Seite - 144 - in Emerging Technologies for Electric and Hybrid Vehicles

Energies 2017,10, 1503 ampere-hour (Ah)methods,aswellas theopen-circuitvoltage (OCV)andimpedancemeasurement methods, giveamore intuitiveandreliable estimation [19,20]. However, this approach isprone to errors: errorsrelatedto initialSOCdeterminationandtheaccumulativeerrors fromsensorsduring themeasurement of current and time. Secondly, themachine learning-based estimationmethods (alsocalleddata-drivenapproaches), suchas theartificialneuralnetwork–fuzzylogic (FL) [16]and support vectormachine [17]methods. However,machine learning-basedmethods require ahigh computationaleffortdueto large trainingdatasets for trainingthemodel, althoughtheyconsider the nonlinearitiesof thebatterymodel. Inaddition,mostmachine learning-basedSOCestimationmodels wereestablishedoffline. Theyarenotsuitable forapracticalandlow-costembeddedbattery-power systemapplication.Lastly, thestate-spacemodel-basedestimationmethods(suchasusingtheextended Kalmanfilter(EKF))reducetheconvergenttimebutincreasethecomputational loadoftheBMS[21,22]. Theaforementionedliteraturehas itsowndisadvantagesandadvantages.However,what isunusual is thatmostSOCimplementation isonlymeant forasinglebatterycell. It isnotuseful for researchers whowould like to implement itonmulti-cellbatteries foractualapplications. Due to the characteristics ofLIBs, the faults of thebattery-power systemmay lead to serious safety issues, suchascatchingfireandexplosion. For instance,a lithium–cobaltoxidebatterybackup powersystemcaughtfire inaBoeing787of JapanAirlines in2013 [23].Hence, thecapabilitiesof fault diagnosis andprotectionare important andnecessary in abattery-power system. Fault-diagnosis technology is an interdisciplinaryfield that combines control theory, computernetwork,database, artificial intelligence and other technologies. In the past fewyears, thereweremany researchers focusedonbattery-powersystems. Bohlenetal. [24] investigatedtheinternal-resistancefaultdiagnosis ofbatteriesbyamodel-based identificationmethod.D.P.Abrahametal. [25]provedthat thechanges of battery electrodes are the cause of the sudden increase in the battery internal resistance and the power degradationmechanism in the power battery pack. X.J. Liu [26] tried to diagnose the battery faultsbyusing the fuzzy-logicmethod. Although the research involveddifferentmethods ormodels that produced good results, most of the authorswere focusing on overall theoretical aspectsof faultdiagnosisusingnonlinearmodel-basedor intelligentapproaches, suchas fuzzy-logic andneural networkmethods, to determine the battery faults. However, these required a higher computational timeand further resources toperform the fault diagnosis. Therefore, in this study, oneof theresearchobjectives is touseacomputational, inexpensiveandintuitiveapproachtodetect and diagnose the faults in the battery and provide corresponding remedy actions for the faults. Inaddition, theself-recoveryschemeisalsoproposed. Firstly, four typesofcritical faults [9,10,26–32], beingtheover-chargedfault,over-dischargedfault,over-current faultandexternal short-circuit fault, are considered. Secondly, different fault diagnosis algorithmsare studied, and the corresponding solutionsareproposed. Lastly, theproposedfaultdiagnosisandself-recoveryschemesareappliedto the12-cellbatterypackprototypeandvalidatedexperimentally. Insummary,astructureofasmartmulti-cellbattery-powersystemisproposedto improvethe safety and its operational intelligence. Hence, the smart battery-power systemhas the following features: (1) compatibilityandflexibilitywithdifferentkindsofLIBsandbatterypackconfigurations; (2) capability for SOCself-initialization and self-adjustment; (3) capability for fault diagnosis and self-recovery; and (4) ability to provide a human–machine interface for status report and system configuration, locally or in the cloud. A fewdesignedmodules, such as batterydata acquisition, batterypackSOCestimation, faultdiagnosis, data communication,user-interfacemodule fordata displayandsystemconfiguration,anddual-pathswitchingforcharginganddischarging,areproposed. In theproposedsmartbattery-powersystem, theSOCself-initializationschemecoupledwith fault diagnosisandtheself-recoveryalgorithmwere investigatedandimplemented ina12-cell series (12S) battery-packprototype. Theexperimental resultsshowthat thesystemcandiagnose thefaultsand carryout thecorrespondingprotectionandrecoveryactions. Inaddition, theproposedbatterypack hasbeenshowntoestimate theSOCsuccessfullyunder theactual loadprofile fromanelectricvehicle. 144