1. RIUdeG
  2. Tesis
  3. Doctorado
  4. CUCEI
Please use this identifier to cite or link to this item: https://hdl.handle.net/20.500.12104/90884
Full metadata record
DC FieldValueLanguage
dc.contributor.authorSandoval Guzmán, Betsy
dc.date.accessioned2022-09-12T22:10:06Z-
dc.date.available2022-09-12T22:10:06Z-
dc.date.issued2022-04-07
dc.identifier.urihttps://wdg.biblio.udg.mx
dc.identifier.urihttps://hdl.handle.net/20.500.12104/90884-
dc.description.abstractThis thesisexplorestheadvantagesoforganizingpowersystemsdataintensorar- raysandusingtensordecompositionstoprocesstheinformationoverthetraditional wayin2Darrays(matrices).Todoso,thisresearchworkreformulatessomewell- knownpowersystemproblems,suchasdataclustering,datavisualizationanddata compression forsmartmeterandPMUdata,ScreeningContingencyandCoherence Identication.This,undertheassumptionthatmanymeasureddatainthepower system canhaveamultidimensionalrepresentationintoatensor,whichcananal- ysed usingmethodsbasedontensordecomposition.Todemonstratetheadvantages of workingwiththesemethods;alongthisthesisfourdataminingmethodologies are proposedtounderstandingofthepowersystemdatacollectedbythemoderns monitoring systems: 1)An unsuperviseddataminingalgorithmisproposedforhomogeneoussmartgrid data, particularlyforelectricalloadprolesusingParallelFactorAnalysis(PARAFAC) tensor decomposition.Sincetheproposedtensorrepresentationallowstoassigna givendimensiontoaparticularvariableinvolved;datareduction,datacompression, data visualizationanddataclusteringarearchivedseparatelyforeveryvariable. 2)Data CompressionofPowerSystemsInformationfromAdvancedMonitoringIn- frastructure suchasPhasorMeasurementUnits(PMUs)andtheSmartMeter(SMs) is proposedusingTuckertensordecompositiontoachieveahighcompressionratio and alowreconstructionerror. 3)Screening ContingencyandCoherencyIdenticationarefacedsimultaneously ii based onPARAFAC-2tensordecomposition.Todoso,thetemporalandgeographical information isextractedfromatensorarraywithmultiplescontingencies. 4)A hybridclusteringalgorithmforelectricalloadprolesconsideringweather variablesisproposed.Theproposedalgorithmisbasedontheideaofunifyingcost functions resultingfromthedimensionalityreductionformulationofmultidimensional arrays. The resultsinthefourproposedmethodologiesshowthatthereformulationof the problemsyieldsmoreinformationabouttheeventstudiedthanwithtraditional matrix-based approaches.Thusdemonstratingthattensordecompositionisatool with greatpotentialfordataanalysisinelectricpowersystems.
dc.description.tableofcontents1 Introduction 1 1.1 Motivation . ................................ 1 1.2 Objectives . ................................ 4 1.3 Hypothesis . ................................ 5 1.4 StructureoftheThesis . ......................... 5 2 TheoryofTensorDecomposition 7 2.1 Introduction . ............................... 7 2.1.1 NomenclatureandDenitionofTensor . ............ 8 2.1.2 TwoDimensionalDataRepresentation(MatrixRepresentation) 9 2.1.3 MultidimensionalDataRepresentation(TensorRepresentation) 9 2.2 MathematicalBackground . ....................... 10 2.2.1 Kroneckerproduct . ....................... 10 2.2.2 Khatri-Raoproduct . ....................... 11 2.2.3 The n-mode product:Tensormultiplication . ......... 11 2.2.4 UnfoldingProcess . ........................ 11 2.2.5 Rank-onetensor . ......................... 11 2.2.6 Sub-arraysofa3DTensor:FibersandSlices . ......... 12 2.2.6.1 Fibers . ......................... 12 2.2.6.2 Slices . ......................... 13 2.3 SelectedTensorDecompositions . .................... 13 2.3.1 PARAFACTensorDecomposition . ............... 13 2.3.1.1 AnalysisofthePARAFACtensordecompositionand the \PrincipleofParallelProportionalProles . .. 15 2.3.2 TuckerTensorDecomposition . ................. 16 2.3.2.1 AnalysisoftheTuckertensordecompositionandthe core tensor . ...................... 18 2.4 PARAFAC-2Decomposition . ...................... 19 2.5 ConclusionsofChapter . ......................... 20 3 AnUnsupervisedDataMiningApproachforHomogeneousElectri- cal LoadProles 21 3.1 Introduction . ............................... 21 3.2 StateoftheArt . ............................. 23 3.3 InterpretationofPARAFACtensordecompositionforSmartMeterdata 25 3.3.1 AnalysisofslicesandbersasresultofPARAFACmodelde- composition . ........................... 26 3.4 AddedvalueofPARAFACinElectricalPowerSystems . ....... 27 3.5 MainResults . .............................. 28 3.5.1 TensordesignfortheERCOTsystem . ............. 29 3.5.2 PARAFACTensordecompositionoftheERCOTsystem . .. 30 3.5.2.1 DataCompression . .................. 30 3.5.2.2 Datavisualizationanddataclustering . ....... 31 3.5.3 DataReconstruction:Missingdata . .............. 35 3.6 AnalysisofComputationalComplexity . ................ 37 3.7 Conclusion . ................................ 39 4 DataCompressionforAdvanceSensingandCommunicationTech- nology inSmartGrids 40 4.1 Introduction . ............................... 40 4.2 StateoftheArt . ............................. 42 4.2.1 CompressionmethodsforPMUdata . ............. 43 4.2.2 CompressionMethodsforSmartMeterdata . ......... 45 4.3 DatacompressionbasedonTuckerdecomposition . .......... 46 4.3.1 ComputingTuckertohandlemissingdatausingHigher-Order Orthogonal Iteration(HOOI) . ................. 48 4.4 ApplicationsusingSmartGridData . ................. 49 4.4.1 CaseStudy1:datacompressionofPMUmeasurments . ... 50 4.4.1.1 PMUdatacompressionbasedonTuckerdecomposi- tion . .......................... 53 4.4.1.2 PMUdatacompressionbasedonSVD . ....... 55 4.4.2 CaseStudy2:datacompressionofSmartMeterdata . .... 56 4.4.2.1 SMdatacompressionbasedonTuckerdecomposition 57 4.4.2.2 SMdatacompressionbasedonSVD . ........ 59 4.5 ConclusionsoftheChapter . ....................... 61 5 ScreeningContingencyandCoherencyIdenticationinPowerSys- tems 62 5.1 Introduction . ............................... 62 5.2 TwoDimensionalDynamicFeatureExtraction . ............ 64 5.3 MultidimensionalDynamicFeatureExtraction . ............ 65 5.3.1 MultidimensionalRepresentation(TensorRepresentation) . . 65 5.3.2 Extractionofthespatialandtemporalinformationfromthe proposedtensor . ........................ 65 5.3.3 SeverityIndexbasedontemporalinformation . ........ 67 5.4 NumericalResults . ............................ 67 5.4.1 TestCase . ............................ 68 5.4.2 ContingencyScreening . ..................... 69 5.4.3 IdenticationofCoherentAreas . ................ 71 5.5 Discussion . ................................ 72 5.6 ConclusionsChapter . .......................... 73 6 ClusteringofCoupledArraysonPowerSystemsUsingaHybrid DecompositionApproach 75 6.1 Introduction . ............................... 75 6.1.1 Summaryofcontributions . ................... 76 6.2 ClusteringofIndividualArrays . .................... 77 6.2.1 Clusteringofindividualvariablesstoredinmatrixarrays:In- trinsic approach . ......................... 77 6.2.2 Clusteringofcombinedvariablesstoredintoatensor:Extrinsic approach . ............................. 79 6.3 ClusteringofCoupledArrayswithCombinedVariables:HybridAp- proach . .................................. 80 6.3.1 Reconstructionofvariablesafterdecomposition . ....... 82 6.4 IllustrativeExampleUsingSyntheticData . .............. 82 6.5 DemonstrationUsingSmartGridData . ................ 85 6.5.1 Testcaseandconstructionofcoupledarrays . ......... 86 6.5.2 Mainresults . ........................... 86 6.5.3 Comparisonusingclusteringofindividualarraysanddetection of Atypicaldays . ......................... 89 6.5.3.1 Comparisonusingonlyinternalvariables . ...... 89 6.5.3.2 Comparisonusingonlyexternalvariables . ...... 90 6.5.3.3 Detectionofatypicaldays . .............. 92 6.6 ConclusionsoftheChapter . ....................... 93 7 ConclusionsoftheThesis 94 7.1 GeneralConclusions . .......................... 94 7.2 GeneralKeyContributions . ...................... 94 7.3 FutureWork . ............................... 96 A Appendix 97 A.1 SingularValueDecomposition(SVD) . ................. 97 A.2 PrincipalComponentAnalysis(PCA) . ................. 98
dc.formatapplication/PDF
dc.language.isospa
dc.publisherBiblioteca Digital wdg.biblio
dc.publisherUniversidad de Guadalajara
dc.rights.urihttps://www.riudg.udg.mx/info/politicas.jsp
dc.titleANÁLISIS DE DATOS EN SISTEMAS ELÉCTRICOS DE POTENCIA MEDIANTE TÉCNICAS BASADAS EN MÉTODOS DE DESCOMPOSICIÓN TENSORIAL
dc.typeTesis de Doctorado
dc.rights.holderUniversidad de Guadalajara
dc.rights.holderSandoval Guzmán, Betsy
dc.coverageGUADALAJARA, JALISCO
dc.type.conacytdoctoralThesis
dc.degree.nameDOCTORADO EN CIENCIAS DE LA ELECTRONICA Y LA COMPUTACION CON ORIENTACIONES
dc.degree.departmentCUCEI
dc.degree.grantorUniversidad de Guadalajara
dc.rights.accessopenAccess
dc.degree.creatorDOCTOR EN CIENCIAS DE LA ELECTRONICA Y LA COMPUTACION CON ORIENTACIONES
dc.contributor.directorBarocio Espejo, Emilio
Appears in Collections:CUCEI

Files in This Item:
File SizeFormat 
DCUCEI10078FT.pdf5.28 MBAdobe PDFView/Open
Show simple item record


Items in RIUdeG are protected by copyright, with all rights reserved, unless otherwise indicated.