简介:SeveralARMAmodelingapproachesareaddressed.Inthesemethodsonlypartofacorrelationsequenceisemployedforestimatingparameters.Itissatisfying,ifthegivencorrelationsequenceisofrealARMA,sinceanARMAprocesscanbecompletelydeterminedbypartofitscorrelationse-quence.Butforthecaseofameasuredcorrelationsequencethewholesequencemaybeusedtore-ducetheeffectoferroronmodelparameterestimation.Inaddition,thesemethodsnowdonotguar-anteeanonnegativespectralestimate.Inviewoftheabove-mentionedfact,aconstrainedleastsquaresfittingtechniqueisproposedwhichutilizesthewholemeasuredcorrelationsequenceandguar-anteesanonnegativespectralestimate.
简介:Stochasticadaptivecontrolisconsideredforthediscrete-timemulti-inputandmulti-outputsystemofmulti-delaywithnoiseexpressedbyanARMAprocess.TheCARIMAmodelisaspecialcaseofthesysteminquestion.Theoptimaladaptivecontrollawisgivenanditisshownthataquadraticcostfunctionisminimizedandtheclosed-loopsystemisstable.Further,whenthesystemisofminimumphase,theconvergenceratesofparameterestimatesandofthecost-functionarealsoderived.
简介:Atimeseriesx(t),t≥1,issaidtobeanunstableARMAprocessifx(t)satisfiesanunstableARMAmodelsuchasx(t)=a1x(t-1)+a2x(t-2)+…+a8x(t-s)+w(t)wherew(t)isastationaryARMAprocess;andthecharacteristicpolynomialA(z)=1-a1z-a2z2-…-a3z3hasallrootsontheunitcircle.Asymptoticbehaviorofsumform1ton(x2(t))willbestudiedbyshowingsomeratesofdivergenceofsumform1ton(x2(t)).ThiskindofpropertiesWillbeusedforgettingtheratesofconvergenceofleastsquaresestimatesofparametersa1,a2,…,a?
简介:Basedonthemulti-sensoroptimalinformationfusioncriterionweightedbymatricesinthelinearminimumvariancesense,usingwhitenoiseestimators,anoptimalfusiondistributedKalmansmootherisgivenfordiscretemulti-channelARMA(autoregressivemovingaverage)signals.Thesmoothingerrorcross-covariancematricesbetweenanytwosensorsaregivenformeasurementnoises.Furthermore,thefusionsmoothergiveshigherprecisionthananylocalsmootherdoes.
简介:StatisticalpropertiesofwindsneartheTaichungHarbourareinvestigated.The26years′incompletedataofwindspeeds,measuredonanhourlybasis,areusedasreference.ThepossibilityofimputationusingsimulatedresultsoftheAuto-Regressive(AR),Moving-Average(MA),and/orAuto-RegressiveandMoving-Average(ARMA)modelsisstudied.Predictionsofthe25-yearextremewindspeedsbasedupontheaugmenteddataarecomparedwiththeoriginalseries.Basedupontheresults,predictionsofthe50-and100-yearextremewindspeedsarethenmade.