-
2448 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 6,
JULY 2011
Improved Performance of Serially ConnectedLi-Ion Batteries With
Active Cell Balancing
in Electric VehiclesMarkus Einhorn, Student Member, IEEE, Werner
Roessler, and Juergen Fleig
AbstractThis paper presents an active cell balancing methodfor
lithium-ion battery stacks using a flyback dc/dc convertertopology.
The method is described in detail, and a simulation isperformed to
estimate the energy gain for ten serially connectedcells during one
discharging cycle. The simulation is validatedwith measurements on
a balancing prototype with ten cells. It isthen shown how the
active balancing method with respect to thecell voltages can be
improved using the capacity and the stateof charge rather than the
voltage as the balancing criterion. Forboth charging and
discharging, an improvement in performance isgained when having the
state of charge and the capacity of the cellsas information. A
battery stack with three single cells is modeled,and a realistic
driving cycle is applied to compare the differencebetween both
methods in terms of usable energy. Simulations arealso validated
with measurements.
Index TermsBatteries, battery management systems, dc-dcpower
converters, electric vehicles, energy storage.
I. INTRODUCTION
S INGLE battery cells are usually connected in parallel andin
series to achieve higher capacity and voltage. The par-allel
connection is simple to handle since the cells appear as abig
single cell, and no management is necessary (similar to theparallel
connection of capacitors). The current splits accordingto the
internal impedance of the single cell, and the terminalvoltage of
each cell is equal.
Overcharging, as well as overdischarging, of lithium-ion(Li-ion)
cells causes irreversible damage and is also a majorsafety issue
[1][3]. Therefore, reliable monitoring of each cellvoltage is
necessary. The range between the charging voltagelimit (CV L) and
the discharge voltage limit (DV L), wherein
Manuscript received December 14, 2010; revised March 23, 2011;
acceptedMay 4, 2011. Date of publication May 12, 2011; date of
current versionJuly 18, 2011. This work was supported by the
Austrian Research PromotionAgency (Oesterreichische
Forschungsfoerderungsgesellschaft mbH, FFG) un-der research project
8219115: Active Balancing fuer Lithium-Ionen-Batterienin
Automobilanwendungen (BALI). The review of this paper was
coordinatedby Dr. A. Davoudi.
M. Einhorn is with the Mobility Department, Electric Drive
Technolo-gies, AIT Austrian Institute of Technology, 1210 Vienna,
Austria (e-mail:[email protected]).
W. Roessler is with the System Engineering Automotive, Infineon
Technolo-gies, 85579 Neubiberg, Germany.
J. Fleig is with the Institute of Chemical Technologies and
Analytics, ViennaUniversity of Technology, 1060 Vienna,
Austria.
Color versions of one or more of the figures in this paper are
available onlineat http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TVT.2011.2153886
Fig. 1. Typical cell voltage of a Li-ion battery during
charging/discharging [4].
Fig. 2. Cell voltages of three serially connected cells with a
capacity of 35,40, and 45 Ah when applying a 40-A constant
discharging current.
a Li-ion cell can be utilized, is shown in Fig. 1, and cell
voltagemust not get in the shaded area.
In a serially connected battery stack, the discharging, as
wellas the charging, process has to be stopped immediately as
soonas one of the terminal cell voltages fall below DV L or
exceedsCV L. The current through serially connected cells is the
same.Therefore, when the cells initially have the same state of
charge(SOC), the cell with the lowest capacity is the first one
thatreaches DV L and CV L when being charged and
discharged,respectively. Fig. 2 shows the simulated cell voltages
of three
0018-9545/$26.00 2011 IEEE
-
EINHORN et al.: PERFORMANCE OF SERIALLY CONNECTED LI-ION
BATTERIES WITH ACTIVE CELL BALANCING 2449
serially connected cells with a capacity of 35, 40, and 45
Ahwhen applying a 40-A constant discharging current. The
cellvoltage of the cell with the lowest capacity (35 Ah) is the
firstthat reaches DV L.
The capacity of the whole battery stack is thus limited bythe
weakest cell in the stack. Charging of the battery stackcannot be
continued when one cell (usually the cell with thelowest capacity)
is completely charged although some cellsare not. Discharging of
the battery stack must stop when onecell (usually again the cell
with the lowest capacity) is empty,although the others still have
some charge left [5].
There are many reasons why the capacities of the cells ina
battery stack are not identical. One of them is the variationwithin
the manufacturing process due to technical and eco-nomical
limitations. Hence, the cell capacities are initially notequal, and
moreover, there is a different capacity drift over thelifetime.
When single cells are built together to a battery stack,
eachcell has a different temperature even with a
well-designedcooling system [6]. Therefore, the cells age unequally
fast.This is a second reason why a cell capacity diversification
ina battery stack can occur [7], [8].
The performance of a battery stack with different
single-cellcapacities can significantly be increased when the
charge fromthe cells is equalized with an electronic circuit [9],
[10]. This iscalled cell balancing. It can basically be divided
into two maingroups: passive cell balancing and active cell
balancing [11][19].
Passive cell balancing uses a resistor to discharge the cellwith
the highest cell voltage so that charging can be continuedtill all
cells are fully charged. This method is only suitableduring the
charging process and not efficient due to powerdissipation and
energy waste. With active cell balancing, chargecan be transferred
between the cells in a battery stack using ashort time storage
element, which can be either a capacitor or aninductor [20], [21].
This paper focuses on a promising topologyusing a flyback converter
as storage element and particularlyaddresses the performance gain
[22][25].
The balancing circuit is described in detail, and the
per-formance gain with ten serially connected cells with a
largecapacity diversification is simulated and measured.
Moreover,it is shown how the balancing strategy can be improved
whenhaving each cell capacity and each SOC instead of the
cellvoltage as the balancing criterion. This method is simulated
andvalidated with three cells.
II. BALANCING CIRCUIT
The balancing circuit principle is based on a flyback
con-verter. The key component is a transformer with a winding
foreach cell and a winding for the whole battery stack [26]. Fig.
3shows the balancing circuit for ns serially connected cells.
Thebidirectional use of the multiple winding transformer allowstwo
different balancing strategies. Energy from one single cellcan be
transferred to the whole stack (top balancing), andenergy from the
whole stack can be transferred to one single cell(bottom
balancing), as shown in Fig. 4. A detailed descriptionof the
balancing circuit can be found in [24], [25], and [27].
Fig. 3. Balancing circuit for ns serially connected cells with
the multiplewinding transformer T as key component of the flyback
converter structure.The microcontroller C operates the switches S
and S1 . . . Sns (typically low-voltage MOSFETs) according to the
balancing strategy.
Fig. 4. Charge transfer from one cell to the whole stack (top
balancing) andfrom the whole stack to one cell (bottom
balancing).
Top balancing is typically applied during charging to
avoidovercharging of a cell. When the voltage of a cell is closeto
CV L, charge can be transferred to the other cells, and thecharging
process can be continued. With this method, each cellcould be
completely charged, although it would take much time,depending on
the charging current, the balancing current, theSOC, and the
capacity of each cell.
Bottom balancing is typically applied during discharge modeto
increase the usable energy. When the voltage of a cell isclose to
DV L, charge can be transferred to this cell, and thedischarging
process can be continued. With this method, thebattery stack can be
discharged until all cells are completelydischarged (depending on
the discharge current I , the balancingcurrent, the SOC, and the
capacity of each cell).
The influence of the balancing current on the capacity of awhole
battery stack can be approximated. Without balancing,
-
2450 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 6,
JULY 2011
TABLE ICELL CAPACITIES C AND REMAINING CAPACITY RELATED TO
THE
INITIAL CN = 2.3 Ah FOR THE BALANCING SCENARIO [28]
the capacity of the cell with the lowest capacity Cx in a
batterystack with ns serially connected cells defines the capacity
of thewhole battery stack
C = Cx = min{C1, C2, . . . Cns}. (1)If just cell x is supported
during discharging, the usable
capacity of the battery stack is increased to
C = Cx + Ixbal t (1 1
ns
)(2)
with the total balancing time t and an ideal converter. A
batterystack with 11 40-Ah and one 35-Ah cells, a balancing
currentof Ibal = 5 A, and a balancing time of t = 1 h would have
acapacity of
C = 35 Ah + 5 A 1 h (1 1
12
)= 39.6 Ah (3)
which is a higher value than that without balancing.If there is
more than one weak cell, the balancing time ratio of
each cell to the total balancing time ti/t needs to be
considered.The capacity of the battery stack is then increased
to
C = min{
Ci + Ibal t (
tit 1
ns
)}i=1...ns
. (4)
III. SIMULATION AND EXPERIMENTAL VALIDATIONOF THE VOLTAGE
BALANCING METHOD
Ten Li-ion cells have diversely been cycled to decrease
thecapacity and are then serially connected to a battery stack.
Theremaining capacities are shown in Table I.
The theoretical stored energy in the battery stack fromTable I
is
W =nsi=1
Wi =
10
OCV dSOC 10
i=1
Ci = 57.2 Wh (5)
with the SOC versus open circuit voltage (OCV ) curve ex-tracted
from the cell datasheet [28]. Fig. 5 shows the SOCversus OCV curve
exemplarily for a Li-ion battery cell witha Li[NiCoMn]O2-based
cathode and a graphite-based anode.
The balancing circuit from Fig. 3 is modeled in Modelica/Dymola
[29], [30], using the electrical energy storage library
Fig. 5. Linear interpolation of the measured OCV at different
SOC valuesfor the Li-ion polymer cell.
Fig. 6. Simulation arrangement.
[31]. Together with the validated battery model from [32],
theincrease in energy by using an active balancing circuit withthe
ten serially connected cells from Table I is simulated.
Anextraction of the simulation arrangement is shown in Fig. 6.A
constant discharging current (Load) is applied to the batterystack
(Batterypack), and the balancer (Balancer) equalizes thecell
voltages with a balancing current (single-cell side of thebalancing
circuit) of Ixbal = 4 A. This operation mode is calledvoltage
balancing. The balancer is in operation mode only whenthe
difference voltage between the cell with the lowest and thatwith
the highest voltage is greater than 20 mV. Fig. 7 shows
theschematic operation mode during voltage balancing for threecells
with different cell capacities. The efficiency and powerloss in
control are taken into consideration with an assumedefficiency of
the dc/dc converter in the balancer of = 90%.The energy and the
charge from the battery stack are estimated(Energy).
This scenario has also been measured using the cells fromTable I
and the active cell balancing prototype shown in Fig. 8.The
measured cell voltages during discharging without balanc-ing and
with a balancing current of Ixbal = 4 A are shownin Fig. 9. Without
balancing (top chart), cell 1 is the one
-
EINHORN et al.: PERFORMANCE OF SERIALLY CONNECTED LI-ION
BATTERIES WITH ACTIVE CELL BALANCING 2451
Fig. 7. Schematic operation of voltage balancing. A positive
cell currentdischarges the cell.
Fig. 8. Active cell balancing prototype with ten serially
conected Li-ion cells.
that first reaches DV L because of its lowest capacity, andthe
discharging process has to stop. The battery stack cannotfurther be
discharged, although cells 210 are not completelydischarged. With
balancing (bottom chart), all cells reach DV Ltogether because the
weak cells are supported during the wholedischarging process. The
cell voltages of the balanced cellchange during balancing according
to its balancing currentand its internal impedance (oscillations).
Fig. 10 shows thesimulated and measured gain in terms of energy
[see Fig. 10(a)]and charge [see Fig. 10(b)] of this balancing
experiment. Withthis scenario, the discharge energy is increased
with activevoltage balancing by 15%.
Compared with the rated capacities of the used cells (CN =2.3
Ah) and the constant discharge current (I = 1.8 A), thebalancing
current of Ixbal = 4 A is very large. For cells with ahigher
capacity and a larger discharging current, the balancingcurrent
might be too small to equalize the cell voltages com-pletely in
real time. It can also happen that the wrong cells arebalanced if
just the cell voltages are considered, as discussed inthe next
section.
IV. CAPACITY BALANCING
Up to now, charge is transferred from the battery stack tothe
cell with the lowest voltage during discharging. Duringcharging,
charge is transferred from the cell with the highestvoltage to the
battery stack. In this so-called voltage balancingstrategy, only
the cell voltages are considered. However, it isshown in the
following that these criteria do not always lead toan optimal
decision that cells have to be balanced [33]. Whenthe cells are not
balanced in an optimal manner, the usableenergy in the battery
stack is lower than that with an optimalbalancing strategy.
The main aspects of the balancing procedure can alreadybe
analyzed with a three-cell battery stack because just thecells with
the highest and the lowest voltage are crucial. Threeserially
connected cells are assumed. Cell 1 has the lowestcapacity (e.g.,
35 Ah), cell 2 has an intermediate capacity (e.g.,40 Ah), and cell
3 has the largest capacity (e.g., 45 Ah).
Fig. 11(a) shows the schematic charge transfer with
voltagebalancing during one typical charging and discharging
period.The top diagram shows the cell voltage of each cell, the
chartin the middle shows the SOC of each cell, and the
bottomdiagram shows the energy in each cell. In phase I, the cells
arecharged, starting from a different SOC, and charge is takenfrom
cell 3 because of its highest voltage and transferred tocells 1 and
2. Indeed, the energy from cell 3 is transferred tothe whole
battery stack, and since only three cells are present,the energy is
split into three equal parts and spread to cells1, 2 and 3. The net
charge transfer though is from cell 3 tocell 1 and to cell 2. Cell
3 also has the highest capacity, andso, it will take more time to
fully charge it than the othercells when the current for each cell
is equal. The slopes of thecurves in the middle and bottom diagrams
indicate how fastthe cells are charged. When the cell voltage of
cell 1 exceedsthe others at the beginning of phase II (it is
assumed thatall cells have the same voltage and, therefore, have
the sameSOC = SOC at this moment), charge is removed from cell 1and
transferred to the other cells 2 and 3 because cell 1 now hasthe
highest voltage. When the first cell is fully charged (cell1
because of its lowest capacity) ,the charging process muststop
immediately to avoid overcharging this cell. Beyond thispoint, the
charging process could be continued with a severelyreduced charging
current and an active balancing system untilall cells are
completely full. This would take much more timeand is not
considered here. In phase III, charge is transferred tocell 3;
although stored in cell 3 is the largest amount of energy,it has
the lowest cell voltage. In phase IV, cell 1 has the lowestvoltage,
limits the duration of the discharging process, and istherefore
supported. Voltage balancing works inefficiently in
-
2452 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 6,
JULY 2011
Fig. 9. Measured cell voltages of the serially connected cells
from Table 1 during discharging (top) without balancing and
(bottom) with voltage balancing. Thebalancing current in the bottom
chart is 4 A using the active cell balancing prototype from Fig.
8.
Fig. 10. Simulated and measured (a) discharging energy and (b)
discharging capacity without balancing and with active
balancing.
phases I and III because the wrong cells are balanced, andthe
transferred charge is partially retransferred in phases IIand
IV.
By using the SOC and the capacity of each cell as the bal-ancing
criterion, the drawback of voltage balancing in phases Iand III can
be eliminated, as shown in the stack in Fig. 11(b).This is called
capacity balancing.
During charging, energy from the cell with the lowest energyto
full charge is taken. The energy to full charge of each cellis the
difference between the total capacity of each cell (C1,
C2, and C3) and the corresponding curves in the bottom chartof
Fig. 11. This is cell 1 during the whole charging
process.Therefore, charge is transferred from cell 1 to cells 2 and
3(phases I and II), and the slopes of the curves in the middleand
bottom diagrams do not change. During discharging, thecell with the
lowest amount of energy is supported. This is forthe whole
discharging process (phases III and IV) cell 1, andcharge is
transferred to this cell. Hence, the slopes of the curvesin the
middle and bottom diagrams do not change in phases IIIand IV.
-
EINHORN et al.: PERFORMANCE OF SERIALLY CONNECTED LI-ION
BATTERIES WITH ACTIVE CELL BALANCING 2453
Fig. 11. Schematic balancing (a) based on the cell voltage and
(b) based on the cell capacity and SOC. The arrows indicate the
charge transfer betweenthe cells.
TABLE IIRATED CELL CAPACITIES CN , MEASURED CAPACITIES C, AND
INITIAL
SOC FOR THE CAPACITY BALANCING SCENARIO
With capacity balancing, the charge time can be decreasedbecause
charge is not retransferred as with voltage balancing.The usable
energy of the battery stack during discharging is in-creased
because the weak cell is supported from the beginning.In addition,
the cell with the lowest capacity is utilized with alower current
for both charging and discharging. In the long run,this could have
a positive effect on the aging since the capacitydecrease is
related to the cell current.
V. SIMULATION AND EXPERIMENTAL VALIDATIONOF THE CAPACITY
BALANCING METHOD
Three Li-ion polymer cells with different capacities andSOC, as
shown in Table II, are serially connected to a battery
stack [34]. Cell 1 with CN1 = 20 Ah is a single cell, cell 2
withCN2 = 40 Ah is two single cells in parallel, and cell 3 withCN3
= 60 Ah is three single cells in parallel.
The battery stack with the configuration from Table II has
atheoretical stored energy of
W =W1 + W2 + W3
=C1 1
0
OCV dSOC
+ C2 0.90
OCV dSOC + C3 0.80
OCV dSOC
=406.17 Wh. (6)The OCV is a function of SOC and can be extracted
from
the cell datasheet. This battery stack is discharged until one
cellreaches DV L (typically cell 1 with the lowest capacity). For
thedischarging process, the current profile gained from the
FTP72driving cycle, as shown in [35, Fig. 12], is continually
appliedto the battery stack, as well as to the simulation.
-
2454 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 6,
JULY 2011
Fig. 12. Definition of the FTP72 driving cycle, power
consumption of atypical compact electrical vehicle, and current
requirement from a batterystack with 100 serially connected single
cells with a cell voltage of 3.6 V,respectively [35].
Fig. 13. Test circuit to validate the capacity balancing
simulation.
Fig. 14. Experiment setup for the circuit from Fig. 13.
The simulation from Fig. 5 is used and extended by thecapacity
balancing operation mode, and the battery stack isconfigured
according to the stack from Table II. The active bal-ancing system
is connected to the battery stack with a balancing
Fig. 15. Cell voltages during the FTP72 discharging cycle from
Fig. 12 [topchart of (a)] without balancing, [middle chart of (a)]
with voltage balancing, and[bottom chart of (a)] capacity
balancing. With voltage balancing during t = 0and t 71 min, cell 1
is supported. After t 71 min, cell 3 is supported.(b) With capacity
balancing, cell 3 is supported during the entire
dischargingprocess.
current of 3 A (single cell side of the dc/dc converter).
Theavailable charge and energy over the whole discharging processis
calculated. The simulated capacity balancing scenario is
thenvalidated with the experimental results related to the
circuitfrom Fig. 13 since the capacity balancing strategy has yet
to beimplemented in the prototype in Fig. 8. Instead of the
flybackconverter with the prototype, a current source with a
power(P ) coupled electronic load is used to perform the
chargetransfer assuming an efficiency of 0.9. During the whole
test,the cells are in a climate chamber to minimize
temperatureeffects. Fig. 14 shows (from left to right) the battery
test benchfor discharging the stack with the FTP72 cycle, the
climatechamber with the battery stack in it, an electronic load
coupledwith a current source, and a PC for the measurement.
In Fig. 15, the measured cell voltages during the FTP72 cur-rent
profile are shown for the three different balancing scenarios(no
balancing, voltage balancing, and capacity balancing). With
-
EINHORN et al.: PERFORMANCE OF SERIALLY CONNECTED LI-ION
BATTERIES WITH ACTIVE CELL BALANCING 2455
Fig. 16. Measured and simulated (a) discharging energy and (b)
discharg-ing capacity without balancing, with voltage balancing and
with capacitybalancing.
no balancing (top chart), cell 1 with the lowest capacity isthe
first that is completely discharged, although it started withSOC =
1. With voltage balancing (middle chart), cell 1 is stillthe one
that first reaches DV L, but the discharging processis
significantly longer (71 min) than that without balancing.The first
50 min of the discharging process correlate withphase I in Fig.
11(a), where cell 3 with the highest amount ofstored energy is
supported (not optimal). In the bottom chart ofFig. 15, the
capacity balancing strategy is applied, and cell 1is supported
during the whole discharging process. Cells 1 and2 reach DV L
almost at the same time, just cell 3 has energystill left because
of its much larger capacity. If the balancingcurrent were higher,
all three cells would reach DV L simulta-neously. This would be
optimal because the battery stack is thencompletely discharged, and
all energy in the cells could be used.
The measured and simulated discharging energy for thedifferent
balancing scenarios are shown in Fig. 16(a). Theavailable
discharging capacity for different balancing scenariosis shown in
Fig. 16(b). Since the measured discharging currentis applied to the
simulation, there is no difference between themeasured and the
simulated discharging charge. Without anybalancing, the capacity of
the battery stack is as weak as the
smallest cell. In this case, the battery stack has a
maximumcapacity of 21.9 Ah, which correlates with a usable energyof
240 Wh. Voltage balancing increases the capacity by 27%to 28.3 Ah
and 306 Wh, respectively. The best performancefor this scenario is
accomplished when balancing the availablecapacity. The capacity of
the battery stack can be increasedby 32% to 29.1 Ah and 318 Wh,
respectively. Even withcapacity balancing, the usable energy is
just around 79% ofthe theoretical value from (6). For improving the
result, thebalancing current must be increased as mentioned
earlier.
When all cells are fully charged, there is no differencebetween
voltage and capacity balancing during discharging (forcells with
the same chemistry). The cell with the lowest voltagealso has the
lowest amount of stored energy, and therefore, thebalanced cells
for both voltage and capacity balancing are thesame. There is also
no difference between voltage and capacitybalancing during charging
if all cells are completely dischargedbefore starting the charging
process (the cell with the highestvoltage is also the cell with the
lowest energy to full charge,and therefore, the balanced cells for
both voltage and capacitybalancing are the same). When the cells
are not all completelycharged before discharging, capacity
balancing increases theamount of usable energy of the battery
stack. When the cellsare not all completely discharged before
charging, the batterystack can be charged in a shorter time, and
more energy can beloaded into the battery stack when using capacity
balancing.
To use the active balancing system in capacity balancingmode, a
method to estimate the cell capacities and the SOCduring operation
is necessary. The SOC can, for example,be estimated by measuring
the OCV and by using Fig. 6 or[36][39]. Since the capacity of a
battery cell changes over itslifetime due to aging [8], [40], it is
not enough to estimate thecapacity just once (e.g., after
production). In general, the capac-ity of a single battery cell can
be estimated by fully discharg-ing it and integrating the measured
current (charge counting)[41], [42]. However, this approach is very
difficult in a seriallyconnected battery stack because the cell
with the lowest capac-ity is the first one that is completely
discharged, and the batterystack cannot be further discharged.
Hence, with charge countingonly, the capacity of the cell with the
smallest capacity in aserially connected battery stack can be
estimated. Methods toestimate the capacity of each battery cell in
a serially connectedbattery stack are, for example, presented in
[43] and [44].
If the capacities and the SOC of the cells are not well
esti-mated during capacity balancing (e.g., due to a drift of the
val-ues over time), it could happen that a cell is
charged/dischargedover/below CV L/DV L. Therefore, it is
recommended to ad-ditionally monitor each cell voltage to prevent
the cells fromovercharging/overdischarging.
VI. CONCLUSION
If several Li-ion cells are serially connected to a
batterystack, the worst cell defines the limit of the whole
battery. Whenthe charging/discharging voltage limit of one cell is
reached,charging/discharging has to be stopped, regardless of how
muchenergy is left in the other cells. If the cell capacities in a
batterystack are different (due to production diversification or
aging),
-
2456 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 60, NO. 6,
JULY 2011
active balancing remarkably improves the performance of abattery
stack.
A flyback dc/dc converter topology has been presented tobalance
the cell voltages of a battery stack with ten seriallyconnected
cells. The cell capacities are considerably different,and with
active balancing, the usable energy of the battery stackcan be
improved by 15%.
Balancing the cell voltages is not always the most effectiveway
to improve the usable energy in a battery stack withdifferent cell
capacities. When the cells are not completelycharged/discharged
before discharging/charging, balancing thevoltages would lead to a
suboptimal result. The usable energyof a battery stack can
significantly be improved when the storedamount of energy of each
cell, and not the cell voltages, isconsidered (balancing the
available capacity).
Further work will focus on implementing the capacity bal-ancing
strategy to the prototype and estimating the cell capaci-ties and
the SOC for cells in a battery stack during operation.
ACKNOWLEDGMENT
The authors would like to thank C. Kral and F. V. Contefrom the
AIT Austrian Institute of Technology for reviewingthis paper.
REFERENCES[1] D. Belov and M.-H. Yang, Investigation of the
kinetic mechanism in
overcharge process for Li-ion battery, Proc. 16th Int. Solid
State Ion.Conf., vol. 179, no. 2732, pp. 18161821, Sep. 2008.
[2] H. Maleki and J. N. Howard, Effects of overdischarge on
performanceand thermal stability of a Li-ion cell, J. Power
Sources, vol. 160, no. 2,pp. 13951402, Oct. 2006.
[3] J. McDowall, P. Biensan, and M. Broussely, Industrial
lithium ion bat-tery safetyWhat are the tradeoffs?, in Proc. IEEE
29th INTELEC,Sep. 2007, pp. 701707.
[4] Li-Tec HEA40 High Energy Cell Datasheet, 2008.[5] S. M.
Lukic and A. Emadi, Charging ahead, IEEE Ind. Electron. Mag.,
vol. 2, no. 4, pp. 2231, Dec. 2008.[6] C. Mi, B. Li, D. Buck,
and N. Ota, Advanced electro-thermal modeling
of lithium-ion battery system for hybrid electric vehicle
applications, inProc. IEEE VPPC, Sep. 912, 2007, pp. 107111.
[7] M. Broussely, P. Biensasn, F. Bonhomme, P. Blanchard,S.
Herreyre, K. Nechev, and R. J. Staniewicz, Main aging mechanismsin
Li ion batteries, J. Power Sources, vol. 146, no. 12, pp. 9096,Aug.
2005.
[8] M. Broussely, S. Herreyre, P. Biensan, P. Kasztejnac, K.
Nechev, andR. J. Staniewicz, Aging mechanism in Li ion cells and
calendar lifepredictions, J. Power Sources, vol. 97/98, pp. 1321,
Jul. 2001.
[9] S. Wen and T. Instruments, Cell balancing buys extra run
time and batterylife, Analog Appl. J., 2009, Dallas, TX, 1Q.
[10] M. Kultgen, Managing high-voltage lithium-ion batteries in
HEVS,EDN: Information, News, and Business Strategy for Electronics
DesignEngineers, Milpitas, CA, Apr. 2009.
[11] J. Guerin and W. Liu, Cell balancing algorithm verification
through asimulation model for lithium ion energy storage systems,
Soc. Automat.Eng. Int. Conf., 2010, Detroit, MI.
[12] X. Wei and B. Zhu, The research of vehicle power: Li-ion
battery packbalancing method, in Proc. IEEE 9th Int. Conf.
Electron. Meas. Instrum.,Aug. 2009, pp. 24982502.
[13] N. S. Jian Cao and A. Emadi, Battery balancing methods: A
comprehen-sive review, in Proc. IEEE VPPC, 2008, pp. 16.
[14] J. W. Kimball, B. T. Kuhn, and P. T. Krein, Increase
performanceof battery packs by active equalization, in Proc. IEEE
VPPC, 2007,pp. 323327.
[15] S. W. Moore and P. Stevens, An investigation of cell
balancing andequalization for lithium-ion battery packs, in Proc.
21st Annu. EVS,2005, pp. 19.
[16] J. Cao, N. Schofield, and A. Emadi, A review of cell
equalization meth-ods for lithium ion and lithium polymer battery
systems, Soc. Automat.Eng. Int. Conf., Jan. 2001, Detroit, MI.
[17] W. Bentley, Cell balancing considerations for lithium-ion
battery sys-tems, in Proc. 12th Annu. Battery Conf. Appl. Adv.,
1997, pp. 223226.
[18] N. Kutkut and D. Divan, Dynamic equalization techniques for
series bat-tery stacks, in Proc. IEEE 18th INTELEC, Oct. 610, 1996,
pp. 514521.
[19] D. Hopkins, C. Mosling, and S. Hung, Dynamic equalization
duringcharging of serial energy storage elements, IEEE Trans. Ind.
Appl.,vol. 29, no. 2, pp. 363368, Mar./Apr. 1993.
[20] A. Xu, S. Xie, and X. Liu, Dynamic voltage equalization for
series-connected ultracapacitors in EV/HEV applications, IEEE
Trans. Veh.Technol., vol. 58, no. 8, pp. 39813987, Oct. 2009.
[21] K. Nishijima, H. Sakamoto, and K. Harada, A PWM controlled
simpleand high performance battery balancing system, in Proc. IEEE
31stAnnu. PESC, 2000, vol. 1, pp. 517520.
[22] C. Bonfiglio and W. Roessler, A cost optimized battery
managementsystem with active cell balancing for lithium ion battery
stacks, in Proc.IEEE VPPC, Sep. 710, 2009, pp. 304309.
[23] X. Wei, X. Zhao, and D. Haifeng, The application of flyback
DC/DCconverter in Li-ion batteries active balancing, in Proc. IEEE
VPPC,Sep. 710, 2009, pp. 16541656.
[24] W. Roessler, Aktiver ladungsausgleich fuer lithium-ionen
batterien, ATZElektronik, vol. 3, no. 2, pp. 6266, Feb. 2008,
Infineon Technologies.
[25] W. Roessler, Boost battery performance with active
charge-balancing,EE Times-India, Munich, Germany, Jul. 2008.
[26] W. Roessler, Circuit Arrangement and Method for
Transferring ElectricalCharge Between Accumulator Arrangement, U.S.
Patent 2008/0 272 735A1, Nov. 6, 2008.
[27] Application NoteBattery Supervision and Cell Balancing
Module for Li-Ion Battery Packs, Battery Management Board 4.0,
Munich, Germany,Jun. 2010. [Online]. Available:
http://www.infineon.com/cms/en/product/promopages/aim-mc/ATV_SYS/Battery_Management_Slave_Board_4.0
[28] ANR26650 High Power Lithium Ion Cell Datasheet, p. A123,
2006.[29] Modelica Association, ModelicaA Unified Object-Oriented
Language
for Physical Systems Modeling, Language Specification Ver. 3.2,
2010.[30] P. Fritzson, Principles of Object-Oriented Modeling and
Simulation With
Modelica 2.1. Piscataway, NJ: Wiley-IEEE Press, 2004.[31] M.
Einhorn, F. V. Conte, C. Kral, C. Niklas, H. Popp, and J. Fleig,
A
modelica library for simulation of elecric energy storages, in
Proc. 8thInt. Modelica Conf., 2011, pp. 110.
[32] M. Einhorn, F. V. Conte, C. Kral, J. Fleig, and R. Permann,
Parameter-ization of an electrical battery model for dynamic system
simulation inelectric vehicles, in Proc. IEEE VPPC, Sep. 2010, pp.
17.
[33] M. Einhorn and J. Fleig, Improving of active cell balancing
by equalizingthe cell energy instead of the cell voltage, in Proc.
25th Int. Battery,Hybrid, Fuel Cell EVS, Shenzhen, China, Nov.
2010.
[34] EIG ePLB C020B Lithium Ion Polymer Cell Datasheet,
2010.[35] FTP72 Urban Dynamometer Driving Schedule (UDDS).[36] J.
Lee, O. Nam, and B. Cho, Li-ion battery SOC estimation method
based on the reduced order extended Kalman filtering, J. Power
Sources,vol. 174, no. 1, pp. 915, Nov. 2007.
[37] S. Lee, J. Lee, and B. H. Cho, State-of-charge and capacity
estimationof lithium-ion battery using a new open-circuit voltage
versus state-of-charge, J. Power Sources, vol. 185, no. 2, pp.
13671373, Dec. 2008.
[38] I.-S. Kim, The novel state of charge estimation method for
lithiumbattery using sliding mode observer, J. Power Sources, vol.
163, no. 1,pp. 584590, Dec. 2006.
[39] J. Chiasson and B. Vairamohan, Estimating the state of
charge of abattery, IEEE Trans. Control Syst. Technol., vol. 13,
no. 3, pp. 465470,May 2005.
[40] J. Vetter, P. Novak, M. R. Wagner, C. Veit, K.-C. Mller, J.
Besenhard,M. Winter, M. Wohlfahrt-Mehrens, C. Vogler, and A.
Hammouche, Age-ing mechanisms in lithium-ion batteries, J. Power
Sources, vol. 147,no. 1/2, pp. 269281, Sep. 2005.
[41] S. Park, A. Savvides, and M. Srivastava, Battery capacity
measurementand analysis using lithium coin cell battery, in Proc.
Int. Symp. LowPower Electron. Des., 2001, pp. 382387.
[42] P. Pascoe, H. Sirisena, and A. Anbuky, Coup de fouet based
VRLAbattery capacity estimation, in Proc. 1st IEEE Int. Workshop
Electron.Des., Test, Appl., 2002, pp. 149153.
[43] M. Einhorn, V. Conte, C. Kral, and J. Fleig, A method for
online capacityestimation of lithium ion battery cells using the
state of charge and thetransferred charge, in Proc. 2nd IEEE ICSET
, 2010, pp. 16.
[44] X. Zhang, X. Tang, J. Lin, Y. Zhang, M. Salman, and Y.-K.
Chin,Method for Battery Capacity Estimation, U. S. Patent 2009 322
283,Dec. 31, 2009.
-
EINHORN et al.: PERFORMANCE OF SERIALLY CONNECTED LI-ION
BATTERIES WITH ACTIVE CELL BALANCING 2457
Markus Einhorn (S11) was born in Vienna,Austria, in 1984. He
received the B.Sc. and Dipl.-Ing. degrees (with distinction) in
electrical engineer-ing from the Vienna University of Technology
in2008 and 2009, respectively, where he is currentlypursuing the
Ph.D. degree.
He is currently with the Mobility Department,Electric Drive
Technologies, AIT Austrian Instituteof Technology, Vienna, as a
Research Associate. Hisrecent work has focused on the design and
modelingof power electronics and battery systems.
Mr. Einhorn is a member of the Modelica Association and of the
OVEAustrian Electrotechnical Association.
Werner Roessler was born in Bavaria in 1958. Hereceived the
Dipl.-Ing. degree from the TechnicalUniversity Munich, Mnchen,
Germany, in 1983,after studying communications engineering.
Since 1983, he has been with the SemiconductorDivision, Siemens,
Neubiberg, Germany (in 1999,it became Infineon Technologies). Until
2000, hewas an Application Engineer for television engineer-ing
with focus on power supply, microcontrollers,and teletext. He is a
member of the InternationalStandardization Group for High Level
Teletext. After
two years of Hardware development for a speech recognition chip,
he moved tothe Automotive Division as Application Engineer for
automotive sensors whichfocus on magnetic and pressure sensors.
Since 2006, he has been a SystemEngineer for automotive hybrid
applications with focus on battery managementsystems.
Juergen Fleig received the Diploma degreein physics from the
University of Tuebingen,Tuebingen, Germany, in 1991 and the Ph.D.
degreein chemistry from the Max-Planck-Institute ofSolid-State
Research, Stuttgart, Germany, in 1995.
After working as a Researcher with the Max-Planck-Institute of
Solid-State Research for severalyears, he accepted a position as
Professor of electro-chemistry with the Vienna University of
Technology,Vienna, Austria, in 2005. His main research subjectsare
electroceramics and materials for electrochemi-
cal energy conversion devices, including basic investigations on
the physicaland chemical processes determining cell
efficiencies.
/ColorImageDict > /JPEG2000ColorACSImageDict >
/JPEG2000ColorImageDict > /AntiAliasGrayImages false
/CropGrayImages true /GrayImageMinResolution 300
/GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true
/GrayImageDownsampleType /Bicubic /GrayImageResolution 300
/GrayImageDepth -1 /GrayImageMinDownsampleDepth 2
/GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true
/GrayImageFilter /DCTEncode /AutoFilterGrayImages false
/GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict >
/GrayImageDict > /JPEG2000GrayACSImageDict >
/JPEG2000GrayImageDict > /AntiAliasMonoImages false
/CropMonoImages true /MonoImageMinResolution 1200
/MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true
/MonoImageDownsampleType /Bicubic /MonoImageResolution 600
/MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000
/EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode
/MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None
] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false
/PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000
0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true
/PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ]
/PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier ()
/PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped
/False
/Description > /Namespace [ (Adobe) (Common) (1.0) ]
/OtherNamespaces [ > /FormElements false /GenerateStructure
false /IncludeBookmarks false /IncludeHyperlinks false
/IncludeInteractive false /IncludeLayers false /IncludeProfiles
false /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe)
(CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector
/DocumentCMYK /PreserveEditing true /UntaggedCMYKHandling
/LeaveUntagged /UntaggedRGBHandling /UseDocumentProfile
/UseDocumentBleed false >> ]>> setdistillerparams>
setpagedevice