Anda di halaman 1dari 20

Nonuniform Temperature Performance Eects in Lithium-ion

Batteries
Matthew Klein
Department of Mechanical and Aerospace Engineering
University of California, Davis, USA
mpklein@ucdavis.edu
Jae Wan Park
Department of Mechanical and Aerospace Engineering
University of California, Davis, USA
jwpark@ucdavis.edu
September 19, 2014
Abstract
The purpose of this research is to provide experimental data for batteries under constant controlled
nonuniform temperature proles and to develop modeling methodologies with computational complexity
suitable for relatively rapid design optimization studies. A test system has been developed control
the average temperature and temperature gradient of single pouch-style lithium-ion cell. The testing
of multiple parallel-connected cells will be performed in this system as well. A 12-point test matrix
consisting of average cell temperatures of 15, 25, and 35

C, and constant lateral temperature dierences


of 0, 5, 10, and 20

C is used. The cells are tested using automotive driving schedules as well as Hybrid-
Pulse-Power-Characterization (HPPC) testing proles. Preliminary results show that electrochemical
performance tends to degrade with an increase in temperature gradient over the two-hour long driving
schedule. However, for pulse testing the increased gradient causes an improvement in performance. For
both test conditions, the eect of the gradient was greater at lower average temperatures. These results
are motivating new testing approaches and will be very useful in the validation of new multidimensional
models intended to simulate the observed behaviors. Ultimately, the results will aid in the optimization
of thermal management system design and control.
1
1 Introduction
In recent years lithium-ion batteries have become a popular choice for energy storage in automotive appli-
cations. This has been largely due to the increase in heavily-electried vehicles, e.g. full Battery-Electric-
Vehicles (BEVs) and Plugin-Hybrid-Electric Vehicles (PHEVs). As of December 2013 a total of 7.5M Nickel-
Metal-Hydride (NiMH) based battery hybrid-electric vehicles had been sold world-wide [1]. This represents
approximately 11.25GWh of deployed battery packs. The rst lithium-ion based vehicle was released in
December 2010. Four years later, an estimated 8.5GWh of lithium-ion based vehicle storage was produced;
i.e. Li-ion production met 75% of NiMh production in 33% of the time [2, 3]. Meanwhile Tesla Motors is
planning an estimated annual production of 85GWh (or 500,000 vehicles) of lithium-ion batteries by 2020
with an initial investment of over $4B in new plant construction [4]. Therefore, lithium-ion battery pack
development is and will continue to be a topic of great research interest.
A critical subsystem of a battery pack is the Thermal Management System (TMS). The primary function
of the TMS is to control the temperature of the battery to ensure eective power delivery, lifetime and safety
of the battery pack. The optimization of this component is nontrivial due to the necessary understanding
of the thermal-electrochemical coupling between the TMS and batteries. Lithium-ion batteries tend to be
quite sensitive to temperature. For example, for approximately every 10

C increase in the lifetime average


temperature decreases the life expectancy be a factor of two. Conversely, for every 20

C temperature drop the


performance can degrade by a factor of two. This study aims to provide novel information with regards to how
commercial lithium-ion cells react to nonuniform internal temperature distributions similar to those caused
by TMS heat exchangers. This report is outlined as follows: Section 2 provides background information with
regards to the development of a TMS, Section 3-5 detail the project description and preliminary results.
Finally, Section 6 will present a proposed project time-line.
2 Background
2.1 Lithium-ion Cell Construction
A battery pack generally will consist of a set of electrochemical cells, grouped together in parallel and serial
electrical congurations to meet the voltage and ampere-hour capacity targets; packaging, which is used
for constraining the cells and other components; control/management electronics, used for monitoring the
battery closely; and some form of heat exchange system, commonly referred to as the Thermal Management
System (TMS). Ideally one designs all components to be as light and tightly packaged as possible, due to
the relatively low specic energy of batteries compared to hydrocarbon fuels. The cells usually account for
60-80% of the pack volume/mass [5]. Therefore, improving TMS design can eectively improve the battery
pack energy density without improving the battery chemistry.
The pack design process begins with cell selection, and this dictates much of the following architectural
decisions. Cells are produced most commonly in either a cylindrical or prismatic architecture, whereby the
basic cell sandwich is either rolled and placed into a cylindrical steel canister, or stacked and encased in a
vacuum sealed, mylar-coated, aluminum pouch. Prismatic cells are also available in steel canisters similar
to cylindrical cells. Commercial automotive cells have voltages and capacities ranging between 2.5-4.2V and
1-20Ah, respectively. These either appear in cylindrical or prismatic formats.
At the cell sandwich scale both cells are identical. The positive electrode consists of an aluminum foil
coated on both sides with a slurry containing the active material, which in modern commercial cells is either
a variant of LiMO
2
where M may be pure Co, Ni or Mn or some combination of all three. Another common
cathode active material is LiFePO
4
. Nearly all commercial cells utilize graphite as the negative electrode
active material. The active material on each electrode will be diluted with binders and conductive additives to
optimize longevity, performance and energy density. The electrodes are porous (to increase reaction surface
area) and this is an important design parameter for cell performance and lifetime. A plastic separator,
providing electrical isolation, is placed between the electrodes. Finally, a liquid electrolyte occupies all pores
throughout the entire cell sandwich, providing a conductive pathway for Lithium-ions to transfer from one
electrode to another, when the electrical circuit is closed.
Due to the layering of the cell the eective bulk thermal conductivity is anisotropic and typically is on
the order of 25 W/mK in-plane and 1 W/mK through-plane. The anisotropy of the cell along with the form
2
factor play a critical role in the design of the heat exchange mechanism that is placed in direct contact with
the cells.
2.2 Thermal Management System Overview
A basic overview of battery module architecture, including the integration of thermal management heat-
exchangers, will be discussed briey. Lithium-based automotive battery packs commonly supply a vehicles
entire power demand, which can easily exceed 100kW. These battery packs tend to consist of hundreds,
if not thousands of cells in order to support the power and energy requirements. All production battery
packs integrate a subset of these cells into modules containing anywhere from a few to a few hundred. This
allows for a more manageable design and assembly of the nal pack. It is also quite common for the thermal
management system (TMS) heat exchange mechanism to be placed in contact with the cells at the module
level, to enable simpler interface at the pack level. Due to the signicant sensitivity of electrochemical
performance, degradation rate, and safety of a lithium ion battery pack, eective design and integration of
the TMS is critical.
Cylindrical cells are commonly stacked into square or hexagonal arrays when integrated into a module.
The remaining volume is used for either air cooling/heating or some sort of liquid-based heat exchanger is
routed through the module and placed in direct contact with the cells. Figure 1 illustrates the cylindrical
cell heat exchange congurations. Either these are directly cooled/heated by the uid or indirectly. Direct
methods nearly always use air as the uid (liquid ooding may be used in extreme situations, i.e high
discharge/heat gen. rates). Indirect methods pump a liquid through a heat exchanger, which is then placed
into contact with the cells. Pouch cells are stacked into larger blocks with a mix of serial and parallel
electrical connections to neighboring cells.
Heat exchange architectures for pouch cells are illustrated in Figure 2. Direct (air) cooling and three
other indirect techniques are employed. Figure 2b displays the Edge-Fin system, where thermally conductive
ns are interleaved between the pouch cells. Both A123 and EiG (battery manufacturers) have employed
this architecture for some of there module designs in the past. A liquid carrying chill-plate will be placed
into contact with at least one side of the ns. This system tends to be both thermally inecient and space
inecient. Figure 2c displays a method where a thin (<1mm) chill-plate is interleaved between the pouch
cells in the module. These chill-plates are connected hydraulically in parallel for uniform heat transfer across
cells. This system has been employed in the Chevrolet Volt battery pack. Finally Figure 2d shows a purely
Edge system where a chill-plate would be attached directly to the edge of the cells. To the authors knowledge,
this system has not been commercially employed. It performs somewhere between the Edge-Fin and Face
systems in terms of thermal performance, but provides enhanced packaging and design simplicity benets
over those systems. Therefore, system architecture is important, however not enough knowledge exists today
to allow an engineer to perform tradeo studies to compare these systems accurately.
Figure 1: Cylindrical cell thermal management. Left: Direct contact uid heat exchange. Right: Indirect
contact uid heat exchange.
It is the job of the TMS engineer to optimize the geometry (i.e. thickness, width, etc of the heat
exch.) and performance of the heat exchanger being integrated into the module. Common industry practice
3
(a) (b) (c) (d)
Figure 2: Pouch cell thermal management. (a) Direct uid heat exchange. (b) Indirect uid heat exchange.
Edge-Fin style. (c) Indirect uid heat exchange. Face style. (d) Indirect uid heat exchange. Edge style.
maximizes the packaging eciency and Coecient of Performance (COP, ) of the heat exchangers, while
placing a constraint to maintain less than 5

C temperature dierences throughout the entire battery pack.


This also inherently limits the heat transfer rate as internal cell temperature distributions cannot exceed said
limit. Packaging eciency (for a module) is dened simply as the ratio of cell volume to module volume in
Equation 1. The COP is dened as the ratio between the heat transferred into the module versus the work
required to cause said heat transfer in Equation 2. The required work takes the form of fan, pump, and/or
compressor losses. The TMS can also be categorized based on either being active or passive in nature. In
this context a passive system may still have pump/fan requirements as, for battery packs, passive tends to
refer to the use of ambient uids to provide heat exchange. Whereas, for active systems the heat exchange
uid is actively controlled to a desired temperature and then applied the battery. Generally, for vehicles
with a battery greater than 20kWh (+/- 5kWh), an active TMS is employed.

pkg,mod
=
V
cell
V
module
(1)
COP
hx,mod
=
Q
mod
W
mod
(2)
The stringent temperature distribution limit is motivated by the fact that bulk resistance of a battery tends
to follow an Arrhenius relation (Eq. 3) with respect to temperature, Figure 3. Therefore, one might think
that if a cell had a nonuniform temperature prole the colder portions of the cell would be cycled less than
the hot portions. The time integrated nonuniform current distribution would result in a nonuniform ampere-
hour-throughput (AHT) in the cell/module/pack, and therefore cause uneven degradation. It is assumed in
industry that this is negative result. The severity of the situation will of course depend on many factors,
one being the magnitude of the temperature distribution relative to the sensitivity. For example, if dR/dT
were small, one might hypothesize that temperature distributions have little impact on performance. To
date, however, very limited literature exists in which the actual sensitivity to temperature distributions has
been experimentally studied. By studying this more it could enable smarter integration of the TMS in the
battery with the potential to signicantly reduce system costs and complexity, while improving the battery
packs eective energy density. A review of the available literature will now follow.
R = R
ref
exp

E
a
R

1
T
ref

1
T

(3)
4
Figure 3: Arrhenius relationship for resistance relative to temperature.
2.3 Literature Review
The design of a battery thermal management system requires knowledge of the heat generation coming from
the battery. The literature has trended in developing complexity toward rened modeling capabilities in
four main areas: i) electrochemical heat generation modeling, ii) thermal modeling of various geometries,
iii) coupling of electrochemical and thermal models, and iv) multidimensional models applied to the critical
physical domains. Experimental based literature in all areas has followed in an eort to provide data for
model validation. Model and experimental work will be presented.
2.3.1 Electrochemical Heat Generation
Irreversible heat sources in batteries come from concentration, activation and Ohmic losses in the electro-
chemistry. Reversible heating eects propogate from entropic eects due to changes in the Open Circuit
Voltage (OCV) of the battery w.r.t. temperature. Early work in heat generation modeling mostly focused
on the development of energy balance equations derived from the losses modeled using 1D electrochemical
models, e.g. Doyle et al [6]. Bernardi et al. [7] proposed the use of Equation 4 where the heat source
term is a sum of the entropic (reversible) eects, Ohmic (irreversible) eects and mixing and phase-changes.
They recommended that the mixing and phase-change terms may be negelected for a simplication as that
component of heat contribution is usually small. This equation is most commonly applied in thermal studies
where an electrochemical model is implemented to derive the heat generation.
C
p
dT
dt


Q =

1
I
1
T
dU
avg
1
dT
+

I
1
U
avg
1
IV (4)
Rao et al. [8] and Thomas et al. [9, 10] developed heat equations for insertion style electrode batteries
(e.g. Li-ion) and revisited the work of Bernardi. They mostly attempted to deal with the nonuniformity of
reaction in the cell sandwich thickness direction that can occur, particularly at higher currents, in lithium-ion
batteries. Thomas work also includes calormetric experimental studies for the validation of the models. To
complement the modeling eorts others have attempted to develop methods whereby the heat generation is
accounted for via experimental data recorded from constant-current charge/discharge cycles at various rates.
5
This method has also been validated through calorimetry. The work of Hong et al. [11] and others therein
are an example of this work. Work which focuses on implementing these modeling techniques for studying
coupled thermal-electrochemical eects will be presented next.
2.3.2 Coupled Thermal-Electrochemical Investigations
Both rst-principle and empirical models have been applied with coupled thermal and electrochemical do-
mains. The work of Pals et al. [12, 13] is one of the rst examples of rst-principles based coupled thermal-
electrochemical modeling. They use the heat generation model developed by Bernardi et al. [7] and a lumped
(single temperature point, ODE) thermal model to feedback the updated temperature of the cell into tem-
perature sensitive reaction kinetics and transport parameters. Pals et al. applies this technique to both a
single cell (lumped thermal model) and a cell stack (set of thermal lumps). For the heat generation of the
individual cells they derive a lookup table through isothermal simulation of the detailed single cell model over
a range of temperatures. This greatly simplied the simulation of the heat sources for the stack level model.
This approximation worked well at predicting the cell temperature conditions at the center of the stack, but
had poor agreement at the edges. They found that including nonuniform thermal eects is important in heat
generation modeling as the heat generation rate is dependent on cell temperature, as losses are reduced at
higher temperatures. This work marks one of the rst attempts at understanding how battery performance
under nonuniform thermal conditions is eected.
A more rigorous modeling eort to describe thermal-electrochemical coupling was applied by Gu and
Wang [14]. Srinivasan and Wang [15] continued that work and applied it to LiMnO
2
cathode based cells.
They also evaluate the accuracy of an experimentally based heat generation model, but this was seen to
introduce as much as 15% error in the thermal model. Other approaches where a simplied rst-principles
electrochemical model was implemented may be reviewed in Kumaresan et al. [16], Cai and White [17], and
Guo et al. [18].
Finally, empirical models, derived from frequency response measurements of batteries under various
temperatures, have been used. These models are most commonly referred to as Equivalent Circuit Models
(ECM), as they are commonly represented by an equivalent electrical circuit. Gomez et al. [19] utilize a 7th
order ECM with temperature sensitive parameters to model the thermal-electrochemical coupling. All of the
coupled models described in this section used lumped thermal models, and therefore were unable to quantify
spatial current-distribution eects that would occur in large format cells under nonuniform temperature
conditions. Work related to spatial eects is discussed next.
2.3.3 Multidimensional Investigations
There are three classes of multidimensional investigations developed thus far in the literature: i) multidi-
mensional thermal modeling with one-way coupling to a heat generation model; ii) multidimensional models
with two-way coupling between discretely distributed electrochemical sub-models to provide nonuniform
heat generation; iii) experimental investigations in which batteries are placed under nonuniform thermal
conditions and the eect of thermal nonuniformity relative to performance metrics are tested.
The rst is a class of models that use multidimensional thermal models, but only one-way coupling
with a heat generation model, i.e. the heat generation does not have temperature feedback. Therefore the
heat generation rate inside the cell is considered uniform, and is either the output of an electrochemical
model (of varying complexities) or of experimental data. Chen and Evans [2022] originally led this eort.
Later work here focused on geometrical renement of the meshes used for the thermal modeling, i.e. more
realistic representations of the cell construction [23, 24]. However, as seen in the previous section discussing
coupling, true two-way thermal-electrochemical coupling is required to accurately predict the temperature
distributions.
Mainly due to the commercialization of large-format (<2Ah) cells placed into high power (automo-
tive) applications, research into coupled-distributed models has been more recently motivated. Gerver and
Meyers [25] develop a discretely distributed 2D network of electrochemical models that captures current
distribution eects across the electrode plane in a pouch cell. At each node of the network they use an
approximation the Doyle model developed in [6]. No experimental validation of the model was performed for
this work. Kim et al. [26] developed their Multi-Scale-Multi-Domain (MSMD) modeling framework. This
framework breaks the battery into the following domains: i) Cell domain, ii) electrode domain, and iii)
6
particle domain. The cell domain model consists of electrical and thermal models with two-way coupling to
the smaller scales. The Doyle model is an example of the electrode domain model that may be used. Finally,
the particle model treats solid phase lithium ion diusion. This technique predicts current and temperature
distribution throughout the cell and is therefore able to model State of Charge (SoC) imbalances throughout
the cell, which is important.
1
It does not, however, provide true three-dimensional modeling at the electrode
scale, i.e. discrete instances of electrode models are used and therefore no lithium transport is modeled
perpendicular to the cell sandwich. Additionally, no experimental validation was provided. Other examples
where electrochemical-thermal coupling exists in multidimensions are provided by Awarke et al. [27] where
they attempted to model nonuniform degradation across the plane of a pouch-cell. Christensen et al. [28] and
Allu et al. [29] provide other general modeling frameworks similar to Kim et al. An example where low order
empirical electrochemical models are implemented spatially is from Yi et al. [30]. They implement a very
simple 2D distributed resistance model where the current density/overpotential has an Ohmic relationship,
as in Equation 5. This model is commonly known as the NTG model as it was rst used by Newman and
Tiedemman as well as Gu during the 1980s to study spatial eects. Yi et al. does provide infrared imaging
on constant-current discharges for validation of the thermal model. Allu et al. show experimental validation
for the case of a pouch-cell using the NTG model as well.
J = Y (U V
p
V
n
) (5)
None of the coupled models presented in this section treat the electrochemical domain in a rigorous
3D way, rather they implement discrete electrochemical zones. It is generally assumed that for the SoC
imbalances developed in commercial cells, and due to the nature of the electrode, ion diusion in the planar
direction is very small, and therefore this assumption should be valid.
The third class of multidimensional investigations focuses on experimental studies. A few recent experi-
mental investigations by Fleckenstein et al. [31], Troxler et al. [32], and this author [33] study the performance
eects caused by nonuniform temperature conditions on lithium-ion batteries. Fleckenstein et al. focus on
understanding the amount of SoC nonuniformity that is developed in a LiFePO
4
18650 cylindrical cell. To
test for this they setup three of these cells with varying degrees of insulation. One is placed in the open
ambient condition, a second in a partially insulated condition, and a third in a fully insulated condition.
These are meant to model the outer, middle, and inner layers of a cylindrical cell. Finally, the three cells are
connected in parallel, with current measuring shunts placed on each cell and the the cells are tested under
high C-rate constant-current charge/discharge pulses.
2
Because the OCV curve vs. SoC for LiFePO
4
is
relatively at it was determined that there was almost no charge equalization that occurred between the
three cells after pulse testing was completed, therefore nonuniform degradation would be expected to occur
throughout the lifespan of the cell. Or in other words, this chemistry is possibly more temperature sensitive
due to the at OCV curve. This study was a performance based study; no lifetime eects were studied.
Troxler et al. placed a pouch-cell sandwiched between two thermo-electric coolers (TEC) devices such
that the face of the coolers we in direct contact with the broad faces of the pouch-cell and set the coolers
at two separate temperatures to drive a controlled internal temperature gradient through the cell. Under
various conditions they applied Electrochemical Impedance Spectroscopy (EIS) in order to test the frequency
response of the cell and determine how that was eected by the nonuniform thermal conditions. They
determined that the increased gradients improved the battery performance (lowered the impedance). This
was a performance (power) based study; no lifetime eects were presented.
In this authors work a pouch-cell was tested under average temperatures of 15, 25, and 35

C with
0, 5, 10, and 20

C temperature deltas across the 100mm cell width. In this study TECs were placed
in contact with the edge of the pouch cell, rather than the main faces in Troxler, et al. This study was
separately motivated to understand how performance would be eected due to edge based heat transfer
such as presented by the TMS architecture in Figure 2d. The cell was tested under dynamic power proles
following the US EPA US06 Driving Schedule, scaled for a two-hour long discharge from 90-10% SoC.
1
SoC is the ratio between the presently available ampere-hour (Ah) capactiy of the battery to the total available Ah when
fully charged. If a cell has a nonuniform current distribution this can cause portions of the cell to have nonuniform capacity
levels, which is termed a SoC imbalance.
2
C-rate is the ratio between the total Ah capacity of the cell and current load (in amps) being applied to the battery. For
example, a 1C-rate for a 10Ah battery will be a charge or discharge of 10A. Also a 1C is results in a 1 hour discharge. A 2C
rate is a 20A discharge, and therefore a 30 minute long discharge.
7
Maximum discharge/charge rates were 3C/1C, respectively. Eective bulk resistances for the discharge and
charge behavior were calculated over the entire SoC range and it was determined that the discharge resistance
increased with increased gradient, however, charge resistance decreased. For both case the magnitude of the
gradient eect was larger at lower average temperatures. In an attempt to elucidate the dierence between
discharge/charge behavior pulse tests were then carried out. For these it was found that for both charge
and discharge the performance improved with increased gradient. For the pulse case the behavior was the
same as for Troxler et al.s work. There are two important points to make here. First, the gradient eect on
discharge performance behavior in Klein et al.s work was opposite to that observed by Troxler et al. for the
longer time test, but the same for the short time pulses. It is believed that this is the case due to the the EIS
testing that Troxler used, which is essentially a rapid pulse test method. Second, the reversed behavior for
long time dynamic discharges by Klein may be functions of the type of test that was run, however, ongoing
work is attempting to elucidate this.
2.3.4 Summary of Review
In summary, much work has been accomplished in thermal testing and modeling of lithium ion batteries.
However, for the case of large format cells, where nonuniform temperature conditions can greatly eect the
current distributions much work remains in developing a clearer theory for understanding the key dynamics
at play. In particular, the recently developed multidimensional models that provide promise in predicting
nonuniform thermal eects still require a signicant experimental validation eort. This provides the primary
motivation for the experimental work to be presented next. The next section will detail the preliminary
results and research plans for this work.
3 Phase 1: Nonuniform Thermal Performance Eects on a Lithium
ion Pouch Cell
This section overviews an experimental investigation that has been published in [33]. The methods used,
results obtained, conclusions and plans for future work will be discussed.
3.1 Methods
3.1.1 Experimental System
An experimental system was built to enable the generation of physically realistic temperature distributions
on a single cell scale to study performance eects experimentally. Thermo-electric chill plates (TECs) were
used to control the side edge boundary temperatures of the cell. A schematic of the Temperature Gradient
System (TGS) designed and built for this testing purpose along with an image of the actual system may be
found in the appendix. A 10Ah pouch type Nickel-Cobalt-Manganese cathode/graphite anode cell was used
for this investigation. The approximate cell dimensions are H:120mm, W:100mm, T:8mm.
3.1.2 Experimental Test
The experiment was motivated such that it would investigate a range of thermal conditions a battery may
encounter in an actual automotive application. The cell was tested over twelve dierent treatments whereby
three mean cell temperatures were used: 15, 25, and 35

C, and at each mean temperature four dierent


edge-to-edge temperature dierences were applied: 0, 5, 10, and 20

C, across the width of the cell. The


0

C dierence was used as the control for comparison of the impact caused by increased edge-to-edge dif-
ferences. The range of mean temperatures selected here is based on the expected range of operation that a
lithium-based electried vehicle would be controlled to for meeting lifetime and performance targets. The
temperature dierences cover the expected range that a TMS might be able to induce. For the Edge Chill
Plate (ECP, Figure 2d) conguration a cell could be heated or cooled at a rate of approximately 1

C per
minute with the 20

C dierence, assuming a planar bulk conductivity of 25 W/m-K and what are presently
typical pouch cell dimensions.
8
At each temperature treatment the cell was dynamically discharged using twelve back-to-back US06
proles, equivalent to 96 miles of driving. The power prole was developed for a 1500kg vehicle with a
100kW electric motor and a 40kWh battery pack. This simulated power prole was scaled to the single
cell level and was performed using an Arbin BT2000. A single US06 power prole, scaled to the full vehicle
battery pack may be seen in Figure 4. Each test treatment was randomly repeated three times, averaging the
measurements. A standard deviation of 5.9mV between the three cycles for each treatment was observed,
hence the averaging of the data. The current and voltage data of the cell during the dynamic discharge was
used to parameterize a simple battery model in order to gain an insight into how the bulk performance of
the cell was changed as a function of the temperature dierence and mean temperature over the entire SOC
range of operation. The variation of the actual average temperature and temperature prole within the cell
during dynamic testing is discussed next.
Figure 4: US06 Driving Schedule power prole. A single prole is 600s. A total of 12 were repeated from
90-10%SoC
3.1.3 Temperature Measurements
Thermocouples (TC) were used to measure seven temperatures on the cell face, spanning the 100mm cell
width. Two additional TCs were used to measure the faces of the chill plates used to control the edge
temperatures of the cell. The thermocouples were T-type and had a resolution of +/- 1

C. These TCs were


connected to a National Instruments NI-9213 TC module, which was then coupled to a computer through a
cDAQ-9174. Temperatures were collected at a 5-second time interval.
The temperature treatment prescribed cannot be held perfectly constant based upon the physical nature
of the experimental system. This is a tradeo between needing to insulate the cell faces to have the ability to
maintain a linear temperature prole and the problem this causes by preventing expulsion of heat generated
within the cell. This internally generated heat does perturb the temperature prole and the average and
standard deviations of the temperature measurements are provided in the appendix in Tables A and B,
respectively. In Table A the Mean Cell Temperature (MCT), Mean Temperature Dierence (MTD) and
Mean Prole Error (MPE) are provided. The MCT described in Equation A computes the average of
the data collected from the seven thermocouples in direct surface contact with the cell. First an average
between the seven channels is made, followed by averaging over the US06 dynamic discharging test time. The
MTD, presented in Equation B, is an average of the temperature dierence between the right and left-most
thermocouples on the cell over the US06 test time. The MPE of Equation C computes, for each channel,
an error from the ideal temperature value for the case of a perfectly linear temperature prole from edge-
9
to-edge. At each time instant the errors for each channel are averaged and then this vector is averaged over
the US06 test time. These statistics provide an insight into how closely the actual temperature treatment
was to the intended treatment.
The standard deviation of the average CT, TD, and PE over the US06 test time were computed for all
test treatments and provided in Table B. The standard deviation of the average CT gives a sense of how
much the average cell temperature deviates over the entire test from the intended treatment, and likewise for
the standard deviation of the Temperature Dierence. The standard deviation of the Prole Error provides
a sense of how the temperature prole deviated from the intended linear prole over the test time. It is
interesting to note that the prole errors are consistently the largest for the 15

C average temperature
testing cases. This is due to the increased resistance of the cell, which will cause higher heat generation, and
therefore displace the cell temperature prole further from the desired treatment. It may be possible that
through the use of active boundary control of the TECs the mean prole error could be reduced further for
future testing.
3.1.4 Model
The simple NTG model (Eq. 6) is parameterized using a Least Squares approach on the test data. Charge
and discharge resistance parameters are solved for over 8% SoC batches of data. These values are compared
over all thermal conditions and also used to compute instantaneous power capabilities as another comparison
metric.
V
cell
= R i(t) +OCV (6)
R = f(SoC,

T, T
diff
, sign(i)) (7)
OCV = g(SoC,

T, T
diff
) (8)
A specic test procedure was developed for the Open Circuit Voltage in order to allow for calculation of
the overpotential at each time step. Using the known overpotential and input current, the bulk resistance
values could be computed using Least Squares to approximate Equation 9.

0
.
.
.

i
.
.
.

= R

i
0
.
.
.
i
i
.
.
.
i
n

(9)
3.1.5 Open Circuit Voltage Test Procedure
The open circuit voltage of lithium ion cells is a function of SoC and temperature, necessitating specic
testing which may yield these values over the entire operating range of interest. It is critical to accurately
determine the OCV as a function of SoC as errors developed here will be absorbed into the computed R
values in order to maintain a proper tment. This is problematic as the R values are to be used to determine
how the pulse performance is eected by the applied temperature dierence.
There are many test methods for determining the OCV of a cell [34], however, these methods tend to
provide varying results beyond the accuracy desired here. The method developed here applies one second
long CC discharge followed by charge pulses, voltage and current data were collected at a 20Hz sample rate.
The four pulse sets applied back-to-back were: 1C, C/2, C/10, and C/100. A removal of 5% SOC at C/2
was performed immediately after the pulses, which is then followed by a 60 second rest, Figure 5a. The rest
was added in order to allow for the TGS to maintain the prescribed treatment prior to the application of the
next pulse set. The rst voltage points collected (0.05 s after pulse application), for the charge and discharge
pulses, were averaged to determine the OCV, these points may be seen in Figure 5b, labeled as A1, A2, etc.,
where A1 is the rst voltage point for the 1C discharge pulse and A2 is the rst recorded voltage point for
10
the 1C charge pulse. This test method was run for all temperature treatments to rst collect the data needed
for (3). Figure 5c shows the prescribed current prole for stepping sequence. The rst 1C charge pulse was
reduced at high SOCs due to an intentional implementation of a voltage clamp in the Arbin program at the
manufacturers maximum recommended cell voltage of 4.2 V.
Figure 5: Stepped pulse OCV test method. a) Voltage measurements during one iteration of the test method.
b) Voltage response shown in section highlight by the red box in a). c) Current prole used for step sequence.
Multiple magnitude pulses were used to determine the region of linear and symmetric voltage response as
this is the only region where averaging would result in a valid OCV. This method was conrmed via separate
testing in two ways: 1) Additional one hour long rests were applied prior to and after the stepping sequence
and the voltage values collected at the end of those rests were compared to each other as well as the averaged
values and 2) a more traditional method of CC capacity removal of 5% SoC increments followed by a two
hour rest was performed. These values were compared and it was concluded that the C/100 voltage averaging
provided equivalent values to the long rest test method. The new step method used here can provide 5%
SoC interval OCV data in less than three hours including charging the cell. This is a 93% reduction in test
time compared to the two hour long rest test method performed at 5% SoC increments.
11
3.2 Results
3.2.1 Open Circuit Voltage
The OCV test results indicated a very small sensitivity to average temperature and almost no sensitivity to
gradient. Additionally, it is dicult to completely remove transport related eects, particularly at reduced
temperatures, and therefore the higher temperature OCV curve was adopted as the only data used for
computing overpotential for all thermal treatments.
3.2.2 Bulk Resistance
Figure 6 shows plots of the bulk discharge and charge resistances computed for all thermal treatments.
Discharge resistance are plotted in the left column of the gure; charge in the right. The three rows indicate
the average cell temperature condition; top:15

C, middle:25

C, bottom:35

C. For each plot, the resistance


parameter computed is plotted over the full SoC range tested here. All temperature dierences for a given
average temperature are plotted together.
Figure 6: Discharge and charge resistance values. Left column: discharge resistances. Right column: Charge
Resistances.
It was observed that the bulk discharge resistance values tended to increase with increased gradient. This
results in performance loss with increased gradient. Conversely, the computed bulk charge resistances were
reduced with increased gradient. A hypothesis for the reversal in behavior may be rooted in the testing
method/model used here. The US06 prole consists predominantly of discharge events. Additionally, the
peak discharge currents are three times larger than the peak charging currents. As a result, concentration
overpotentials in the direction of discharge likely remain present during the limited charging events in this
12
prole. This would lead to an apparently reduced charge resistance being computed via the technique applied
here. The discharge data might also be thought of a being more closely related to steady behavior, and charge
data more like pulsing information. This might also explain why the charging behavior followed the trends
observed by Troxler et al. [32], where they implemented low current pulsing at various frequencies (EIS) to
parameterize a transfer function. To actually describe why the gradients are having these eects will take
continued testing as well as rst-principles model development. This data will be useful for multidimensional
model validation.
3.3 Conclusions
3.3.1 Summary
There is a clear impact on lithium ion pouch cell performance with the application of a one-dimensional
temperature gradient along the lateral face of the cell, caused by edge based thermal management systems.
For the case of the cell being tested here, as much as a 30% reduction in instantaneous discharge power
capability is seen at 15

C mean temperature and 20

C gradient. The performance impact is aected by


the mean temperature as at the 35

C mean temperature and 20

C gradient case the power reduction was


lowered to a maximum of 12% reduction in instantaneous discharge power. The charging power capability
shows an apparent improvement with increased temperature gradient and has shown a 20% improvement
for all mean temperatures at the applied 20

C gradient.
These results have potential implications in the allowable tolerance of temperature gradients caused by
the TMS of an electric vehicle battery pack. For example, an edge based TMS has the possibility of enabling
improved packaging eciency of the battery pack and reduced system complexity. Additionally, the TMS
control strategy may be impacted by this information as the gradient impact is greatly reduced at increased
mean temperatures, possibly allowing for higher heat transfer rates at increased cell mean temperatures.
Therefore, one might devise a TMS control strategy where heat transfer rates are a function of the average
cell temperature in order to reduce the negative performance eects caused by temperature gradients.
3.3.2 Future Work
Finally, the results provided here are preliminary and continued work in this area is needed. In particular,
the following should be studied further: 1) Applying models with improved physical description of the
electrochemical dynamics in order to understand the primary mechanisms aected by the gradient; and
2) Performing cycle life testing to understand whether or not the increased temperature gradients lead
to premature capacity or performance loss of the cell. A new study with parallel connected cells under
nonuniform temperature conditions will be performed next. Individual cell currents will be measured to
enable a quasi-internal measurement of the current-distribution of a cell under nonuniform temperature
conditions. This will improve the understanding of how well the lumped thermal-electrochemical models
predict current distribution in cells with nonuniform temperature proles. Finally, the ultimate goal of this
work is to add new data that aids in the development of models for numerical investigation of control strategy
development and battery module heat exchanger design. The optimization goal being to minimize unwanted
thermal eects caused by the battery pack TMS on performance and lifetime.
4 Phase 2: Comparison Between Nonuniform Thermal Perfor-
mance Eects of Parallel Connected and Single Cells
The planned work to be presented in this section was directly motivated by two key factors: i) add new data
that provides additional multidimensional model validation capabilities, and ii) investigate the impact that
gradients have on dierent Li-ion chemistries.
The validation of the use of discretely distributed electrochemical submodels in the multidimensional
modeling frameworks presented in the literature review section of this report is a primary target of the
advanced model validation that is desired by this new test. Additionally, the work of Fleckenstein et al. [31]
focused signicantly on the measurement of internal SoC distributions caused by nonuniform current distri-
butions, which were caused by nonuniform thermal distributions throughout a cell. Specically, Fleckenstein
13
mentions that the equilization of these SoC distributions during cell resting periods was nearly nonexistent
due the at OCV curve of LiFePO
4
, which therefore provides almost no driving force to equalize the SoC
imbalance. There are, however, an entire set of other Li-ion chemistries that have steeper OCV curves over
the SoC. Therefore, it would be interesting to study the impact that OCV slope has on the internal SoC
equilization and therefore how that plays a role in the sensitivity to nonuniform temperature distributions.
4.1 Planned Methods
4.1.1 Experimental Setup
Two dierent cell chemistries will be tested to enable the study of the eect between temperature gradient
sensitivity and the gradient of the OCV w.r.t. SoC. A manganese rich LiNiCoMn type cathode is selected
for the steeper OCV type chemistry, while LiFePO
4
is the selected at OCV chemistry.
For each chemistry two dierent cell congurations will be tested. Cell Conguration Type 1 (CT1)
will consist of a pouch style cell (10Ah) in which 1D temperature proles will be induced along the width
of the cell, as in [33]. Cell Conguration Type 2 (CT2) will consist of ve parallel connected cylindrical
cells (18650 at 2.5Ah). For a given chemistry the cylindrical and pouch cell basic electrode designs are
matched as closely as is feasible. Primarily this is done by selecting cells from the same manufacturer and of
similar power ratings. For CT2 the cells are clamped bewteen two specially designed aluminum plates and
a 1D temperature prole is placed along the cells. The cells in CT2 are connected electrically in parallel.
Additionally, current measuring shunts are used to allow for the measurement of individual currents of each
of the ve cylindrical cells. The purpose of CT2 is to attempt to enable measurement of internal current
distribution in a single cell under nonuniform temperature conditions. CT1 is then used to check whether a
cluster of parallel connected cells behaves the same electrochemically under the same thermal conditions.
This setup will allow for two primary novel contributions. First, if a cluster of parallel connected cells
does behave similarly to a single under nonuniform temperature conditions this can aid in validating the use
of discrete 1D electrochemical submodels in the multidimensional modeling frameworks recently developed.
If they do not act similarly, then this modeling technique may need to be revisited, and this will be an
equally valuable result. The second novel aspect to this work is the comparison to thermal nonuniformity
between two chemistries and the ability to measure internal relaxation currents in a cell module. Even if
CT1 and CT2 do not behave similarly to enable improved understanding at the cell level, the comparison of
dierent chemistries across CT2 will provide new insight to the concern that should be placed for thermal
homogeneity at the module level where cells are connected in parallel.
Figure 7 shows a schematic and image of the system planned for use in this study. The clamps holding
the cells were designed such that each cell will be held at a uniform temperature, with each cell at dierent
a temperature.
Figure 7: System designed for multi-cell testing in CT2. A linear temperature prole may be placed along
the cells.
14
4.1.2 Testing Procedures
Charge/discharge constant-current pulse tests, randomized frequency input current pulses (0.1-10Hz) and
driving schedule style proles will be used to test the cells under each thermal condition. This will allow for
quantication of behavior across all realistic operating regimes expected of an automotive battery.
5 Phase 3: Multidimensional Empirical Thermal-Electrochemical
Modeling for Thermal Management System Design Optimiza-
tion
The results from the rst two phases of this work will provide crucial insights into valid modeling techniques
and assumptions. Ultimately, the primary reason for carrying out those experiments is to better understand
the limits with which the behavior related to nonuniform temperature distributions may be modeled. This
section of the work will attempt to create a relatively low order semi-empirically based multi-dimensional
coupled thermal-electrochemical model. This model will be developed such that it may be implemented into
a cell-to-heat exchanger geometry design optimization study.
For example, the costs utilized in an optimization may be developed as follows. Equation 10 represents
the packaging eciency of a battery module. Ideally, this value would be unity. Batteries are already both
expensive and limited in energy density relative to hydrocarbon fuels, and thus minimization of the need for
ancillary systems is important for both reasons.

mod
=
V
cell
V
mod
=
V
cell
V
cell
+V
hx
+V
pkg,etc
(10)
The Coecient of Performance, COP, of the module heat exchanger is expressed simply as the ratio of
heat transferred to the amount of work needed to achieve that heat transfer, as is traditionally dened 11.
Ideally, this value will be much larger than 1 and practically it hovers between 1-4.

hx
=

Q
hx
W
tms
=

Q
hx
W
pump
+W
cmpr
+W
fan
(11)
A means of relating the parasitic loads for operating the TMS and the reduction in electried vehicle
driving range may be dened as follows in Equation 12. It is simply the ratio between the capacity of the
battery that was utilized to operate the TMS versus the capacity utilized to transport the vehicle.
=
SoC
tms
SoC
veh
(12)
Finally, the electrochemical eects should be accounted for. Both the performance and lifetime should
be included as the lifetime eects are one of the largest motivators for the inclusion of a TMS in a battery
pack to begin with. Equation 13. For the lifetime this may simply be accounted for by a ratio between the
predicted battery lifetime with versus without a thermal management system. Ideally, we would have the
simplest and absolute minimal volume TMS occupying the battery pack while providing adequate lifetime.
The performance eect could be modeled similarly. Ideally, this ratio would be larger than unity. The
magnitude of this value will change signicantly depending on the operating conditions (ambient thermal
environment) the battery pack is subject to.
=
Life
tms
Life
notms
(13)
Thus an objective function might take the following form of Equation 14, where an integration of the
eects over the total life of the battery could be accounted for. Obviously, the need for treating the lifetime
eects in a distributed way will be important. The work of Awarke et al. [27] and Smith et al. [35] incorporated
a physics and empirical based spatially distributed degrdation models, respectively.
J

= max

mod
+
teol

i=t0

hx,i
+
2
(1
i
) +
3

life,i
+
4

perf,i

(14)
15
In order to reduce the computational demands necessary for integrated lifetime studies, a lower-order
semi-empirical model is planned to be implemented.
5.1 Planned Methods
Equivalent Circuit Models [36] have been utilized successfully for decades in modeling the electrical behavior
of batteries. An initial trial at this technique could be to simulate a discrete grouping of cells, similar to
CT2 from phase 2 of the project, connected in a network with spatial dimensionality that is relevant to the
temperature proles applied to the battery. For example, in our study from phase 1 and the planned work
from phase 2 1D set of ECMs could be modeled in parallel, with temperature sensitivity in the parameters.
This model will be validated using the results from phase 1 and 2. A more rigorous method of applying this
technique in multiple dimensions is being developed in a collaboration with Oak Ridge National Laboratory.
Both methods will be analyzed for their validity. The simplest model capable of capturing the spatial
dynamics accurately will be implemented into the optimization design study.
The optimization design study will analyze the important factors, i.e. Equation 14, of several common
TMS architectures in an eort to provide insights into how, for various architectures, one might reduce
undesireable thermal eects as well as packaging volume requirements caused by the TMS.
6 Project Time-line
Table 1 shows the plan for the remainder of the work proposed herein. Thus far, the work is on track.
Table 1: Remaining schedule for this work.
References
[1] Toyota USA. Worldwide Sales of Toyota Hybrids Top 6 Million Units. http:
//corporatenews.pressroom.toyota.com/releases/worldwide+toyota+hybrid+sales+top+6+
million.htm?view\ id=35924, 2014. Accessed: 2014-09-15.
[2] John Voelcker. Tesla Sales On Target, Gigafactory Groundbreaking Next Month: FINAL UP-
DATE. http://www.greencarreports.com/news/1091939\ tesla-sold-6450-model-s-electric-
cars-built-7530-in-q1, 2014. Accessed: 2014-09-15.
16
[3] Antony Ingram. Plug-In Electric Car Sales In June: Leafs Best-Ever June, Volt Up On May (FINAL
UPDATE). http://www.greencarreports.com/news/1093059\ plug-in-electric-car-sales-in-
june-leafs-best-ever-june-volt-up-on-may-final-update, 2014. Accessed: 2014-09-15.
[4] Tesla Motors. Gigafactory Plans. http://www.teslamotors.com/sites/default/files/
blog\ attachments/gigafactory.pdf, 2014. Accessed: 2014-09-15.
[5] Tesla Motors. Increasing Energy Density Means Increasing Range. http://www.teslamotors.com/
roadster/technology/battery, 2014. Accessed: 2014-09-16.
[6] Marc Doyle, Thomas Fuller, and John Newman. Modeling of Galvanostatic Charge and Discharge of
the Lithium/Polymer/Insertion Cell. Journal of Power Sources, 140(6):15261533, 1993.
[7] D. Bernardi, E. Pawlikowski, and J. Newman. A General Energy Balance for Battery Systems. Journal
of The Electrochemical Society, 132(1):512, 1985.
[8] Lin Rao and John Newman. HeatGeneration Rate and General Energy Balance for Insertion Battery
Systems. Journal of The Electrochemical Society, 144(8):26972704, 1995.
[9] Karen E. Thomas, Christian Bogatu, and John Newman. Measurement of the Entropy of Reaction
as a Function of State of Charge in Doped and Undoped Lithium Manganese Oxide. Journal of The
Electrochemical Society, 148(6):A570A575, 2001.
[10] Karen E. Thomas and John Newman. Thermal Modeling of Porous Insertion Electrodes. Journal of
The Electrochemical Society, 150(2):A176A192, 2003.
[11] JongSung Hong, H. Maleki, S. Al Hallaj, L. Redey, and J. R. Selman. ElectrochemicalCalorimetric
Studies of LithiumIon Cells. Journal of The Electrochemical Society, 145(5):14891501, 1998.
[12] Carolyn Pals and John Newman. Thermal Modeling of the Lithium/Polymer Battery I . Discharge
Behavior of a Single Cell. Journal of Power Sources, 142(10), 1995.
[13] Carolyn Pals and John Newman. Thermal Modeling of the Lithium/Polymer Battery II . Temperature
Proles in a Cell Stack. Journal of Power Sources, 142(10):32823288, 1995.
[14] W. B. Gu and C. Y. Wang. ThermalElectrochemical Modeling of Battery Systems. Journal of The
Electrochemical Society, 147(8):29102922, 2000.
[15] Venkat Srinivasan and C. Y. Wang. Analysis of Electrochemical and Thermal Behavior of Li-Ion Cells.
Journal of The Electrochemical Society, 150(1):A98A106, 2003.
[16] Karthikeyan Kumaresan, Yuriy Mikhaylik, and Ralph E. White. A Mathematical Model for a Lithium
Sulfur Cell. Journal of The Electrochemical Society, 155(8):A576A582, 2008.
[17] Long Cai and Ralph E. White. An Ecient ElectrochemicalThermal Model for a Lithium-Ion Cell
by Using the Proper Orthogonal Decomposition Method. Journal of The Electrochemical Society,
157(11):A1188A1195, 2010.
[18] Meng Guo, Godfrey Sikha, and Ralph E. White. Single-Particle Model for a Lithium-Ion Cell: Thermal
Behavior. Journal of The Electrochemical Society, 158(2):A122A132, 2011.
[19] Jamie Gomez, Ruben Nelson, Egwu E. Kalu, Mark H. Weatherspoon, and Jim P. Zheng. Equivalent
circuit model parameters of a high-power Li-ion battery: Thermal and state of charge eects. Journal
of Power Sources, 196(10):48264831, 2011.
[20] Yufei Chen and James W. Evans. Heat Transfer Phenomena in Lithium/PolymerElectrolyte Batteries
for Electric Vehicle Application. Journal of The Electrochemical Society, 140(7):18331838, 1993.
[21] Yufei Chen and James W. Evans. ThreeDimensional Thermal Modeling of LithiumPolymer Batteries
under Galvanostatic Discharge and Dynamic Power Prole. Journal of The Electrochemical Society,
141(11):29472955, 1994.
17
[22] Yufei Chen and James W. Evans. Thermal Analysis of LithiumIon Batteries. Journal of The Electro-
chemical Society, 143(9):27082712, 1996.
[23] S.C. Chen, C.C. Wan, and Y.Y. Wang. Thermal analysis of lithium-ion batteries. Journal of Power
Sources, 140(1):111124, 2005.
[24] Shin-Chih Chen, Yung-Yun Wang, and Chi-Chao Wan. Thermal Analysis of Spirally Wound Lithium
Batteries. Journal of The Electrochemical Society, 153(4):A637A648, 2006.
[25] Rachel E. Gerver and Jeremy P. Meyers. Three-Dimensional Modeling of Electrochemical Performance
and Heat Generation of Lithium-Ion Batteries in Tabbed Planar Congurations. Journal of The Elec-
trochemical Society, 158(7):A835A843, 2011.
[26] Gi-Heon Kim, Kandler Smith, Kyu-Jin Lee, Shriram Santhanagopalan, and Ahmad Pesaran. Multi-
Domain Modeling of Lithium-Ion Batteries Encompassing Multi-Physics in Varied Length Scales. Jour-
nal of The Electrochemical Society, 158(8):A955A969, 2011.
[27] Ali Awarke, Stefan Pischinger, and J urgen Ogrzewalla. Pseudo 3D Modeling and Analysis of the SEI
Growth Distribution in Large Format Li-Ion Polymer Pouch Cells. Journal of The Electrochemical
Society, 160(1):A172A181, 2013.
[28] Jake Christensen, David Cook, and Paul Albertus. An Ecient Parallelizable 3D Thermoelectrochemical
Model of a Li-Ion Cell. Journal of The Electrochemical Society, 160(11):A2258A2267, 2013.
[29] Srikanth Allu, Sergiy Kalnaus, Wael Elwasif, Srdjan Simunovic, John A. Turner, and Sreekanth Pan-
nala. A new open computational framework for highly-resolved coupled three-dimensional multiphysics
simulations of Li-ion cells. Journal of Power Sources, 246(0):876886, 2014.
[30] Jaeshin Yi, Ui Seong Kim, Chee Burm Shin, Taeyoung Han, and Seongyong Park. Three-Dimensional
Thermal Modeling of a Lithium-Ion Battery Considering the Combined Eects of the Electrical and
Thermal Contact Resistances between Current Collecting Tab and Lead Wire. Journal of The Electro-
chemical Society, 160(3):A437A443, 2013.
[31] Matthias Fleckenstein, Oliver Bohlen, Michael A. Roscher, and Bernard Baker. Current density and state
of charge inhomogeneities in Li-ion battery cells with LiFePO4 as cathode material due to temperature
gradients. Journal of Power Sources, 196(10):47694778, 2011.
[32] Yannic Troxler, Billy Wu, Monica Marinescu, Vladimir Yut, Yatish Patel, Andrew J. Marquis, Nigel P.
Brandon, and Gregory J. Oer. The eect of thermal gradients on the performance of lithium-ion
batteries. Journal of Power Sources, 247(0):10181025, 2014.
[33] Matthew Klein, Shijie Tong, and Jae Wan Park. The Performance Eects of Edge-Based Heat Transfer
on Lithium-Ion Pouch Cells Compared to Face-Based Systems. In SAE Technical Paper 2014-04-01,
SAE World Congress, Detroit, MI, USA, 2014.
[34] Suleiman Abu-Sharkh and Dennis Doerel. Rapid test and non-linear model characterisation of solid-
state lithium-ion batteries. Journal of Power Sources, 130(12):266274, 2004.
[35] Long Beach, CA, 2009.
[36] Xiaosong Hu, Shengbo Li, and Huei Peng. A comparative study of equivalent circuit models for Li-ion
batteries. Journal of Power Sources, 198(0):359367, 2012.
Appendix
MCT =
N
ch

ch=1

Nt
i=1
T
ch,i
Nt
N
c
h
(A)
18
MTD =
Nt

ch=1
T
7,i
T
1,i
N
t
(B)
MPE =
N
ch

ch=1

Nt
i=1
T
ch,i
T
ch,ideal
Nt
N
c
h
(C)
Table A: Average measured temperatures during testing procedures.
19
Table B: Standard deviation of measured temperatures during testing procedures.
Figure A: Schematic and photograph of the TGS used for controlled temperature distribution testing in this
work.
20

Anda mungkin juga menyukai