Seite - 171 - in Emerging Technologies for Electric and Hybrid Vehicles

Energies 2017,10, 5 6R& K K K K 6R& K K K K (a) (b) Figure10.Timeconstantestimationresultswithdifferent lengthsof theexperimentaldataset: (a)τshort; (b)τlong. Afterdeterminingthe lengthof thefittedexperimentaldataset,wecansubsequentlyobtain the resistances. Figure 11 shows theRlong estimation results by the conventional fitting function and the improvedfitting function. It canbe concluded fromFigure 11 that theRlong estimatedby the conventionalfitting function isgenerally less thantheoneestimatedbythe improvedfitting function, because itneglects the (1−e−D/τlong)part. Inorder todemonstrate theadvantageof the improved fittingfunction,data fromthe20thcycleof thedischargingpulse-rest testareadopted. In thiscycle, SoCchangesfrom62%to60%duringthepulse-dischargingperiod, thenkeepsthevalueof60%during the followingrestperiod. Thecurrentprofileof the20thdischargingpulse-rest test isappliedonthe ECMMATLAB/SIMULINKmodel asanexcitation. Figure12a,b shows themodeloutputvoltage responseswith twosetsofestimatedmodelparameters. It canbeseenthat themodelwithparameters estimatedbytheproposedfittingfunctionoutputsbetterestimationresults. The lowervoltageerror is mainlycontributedbythehighervoltagedropacross the long-termRCnetwork,asplottedinFigure12c. Inaddition, therootmeansquareerrors (RMSEs)betweenthemeasuredvoltageandthemodeloutput voltageatdifferentSoCsaregiveninTable3. It canalsobeseenthat themodelparametersestimatedby theproposedfittingfunctionshowabetterperformanceforawiderangeofSoC. 6R& FRQYHQWLRQDO ILWWLQJ IXQFWLRQ LPSURYHG ILWWLQJ IXQFWLRQ Figure11.Rlong estimationresults. Table3.ComparisonofRMSEatdifferentSoC. SoC (%) 10 20 30 40 50 60 70 80 90 RMSE(mV) Conventionalfitting function 1.802 1.714 2.167 1.540 1.268 2.803 2.416 1.558 1.444 Improvedfittingfunction 0.7658 0.7582 0.9707 0.7643 0.5000 1.202 1.242 0.7104 0.6482 171