelet to pre process branch current measurements sampled via VI Bus measurement for extracts statistical features. Then all available data are entered into deep neural networks to generate false data. The proposed scheme can offer significantly improved fault type classification precision compared to previous work. In addition, the system can detect fault locations that are not available in previous work. A comprehensive evaluation study on the CERTS microgrid and the IEEE 5 bus system is undertaken to evaluate the performance of the proposed failure detection scheme. The results of the simulation demonstrate the effectiveness of the proposed scheme in terms of detection accuracy, computer speed and measurement uncertainty robustness.
Index Terms Fault detection, fault location, microgrid protection, wavelet transform, neural network.
Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008
Mr. Payal Ingole1, Assst .Prof. A. V. Mohod2 1
INTRODUCTION:
Data and digital signal processing methods for microgrid fault type or phase detection are used in recent years by a growing number of researchers. For example, random tree and random forest are often used to detect faults in both the networked and insulated microgrid systems (see example for [4], [10] [12]). For the detection of defects in the literature[3] [12] also have been used other machine learning techniques e.g. vector support and algorithms of neighbouring k nearest. Due to these data driven approaches' high calculation speed, satisfactory fault classification can be developed nearly in real time. Furthermore, in order to better exploit the time frequence properties for analyzes [3],[4],[12],digitalsignalsprocessingapproachessuchasdiscreteFouriertransformanddiscreteWavelet(DWT)areused for"pre processing"inputsignals.Forinvestigationsofmicrogridfaultdetectionplansintheliterature,interestedreaders mayreferto[1]and[4].Inthedevelopmentofmicrogridfaultdetectionprogrammes,however,thereremainsaresearch gap.Someexistingresearchcannotprovideinformationaboutthefaulttypeandcannotthereforebeproperlyadoptedin the trip paradigm of the single phase (see [12], [13] for examples). In addition, current work on detecting fault locations focuses on microgrids with low voltage CDs, e.g. [14], [15]. The location fault can normally be identified by travelling or injection driven algorithms in AC distribution network [9], [16]. Traveling wave algorithms, however, are affected by reflected wave detection problems and problems of discrimination[14],[17], some of which require synchronised communicationlinkdata[18]Nofaultlocatingperformanceonislandedmicrogridsandloop/ringtopologynetworkswas demonstratedbyeither.Meanwhile,phase to groundfaultsarethesoleapplicableinjection basedalgorithmsintheradial
ArtificialMethodology:Neural Network
We conduct extensive case studies to analyse the performance of the mechanism proposed and compare the resultswiththelatesttechnology.
ThestructureofanANNmustbesimpleandefficientforsolvingtheproblemtoimplementanANNinachip.TheANNis carriedoutbecauseithasthecapacitytolearn.Theworkspresentedin[2],[8],[10],[11],[16],[17],[19]cantherefore be avoided in their demerits. The ANN used in the paper is a three faceted network with seven inputs [28]: difference betweenthetwo andthree phase VA−B,VB−C,VA−C , current (IA 00B, IB−C, IA−C , he neutral current entropy wavelet; (Ig). The ANN has a cached layer and a layer of output.Fig.4displaysthestructureoftheneuralnetwork.Theneuralnetworkisabletoidentify11types(single phaseto ground,unbalancedtwo phase,andbalancedfaults)offaultsoccurringatthegridside,ZoneA,orZoneB.AsshowninFig. 4,thereareoneneuronattheoutputlayer:tolocalize,classifyandidentifythefaultinthispaper3ANNisconnect
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056
The proposed detection results can be accurate and quick (type, phase, and location) without communications compared with existing defect detection mechanisms. It can be adapted to various operating modes, network topologiesandradialeandlooptopology,forexamplegrid connected,isolatedmode.Powerdynamicscanalsobe handledwithbothconventionalandIIDGsynchronousgenerators.DiscreteWaveletTransform(DWT). The FT does not provide direct data for a signal oscillating and is suitable for investigating the problems of the constant state. STFT divides the full time interval into several small/equal intervals, which will be analysed separately by FT. However,itisineffectivetodetectveryshortandhighfrequencysignalsviaSTFT.Duetoitsdiversewindowfunctionfor the time domain, the WT has been used widely to analyse passing signals.
T WT FT STFT T DWT WT T DWT T DWT anddjfilterbanks(k).Alow passfilterHanda high passfilterg[25]decomposethesignal.Afactorof2Aisthensampledforthefilteredsignal.
WeproposeanIEEE5BUSsystemfailuredetectionmechanismtoprovideaccurateandexpeditedinformationaboutthe defecttype,phaseandlocation; • ANNandANFISarecombinedwithDWTinordertoresolvethedata drivenmicrogridfailuredetectionproblem;
Fig:14levelThree scalesignaldecomposition
Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page631 networks[14]. Thisarticle presentsa smart failure detection systembasedonDWTandthe recentdevelopmentof deep neuralnetworks(DNN),aclassofmachinelearningtechniquesdrivenbydatatoovercomeafaultdetectiongapforIIDG enabledmicrogrids.WhereasDWTislikelytocausenoiseandpowerdisturbances,DNNisintroducedwiththeexceptional abilityofDNNtohandlenoisedata[19][20]toimproveitsrobustness Theschemeoffaultdetectionschemeispresented in Fig. 1. Fig. 1. The system uses current magnitudes of three phases in a cycle, which are sampled as input data via protectiverelays.DWTprocessesthemeasurementstoremovethetime frequencydomainfunctionalities.Inaddition,the measurements will then be included in three DNNs for classification of fault type, identification of fault phase and detectionoffaultlocations.Finally,informationaboutthefaultisdeveloped,whichcanbeusedinsubsequentprotection andremedialcontrolmeasures.Thisreportsummarizesthecontributionsof:
Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 2022, IRJET Impact Factor value: 7.529 | 9001:2008
©
Where, the membership functions Ai, Bi, i = 1, 2, are specified in some parametric way from a family of membership functions, such as triangular or Gaussian. Layer 2 consists of fuzzy neurons with an aggregation operator being some t norm. We use the product t norm in this example, in view of the way that product is used in Sugeno T ’ procedure.TheoutputofLayer2is (O21,O22)=(A1(x1)B1(x2),A2(x1)B2(x2)) Layer3isanormalize.TheoutputofLayer3is
Certified Journal | Page632 ANN1HAS10INPUTAND10OUTPUTS ANN2HAS10INPUTAND5OUTPUTS, ANN3HAS5INPUTAND2OUTPUTS
Adaptive network fuzzy inference systems
Tofixtheideas,considertheproblem ofgraphicallyrepresentingthewayfuzzycontrol isachievedintheSugeno Takagi model.Forasimpleexample,considerafuzzyrulebaseconsistingofonlytworules:.
|
ISO
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056
Thus, the studied faulty locations are only three and are given in Fig. 2. The architecture of the ANN (e.g., the number of hiddenlayers)isdecidedgenerallybyheuristics[28].Morehiddenlayerstendtoattainagoodfittingbutaremostlikelyto beoverfitting.ThisimpliesthattheANNfitstoowelltothetrainingsetbutdoesnotperformwellonthetestingset.
To illustrate the use of neural networks for fuzzy inference, we present some successful adaptive neural network fuzzy inferencesystems,alongwithtrainingalgorithmsknownasANFIS.Thesestructures,alsoknownasadaptiveneuro fuzzy inference systems or adaptive network fuzzy inference systems, were proposed by Jang [35]. It should be noted that similarstructureswerealsoproposedindependentlybyLinandLee[40]andWangandMendel[77].Thesestructuresare usefulforcontrolandformanyotherapplications.
Where,AiandBiarefuzzysetsand f1(x)=z11x1+z12x2+z13 f2(x)=z21x1+z22x2+z23 Recallthatwhennumericalinputx=(x1,x2)ispresented,theinferencemechanismwillproducethenumericaloutput Afuzzy neuralnetworkforimplementingtheaboveisshowninfigure3.2.Theobservedinputx=(x1,x2)ispresentedto Layer1byInputLayer0.TheoutputofLayer1is (O11,O12,O13,O14)=(A1(x1),A2(x1),B1(x2),B2(x2))
R1:Ifx1isA1andx2isB1theny=f1(x)
R2:Ifx1isA2andx2isB2theny=f2(x)
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page633 Figure2:First orderSugenofuzzymodelwithtworules ThefuzzyneuronsinLayer4outputthevalues Finally,theoutputlayercalculatesthecontrolactionbysumming: Of course, the above neural network type for representing the inference procedure for a rule base of two rules can be extendedinanobviouswaytoanarbitrarynumberofrules. Fig.3FullyconnectedANFIS1structure.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page634 Proposed workFig4.:IEEE5BUSsystemmodel fig5.wavelet basedANNandANFISmodel Fig6.waveletsubsystemmodel fig.7ANFISsubsytemmodel
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page635
InthisprojectIEEE5bussystemissimulatedinmatlab2016,foranlysisoffaultwavelet basedANNandANFISisapplied on model. Fig.4 shows IEEE5 bus system. Fig.5 shows complete model with wavelet subsystem model, ANN0 model and ANFIS model. In fig. 6 wavelet subsystem is connected continues waveform of Vabc and Iabc is converted into discrete waveform and after that DWT (discrete wavelet transform) is applied. But after DWT waveform is converted into continuessignaltocalculatetheresultfromDWT
Fig8.ANNsubsystemmodel TABLE1DATASETFORANNANDANFIS
NOEMALRESULT: CONDITON: FIG.9VOLTAGEVabcANDCURENTIabcwaeform Fig.10waveletDWTofvoltageandcurrent
Certified Journal | Page636
©
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056
|
Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 2022, IRJET Impact Factor value: 7.529 ISO 9001:2008
|
ForgivingasteadyconstantresulttoANNandANFIS,DWTisconvertedenergysignal.Fig.7showsANFISmodelinthat3 ANFISisapplied.Onwhich1st ANFISisforidentifylocationoffault,2nd isforidentifythefaultonsystemand3rd ANFISis fordiagnosedfaultornormalcondition.InsamewayANNisconnectedwhichisshowinfig.8
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page637 Fig.11 energywaveformofcurrentandvoltage Fault condition Fig.12 VabcandIabcatzone1LGfault
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056 Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page638 FIG.13ANNSHOWSLGFAULTANDFAULTATBUS5 FIG.14ANFISSHOWLLGfaultandonbus5.
CONCLUSION:
[15]N G ,Y M Y , A U ,“E VSC DC ,”IEEE T I E ,” 65, 6, 4799 4809,Jun.2018.
International Research Journal of Engineering and Technology (IRJET) e ISSN: 2395 0056
References: [1]S P z,H L ,A K , S B ,“S of the :A ,”IEEEA , vol.3,pp.890 925,2015.
[2] Y. Y. Ho , Y H W , Y R C , Y D L , P L , “F usingwavelettransformandadaptivenetwork zz ,”E , 7, 4, 2658 2675,2014. [3]S.KarandS.R.S ,“H ,” P 6 I C E P E S ,P ,F ,2016, 258 263. [4] B. K. Panigrahi, P. K. Ray, P. K. R , S K S , “D ,”P I C C ,P C T ,K ,I ,2017 [5] T C , W C , E G C z , “U ,” Proc.Grid InteropForum,Irving,TX,USA,2012.
fig 9 to fig. 14 shows result waveform of fault condition and normal condition. From the result we get 51 different cases viz.LGFAULT(3TYPES),LLFAULT(3TYPES),LLGFAULT(3TYPE),LLLLG/LLLLFAULTANDNORMAL CONDITION.All thisfaultis taken on 5 buses soit will betotal 51 cases.All these51cases are experimented on IEEE5BUS system with discretebasedANNandANFIS.
[7] M M , Y T , L C , Y Z , P J , “A ,” P IEEEPESI S G T ,T j ,C ,2012 [8] M. Mishra P K R , “D grid faults based on HHT and machine learning ,”IETG T D , 12, 2, 388 397,2018.
Theproposed methodisbasedon wavelettransformanda newclassifier namedasArtificial Neural Network (ANN)and wavelettransform.Theproposedmethodisimplementedona5busIEEEbusgridinMATLAB/SIMULINKsoftware. The results show the high accuracy of fault detection of the proposed method. In this paper, 5 bus is considered, in which differentfaultistakenandthisfaultislocalized,identifiedbythewaveltanalysis,ANNtechniqueandANFIStechnic.This methodimplementedforclassifyordetectionoffaultinIEEE14bussystem.Inthispaperwewereclassifythefaultusing ANNandANFISclassifier.ItisobservedthatANNclassifytheeventofislanding upto90%whileANFISclassifytheevent ofFAULTupto74.3%. HenceitisclearthatANNprovidebestclassificationresultsfordetectionofFAULTnumberevent.
[6]D P M ,S R S , G J ,“A miningbasedintelligentprotectionscheme ,”IEEET S G , 7, 5,pp.2295 2304,Sep.2016.
[9] B. Kroposki , “D speed static switch for dis ,” Proc.PowerConvers.Conf.,Nagoya,Japan,Apr.2 5,2007,pp.1418 1423.
[10] Y Y H , J L G , F Y H , “D z roller for static switch in microgrid using wavelet TSK ,”IEEET I I , 14, 11, 4864 4872,Nov.2018.
Volume: 09 Issue: 03 | Mar 2022 www.irjet.net p-ISSN: 2395-0072 © 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page639
[11]J JQ Y ,Y H ,AYS L , VO K L,“I let baseddeep ,”IEEET S G , 10, 2, 1694 1703,Mar.2019. [12] M A J , H S , T G , “F ,” IEEE Syst.J.,vol.13,no.2,pp.1725 1736,Apr.2018.
[13] Z Y J , Z W L, N Q W , M C Z , “A P proach to fault diagnosis and restoration for power ,” IEEE S J, 12, 3, 2566 2576, Sep. 2018. [14] R D , S R S , B K P , V G V , “K area back up protection and faulted line identification for series ,” IEEE S J, 12, 3, 2634 2644,Sep.2018.