详细介绍了模糊人工神经网络( fuzzy neural network)在电池SOC(State of charge)中的应用。

Kalman lteringstateofchargeestimationforbatterymanagementsystembasedonastochasticfuzzyneuralnetworkbatterymodel
XuLonga,WangJunpingb, ,ChenQuanshic
a
CollegeofChemicalEngineering,NorthwestUniversity,Xi’an710069,PRChina
SchoolofMechanicalEngineering,Xi’anJiaotongUniversity,Xi’an710049,PRChinac
StateKeyLaboratoryofAutomobileSafety&EnergyConservation,TsinghuaUniversity,Beijing100084,PRChina
b
articleinfoabstract
ExtendedKalman lteringisanintelligentandoptimalmeansforestimatingthestateofadynamicsys-tem.InordertouseextendedKalman lteringtoestimatethestateofcharge(SOC),werequireamath-ematicalmodelthatcanaccuratelycapturethedynamicsofbatterypack.Inthispaper,weproposeastochasticfuzzyneuralnetwork(SFNN)insteadofthetraditionalneuralnetworkthathas lteringeffectonnoisyinputtomodelthebatterynonlineardynamic.Then,thepaperstudiestheextendedKalman l-teringSOCestimationmethodbasedonaSFNNmodel.Themodelingtestisrealizedonan80AhNi/MHbatterypackandtheFederalUrbanDrivingSchedule(FUDS)cycleisusedtoverifytheSOCestimationmethod.ThemaximumSOCestimationerroris0.6%comparedwiththerealSOCobtainedfromthedis-chargingtest.
Ó2011ElsevierLtd.Allrightsreserved.
Articlehistory:
Received2June2010
Receivedinrevisedform16June2011Accepted22June2011
Availableonline21September2011Keywords:
ElectricvehiclesBatterypack
Stateofcharge(SOC)
StochasticfuzzyneuralnetworkExtendedKalman lter(EKF)
1.Introduction
Batteriesaremajorenergysourcesforelectricvehicles(EV)inwhichplayasigni cantrole.Toensurethebatteriesbereliableandcapableofdeliveringpowerandenergywhenrequired,anaccuratedeterminationofstateofcharge(SOC)isnecessary.Thus,theaccurateestimationoftheSOCofthebatterypackisthekeyfactorformanagingbatteriesef ciently.
Estimatingthebattery’sSOCiscomplicatedbythefactthattheSOCdependsonmanyfactorssuchastemperature,batterycapaci-tance,andinternalresistance.SomeSOCestimationapproachesarepresentedwiththedevelopmentofEVtechnology.Pilleretal.[1]describedthecommonlyusedmethodsforSOCdeterminationincludingchemistry-dependentmethods,open-circuitvoltage(OCV)measurements,electrochemicalmodeling,impedancespec-troscopy,andcircuitmodels.Thesemethodsusuallyrelyonthespecialconcretemodels,whichdifferindifferentbatterysystemsandareonlyutilizedforof inetest.Ahcountingapproachisthestraightforwardapplicationofthede nitionofSOCandisthesim-plestmethod.Thecurrententeringandleavingthebatteryismea-suredperiodically.Then,theSOCofthebatteryisupdatedbyaddingorsubtractingthelastperiod’snetcumulativecharge.Ifasuf cientlyaccuratecurrentsensorisused,thismethodisreason-ablyaccurateandinexpensivetoimplement.However,itsuffersCorrespondingauthor.Tel.:+862982668615;fax:+862982665331.
E-mailaddress: y_wang1110@sina.com(J.Wang).

0196-8904/$-seefrontmatterÓ2011ElsevierLtd.Allrightsreserved.doi:10.1016/j.enconman.2011.06.003
severaldrawbacks.Firstly,itisdif culttoestimatecoulombicef -ciencygaccurately,thusLinetal.[2]introducedtheequivalentcoulombicef ciency.Secondly,thisapproachitselfisnotabletodeterminetheinitialSOCdenotedasSOC0andsolvethevariationofSOC0resultedfromself-dischargeandotherfactors.Thirdly,theerrorislargerwhenthebatteryworksatahighandlowtemper-atureorwhenthecurrent uctuatesdramatically.Finally,themethodisanopenloopSOCestimator,anderrorsinthecurrentdetectorareaccumulatedbytheestimator.Thelongertheestima-torisoperated,thelargerthecumulativeerrorbecomes.Shen[3]proposedanarti cialneuralnetworkestimationmethodthatcouldpredictSOCforallbatterysystemswithgreataccuracywithoutknowledgeofcellinternalstructure.Kalman ltersareanintelli-gentandsometimesoptimalmeansforestimatingthepresentva-lueofthetime-varyingstateofadynamicsystem.Chengetal.[4]designedathree-layerfeed-forwardnetworkandanenhancedim-munealgorithmandtoestimateSOC.Plett[5–7]appliedtheKal-man ltertheorytoestimatetheSOCbymodelingbatterysystemtoincludethewantedunknownquantitiesinitsstatedescription.Anadditionalbene toftheKalman lteristhatitautomaticallyprovidesdynamicestimationerrorboundsontheestimationaswell.Theyexploitthisfacttogiveaggressiveperformancefrombat-terypack,withoutfearofcausingdamagebyover-chargeorover-discharge.Researchershaverealizedthatintegrationofseveralindividualapproachesmaygiverisetobetterperformance.Plett[7]proposedanewcombinedSOCestimationmethodinwhichanextendedKalman lterisusedtoestimatetheopen-circuit
Kalman filtering soc e... 暂无评价 7页 1下载...An SOC estimation model based on a radial basis... S. Recursive Bayesian Inference on Stochastic ...
State of Charge Estimation Based on a Combined ...Keywords- battery SOC; Kalman filtering; neural ...are zero-mean white Gaussian stochastic capacity. ...
SOC Estimation Method based on the Reduced Order Extended Kalman Filtering...such as neural network, fuzzy logic, adaptive observer and extended Kalman ...
Estimation Strategies Based On Equivalent Circuit_...The Kalman filtering algorithm is divided into two...24. Haykin S., Kalman Filtering and Neural ...
We developed a SOC estimation algorithm based on ...Computational intelligence techniques such as neural ...Keywords- battery SOC; Kalman filtering; ...
cial neural networks, fuzzy logic–based models, ...(EKF) into BMS to implement SOC estimation of ...The Unscented Kalman Filter, Kalman Filtering and,...
Kalman Filter (EKF) technique is then adopted as the SOC estimation ...cial neural networks based models [5e7], fuzzy logic models [8,9] and ...
Kalman Filter (EKF) technique is then adopted as the SOC estimation ...cial neural networks based models [5e7], fuzzy logic models [8,9] and ...
热门文档