Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
J. Power Sources, in press (1998)
The Use of Computer Simulation in the Evaluation of
Electric Vehicle Batteries
W.B. Gu and C.Y. Wang
1
Department of Mechanical Engineering and Pennsylvania Transportation Institute
The Pennsylvania State University
University Park, PA 16802
email: cxw31@psu.edu
tel: (814)8634762
fax: (814)8634848
and
B.Y. Liaw
Hawaii Natural Energy Institute, SOEST
University of Hawaii at Manoa
Honolulu, HI 96822
____________________________________
1
Corresponding author.
2
ABSTRACT
An integrated simulation and testing approach is presented to evaluate batteries for
electric vehicle (EV) applications. This new approach combines traditional experimental
testing with computer simulations to create a costeffective means to evaluate EV batteries
and provide important information that is difficult or impossible to obtain from purely
experimental measurements. The present simulators for the leadacid and NiMH batteries
are developed based on the fundamental principles governing their electrochemical
behaviors and are created using an advanced computational fluid dynamics (CFD)
technique. Computer simulations are validated by experimental data under the dynamic
stress test (DST) procedure for a leadacid battery module and a NiMH cell with good
agreement. Moreover, computer simulations reveal that the studied leadacid battery under
utilizes the active material by as much as 70% and the MH electrode of the NiMH cell is
overdesigned by about 30% under the simulated EV duty. Therefore, there is good
potential of increasing the specific energy and reducing the cost if batteries are optimized
for electric vehicles using a simulationbased design approach.
Keywords: Electric vehicle batteries, Dynamic testing, Computer simulation.
3
1. Introduction
One way to improve air quality in urban areas is to replace conventional ICengine
automobiles with electric vehicles (EVs) powered by rechargeable batteries. However, the
success of such a replacement depends greatly on the development of advanced batteries for
EV use. Requirements for EV batteries include: (1) a high specific energy for a driving
range comparable to that of a conventional ICengine powered vehicle; (2) a high specific
power for accelerating and hillclimbing capabilities; and (3) a long cycle life to assure an
acceptable cost. The performance of an EV is mainly controlled by the performance of its
battery system; hence, the evaluation of battery performance becomes essential in the
development of EVs.
Traditionally, the evaluation of EV batteries relies on experimental testing at both
laboratory and field scales. For example, to estimate the driving range of an EV, one can
use the peak power vs. depth of discharge (DOD) plot and the specific energy vs. specific
power plot (Ragone plot), provided the power vs. speed relationship for the vehicle is
known [1]. However, creating these plots involves a range of constant current and constant
power discharge tests, and thus, is timeconsuming and costly.
Various test schemes have been developed that aim at providing a direct measure of the
battery performance in an EV environment. Among them, the Simplified Federal Urban
Driving Schedule (SFUDS) [2] and the Dynamic Stress Test (DST) [3] are two power
based test procedures to evaluating the power delivering capability of EV batteries [48].
The difference between the DST and SFUDS lies in that the DST is defined by the
percentage of the peak power versus time, while the SFUDS employs the specific power
with the maximum of 79 W/kg versus time (see Fig. 1).
The facility needed to implement these test schemes for EV batteries is a computer
controlled battery cycler capable of controlling various modes of charge and discharge
regimes on the battery. Since the simulated driving cycles like DST and SFUDS are
characterized by high power loads and short pulses, and the power as a product of current
and voltage is usually measured only indirectly, accurate laboratory testing requires
sophisticated equipment [8]. In some circumstances, an EV battery may not be
comprehensively tested due to the insufficient power resolution or limitation of the battery
cycler [1]. Moreover, purely experimental testing does not enable innovative design and
optimization of batteries.
To overcome the abovementioned limitations in battery testing, the alternative approach
of assessing EV battery performance by a numerical modeling has been recognized for a
long time [9]. A numerical simulator based on the first principles can not only help to
improve the understanding of underlying mechanisms by which batteries work, but also
provide internal information that is difficult to obtain even from experiments of stateofthe
art. Such information is particularly valuable for the design and optimization of battery
systems. For instance, acid depletion in leadacid batteries and solid state diffusion in Ni
MH batteries are usually the limiting factors of battery performance, but it is difficult to
measure them in situ. On the other hand, they can be easily computed from simulations
provided battery simulators are fully validated by experimental data.
Attempts were made to use mathematical models to evaluate battery performance under
simulated driving cycles [913]. Tiedemann and Newman [9] numerically investigated the
thermal response of a leadacid battery under a simulated driving cycle similar to SAE J227
4
a/D (a currentbased test procedure). Ekdunge [10] simulated the RWEdriving cycle (also
a currentbased test procedure) using a simplified mathematical model for leadacid batteries
which treats the battery as three separate regions (i.e., the positive electrode, the separator,
and the negative electrode) and calculates electrode potentials empirically in each region to
significantly reduce computational time. Chen and Evans [11] numerically studied the
thermal behavior of a lithiumpolymer battery under the SFUDS profile. They employed
empirical equations for the electrochemical reaction rate. Doyle and Newman [12] outlined
a mathematical approach to evaluating the limitations of lithiumpolymer battery systems
but did not give simulation results for driving cycles. Most recently, Karden et al. [13]
presented a mathematical model for a valveregulated leadacid battery based on
fundamental principles. Their simulations were made for a highly dynamic currentbased
discharge cycle.
In this article, we present a new approach to evaluate EV batteries by integrating
computer simulation into experimental testing. While experimental testing is able to provide
battery performance data (e.g. peak power and capacity) under simulated driving cycles,
computer simulations can, in addition to validating these performance data, provide
information that is difficult or impossible to obtain from experimental measurements.
Integration of the two thus promises a costeffective but comprehensive approach to
evaluating EV batteries.
Section 2 presents numerical simulators for leadacid and NiMH batteries,
respectively, based on the fundamental principles of thermodynamics, electrochemistry,
species transport, solid state diffusion, and charge transfer [1417]. These firstprinciple
models are numerically solved using a Computational Fluid Dynamics (CFD) technique.
Boundary conditions in terms of power as required in both DST and SFUDS profiles can
be directly implemented in the present simulators, a unique feature different from virtually
all previous codes where currentbased conditions are required and simulating power
boundary conditions requires iterations. Section 3 briefly describes the traditional
experimental testing conducted in a laboratory. Section 4 demonstrates the use of these
CFD battery codes, in conjunction with the laboratory testing, for the evaluation of two EV
batteries under the standard procedure. Conclusions and areas of future research are
summarized in Section 5.
2 Computer Simulation
2.1 Description of Cell Models
A battery module is composed of cells connected in series and in parallel to satisfy the
requirement of voltage and capacity. The battery modeling therefore rests on the cell
modeling for the fundamental phenomena of electrochemistry and species transport. A
general micromacroscopic model for batteries and fuel cells based on these first principles
was recently developed by Wang et al. [16], which not only incorporates species and
charge conservation in the solid matrix and electrolyte phase, respectively, but also
accounts for the interfacial phenomena occurring at the electrode/electrolyte interface. The
model equations were derived based on the concentrated electrolyte theory and the volume
averaging approach. These equations address such physicochemical phenomena as: (1)
5
electrochemical reactions via the ButlerVolmer equation; (2) transport of species in each
phase due to diffusion, migration, and/or convection; (3) potential drop across either
electronic and/or ionicconducting phase; and (4) local changes in electrode structures.
Applications of the general model to leadacid, NiCd and NiMH batteries were
demonstrated in the previous work [1417], and model predictions were compared with a
number of previous theoretical and experimental results available in the literature.
In the present work, focus is placed on leadacid and NiMH batteries as they are
chosen to represent a wide array of EV batteries for demonstration of the unique capabilities
offered by the integrated simulation and testing approach.
There are vast differences in the operating mechanisms between leadacid and NiMH
batteries. While leadacid batteries undergo a solid phase transformation during the
electrochemical reaction, NiMH batteries involve the solid state diffusion of hydrogen in
the MH particles and the diffusion of protons in the solid nickel electrode. Moreover, the
acid is consumed during discharge and reproduced during charge at both positive and
negative electrodes in the leadacid cell. Hence, acid depletion is usually a limiting factor of
cell discharge, and the transport of acid from the reservoir and/or separator into the porous
electrode is key to a successful cell operation. In contrast, the alkaline solution serving as
the electrolyte in the NiMH cell is overall conserved during a cell operation, and the cell
performance is, instead, limited by the species diffusion in the solid phase. The diffusion
coefficients of active materials at both electrodes are therefore important parameters
dictating the performance of intercalative battery. More details of the cell models for lead
acid and NiMH batteries can be found in References [1417] and thus are not repeated
here.
2.2 Battery Models
Construction of a battery model from a single cell electrochemical model is necessary in
order to develop a simulator for EV battery modules. Battery modeling can be readily
accomplished by considering the battery configuration consisting of cells connected in
series and in parallel as well as accounting for the electrical resistances contributed by
intercell connectors, terminals, and cell grids, as schematically shown in Fig. 2. Let V, I,
and R denote the voltage, current, and electrical resistance, with subscripts b and c
referring to battery and cell, respectively. The battery voltage can then be expressed in
terms of the cell voltage as follows
V
b
= N
s
V
c
+ I
b
¸
¸
,
_
(N
s
1)R
ic
+2R
t
+
2N
s
R
g
N
p
(1)
where N stands for the number of cells, with subscripts s and p denoting cells in series and
in parallel, respectively. The current through the whole battery, I
b
, is related to the cell
current simply by
I
b
= N
p
I
c
(2)
Various electrical resistances appearing in Eq. (1) are schematically illustrated in Fig. 2.
While the intercell resistance R
ic
and the terminal resistance R
t
can be measured, the
resistance of cell grids, R
g
, is geometrydependent and usually can only be estimated from
battery test data [18].
6
Using Eqs. (1) and (2), and noting that power is the product of current and voltage, the
cell power is then related to the battery power by
P
c
=
P
b
N
p
N
s

¸
]
1
1 N
p
N
s
¸
¸
,
_
(N
s
1)R
ic
+2R
t
+
2N
s
R
g
N
p
I
2
c
= C
1
 C
2
I
2
c
(3)
where
C
1
=
P
b
N
p
N
s
(4)
and
C
2
=
N
p
N
s ¸
¸
,
_
(N
s
1)R
ic
+2R
t
+
2N
s
R
g
N
p
(5)
When the power applied to the battery, P
b
, is given, C
1
and C
2
become constant. Equation
(3) thus provides a link between the cell model and the module simulator. In other words,
given the power profile for the battery and various internal resistances, coefficients C
1
and
C
2
can be computed via Eqs. (4) and (5), and Eq. (3) then represents a boundary condition
for the cell model.
2.3 Challenge to Simulate Driving Cycles
The boundary condition for the cell model, Eq. (3), represents a numerical difficulty and
requires a special treatment, since the battery current and voltage are not known a priori.
Most previous models in the literature are based on the multiregion approach, that is, the
governing equations are formulated and solved separately in the positive electrode,
separator, and negative electrode. The solution for the whole cell is obtained by requiring a
common current density across each interface between two distinct regions. Such a solution
methodology is convenient when the current density is prescribed, but necessarily calls for
iterations in cases where the boundary condition is not currentbased.
The second numerical difficulty arises from the highly dynamic characteristics of
driving cycles. The battery current changes direction in seconds so that a small time step is
needed for a sufficient temporal resolution in order to obtain convergent solutions. These
features make the dynamic simulation of driving cycles much more timeconsuming than
that for a single constantcurrent charge or discharge.
2.4 Numerical Procedures
To overcome the abovementioned numerical difficulties, we employ an advanced
computational fluid dynamics (CFD) technique, in which the model equations are written in
the following general form
∂Φ
∂t
+ ∇⋅(vΦ) = ∇⋅(Γ∇Φ) + S (6)
transient convection diffusion source
where Φ stands for a general variable to be solved and can represent species concentration,
potential, phase fraction and so on in a battery model. Γ is a diffusion coefficient, and S a
source term which includes all terms that cannot be included in the previous terms. Some
examples of Sterm can be found in Ref. [14] for leadacid batteries.
7
The initial condition is set in terms of the initial state of charge (SOC). The boundary
condition can be in various modes: constant current, constant voltage, constant power, and
pulsed power loads, depending on practical applications. The current and voltage
conditions can readily be implemented, as done similarly in previous battery models.
Simulation of the power boundary condition in the literature was done by using an iterative
method in which a guess is taken for the current and a simulation of the cell is carried out to
compute the resulting voltage. The voltage is then used to determine an updated current
from the given power. The process is repeated until convergence. In the present work, we
describe an efficient approach to directly implement the power boundary condition without
needing the abovementioned iterative procedure. Our approach is inspired by the
Computational Fluid Dynamics techniques discussed in Ref. [19].
First, the mathematical definition of the current into or out of a cell is given as
follows
σ
∂φ
s
∂x
x=0
=
I
c
A
e
(7)
where σ is the electrical conductivity of the solid phase, φ
s
the solid phase potential, x the
coordinate defined along the cell width, and A
e
the electrode area. Discretize the cell into a
number of control volumes as schematically shown in Fig. 3, where 1 starts from the
current collector in the positive electrode and n designates the current collector in the
negative electrode. The cell voltage V
c
is then given by
V
c
= φ
s
(1)  φ
s
(n) (8)
Let φ
s
(n) = 0 as the reference potential, and apply the Taylor series expansion to
approximate the potential at node 1 in terms of the potential at node 2, Eq. (8) can be recast
into
V
c
= φ
s
(1) = φ
s
(2) 
∂φ
s
∂x
x=0
∆x (9)
where ∆x is the half thickness of the control volume adjacent to the current collector in the
positive electrode (see Fig. 3). Eliminating
∂φ
s
∂x
x=0
from Eq. (9) by use of Eq. (7), we
have
V
c
= φ
s
(2) +
∆x
σA
e
I
c (10)
Noting that the cell power is the product of cell current and voltage, and combining Eq. (3)
and (10) gives
¸
¸
,
_
C
2
+
∆x
σA
e
I
2
c
+ φ
s
(2) I
c
 C
1
= 0 (11)
Solving Eq. (11) for I
c
and substituting it into Eq. (7) yields
σ
∂φ
s
∂x
x=0
=
φ
s
(2)+
¸
]
1
1
(φ
s
(2))
2
+4C
1
¸
¸
,
_
C
2
+
∆x
σA
e
1/2
2
¸
¸
,
_
C
2
A
e
+
∆x
σ
(12)
Equation (12) is a current flux into the control volume immediately adjacent to the boundary
and thus can be equivalently implemented as a source term for the control volume around
node 2. Its nonlinear dependence upon the potential φ
s
of the control volume, however,
requires a "linearizing" procedure in order to speed up computations and convergence [19];
namely
8
σ
∂φ
s
∂x
x=0
= S
c
+ S
p
φ
s
(2) (13)
where S
c
stands for the constant part of the current flux, while S
p
is the linear coefficient
of φ
s
(2). There are different ways to linearize Eq. (12) into the form given by Eq. (13), but
S
p
must always be negative. Otherwise, a divergent solution or physically unrealistic
solutions would arise; see detailed discussion by Patankar [19]. For this reason, we use the
following linearization scheme:
S
p
= 
1
2
¸
¸
,
_
C
2
A
e
+
∆x
σ
(14)
S
c
=
¸
]
1
1
(φ
o
s
(2))
2
+4C
1
¸
¸
,
_
C
2
+
∆x
σA
e
1/2
2
¸
¸
,
_
C
2
A
e
+
∆x
σ
(15)
where φ
o
s
is the potential value in the previous iteration. Apparently, S
p
given by Eq. (14)
is always negative. Numerical tests found that implementation of the power boundary
condition using Eq. (13) not only avoids iterating the cell current but also significantly
reduces computational time and improves the computational stability.
The general differential equation (6) is discretized by the control volumebased finite
difference method [19], and the resulting set of linear algebraic equations is iteratively
solved. A single numerical solver for the general differential equation is repeatedly applied
to each scalar field over a control volume mesh. Furthermore, the present model equations
derived from the volumeaveraging technique are equally applicable in various regions such
as electrodes, electrolyte reservoir, and separator. Thus, matching conditions between
different regions are not necessary. Such a singledomain formulation offers considerable
simplifications in numerical simulations. While our simulators are capable of simulating
multidimensional behaviors (see e.g. [14]), only onedimensional simulations are needed
and carried out in the present work.
Stringent numerical tests were performed in every case to ensure that the solutions
were independent of the grid size and time step. For the cases to be illustrated below, it was
found that the typical number of grid lines across the cell width was about 60. The
equations were solved as a simultaneous set, and the convergence was considered to be
reached when the relative error in each field between two consecutive iterations is less than
10
6
. A typical DST cycle simulation required approximately 10 minutes of CPU time for a
leadacid battery and a NiMH cell on an HP B160L workstation. In comparison, the real
time testing of the batteries takes 77 minutes and 3 hours, respectively.
2.5 Case Studies
While a number of case studies have been performed for various types of batteries and
various brands of the same type of batteries, only two batteries are chosen for presentation
in this article to demonstrate the unique capability of the integrated simulation and testing
approach: a commercially available leadacid battery and a prototype NiMH cell. Other
case studies exhibited similar results.
9
The battery specification and operating conditions for the leadacid battery and the Ni
MH cell, respectively, are listed in Table 1. Batteryspecific parameters are given in Table
2, while kinetic and transport parameters also needed in simulations are taken from Ref.
[18] for the leadacid battery. The NiMH cell to be simulated is taken from Ref. [8] along
with the experimental data given therein. Since no detail on this cell geometry is given, the
electrode thicknesses are estimated from the given capacities with assumed typical electrode
capacities given by De Vidts et al. [21] for the metal hydride electrode and De Vidts and
White [22] for the nickel electrode. The sizes of active material particles and other
parameters including those of electrode kinetics used in numerical simulations are also
taken from their work [21, 22].
3 Experimental Testing
Experimental testing of the leadacid battery described in Table 1 was conducted using an
AeroVironment ABC150 battery diagnostic system, and the detailed test procedures were
given in Ref. [20] and thus not repeated here. The test equipment is versatile and can
handle battery modules as well as packs.
4 Results and Discussion
4.1 LeadAcid Battery
Figure 4(a) shows both experimental and simulated voltage profiles of the leadacid battery
under the DST cycles. The voltage drops to the lowest point during the 100% peak power
discharge and reaches the highest value during 50% peak power charge within each
subcycle of 360 s. The difference between the two extremes is about 4 V. Overall, the
voltage decreases with the number of subcycles, while it oscillates in phase with the power
pulses. The current gradually increases with DOD due to the drop in the battery voltage so
that the same power level is maintained; see Figure 4(b). A good agreement can be seen
between the experimental data and simulated results.
More importantly, computer simulations can provide detailed information to
understand the battery state in testing. As shown in Fig. 4, the experimental DST cycling is
terminated at the 80% DOD in order to avoid battery overdischarge and hence permanent
damage. However, the simulation can continue the DST cycling beyond the 80% DOD up
to a point when the battery is fully discharged (at t ≈ 100 min). This information permits us
to determine the limit of the battery under the DST cycle or in a real driving cycle in a non
destructive fashion.
Figure 5 shows the electrolyte concentration distributions across the leadacid cell at the
end of the experimental and simulated DST cycles, respectively. At the end of the
laboratory testing (solid line), the acid concentration exhibits a nonuniform profile, but
there is still a sufficient amount of acid inside the battery cell for further discharge.
However, the acid concentration at the end of the simulated DST cycle (dotted line) is
virtually zero at the positive electrode, thus causing a drastic drop in the cell voltage. There
also exists a sharp decrease of the acid concentration at the interface between the positive
electrode and the separator. This is because the local porosity is smallest corresponding to
10
the largest active material utilization (see Fig. 6) and the acid transport from the separator to
the positive electrode is somewhat choked. It appears beneficial for the leadacid battery to
increase its initial porosity of the positive electrode as much as practically possible.
The simulator can also provide data that can not be easily measured during the battery
testing. For instance, the local state of charge, which reflects the active material utilization,
is a key parameter to evaluate battery performance under a certain duty. Unfortunately, it is
difficult to measure. In contrast, a simulator based on first principles can generate such
information concurrently with the voltage and current profiles. Therefore, additional
information can be offered by the simulation to help evaluate the battery performance.
Figure 6 displays the active material utilization profiles in both electrodes of the lead
acid cell at the end of testing and simulation, respectively. The active material utilization
here is defined by the percentage of the maximum charge capacity as determined by
comparing the amount of active material participating in the electrode reaction to the initial
amount. It can be seen that the active material utilization distribution is severely nonuniform
at the positive electrode at the end of the DST cycle testing, with the active material near the
interface between the positive electrode and the separator being almost used up. The
utilization of active material at the negative electrode generally remains low, indicating a
large amount of active material at the negative electrode is unused. This result shows that
for the simulated driving cycle, the capacity of the negative electrode is overdesigned. It
should be possible to increase the specific energy of the leadacid battery simply by
reducing the size and hence weight of the negative electrode. In general, Figure 6 reveals
that the tested leadacid battery underutilized the active material by as much as 70% under
the simulated EV duty, indicating great potential of increasing the specific energy of the
leadacid battery and reducing the cost through the simulationbased design.
4.2 NiMH Cell
The DST cycle curves for the NiMH cell produced by experiment and simulation are
shown in Fig. 7. The general shape of the voltage curves is similar to that of the leadacid
battery, except for the initial quick drop which may be attributed to the hydrogen
absorptionadsorption equilibrium inside the metal hydride particles. The equilibrium or
opencircuit potential of intercalative cells such as NiMH and Liion is strongly dependent
upon the local state of charge. More specifically, it is a function of the surface concentration
of reactant species. However, such thermodynamic information is not available for the
studied cell, and thus this dependence is not included in the present simulation.
Nevertheless, a qualitative comparison between experimental and predicted data can be
made and a reasonable agreement is seen in Fig. 7. In particular, the end of the DST cycle
was measured at 2.97 hours versus predicted at 2.87 hours. Some discrepancies between
the measured and simulated voltages, as can be observed in Fig. 7, may possibly be
reduced if the input parameters are all given.
Figure 8 shows the predicted proton and hydrogen concentration profiles across the Ni
MH cell during the DST cycle. The cell voltage drops very quickly at the end of DST
simulation, which indicates the exhaust of the active material at the reaction site (i.e., the
electrode/electrolyte interface). The proton concentration reaches the maximum value at the
surface of nickel active material so that the cell can no longer be discharged, while the
11
average proton concentration within the active material particles are still below the
maximum value meaning more capacity available to incorporate protons and hence
discharge. Therefore, the performance of this NiMH cell is dictated by the microscopic
proton diffusion from the surface to the bulk of NiOOH particles. The availability of the
active material at the surface of metal hydride particles is also strongly dependent on the
hydrogen diffusion from the bulk to the surface as the surface hydrogen concentration
continually diminishes with discharge. The departure of the interfacial from the bulk
concentrations in microscopic particles is proportional to the transfer current density
produced at the reaction interface. As the battery goes deeper in DOD and withdraws a
larger current under the constantpower discharge, the difference between the bulk and
interfacial concentrations becomes larger and reaches the maximum at the end of the DST
cycle. This maximum difference could be used to identify the limiting mechanisms for the
operation of a specific cell.
Because the capacity ratio of the metal hydride electrode over the nickel electrode was
designed to be 1.5 [8], this NiMH cell is positive electrode limiting. It can be seen from
Fig. 8 that there is still plenty of hydrogen (ca. 40%) available in the bulk of the metal
hydride particles at the end of the DST cycle. Because the metal hydride material is
expensive, one may design a more compact and economical NiMH cell without degrading
the performance.
Unlike the leadacid battery which is limited by the depletion of acid at the positive
electrode, the electrolyte concentration in NiMH batteries has little effect on the
performance. During discharge, the alkaline is produced at the nickel electrode and
consumed at the MH electrode. The total alkaline is, however, conserved during battery
operation. In addition, in the DST cycle a discharge pulse is followed immediately by a rest
period and/or a charge pulse, therefore, the electrolyte concentration distribution across the
cell is expected to be more uniform than in a constant load discharge. Figure 9 shows that
the alkaline concentration is indeed close to the initial value during the DST cycle and does
not become a limiting factor for the performance of the NiMH cell.
5 Conclusions
An integrated simulation and testing approach is demonstrated for the evaluation of EV
batteries. This novel approach combines stateoftheart battery simulators and test facilities
to yield valuable information that enables innovative battery design and improvement.
The developed computer simulators are capable of dealing with various charge and
discharge regimes, including the powerbased test regimes like the DST and SFUDS
cycles. Comparisons were made between the simulation results and the experimental data
for a commercially available leadacid battery and a prototype NiMH cell. Good agreement
was obtained between the measured and simulated voltage curves under the DST cycle.
Moreover, simulated results reveal the potential of increasing specific energy and reducing
costs as well as the areas for design improvements. Here lies the significance of this
integrated simulation and testing approach.
Thermal effects during a simulated or actual driving cycle can be important to the
battery performance. For the case of the leadacid batteries, the temperature rise within the
driving cycle may result in the loss in capacity [23], and hence the reduction of cycle life.
12
The high thermal stresses imposed by the driving cycle will be incorporated in our
simulators in the future work.
ACKNOWLEDGMENTS
This work is supported by the Pennsylvania State University, University of Hawaii,
and the Defense Advanced Research Projects Agency (DARPA), Tactical Technology
Office, Electric Vehicle Technology Program, under the cooperative agreement No.
MDA9729520009.
List of symbols
A
e
electrode area,
cm
2
C
1
, C
2
coefficients introduced in Eqs. (4) and (5)
I current density, A/cm
2
N number of cells
P power density, W/cm
2
R electrical resistance, Ω
S source term
t time, s
V voltage, V
v velocity vector, cm/s
x coordinate in the direction of cell width, cm
Greek Symbols
Γ diffusion coefficient, cm
2
/s
∆x half thickness of a control volume along xdirection, cm
σ electronic conductivity of the solid matrix, S/cm
φ
s
electric potential in the solid active material phase, V
Φ a general variable in Eq. (6)
Subscripts
b battery
c cell
g grid
ic intercell connector
p cells in parallel
s cells in series
t terminal
13
REFERENCES
[1] B.E. Dickinson, D.H. Swan and T.R. Lalk, "Comparison of advanced battery
technologies for electric vehicles," in Electric Vehicle Power Systems
Hybrids/Batteries/Fuel Cells, SAE SP984 (1993) 111.
[2] EHP Battery Test Working Task Force, USDOE, Report No. DOE/ID 10146, Aug.
1988.
[3] G. Hunt, USABC Electric Vehicle Battery Test Procedures Manual, Rev. 2, USABC,
Jan. 1996.
[4] D.D. Brandt, "Driving cycle testing of electric vehicle batteries and systems", J.
Power Sources, 40 (1992) 7379.
[5] K.A. Striebel, F.R. McLarnon and E.J. Cairns, "Laboratoryscale evaluation of
secondary alkaline zinc batteries for electric vehicles," J. Power Sources, 47 (1994) 111.
[6] A.F. Hollenkamp, L.T. Lam, C.G. Phyland and N.C. Wilson, " Evaluation of
lead/acid batteries under simulated electricvehicle duty: development of design parameters
on the basis of SFUDS performance," J. Power Sources, 59 (1996) 177183.
[7] N. Ito, H. Inoue, K. Sato, N. Morishita and M. Ikoma, "Development of sealed
type Nickel/MetalHydride Battery for Electric Vehicle," in EVS13 (1996) 3136.
[8] D.E. Reisner, J.H. Cole and M. Klein, "Bipolar nickelmetal hydride EV battery,"
in EVS13 (1996) 3744.
[9] W. Tiedemann and J. Newman, "Mathematical modeling of phenomena
contributing to thermal rise in leadacid batteries used in electric vehicles," in Advances in
LeadAcid Batteries, K.R. Bullock and D. Pavlov, Editors, PV 8414 (1984) 336347.
[10] Per Ekdunge, "Simplified model for the leadacid battery," J. Power Sources, 46
(1993) 251262.
[11] Y. Chen and J.W. Evans, "Threedimensional thermal modeling of lithiumpolymer
batteries under galvanostatic discharge and dynamic power profile," J. Electrochem. Soc.,
141 (1994) 29472955.
[12] M. Doyle and J. Newman, "The use of mathematical modeling in the design of
lithium/polymer battery systems," Electrochimica Acta, 40 (1995) 21912196.
[13] E. Karden, P. Mauracher and F. Schope, "Electrochemical modeling of lead/acid
batteries under operating conditions of electric vehicles," J. Power Sources, 64 (1997)
175180.
14
[14] W.B. Gu, C.Y. Wang and B.Y. Liaw, "Numerical modeling of coupled
electrochemical and transport processes in leadacid batteries," J. Electrochem. Soc., 144
(1997) 20532061.
[15] B.Y. Liaw, W.B. Gu, and C.Y. Wang, "Numerical simulation of coupled
electrochemical and transport processes in battery system," Paper No. 97124 in
Proceedings of 32nd Intersociety Energy Conversion Engineering Conference, SAE,
Warrendale, PA, 1997
[16] C.Y. Wang, W.B. Gu and B.Y. Liaw, "Micromacroscopic coupled modeling of
batteries and fuel cells. part 1. model development," J. Electrochem. Soc., in press, 1998.
[17] W.B. Gu, C.Y. Wang and B.Y. Liaw, "Micromacroscopic coupled modeling of
batteries and fuel cells. part 2. Application to NiCd and NiMH cells," J. Electrochem.
Soc., in press, 1998.
[18]T.V. Nguyen, R.E. White and H. Gu, "The effect of separator design on the
discharge performance of a starved leadacid cell," J. Electrochem. Soc., 137 (1990)
29983004.
[19]S.V. Patankar, Numerical Heat Transfer and Fluid Flow, Hemisphere, Washington
D.C. (1980).
[20] M. Geng and B.Y. Liaw, Test Report HBTL004, Hawaii Battery Testing
Laboratory, 1997.
[21]P. De Vidts, J. Delgado and R.E. White, " Mathematical modeling for the discharge
of a metal hydride electrode," J. Electrochem. Soc., 142 (1995) 40064013.
[22]P. De Vidts and R.E. White, "Mathematical modeling of a nickelcadmium cell:
Proton diffusion in the nickel electrode," J. Electrochem. Soc., 142 (1995) 15091519.
[23]B.L. McKinney and G.H. Brilmyer, "The impact of urban driving schedules on the
thermal management of leadacid batteries for electric vehicles," in Advances in LeadAcid
Batteries, K.R. Bullock and D. Pavlov, Editors, PV 8414 (1984) 348359.
15
List of Tables
Table 1. Battery specifications and operating conditions
Table 2. Parameters used in the simulations
16
Table 1. Battery specifications and operating conditions
Battery type LeadAcid (module) [20] NiMH (single cell) [8]
Cells in series 6 1
Cells in parallel 8 1
Capacity (C/3 rate), Ah 85 1.5
Capacity ratio (/+) 1.4 1.5
Dimension of electrode 6.5 in. × 6.0 in. 3 in. × 3 in.
Electrode area, cm
2
251.6 58.1
Thickness, cm
positive electrode
separator
negative electrode
0.159
0.159
0.159
0.045
*
0.025
0.040
*
Electrolyte concentration, M 5.1 7.1
Peak discharge power, W 3000 4.4
**
Operating temperature, ˚C 23 25
* calculated from the given capacity and capacity ratio.
** calculated from the given maximum current of 4 A.
17
Table 2. Parameters used in simulations
Battery type leadacid (module) NiMH (single cell) [2122]
Porosity at fully charged state
positive electrode 0.61 0.44
separator 0.92 0.68
negative electrode 0.55 0.30
Volume fraction of inert material
PbO
2
electrode 0.05 N/A
Pb electrode 0.10 N/A
Radius of active material particles, cm
nickel cylinder (substrate) N/A 2.9×10
4
(1.5×10
4
)
MH sphere N/A 10
3
Maximum specific surface area, cm
2
/cm
3
positive electrode 2.3×10
5
3864
negative electrode 2.3×10
4
2100
Electrical resistance, Ω
intercell connector 2×10
4
N/A
terminal 2×10
4
N/A
cell grid 2×10
3
N/A
18
List of Figures
Figure 1 Dynamic stress test (DST) cycle and simplified federal urban driving schedule
(SFUDS) profile
Figure 2 Schematic configuration of a battery module
Figure 3 Schematic of control volumes and computational meshes for a battery cell
Figure 4 DST curves of the leadacid battery module at 3.0 kW peak power and ambient
temperature. Comparison of the experimental data with simulated results.
Figure 5 Electrolyte concentration profiles across a leadacid battery cell at t = 76.06 min
(corresponding to the end of the DST cycle testing) and t =99.97 min (at the end
of the simulated DST cycle).
Figure 6 Electrode active material utilization profiles across the leadacid cell at t = 76.06
min (corresponding to the end of the DST cycle testing) and t = 99.97 min (at
the end of the simulated DST cycle).
Figure 7 Comparison of measured and simulated DST voltage profiles of a NiMH cell
Figure 8 Predicted proton and hydrogen concentration profiles across the NiMH cell
during the DST cycle at t = 1.77 h (corresponding to the cutoff voltage of 1.0
V) and t = 2.87 h (when the cell voltage drops below 0.4 V).
Figure 9 Simulated electrolyte concentration profiles across the NiMH cell during the
DST cycle at t = 1.77 h (corresponding to the cutoff voltage of 1.0 V) and t =
2.87 h (when the cell voltage drops below 0.4 V).
Computer simulation. computer simulations reveal that the studied leadacid battery underutilizes the active material by as much as 70% and the MH electrode of the NiMH cell is overdesigned by about 30% under the simulated EV duty. Computer simulations are validated by experimental data under the dynamic stress test (DST) procedure for a leadacid battery module and a NiMH cell with good agreement. Keywords: Electric vehicle batteries. The present simulators for the leadacid and NiMH batteries are developed based on the fundamental principles governing their electrochemical behaviors and are created using an advanced computational fluid dynamics (CFD) technique. Therefore. Dynamic testing. Moreover. 2 . This new approach combines traditional experimental testing with computer simulations to create a costeffective means to evaluate EV batteries and provide important information that is difficult or impossible to obtain from purely experimental measurements. there is good potential of increasing the specific energy and reducing the cost if batteries are optimized for electric vehicles using a simulationbased design approach.ABSTRACT An integrated simulation and testing approach is presented to evaluate batteries for electric vehicle (EV) applications.
and (3) a long cycle life to assure an acceptable cost. while the SFUDS employs the specific power with the maximum of 79 W/kg versus time (see Fig. Traditionally. However. the alternative approach of assessing EV battery performance by a numerical modeling has been recognized for a long time [9]. Various test schemes have been developed that aim at providing a direct measure of the battery performance in an EV environment. the evaluation of EV batteries relies on experimental testing at both laboratory and field scales. creating these plots involves a range of constant current and constant power discharge tests. the evaluation of battery performance becomes essential in the development of EVs. the success of such a replacement depends greatly on the development of advanced batteries for EV use. to estimate the driving range of an EV. and the power as a product of current and voltage is usually measured only indirectly. accurate laboratory testing requires sophisticated equipment [8]. The facility needed to implement these test schemes for EV batteries is a computercontrolled battery cycler capable of controlling various modes of charge and discharge regimes on the battery. Moreover. but also provide internal information that is difficult to obtain even from experiments of stateoftheart. However. Since the simulated driving cycles like DST and SFUDS are characterized by high power loads and short pulses. depth of discharge (DOD) plot and the specific energy vs. Requirements for EV batteries include: (1) a high specific energy for a driving range comparable to that of a conventional ICengine powered vehicle. the Simplified Federal Urban Driving Schedule (SFUDS) [2] and the Dynamic Stress Test (DST) [3] are two powerbased test procedures to evaluating the power delivering capability of EV batteries [48]. speed relationship for the vehicle is known [1]. purely experimental testing does not enable innovative design and optimization of batteries. A numerical simulator based on the first principles can not only help to improve the understanding of underlying mechanisms by which batteries work. and thus. To overcome the abovementioned limitations in battery testing. hence. Attempts were made to use mathematical models to evaluate battery performance under simulated driving cycles [913]. specific power plot (Ragone plot). they can be easily computed from simulations provided battery simulators are fully validated by experimental data. Such information is particularly valuable for the design and optimization of battery systems. For example. On the other hand. is timeconsuming and costly. but it is difficult to measure them in situ. Tiedemann and Newman [9] numerically investigated the thermal response of a leadacid battery under a simulated driving cycle similar to SAE J227 3 . acid depletion in leadacid batteries and solid state diffusion in NiMH batteries are usually the limiting factors of battery performance. Among them. one can use the peak power vs.1. an EV battery may not be comprehensively tested due to the insufficient power resolution or limitation of the battery cycler [1]. Introduction One way to improve air quality in urban areas is to replace conventional ICengine automobiles with electric vehicles (EVs) powered by rechargeable batteries. The performance of an EV is mainly controlled by the performance of its battery system. The difference between the DST and SFUDS lies in that the DST is defined by the percentage of the peak power versus time. provided the power vs. For instance. In some circumstances. (2) a high specific power for accelerating and hillclimbing capabilities. 1).
Ekdunge [10] simulated the RWEdriving cycle (also a currentbased test procedure) using a simplified mathematical model for leadacid batteries which treats the battery as three separate regions (i. for the evaluation of two EV batteries under the standard procedure. we present a new approach to evaluate EV batteries by integrating computer simulation into experimental testing. which not only incorporates species and charge conservation in the solid matrix and electrolyte phase. Chen and Evans [11] numerically studied the thermal behavior of a lithiumpolymer battery under the SFUDS profile.a/D (a currentbased test procedure). based on the fundamental principles of thermodynamics. in conjunction with the laboratory testing. Karden et al. respectively. Integration of the two thus promises a costeffective but comprehensive approach to evaluating EV batteries. in addition to validating these performance data. They employed empirical equations for the electrochemical reaction rate. Their simulations were made for a highly dynamic currentbased discharge cycle. Section 4 demonstrates the use of these CFD battery codes.g. [16]. solid state diffusion.1 Description of Cell Models A battery module is composed of cells connected in series and in parallel to satisfy the requirement of voltage and capacity. The model equations were derived based on the concentrated electrolyte theory and the volumeaveraging approach. While experimental testing is able to provide battery performance data (e. species transport. Conclusions and areas of future research are summarized in Section 5. a unique feature different from virtually all previous codes where currentbased conditions are required and simulating power boundary conditions requires iterations. These equations address such physicochemical phenomena as: (1) 4 . A general micromacroscopic model for batteries and fuel cells based on these first principles was recently developed by Wang et al. provide information that is difficult or impossible to obtain from experimental measurements. These firstprinciple models are numerically solved using a Computational Fluid Dynamics (CFD) technique. and charge transfer [1417]. Section 2 presents numerical simulators for leadacid and NiMH batteries. Boundary conditions in terms of power as required in both DST and SFUDS profiles can be directly implemented in the present simulators. Section 3 briefly describes the traditional experimental testing conducted in a laboratory. electrochemistry. computer simulations can. The battery modeling therefore rests on the cell modeling for the fundamental phenomena of electrochemistry and species transport. peak power and capacity) under simulated driving cycles. In this article. 2 Computer Simulation 2. and the negative electrode) and calculates electrode potentials empirically in each region to significantly reduce computational time.. [13] presented a mathematical model for a valveregulated leadacid battery based on fundamental principles.e. the positive electrode. Doyle and Newman [12] outlined a mathematical approach to evaluating the limitations of lithiumpolymer battery systems but did not give simulation results for driving cycles. Most recently. respectively. the separator. but also accounts for the interfacial phenomena occurring at the electrode/electrolyte interface.
2. While leadacid batteries undergo a solid phase transformation during the electrochemical reaction. More details of the cell models for leadacid and NiMH batteries can be found in References [1417] and thus are not repeated here. Applications of the general model to leadacid. terminals.2 Battery Models Construction of a battery model from a single cell electrochemical model is necessary in order to develop a simulator for EV battery modules. and model predictions were compared with a number of previous theoretical and experimental results available in the literature. NiCd and NiMH batteries were demonstrated in the previous work [1417]. the acid is consumed during discharge and reproduced during charge at both positive and negative electrodes in the leadacid cell. and/or convection. NiMH batteries involve the solid state diffusion of hydrogen in the MH particles and the diffusion of protons in the solid nickel electrode. Let V. (1) are schematically illustrated in Fig. and (4) local changes in electrode structures. current. and R denote the voltage.and/or ionicconducting phase. is related to the cell current simply by Ib = Np Ic (2) Various electrical resistances appearing in Eq. (3) potential drop across either electronic. Battery modeling can be readily accomplished by considering the battery configuration consisting of cells connected in series and in parallel as well as accounting for the electrical resistances contributed by intercell connectors. 2. acid depletion is usually a limiting factor of cell discharge. The current through the whole battery. There are vast differences in the operating mechanisms between leadacid and NiMH batteries. and the cell performance is. In the present work. In contrast. is geometrydependent and usually can only be estimated from battery test data [18]. instead. I. the alkaline solution serving as the electrolyte in the NiMH cell is overall conserved during a cell operation. as schematically shown in Fig. Hence. (2) transport of species in each phase due to diffusion. with subscripts s and p denoting cells in series and in parallel. focus is placed on leadacid and NiMH batteries as they are chosen to represent a wide array of EV batteries for demonstration of the unique capabilities offered by the integrated simulation and testing approach. with subscripts b and c referring to battery and cell. respectively. The diffusion coefficients of active materials at both electrodes are therefore important parameters dictating the performance of intercalative battery. The battery voltage can then be expressed in terms of the cell voltage as follows 2N s R (1) Vb = NsVc + Ib (N s 1)R ic + 2R t + N g p where N stands for the number of cells. and the transport of acid from the reservoir and/or separator into the porous electrode is key to a successful cell operation. the resistance of cell grids. Ib. respectively. 5 . and cell grids. and electrical resistance. limited by the species diffusion in the solid phase. Rg. 2. Moreover. While the intercell resistance Ric and the terminal resistance Rt can be measured. migration.electrochemical reactions via the ButlerVolmer equation.
represents a numerical difficulty and requires a special treatment. (3) then represents a boundary condition for the cell model. Γ is a diffusion coefficient. Eq. Such a solution methodology is convenient when the current density is prescribed. separator. The battery current changes direction in seconds so that a small time step is needed for a sufficient temporal resolution in order to obtain convergent solutions. but necessarily calls for iterations in cases where the boundary condition is not currentbased. 6 . N (N s 1)R ic + 2R t + N Ic = C1 . given the power profile for the battery and various internal resistances. (3). The solution for the whole cell is obtained by requiring a common current density across each interface between two distinct regions.Using Eqs.3 Challenge to Simulate Driving Cycles The boundary condition for the cell model. Pb. and Eq. 2. Most previous models in the literature are based on the multiregion approach. phase fraction and so on in a battery model. we employ an advanced computational fluid dynamics (CFD) technique. C1 and C2 become constant. 2. and negative electrode. since the battery current and voltage are not known a priori. potential. (4) and (5). Equation (3) thus provides a link between the cell model and the module simulator. and S a source term which includes all terms that cannot be included in the previous terms. Some examples of Sterm can be found in Ref. coefficients C1 and C2 can be computed via Eqs. the cell power is then related to the battery power by 2N s R g 2 P Np 2 Pc = N b . and noting that power is the product of current and voltage. In other words. in which the model equations are written in the following general form ∂Φ + ∇⋅(vΦ) = ∇⋅(Γ∇Φ) + S (6) ∂t transient convection diffusion source where Φ stands for a general variable to be solved and can represent species concentration. These features make the dynamic simulation of driving cycles much more timeconsuming than that for a single constantcurrent charge or discharge. is given.C2Ic (3) Ns s p p where Pb (4) C1 = N N p s and N 2N s R C2 = Np (N s 1)R ic + 2R t + N g (5) s p When the power applied to the battery. The second numerical difficulty arises from the highly dynamic characteristics of driving cycles.4 Numerical Procedures To overcome the abovementioned numerical difficulties. that is. the governing equations are formulated and solved separately in the positive electrode. (1) and (2). [14] for leadacid batteries.
The initial condition is set in terms of the initial state of charge (SOC). where 1 starts from the current collector in the positive electrode and n designates the current collector in the negative electrode. Simulation of the power boundary condition in the literature was done by using an iterative method in which a guess is taken for the current and a simulation of the cell is carried out to compute the resulting voltage. φs the solid phase potential. First. and apply the Taylor series expansion to approximate the potential at node 1 in terms of the potential at node 2. (9) by use of Eq. however. (8) can be recast into ∂φ (9) V c = φs(1) = φs(2) . Its nonlinear dependence upon the potential φs of the control volume. Eliminating ∂xsx=0 from Eq. The process is repeated until convergence. requires a "linearizing" procedure in order to speed up computations and convergence [19]. The current and voltage conditions can readily be implemented. 3). 3. the mathematical definition of the current into or out of a cell is given as follows Ic ∂φ (7) σ ∂xs x=0 = A e where σ is the electrical conductivity of the solid phase. constant voltage. Our approach is inspired by the Computational Fluid Dynamics techniques discussed in Ref. (7) yields ∆x 1/2 φ s (2) + (φ s (2)) 2 + 4C 1 C 2 + σAe ∂φ s σ ∂x x=0 = (12) C A + ∆x 2 2 e σ Equation (12) is a current flux into the control volume immediately adjacent to the boundary and thus can be equivalently implemented as a source term for the control volume around node 2. The cell voltage Vc is then given by Vc = φs(1) . In the present work. depending on practical applications.∂xsx=0 ∆x where ∆x is the half thickness of the control volume adjacent to the current collector in the ∂φ positive electrode (see Fig.φs(n) (8) Let φ s (n) = 0 as the reference potential. we describe an efficient approach to directly implement the power boundary condition without needing the abovementioned iterative procedure. The voltage is then used to determine an updated current from the given power. Eq. x the coordinate defined along the cell width. and Ae the electrode area. [19]. we have ∆x I Vc = φs(2) + (10) σAe c Noting that the cell power is the product of cell current and voltage. and combining Eq. (7). Discretize the cell into a number of control volumes as schematically shown in Fig. namely 7 . (3) and (10) gives C + ∆x I2 + φ (2) I . (11) for Ic and substituting it into Eq. constant power.C = 0 (11) 1 s c 2 σAe c Solving Eq. The boundary condition can be in various modes: constant current. as done similarly in previous battery models. and pulsed power loads.
respectively. a divergent solution or physically unrealistic solutions would arise. electrolyte reservoir. but S p must always be negative. A typical DST cycle simulation required approximately 10 minutes of CPU time for a leadacid battery and a NiMH cell on an HP B160L workstation. only two batteries are chosen for presentation in this article to demonstrate the unique capability of the integrated simulation and testing approach: a commercially available leadacid battery and a prototype NiMH cell. In comparison.g. and the convergence was considered to be reached when the relative error in each field between two consecutive iterations is less than 106. Such a singledomain formulation offers considerable simplifications in numerical simulations. Sp given by Eq. Stringent numerical tests were performed in every case to ensure that the solutions were independent of the grid size and time step. There are different ways to linearize Eq. While our simulators are capable of simulating multidimensional behaviors (see e. and separator. and the resulting set of linear algebraic equations is iteratively solved. For this reason. (12) into the form given by Eq. the realtime testing of the batteries takes 77 minutes and 3 hours. Apparently. we use the following linearization scheme: 1 (14) Sp = C A + ∆x 2 2 e σ o ∆x 1/2 (φ s (2)) 2 + 4C 1 C 2 + σAe (15) Sc = ∆x 2 C 2 A e + σ o where φs is the potential value in the previous iteration. [14]). see detailed discussion by Patankar [19]. Other case studies exhibited similar results. Furthermore. it was found that the typical number of grid lines across the cell width was about 60. For the cases to be illustrated below. A single numerical solver for the general differential equation is repeatedly applied to each scalar field over a control volume mesh. Numerical tests found that implementation of the power boundary condition using Eq. (13) not only avoids iterating the cell current but also significantly reduces computational time and improves the computational stability. only onedimensional simulations are needed and carried out in the present work. Thus. while Sp is the linear coefficient of φs(2).5 Case Studies While a number of case studies have been performed for various types of batteries and various brands of the same type of batteries. (13). The general differential equation (6) is discretized by the control volumebased finite difference method [19]. 2. The equations were solved as a simultaneous set. matching conditions between different regions are not necessary.∂φ σ ∂xs x=0 = Sc + Spφs(2) (13) where Sc stands for the constant part of the current flux. the present model equations derived from the volumeaveraging technique are equally applicable in various regions such as electrodes. (14) is always negative. Otherwise. 8 .
Since no detail on this cell geometry is given. The voltage drops to the lowest point during the 100% peak power discharge and reaches the highest value during 50% peak power charge within each subcycle of 360 s. The current gradually increases with DOD due to the drop in the battery voltage so that the same power level is maintained. A good agreement can be seen between the experimental data and simulated results. 4. while it oscillates in phase with the power pulses. There also exists a sharp decrease of the acid concentration at the interface between the positive electrode and the separator. 4 Results and Discussion 4. the simulation can continue the DST cycling beyond the 80% DOD up to a point when the battery is fully discharged (at t ≈ 100 min). [21] for the metal hydride electrode and De Vidts and White [22] for the nickel electrode. computer simulations can provide detailed information to understand the battery state in testing. [18] for the leadacid battery. However. This information permits us to determine the limit of the battery under the DST cycle or in a real driving cycle in a nondestructive fashion. 3 Experimental Testing Experimental testing of the leadacid battery described in Table 1 was conducted using an AeroVironment ABC150 battery diagnostic system. Overall. 22]. The NiMH cell to be simulated is taken from Ref. the voltage decreases with the number of subcycles. the experimental DST cycling is terminated at the 80% DOD in order to avoid battery overdischarge and hence permanent damage.The battery specification and operating conditions for the leadacid battery and the NiMH cell. while kinetic and transport parameters also needed in simulations are taken from Ref. respectively. but there is still a sufficient amount of acid inside the battery cell for further discharge.1 LeadAcid Battery Figure 4(a) shows both experimental and simulated voltage profiles of the leadacid battery under the DST cycles. The difference between the two extremes is about 4 V. This is because the local porosity is smallest corresponding to 9 . the acid concentration exhibits a nonuniform profile. Figure 5 shows the electrolyte concentration distributions across the leadacid cell at the end of the experimental and simulated DST cycles. are listed in Table 1. and the detailed test procedures were given in Ref. respectively. However. The sizes of active material particles and other parameters including those of electrode kinetics used in numerical simulations are also taken from their work [21. thus causing a drastic drop in the cell voltage. More importantly. At the end of the laboratory testing (solid line). the acid concentration at the end of the simulated DST cycle (dotted line) is virtually zero at the positive electrode. As shown in Fig. Batteryspecific parameters are given in Table 2. see Figure 4(b). [20] and thus not repeated here. The test equipment is versatile and can handle battery modules as well as packs. [8] along with the experimental data given therein. the electrode thicknesses are estimated from the given capacities with assumed typical electrode capacities given by De Vidts et al.
the electrode/electrolyte interface). the end of the DST cycle was measured at 2. 4. a qualitative comparison between experimental and predicted data can be made and a reasonable agreement is seen in Fig.2 NiMH Cell The DST cycle curves for the NiMH cell produced by experiment and simulation are shown in Fig. The general shape of the voltage curves is similar to that of the leadacid battery. In general. Therefore. the capacity of the negative electrode is overdesigned. Some discrepancies between the measured and simulated voltages. as can be observed in Fig. The equilibrium or opencircuit potential of intercalative cells such as NiMH and Liion is strongly dependent upon the local state of charge. such thermodynamic information is not available for the studied cell. and thus this dependence is not included in the present simulation.97 hours versus predicted at 2. additional information can be offered by the simulation to help evaluate the battery performance.the largest active material utilization (see Fig. 7. is a key parameter to evaluate battery performance under a certain duty. it is a function of the surface concentration of reactant species. indicating great potential of increasing the specific energy of the leadacid battery and reducing the cost through the simulationbased design. Figure 6 displays the active material utilization profiles in both electrodes of the leadacid cell at the end of testing and simulation. More specifically. In particular. respectively. However. except for the initial quick drop which may be attributed to the hydrogen absorptionadsorption equilibrium inside the metal hydride particles. which indicates the exhaust of the active material at the reaction site (i. The cell voltage drops very quickly at the end of DST simulation. The proton concentration reaches the maximum value at the surface of nickel active material so that the cell can no longer be discharged. The simulator can also provide data that can not be easily measured during the battery testing. Figure 8 shows the predicted proton and hydrogen concentration profiles across the NiMH cell during the DST cycle. It can be seen that the active material utilization distribution is severely nonuniform at the positive electrode at the end of the DST cycle testing. 6) and the acid transport from the separator to the positive electrode is somewhat choked. For instance. The active material utilization here is defined by the percentage of the maximum charge capacity as determined by comparing the amount of active material participating in the electrode reaction to the initial amount. which reflects the active material utilization. Figure 6 reveals that the tested leadacid battery underutilized the active material by as much as 70% under the simulated EV duty.e. 7. 7. it is difficult to measure. with the active material near the interface between the positive electrode and the separator being almost used up. while the 10 . indicating a large amount of active material at the negative electrode is unused. It should be possible to increase the specific energy of the leadacid battery simply by reducing the size and hence weight of the negative electrode. This result shows that for the simulated driving cycle. a simulator based on first principles can generate such information concurrently with the voltage and current profiles. In contrast. It appears beneficial for the leadacid battery to increase its initial porosity of the positive electrode as much as practically possible. may possibly be reduced if the input parameters are all given. Unfortunately.87 hours. the local state of charge.. Nevertheless. The utilization of active material at the negative electrode generally remains low.
average proton concentration within the active material particles are still below the maximum value meaning more capacity available to incorporate protons and hence discharge. however. 5 Conclusions An integrated simulation and testing approach is demonstrated for the evaluation of EV batteries. This maximum difference could be used to identify the limiting mechanisms for the operation of a specific cell. Unlike the leadacid battery which is limited by the depletion of acid at the positive electrode. Because the metal hydride material is expensive. the temperature rise within the driving cycle may result in the loss in capacity [23]. the electrolyte concentration distribution across the cell is expected to be more uniform than in a constant load discharge. the difference between the bulk and interfacial concentrations becomes larger and reaches the maximum at the end of the DST cycle. Moreover. Figure 9 shows that the alkaline concentration is indeed close to the initial value during the DST cycle and does not become a limiting factor for the performance of the NiMH cell. The total alkaline is. Because the capacity ratio of the metal hydride electrode over the nickel electrode was designed to be 1. Comparisons were made between the simulation results and the experimental data for a commercially available leadacid battery and a prototype NiMH cell. simulated results reveal the potential of increasing specific energy and reducing costs as well as the areas for design improvements. conserved during battery operation. This novel approach combines stateoftheart battery simulators and test facilities to yield valuable information that enables innovative battery design and improvement. the alkaline is produced at the nickel electrode and consumed at the MH electrode. including the powerbased test regimes like the DST and SFUDS cycles. this NiMH cell is positive electrode limiting. During discharge. one may design a more compact and economical NiMH cell without degrading the performance. and hence the reduction of cycle life. in the DST cycle a discharge pulse is followed immediately by a rest period and/or a charge pulse. Therefore. Good agreement was obtained between the measured and simulated voltage curves under the DST cycle. In addition. therefore.5 [8]. The departure of the interfacial from the bulk concentrations in microscopic particles is proportional to the transfer current density produced at the reaction interface. 8 that there is still plenty of hydrogen (ca. the performance of this NiMH cell is dictated by the microscopic proton diffusion from the surface to the bulk of NiOOH particles. It can be seen from Fig. The developed computer simulators are capable of dealing with various charge and discharge regimes. The availability of the active material at the surface of metal hydride particles is also strongly dependent on the hydrogen diffusion from the bulk to the surface as the surface hydrogen concentration continually diminishes with discharge. For the case of the leadacid batteries. 40%) available in the bulk of the metal hydride particles at the end of the DST cycle. 11 . As the battery goes deeper in DOD and withdraws a larger current under the constantpower discharge. Thermal effects during a simulated or actual driving cycle can be important to the battery performance. Here lies the significance of this integrated simulation and testing approach. the electrolyte concentration in NiMH batteries has little effect on the performance.
cm electronic conductivity of the solid matrix. cm/s coordinate in the direction of cell width. cm2 coefficients introduced in Eqs. cm2/s half thickness of a control volume along xdirection. MDA9729520009. C2 I N P R S t V v x electrode area. (6) Subscripts b c g ic p s t battery cell grid intercell connector cells in parallel cells in series terminal 12 . (4) and (5) current density.The high thermal stresses imposed by the driving cycle will be incorporated in our simulators in the future work. List of symbols Ae C 1 . Electric Vehicle Technology Program. V a general variable in Eq. Ω source term time. W/cm2 electrical resistance. ACKNOWLEDGMENTS This work is supported by the Pennsylvania State University. S/cm electric potential in the solid active material phase. A/cm2 number of cells power density. cm Greek Symbols Γ ∆x σ φs Φ diffusion coefficient. Tactical Technology Office. under the cooperative agreement No. and the Defense Advanced Research Projects Agency (DARPA). University of Hawaii. V velocity vector. s voltage.
Klein.REFERENCES [1] B.F. J. Ito. Wilson. J. Brandt. "Threedimensional thermal modeling of lithiumpolymer batteries under galvanostatic discharge and dynamic power profile. F. D. Cole and M. PV 8414 (1984) 336347. Striebel. Electrochem.T. Report No. "Comparison of advanced battery technologies for electric vehicles. [12] M." J. 64 (1997) 175180. DOE/ID 10146. 2." J. Lam. K. Phyland and N. Swan and T. Newman." J.R. [10] Per Ekdunge.E. Newman.C. 59 (1996) 177183. P. Hollenkamp." in EVS13 (1996) 3136. Jan. "Mathematical modeling of phenomena contributing to thermal rise in leadacid batteries used in electric vehicles. [13] E.W. Dickinson." J. Mauracher and F. Bullock and D. [3] G.E.A. USABC. 141 (1994) 29472955. "Laboratoryscale evaluation of secondary alkaline zinc batteries for electric vehicles.H. H. Aug. "The use of mathematical modeling in the design of lithium/polymer battery systems." in Advances in LeadAcid Batteries. "Simplified model for the leadacid battery. Power Sources. Tiedemann and J. Pavlov. 40 (1995) 21912196. Power Sources." in EVS13 (1996) 3744. Morishita and M. Evans. Inoue.D." J. USABC Electric Vehicle Battery Test Procedures Manual. Hunt.H. [2] EHP Battery Test Working Task Force. K. [6] A. [8] D. Ikoma. Lalk. Editors. Sato. [11] Y. 13 .. 1988. Cairns. Power Sources.G. 46 (1993) 251262." in Electric Vehicle Power SystemsHybrids/Batteries/Fuel Cells. SAE SP984 (1993) 111. C. [9] W.J. Doyle and J. " Evaluation of lead/acid batteries under simulated electricvehicle duty: development of design parameters on the basis of SFUDS performance. L. [4] D. 40 (1992) 7379. Reisner. USDOE." Electrochimica Acta. "Driving cycle testing of electric vehicle batteries and systems". Schope. Soc. [5] K. Power Sources. Power Sources. 1996. Karden. Rev.R. N. [7] N. Chen and J. "Bipolar nickelmetal hydride EV battery. "Electrochemical modeling of lead/acid batteries under operating conditions of electric vehicles. "Development of sealedtype Nickel/MetalHydride Battery for Electric Vehicle. 47 (1994) 111. McLarnon and E.R.
" J. Brilmyer.Y.E. [19]S. Liaw. Hawaii Battery Testing Laboratory. part 2. W. SAE. Soc. [22]P. [18]T. De Vidts. Gu." J. Electrochem. J. Soc. White.Y. Gu and B. [21]P.E.Y. and C. in press. Soc.Y. "The effect of separator design on the discharge performance of a starved leadacid cell. Hemisphere. Liaw.H.B.Y. Editors.Y." J.V. PA. Soc. [15] B. Patankar. Gu. Soc. 1997 [16] C. Electrochem. Bullock and D. Liaw.Y.L.R. Gu. Wang and B. "Micromacroscopic coupled modeling of batteries and fuel cells. Application to NiCd and NiMH cells. " Mathematical modeling for the discharge of a metal hydride electrode. Electrochem. C. De Vidts and R. 1997.B. Delgado and R.Y. part 1. "Mathematical modeling of a nickelcadmium cell: Proton diffusion in the nickel electrode.Y. Electrochem. in press.B." J. Numerical Heat Transfer and Fluid Flow. Electrochem. "Micromacroscopic coupled modeling of batteries and fuel cells. Wang.E. (1980). 97124 in Proceedings of 32nd Intersociety Energy Conversion Engineering Conference. [23]B. White and H. 1998. PV 8414 (1984) 348359. Liaw. Geng and B. Wang and B. "Numerical simulation of coupled electrochemical and transport processes in battery system. White. 142 (1995) 40064013. Soc. "The impact of urban driving schedules on the thermal management of leadacid batteries for electric vehicles.C. 1998. "Numerical modeling of coupled electrochemical and transport processes in leadacid batteries..B. Test Report HBTL004. [17] W. Liaw. Electrochem." Paper No. 144 (1997) 20532061. W. 142 (1995) 15091519. R. Gu.. [20] M. Washington D." J.[14] W." J. Warrendale.." in Advances in LeadAcid Batteries... C. Wang. K. 137 (1990) 29983004. 14 . model development. McKinney and G. Pavlov.V.. Nguyen.
Battery specifications and operating conditions Table 2. Parameters used in the simulations 15 .List of Tables Table 1.
4 6.5 3 in.045* 0.5 1. 16 . Ah Capacity ratio (/+) Dimension of electrode Electrode area.159 0.1 4.1 Peak discharge power. × 6. Battery specifications and operating conditions Battery type Cells in series Cells in parallel Capacity (C/3 rate).1 Electrolyte concentration. × 3 in.4** 25 LeadAcid (module) [20] 6 8 85 1.025 0.Table 1.159 0. 58. 251. cm positive electrode separator negative electrode 0. M 5.040* 7.0 in. W Operating temperature.6 NiMH (single cell) [8] 1 1 1.5 in. ˚C 3000 23 * calculated from the given capacity and capacity ratio. cm2 Thickness.159 0. ** calculated from the given maximum current of 4 A.
44 0.Table 2. cm2/cm3 positive electrode negative electrode Electrical resistance. Ω intercell connector terminal cell grid 2×104 2×104 2×103 N/A N/A N/A 2.30 leadacid (module) NiMH (single cell) [2122] 17 .61 0.92 0. cm nickel cylinder (substrate) MH sphere Maximum specific surface area.9×104 (1.5×104) 103 0.3×104 3864 2100 N/A N/A 2.68 0. Parameters used in simulations Battery type Porosity at fully charged state positive electrode separator negative electrode Volume fraction of inert material PbO2 electrode Pb electrode Radius of active material particles.05 0.10 N/A N/A 0.55 0.3×105 2.
Electrolyte concentration profiles across a leadacid battery cell at t = 76. Comparison of the experimental data with simulated results.87 h (when the cell voltage drops below 0. Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 18 .77 h (corresponding to the cutoff voltage of 1.List of Figures Figure 1 Dynamic stress test (DST) cycle and simplified federal urban driving schedule (SFUDS) profile Schematic configuration of a battery module Schematic of control volumes and computational meshes for a battery cell DST curves of the leadacid battery module at 3.77 h (corresponding to the cutoff voltage of 1.87 h (when the cell voltage drops below 0. Simulated electrolyte concentration profiles across the NiMH cell during the DST cycle at t = 1.06 min (corresponding to the end of the DST cycle testing) and t =99.0 V) and t = 2. Comparison of measured and simulated DST voltage profiles of a NiMH cell Predicted proton and hydrogen concentration profiles across the NiMH cell during the DST cycle at t = 1.97 min (at the end of the simulated DST cycle).4 V).06 min (corresponding to the end of the DST cycle testing) and t = 99.4 V).0 V) and t = 2.97 min (at the end of the simulated DST cycle).0 kW peak power and ambient temperature. Electrode active material utilization profiles across the leadacid cell at t = 76.
This action might not be possible to undo. Are you sure you want to continue?
Use one of your book credits to continue reading from where you left off, or restart the preview.