Power Distribution Control Coordinating Ultra Capacitors and Batteries for Electric Vehicles (2004)

8/3/2019 Power Distribution Control Coordinating Ultra Capacitors and Batteries for Electric Vehicles (2004)

1/6

Proceeding of the 2004 American Control ConferenceBoston, M assachusetts June 30 - J u l y 2,2004 FrM05.4

Power Distribution Control CoordinatingUltracapacitors and Batteries for ElectricVehicles

Evren O zatay, Ben Zile*, Joel Anstrom and Sean Brennan*,'Member. IEEE

Abstract- Electrical energy storage is a central element to an yelectric-drivetrain technology -wheth er hybrid-electric, fuel-cell,or all-electric. A particularly cost-sensitive issue with energystorage is the high replacement cost of depleted battery hanks.One possibility to ease the power burden on batteries and fuelcells is to use ultra-cap acitors as load-leveling devices. The highpower density of ultra-capacitors allows a significant reductionin the power fluctuations imposed on the remaining electricalsystem; however, t h e same ultra-capacitors have a very lowenergy density and therefore must he used sparingly and withcoordination. A control strategy for coordinated powerdistribution is a central issue for ultracapacitor-supportedsystems. Toward this end, several control methods areimplemented on an electric vehicle equipped with abatterylultracapacitor system with the goal of improving batterylife and overall vehicle emciency. A particular goal is to obtainboth a peaking load control and a frequency-weightedcoordination between capacitor and battery in order to mitigatetransients in the battery current demand. A key control designissue is that th e control objectives vary with respect to vehiclevelocity, driver's power demand, and state-o f-charge of both thebatteries and ultracapacitors.

Index Term-U ltracapacitor, hybrid-electric vehicle, control.I. INTRODUCTION

HIS work is a continuation of a recent project conductedT a t the Pennsylvania Transportation Institute atPennsylvania State University in which ultra-capacitors areused in conjunction with batteries as an energy storage systemfor mass-transit vehicles [I].As part of this prior work, the properties of the ultra-capacitors were extensively characterized and the modelingresults were used to develop an ultra-capacitor energy storage

This work was supported in part by the Mid Atlantic Regional Consortiumfor Advanced Vehicles (MARCAV) and the Depanment of Transportationunder Grant DTRS56-99-T-00 16.E. Ozatay and Ben Zile are graduate students in the Mechanical andNuclear Engineering depamnent at Pennsylvania State U niversity.J. Anstrom is a Research Faculty with the Pennsylvania TransportationInstitute,.S . Brennan (corresponding author) is an Assistant Professor of Mechanicaland Nuclcar Engineering with a joint appointment with the PennsylvaniaTransportation Instibte, 318 Leonhard Building, University Park, PA 16802 .phone: 814-863-2430; fax 814-865-9693 (e-m il: [email protected]).

system model in the Department of Energy's PowertrainSystem Analysis Toolkit (PSAT) hybrid vehicle modelingenvironment. Using this PSAT simulation, the influence ofthe ultra-capacitor storage system were examined in both anelectric and hybrid electric mass transit vehicle. Thesesimulations were validated with experimental data taken fromboth an electric transit bus (without an ultra-capacitor energystorage system), provided by the Chattanooga Area RegionalTransit Authority (CARTA), and on a hybrid electricpassenger vehicle, the Penn S tate Electric Lion, both w ith andwithout the ultra-capacitor energy storage system. The resultsof the project show a significant reduction in peak currentsexperienced by the battery pack in drive cycles with a highnumber o f starts and stops, such as the Manhattan drive cycle.

The benefits of the ultracapacitor system were greatlyaffected not only by the capacity of the ultracapacitors, butalso by the energy managem ent controller. At the completionof the project, it was concluded that additional focus on thisenergy management controller is required in order to achieveoptimization of both the power cycle efficiency ofultracapacitor energy storage system an d o f the battery life.Similar ultracapacitor systems are in developmentelsewhere. Nova Bus Corporation, Quebec, has been testing abus powered by a zinc-air battery pack [2] with a goal ofsupplemen ting the power system with ultracapacitors. HenryBrandhorst and Zheng Chen of Auburn University, Alabamaare currently developing a system that uses an ultra-capacitorto imp rove the power capabilities of a battery-powered system[3]. The results of their tests showed that the addition of theultra-capacitor increased the power level by 40% over thebattery-only system. Solectria Corporation is currentlydeveloping a high power DC-to-DC converter for ultra-capacitors [4]. This ultra-capacitor system was tested on anelectric vehicle - he range of the vehicle was increased 32 %and the energy consumption remained unchanged. XinxiangYan and Dean Patterson of Northern Territory University,Australia have also researched ways to utilize ultra-capacitorsin electric vehicle applications [SI. Their system uses anultra-capacitor bank to complement a zinc-bromine battery;again the goal is to increase the range and performance ofelectric vehicles withou t increasing cost. General ElectricCompany has developed an ultra-capacitoribattery system to

4716-7803-8335-4/041$17.00 02004 AACC


2/6

be used on either electric vehicles or hybrid electric vehicles[6] ,with a specific intention of being used on a hybrid electrictransit bus. Honda has recently come out with a fuel-cellpowered vehicle that uses ultracapacitors for electrical energystorage. The vehicle's po wertrain is capable of producing80hp , can travel at speeds up to 93mpb and for distances of upto 2 20 miles, and carries up to 3.75kg of hydrogen [7].

The remainder of the paper is summarized as follows:Section 2 introduces the dynamic model of the ultracapacitorsystem. Section 3 discusses peaking power control strategiesand experimental results, focusing on the challengesassociated with the ultracapacitor power distribution. Section4 introduces a fiequency-domain control strategy with thegoal of generating a frequency-based allocation of powerbetween the battery and capacitor system. Finally, aconclusions section then summarizes the main researchfmdings.

11. DYNAMICODEL OF THE ULTRACAPACITORThe goal of any multi-mode power storage system is to gain

the advantages of each energy storage mode while avoidingthe disadvantages. The powedenergy relationship for variousenergy storage devices is shown in Fig. 1. The goal is toem ulat e a non-existent "super-device" by coordinating powertransfer between lead-acid batteries and an ultracapacitor.

10.'10.' 10. 100 10' Id to'Energy Osnd y [ m y x g lFig. I . The power-energy relationship for various elechical energy

slorage devices. A high power-density, high energy-density device does notexist, but one can be emulated by carefully wntrolled power distributionbetween electrolytic ultracapacitorsand lead-acid batteries.

F q u a c y . nzFig. 2. Measured frequency response characteristics of various uim.capacitors. The device s are all bandwidth limited, and operation near orfaster than this bandwidth results in severe losses in inputloutput efi iie ncy .

The limitation is that the mass of the vehicle cannot beincreased; the ultracapacitor addition removes 20% of themass of the battery pack, and hence comparison testsconducted later compare an electric vehicle (EV) with 100%battery pack to an EV + ultra-capacitor vehicle with 80%battery pack.To initially characterize the ultracapacitors and c o n f m thedynamic model, both fiequency and time-domain inputhutputanalysis were used. The Econd ultracapacitor for this test wascompared with 3 other ultracapacitors, all developed byMaxwell Technologies. These ultracapacitors all have carbonelectrodes. T he frequency characteristics are shown in Fig. 2.Inspection of the figures reveals that most commerciallyavailable ultracapacitors exhibit a reduction in capacitancestarting about 0.1Hz.Time domain (not shown) and frequency domainmeasurements both confirmed that ultracapacitors are veryefficient for low fiequency use. Both also show that thecapacitance drops (with corresponding decrease in efficiency)for frequencies greater than 0.1 Hz. The time domainmeasurements show that capacitor loss becomes verysignificant (70% for some tests) for fast discharge times. Thislimitation in the frequency-response of the capacitor system isa central design constraint in later control strategies forcoordinating the ultracapacitor usage, i.e. the closed-loopsystem bandwidth must be limited to be roughly no faster than0.1 Hz.

To obtain a mathematical model of the ultracapacitor and

+

Fig. 3. Idealized circuit representationof the ultracapacitor.boost converter, the simplified idealized circuit shown in Fig.3 was chosen. Here R , , and c are the total circuitresistance, total capacitance, and total inductance of thesystem. The ultracapacitor voltage input v, acts as the'control input' to the capacitor system. In implementation,this voltage input is created by a basic bidirectional buckconverter connected to the bus voltage of the vehicle. Thecapacitor input voltage, v, is equal to a duty cycle signal,D ent to a voltage converter times the voltage on the powerbus, "," = DVb, (1)

4717


3/6

Thus, the control input can never exceed the bus voltage.Because the bus voltage is actually a time varying state of thesystem, the mathematical system is in practice bilinear.However, in proper operation the bus voltage may be assumedto change very slowly and over a limited voltage range, i.e. isapproximately constant. At steady state, the duty cycle mustbe equal to the ratio of bus voltage and capacitor voltage, i.e.D = vc / V b , and the inductor current is zero. For powerdistribution purposes, it is convenient to treat the systemcontrol input not as the steady-state duty-cycle D but asdeviations from this value, hD that produce the inductorcurrent, i.e.

D=AD+D, (2)= U + CVb ,The corresponding dynamic equations for the

inductodcapacitor branch are given by Kirchoffs VoltageLaw, where the system states are i , , the current through theinductor and v c , he voltage across the capacitor:

For standard values o f R , L , and C th e systemcontrollability matrix is full rank. Each state is observable,and measured in the system. Therefore the system is bothcontrollable and observable.

The values of R , , nd Cmeasured from the systemare 0.3335 ohms, 0.36 milli-Henries, and 12 Faradsrespectively. With these values, the eigenvalues of thecapacitor's electrical system are -926.1389 and -0.2499rad kec . The fast pole corresponds to the inductor currentdynamics, and it is assumed that this dynamic can be ignoredwith respect to the much slower capacitor voltage change. Inprevious work [I], a reduced-order model was obtained by abalanced truncation. In this study, the method ofresidualization was used to obtain the reduced-order model, achoice made to preserve physical meaning to the systemrepresentation. With di, /dt = 0, an algebraic relationshipexists between i, , vc, an d v,n. Solving for the algebraicrelationship and rearranging, the dynamic equations for thereduced-order model become:

Fig. 4. The proportional contrDl architecture far current comol

v v .j --L+'"R R-dC . - v = jdt

(4)

The choice in using perturbations in the duty cycle as controleffort in Eq. (2) can be made more obvious when Equations(1 ) and (2) are substituted into (4):i, =-- c D ' v b u ~

R +- R

=U- busR

Thus, in the perturbation framework the control effort U isdirectly proportional to inductor current. The correspondingtransfer fun ction for the capacitor system is given by:c-(s) - 2.999 (6)V , ( s ) s+O.2499And in the perturbation states, the change in capacitor voltageis proportional to the integral of the control effort, U. Toobtain a model for the remaining electrical system, data mustbe taken on the system in closed-loop operation, i.e. on asystem with a working control law for the capacitor system.

111. PEAKINGOWERONTROL TRATEGIESWithin this study, there are tw o power distribution control

objectives considered to preserve battery life and both addressoperating conditions that are suspected of producingirreversible damage to the batteries. In order of importance,they are: ( I ) to prevent over-charging or over-drawing thebattery pack, i.e. a need for peaking power control, and (2) toprevent high-frequency power fluctuations on the battery packi.e. a fiequency-domain co ntrol.The first objective is addressed by a combining a current-following control law with a reference-generating algorithmthat partitions power distribution between the batteries andultracapacitors. The current-following control law asimplemented on the system is shown in Fig. 4. It is aproportional control architecture comparing a referencecurrent request to measured inductor current. The gain waschosen manually to achieve satisfactory current followingperformance such that the inductor current time constant isbetween IO and 20 milliseconds.The reference input to the above control law is given by areference generator (feedforward) algorithm that schedulescapacitor usage with respect to (i) the driver's torque request,i.e. throttle input, (ii) the capacitor's state of charge (SOC),and (iii) the vehicle's velocity. The control strategy wasdeveloped with the goal of reducing average and peak battery

4710


4/6

currents over a wide variety of driving cycles, and wasdeveloped and tuned by iterative testing on the demonstratorvehicle.The reference current demanded from the capacitor system,I,, is primarily determined by a calculated estimate of thecurrent needed by the drive motors, from the batteries,capacitors, and any Auxiliary Power Unit's (APU's) - i.e.motorlgenerators or fuel-cells. A look-up table compu tes afeedfonvard current request for the capacitors, i.e. whatfraction of the load current should be carried by the ultra-capacitors. Fig. 5 shows the look-up table used in all finaltesting; the lel? side of the table represents braking, the rightside acceleration. The braking request shows a moresignificant capacitor involvement because:

The batteries cannot charge as rapidly as they candischarge, so regenerative braking current is severelylimited by battery capabilities. Ultra-capacitors offer amuch greater ability to absorb regenerative braking

-1% ....... ; ......:.......:....... : ....... ! .......: ....... ......, , , , , , ., . . . . .. . . . . ., , , , , , ., , , , , , ..% .A .& A A A yocwrmt N.d.d ,"W". Batfansr, m d Capnntom [A,Fig. 5. Nonlinear heuristics added (via look-up tables) to the P-control toachieve targeted state-of-chargeon the ultracapacitor.current.Braking does not generally last as long as acceleration, soultra-capacitor charging rate needs to be higher thandischarging.

Th e reference current I , is also scheduled by capacitorstate-of-charge and vehicle speed. The capacitor's state ofcharge (SOC) is given by the energy that can be stored withinthe capacitor system, a function of the capacitor voltagesquared:

(3)2SOC= K . v ,The gain K is 3.05 x it acts as a normalization constantso that the equation yields SOC = 0 for no voltage on thecapacitor and SOC = 1 for the maximum voltage of 180 V.Thus the SOC is a percentage measure of the total remainingenergy available from the capacitor.The premise for scheduling control effort with SOC andvehicle speed is to have the ultra-capacitor system ready forthe future events (i.e., if the vehicle is stopped , an accelerationis the next thing that can occur, so the ultra-capacitors shouldbe charged and ready). The result of this secondary scheduling

is to produce a small current correction that is then added tothe nominally scheduled current request. This ensures thatwhen traveling at slow speeds, the ultracapacitors are wellcharged for upcoming acceleration transients. Similarly, whentraveling at high speeds, the control ensures that theultracapacitors are sufficiently discharged in preparation for afuture regenerative braking event. The target SOC at zeromph is set to 70% or 151 Volts for an ultracapacitor with amax rating of 180 Volts. At 55 mph, the target SOC is 20%or 81 Volts for an ultracapacitor with a max rating of 180

Fig. 6. The feedfaward and feedback control structure used to determine arefere nce current for the inductor current controller.

Volts. The target SOC is interpolated for speeds in between 0and 55 mph and extrapolated outside this area. A proportionalcontroller then closes the loop on the target SOC to produce acurrent request correction. This correction is limited to add orsubtract up to 20 Amp s to the nominal current reference. Thecomplete formulation of the reference generator is shown in

FigstrategyFig. 6.This controller was implemented on a hybrid electricvehicle as shown in Fig. 7 using an automotive D-Spacesystem. Implemen tation tests were conducted on a chassisdynamometer and on the test track to demonstrate a combinedbatteryhltra-capacitor energy storage system in a hybridvehicle and to obtain data to validate the design models forhybrid-electric transit buses. Fig. 8 is a schematic of theElectric Lion series hybrid vehicle after it was re-configuredwith an ultracapacitoribattery system. The converter isinstalled in the rear cargo area of the vehicle and connected to

4719


5/6

the high voltage bus at the battery pack. The ultracapacitorsare mounted on a small trailer behind the vehicle andconnected to the converter through the rear hatch. Thecapacitors were mounted on the auxiliary trailer for safety so 3wo-

2500t zwo-3

d 15w-9 lo w5c.-0 -

Fig. 8. The Electric Lion series hybrid vehicle used to test the hybrid controlslrategy.that an accidental discharge from the capacitors would notoccur within the vehicle cabin. The prototype vehicle (Fig. 7)was tested both on the Pennsylvania Transportation Institutedynamo meter and test track. This test vehicle has thecapability to act as a pure-electric vehicle (EV) or a hybridelectric vehicle (HEV). Only the EV results are reported inthis work, but both configurations were tested with theultracapacitor; details can be found in [ I ] .

On the dynamometer, the vehicle was driven throughseveral transit bus driving cycles: Manhattan, Orange County,UDDS, and New York. These driving cycles vary in theirdriving type, and are listed in order 60m low-speed urban,multiple stop-start driving to high-speed highway cruising.An example run on the Manhattan cycle, the best cycleperformance, is shown in Fig. 9 with battery and inductor(capacitor) currents shown in the top subplot (sign conventionis that posit ive currents are causin g vehicle to accelerate), busvoltage and capacitor voltage in the middle subplot, andvehicle speed in the bottom plot. At areas of rapidacceleration, clearly the ultracapacitor is discharging currentand reducing voltage and is active in assisting the batterysystem.

-

-

ua

3wo-2500t zwo-

3

d 15w-9 lo w5c.-0 -

ml.ISrlFig. 9. Experiment al results an thedynamom eter, Manhattan cycle

-

-

3500 I I

Fig. 10. Histograms of battery current demand, without uleacapacitors (top)and with uleacapaciton (bottom), for the hybrid vehicle in EV mode,Manhattan driving cycle .The benefit of the peaking controller is evident byexamining histograms of the battery current demand both withand without the ultracapacitors in use, for the same driving

cycle. ?his comparison is show n in Fig. 10. It is evident thatthe ultracapacitors reduced the extre mes in current demand inboth acceleration and in regenerative braking. In acceleration,the peak current demanded 60m the battery changed 6om 140amps without the ultra-capacitors to 90 amps with theultracapacitors. In regenerative brakimg, the activation of theultracapacitors lowered peak battery charging current 6om 60amps to 45 amps. More importantly, the volume of current(i.e. pow er transients) moved through the battery is seen to begreatly reduced over nearly all current levels. This isprimarily due to the ultracapacitors ability to storeregenerative energy and release it during acceleration. Thisbenefit is most evident in stop-start driving such as theManh attan cycle. For highway-like driving, the ultracapacitorsystem performed worse than the battery-only system becausethe ultracapacitors displaced 20% of battery pack.Iv. CURRENT ANDFUTURE ORK: CONTROL STRATEGIES N

THE FREQUENCY DOMAINIn addition to assisting with peaking power demandsdescribed earlier, it is desirable to use a frequency-domaincontrol approach. The concept is to use the ultracapacitors for

4720


6/6

short-duration ransients and use the battery system for steady-state type behavior. This objective is easily achieved by afrequency-weighted separation of control authority for aM I S0 system. Formalized control strategies based onfrequency-domain separation range from frequency-weightedLQR first presented by Gupta [SI, to the now-standard H-infmity synthesis techniques. Such frequency-weightedcontrol strategies have demonstrated utility in devices rangingfrom stacked-actuator disk-drives [9] to frnnthear steeringsystems in automobiles [lo]. An extension to vehicle powerdistribution system is therefore quite natural.The frequency weighting as is being currently studied israther simple: it will be low-pass for the battery and band-pass(at higher frequencies) for the capacito r. At each frequency,

VariableBandwdthLow-Parr Total Current

InductorCurrentw-1 CapacitorVOllaQe (sot)S"pNis0ryvoltageMonitorFig. 1 I . A freq uencydo msin control struchlre to achieve frequency-basedseparation of battery I ultracapacitor usage.

w , he frequency content of the battery usage B(w) andcapacitor usage C (U )must have a w eighting:(7)

However, this weighting is directly analogous to the Bode6equen cy criteria. s( ju)+T O " ) = 1 (8)The refore , if the battery is operating withim a control loop, theinpu t error signal of the battery contro ller should follow thesensitivity function, S( ju ) , nd the output tracking ratio willhave a frequency content following the complementarysensitivity function, Tbu). The notion of combiningsensitivity functions to achieve frequency-domain weightingstrongly suggests the use of a feedback structure, where thetracking errnr of one feedback look is used as a referencegen erato r for another feedback loop. Suc h a current followingcontrol ler is show n in Fig. 11 .The advantage of this structure is that there is a fail-safemechanism between the capacitor and battery system. If thebattery system exhibits current failures, varying dynamics, orother unexpected performance changes , the capacitor system

B(w)+c(u) 1

will inherently compensate for the lost current load.Similarly, if the capac itor system is unable to sup plyadditional current, then the battery performance is onlyminimally affected.An issue with this control structure is in obtaining adynam ic model of the battery system. Standard PSAT batterysimulatio ns in general have an ina dequate characterization ofbattery dynamics. Experimen tal testing is ongoing to modelthe remainder of the electrical system with the goal toimple ment this frequency-dom ain control technique. Resultswill be presented in following publications.

V. CONCLUSIONSThis research presented the controller design and

implementation results for the power coordination between abattery and an ultracapacitor. A control strategy consisting ofcascaded proportional loops was presented, and experimentalresults indicate significant improvements in the battery usageboth in th e peak curr ents experienced by the battery and theoverall power loading of the battery during regenerativebraking and acc eleration current loads.

ACKNOWLEDGMENTThe authors with to recognize the work of Heath Hofinann,

Benjamin Zile, Kandler Smith, Abhishek Neralla, and AmitBatra for their prior work constructing the testbed, developingthe model, and performing experimental validation of theoriginal system. Withou t their prior exte nsiv e work, thisstudy would not h ave been feasible.

REFERENCESI.A. Heath Hohna nn, Kandler Smith, Benjam in Zile, AbhisheckNeralla, and Amit B atra, "Simulation and Field-T esting of HybridUllra-CapacitorIBaUery Energy Storage System s far Electric TransitVehicles," T he Pennsy lvania Transportatio n Institute -MARCAYDOT 00-01, April 15 2003.W. D. ones, "NewYork wantsmore hybrid buses on he city'sstreets," IE EE Spe ctrum , vol. 37, p. 36,2000.H. W. J. Brandhorst and Z. Chen, "Achieving a high pulse powersystem through engineering the ba ttery-capacitor combination."presented at The Sixteenth Annual Battery Confe rence onApplication s and Ad vances, pp. 153-156., 2001.B. . Amet and L. P. Hainer, "High power DC-to-D C converler forsupercapaciton," presented at IEEE lntemational Electric Machinesand Driver Conference, 2000.X. Yan and D. Patterson. "Improvement ofdr ive range, accelerationand d eceleration performance in an electric vehicle propulsionsystem," presented at 30th Annual IEEE Power ElechonicsSpecialists Conference, 1999.King, . ., Salasoo,L., S c h w a ., and Cardinal, M., GeneralElectric Company, Jun 1998.Ross,P., The Top 10 Techno-Cool C ars IEEE Specirwn, vol. 40 ,pp . 30-5, eb, 2003.Gupla, N., "Frequency Shaped Cost Functionalr: Extension of LinearQuadratic Guassian Design Methods Journalof GuidanceandConlrol, vol. 3, Mar, 1982-Apr30, 1982.Schmeck, S. J. a. M. W . C., "On Controller Design fo r Linear Timc-Invariant Dual-Inp ut Single-Output Systems," San Diego, CA, pp.4122-6, Junc 1999.Brennan, S. and Alleyne, A., "Frequency-WeightedCoordinatedFront and T orque Steering," Proceedings of the 2001 AmericanControlConference. Arlington, VA, June 2001.

4721

Documents

Power Distribution Control Coordinating Ultra Capacitors and Batteries for Electric Vehicles (2004)