ArticlePDF Available

A Review of Phase Change Materials for the Thermal Management and Isothermalisation of Lithium-Ion Cells

Authors:

Abstract and Figures

Download published paper at: https://www.sciencedirect.com/science/article/pii/S2352152X19301355?via%3Dihub Li-Ion batteries will play an important role in reaching emission targets by sustaining the further integration of renewable energy technologies and Electric Vehicles (EVs) in society. Their performance however is quite sensitive to temperature, leading to capacity fade, acceleration of ageing effect and possible thermal runaway. A Thermal Management System (TMS) should maintain a battery at an operating temperature within an optimal range and maximise temperature uniformity, i.e. approaching an isothermal condition. Many studies have experimentally investigated the electrical performance of Li-Ion batteries under controlled environmental temperatures. Notably however, these controlled conditions do not impose a uniform temperature or a controlled rate of cooling, as a TMS would. From a review of the relevant literature a ratio of the heat generation to the power production is proposed, i.e. quantifying an equivalent electro-chemical efficiency to advance research in this technological area and as additional TMS design metric. Overall, there is enough evidence that 25–30 °C is the best temperature range to minimise the ageing effect while 25–40 °C is typically reported as the general Li- Ion cells operating range. No specific temperature is identified to optimise the cycle electro-chemical efficiency and minimise the ageing effect. Therefore, a TMS should keep Li-Ion batteries within a specific temperature range according to the need for either higher electro-chemical efficiencies (i.e. higher powers and lower heat generation rates) or higher operating life. There are four main thermal management approaches of Li-Ion batteries: air-cooling, liquid-cooling, boiling and Phase Change Materials (PCM). Air cooling is preferred for safety reasons but is less efficient as the rate of heat transfer achievable is relatively low. Forced air cooling can effectively keep the temperature at a preferred level but fails to guarantee a uniform temperature. Liquid cooling is better in terms of heat transfer performance, but it is less safe and can still result in significant thermal gradients within the pack. Boiling effectively keeps Li-Ion cells temperature constant and uniform but can be quite complex to operate and control. Phase Change Materials (PCMs) as a passive cooling approach are proposed as an effective and low-cost isothermalisation technique. However, when Li-Ion batteries are operated under extreme conditions (high ambient temperature, high discharge rates), PCM are not able to recover all latent energy potential during solidification and this leads to possible thermal runaway. Overall, it is clear that no TMS alone is holistically better than others and the choice between air cooling, liquid cooling, boiling and latent heat PCM systems is entirely linked to the specific combination of temperatures, heat rates, cells capacity and geometry. Active PCM systems however, mainly a combination of liquid cooling and passive PCM, show promising results towards an ideal isothermal condition. Also, they introduce the potential to store the thermal energy and use it as needed, converting a Li-Ion cell from an Electrical Energy Storage System (EESS) to a Combined Heat and Power (CHP) system.
Content may be subject to copyright.
Li-ion batteries Thermal Management Systems: a Literature Review.
S. Landinia,
, J. Leworthyb, T.S. O’Donovana
aHeriot Watt University, School of Engineering and Physical Sciences, Institute of Mechanical, Process and Energy Engineering,
Riccarton, EH11 1HA Edinburgh, United Kingdom
bDukosi Ltd., Unit 4 Bush House Cottages, Edinburgh Technopole, Edinburgh, Penicuik EH26 0BA
Abstract
Li-ion batteries will play an important role in reaching emission targets by sustaining the further integration of renewable
energy technologies and EVs in society. Their performance however is quite sensitive to temperature, leading to capacity
fade, acceleration of ageing effect and possible thermal runaway. As shown in this review, the aims of an effective thermal
management system (TMS) is to maintain a battery at an operating temperature within a certain optimal range and maximise
temperature uniformity. Many studies have experimentally investigated the electrical performance of Li-ion batteries under
controlled environmental temperatures. Importantly however, these controlled conditions do not impose a uniform temperature
or a controlled rate of cooling, as a TMS would. The review of the relevant literature also proposes that the ratio between the
heat generation and the power production, i.e. quantifying an equivalent electro-chemical efficiency would help to advance
research and development in this important technological area. Overall, there is enough evidence that 25-30
oC
is the best
temperature range to minimise the ageing effect. However, no specific temperature is indicated in literature which can optimise
the cycle electro-chemical efficiency and minimise the ageing effect. There are four main approaches to the TMS of Li-ion
batteries: air-cooling, liquid-cooling, boiling and Phase Change Materials (PCM). Air cooling is preferred for safety reasons
but is less efficient as the rate of heat transfer achievable is low. Forced air cooling can effectively keep the temperature at a
preferred level but fails to guarantee a uniform temperature. Liquid cooling is better in terms of heat transfer performance,
but it is less safe and can still result in significant thermal gradients within the pack. Boiling is really effective in keeping
Li-ion cells temperature constant and uniform but is quite complex to operate and control. Phase Change Materials (PCM)
as a passive cooling approach has been proposed as an effective and low-cost TMS. However, when Li-ion batteries are
operated under extreme conditions (high ambient temperature, high discharge rates), PCM are not able to recover all latent
energy potential during solidification and this leads to possible thermal runaway. Therefore, active PCM systems, mainly a
combination of liquid cooling and passive PCM, give promising results by reducing both maximum temperature and in-cell
temperature gradients. Also, they introduce the potential to store the thermal energy and use it as needed, converting a
Li-ion cell from a simple EESS toaCHP system. Overall, it seems clear from the literature that no TMS is holistically better
than others and the choice between air cooling, liquid cooling, boiling and latent heat PCM systems is entirely linked to the
specific combination of temperatures, heat rates, cell’s capacity and geometry.
Keywords: Li-ion batteries, thermal management, air-cooling, liquid-cooling, phase change materials,
electro-chemical efficiency
Corresponding author
Email address: sl30@hw.ac.uk (S. Landini)
Preprint submitted to Elsevier 30th March 2019
Contents
Acronyms 4
Nomenclature 5
1 Introduction 6
2 Lithium-Ion Batteries 7
2.1 Overview ................................................... 7
2.2 Cell Types ................................................... 8
2.2.1 Cylindrical Cells ........................................... 8
2.2.2 Prismatic Cells ............................................ 9
2.2.3 Pouch Cells .............................................. 11
2.3 Chemistries and Electro-chemical Properties ................................ 12
3 Li-ion Batteries Thermal Behaviour 14
3.1 Overview ................................................... 14
3.2 Thermal Properties .............................................. 14
3.3 Electro-chemical Efficiency .......................................... 16
3.3.1 Pouch Cells .............................................. 17
3.3.2 Cylindrical Cells ........................................... 19
3.4 Ageing Effect ................................................. 21
3.5 Batteries Thermal Models .......................................... 24
3.5.1 Lumped Capacitance ......................................... 24
3.5.2 Electro-chemical models ....................................... 26
3.6 Summary ................................................... 27
4 Thermal Management Systems (TMS) 28
4.1 Overview ................................................... 28
4.2 Air-cooling .................................................. 29
4.2.1 Micro-channels ............................................ 30
4.2.2 Heat Pipes .............................................. 31
4.3 Liquid-cooling ................................................. 31
4.4 Boiling ..................................................... 33
4.5 Phase Change Materials ........................................... 36
4.5.1 Overview ............................................... 36
4.5.2 Classification ............................................. 36
4.5.3 Thermal Conductivity Enhancement (TCE) Methods ...................... 41
4.5.3.1 Fins ............................................. 43
4.5.3.2 Carbon and Metallic Additives .............................. 43
2
4.5.3.3 PCM Slurries ........................................ 45
4.5.3.4 Multi-tube Configurations ................................. 47
4.5.3.5 PCM and Heat Pipes ................................... 48
4.5.3.6 Multiple or Cascaded PCM Systems ........................... 48
4.5.4 Li-ion Cells Passive Cooling by PCMs ............................... 48
4.5.5 Li-ion Cells Active Cooling by PCMs ................................ 53
4.6 Summary ................................................... 54
5 Conclusions 55
Reference List 57
Appendix A Efficiency and Heat Ratio 68
Appendix A.1 Pouch Cells .......................................... 68
Appendix A.1.1 Instant Performance ................................... 68
Appendix A.1.2 Overall Performance .................................. 71
Appendix A.2 Cylindrical Cells ........................................ 74
Appendix A.2.1 Instant Performance ................................... 74
Appendix A.2.2 Overall Performance .................................. 77
3
Acronyms
R2Coefficient of Determination 25
ANN Artificial Neural Network 26
CAES Compressed Air Energy Storage 6,55
CENG Compressed Expanded Natural Graphite 51,52
CFD Computational Fluid Dymamics 20,43,51
CHP Combined Heat and Power 1,54,56
CR Charge Rate 2224,27,3335,5456
CTES Cascaded Thermal Energy Storage 41
DOD Depth Of Discharge 1618,20,2527,55,68,74
DR
Discharge Rate 1628,3035,49,51,52,5456,70,
72,73,76
DSM Demand-Side Management 6
EESS Electrical Energy Storage System 1,6,29,5456
EG Expanded Graphite 43,44,52,53
EOL End Of Life 21,22
ESS Energy Storage System 6
EV Electric Vehicle 1,68,13,14,26,27,49,55
HEV Hybrid Electric Vehicle 7
HEV Plug-in Hybrid Electric Vehicle 7
HEX Heat Exchanger 36,42,48
HPS Heat Pipe System 6,29,31,42,48
HR Heat Ratio 1721,72,73
HTF Heat Transfer Fluid 27,36,39,40,42,45,46,48
LCO Lithium Cobalt Oxide 13,14
LCOE Levelised Cost Of Electricity 14
LCOS Levelised Cost Of Storage 55
LFP Lithium Iron Phosphate 13
LH Latent Heat 36,48,53
LMO Lithium Manganese Oxide 13,14
MPCM Multiple/Cascaded Phase Change Material 48
NCA Lithium Nickel Cobalt Aluminium 13
NMC Lithium Nickel Manganese Cobalt 13
O&M Operation & Maintenance 28,49,52
PCM Phase Change Material 1,3,7,28,29,3656
PHS Pump Hydro Storage 6,55
PV Photo-Voltaic 6
RET Renewable Energy Technologies 6
RMSE Root Mean Squared Error 25
S&T Shell & Tube 43,47
SEI Solid-Electrolyte Interface 22,49
SH Sensible Heat 36
SOC State Of Charge 10,26
SOH State Of Health 2123
TCE
Thermal Conductivity Enhancement 7,29,4045,
47,50
TESS Thermal Energy Storage System 36,41,43,48
TMS
Thermal Management System 1,7,14,16,18,20,
2729,31,3335,37,4856
4
5
1. Introduction
A significant reduction in energy-related
CO2
emissions is required to meet the binding commitments of COP21/22
to keep the global temperature rise within 2
°
C of pre-industrial levels. The development and deployed of renewable
energy technology and systems can make a significant contribution to the low-carbon energy mix, exploiting local
energy sources and displacing high carbon fossil fuels. The world primary energy demand has grown by 1.8%
since 2011 [
1
,
2
]. Renewable energy sources accounted for 19.3% of the total final energy consumption in 2015;
biomass/geothermal/solar heat accounted for 4.2% and wind/solar/biomass/geothermal electricity contributed 1.6%.
In terms of global electricity production, renewable energy technologies contributed a total of 24.5% [
1
,
2
]; this
consisted of hydro-power (16.6%), wind (4%), biomass (2%) and solar PV (1.5%). In some cases, ratios of RET
electricity production and demand reached values in the range of 106-140 % (e.g. Denmark, Scotland). The most
developed technologies in this sector were wind (both onshore and offshore) and solar PV (in the EU 86% of new
power additions were from wind and solar PV) [1,2].
Renewable energy sources such as wind and solar are inherently intermittent however and have limited predictability,
therefore any energy systems, irrespective of its scale (domestic, micro-grid, grid level) requires storage to realise
the full potential to match demand with supply [
3
]. Typically Energy Storage Systems (ESS) are chosen for the
specific application in terms of capacity, discharge period, response time and power rating for electrical storage and
output temperature and capacity for thermal storage [
3
]. As reported by REN21 [
1
], the development of electrical
RETs is strictly linked to three main areas of R&D: end-use technologies (e.g. EVs and HPSs), electrical energy
storage systems (EESS) and Demand-Side Management (DSM). Among these, EESS are considered essential to
balance the mismatch between RET electricity production and customers demand (i.e. grid management), permit
the exploitation of local RET sources in off-grid systems, avoid further infrastructure spreading and provide backup
in case of power outages [
1
]. The basic idea is capturing electrical energy during off-peak or RET overproduction
periods and releasing it during high peak and/or high energy price periods. Today, the most developed EESS are
PHS, electro-chemical batteries (mainly Lead Acid, NiCd, NiMH and Li-ion), flywheels and CAES [
1
]. The global
grid-connected and stationary storage power by 2016 was 156
GWe
, with 96% being PHS [
1
,
4
]. In 2016, “advanced”
EESS (not PHS) increased by 0.8
GWe
(+14%), totalling 6.4
GWe
[
1
,
4
]. Of this, 1.7
GWe
(27%) is related to
electro-chemical batteries (mainly Li-ion), which saw a growth of 50% compared to 2015 [
1
]; this is attributed to
the increase in EV battery production in the period 2010-2015, which saw manufacturing costs decrease by 65%.
According to IEA [
3
], 310
GWe
new grid-connected electrical storage power is needed in USA, EU, China and India
to sustain future de-carbonised scenarios.
The operating temperature of a Li-ion battery has a significant influence on its overall electrical performance [
5
];
batteries performance is sensitive to both high and low temperatures [
5
]. At low temperatures, they lose storage
capacity and charge acceptability [
5
]. At high operating temperatures the round trip efficiency decreases; the charge
acceptance, power and energy capacity, reliability and cycle life are all compromised. All of these factors contribute
to the higher capital and operation and maintenance costs when integrated as part of a low carbon energy system
and consequently lead to higher unit price of energy on a full life cycle analysis [5,6].
6
The effective thermal management of Li-Ion batteries which are exothermic during charge and discharge and subjected
to very different environmental conditions depending on their application is therefore essential to reducing the cost
of renewable energy [
5
]. As presented later in this review, the aims of an effective thermal management system
is twofold: 1) to maintain a battery at an operating temperature within a certain optimal range (typically 25-40
°
C) and minimise temperature variation within the battery; a uniform temperature helps to avoid localised cell
deterioration which in turn increases the performance defect throughout all the batteries pack [6].
Many studies [
7
,
8
,
9
,
10
,
11
,
12
,
13
,
14
,
15
,
16
,
17
,
18
] have experimentally investigated the electrical performance
of Li-ion batteries under controlled environmental temperatures. Importantly however, these controlled conditions
do not impose a uniform temperature or a controlled rate of cooling, as a Thermal Management System would. The
review of the relevant literature also proposes that the ratio between the heat generation and the power production,
i.e. quantifying an equivalent electro-chemical efficiency would help to advance research and development in this
important technological area.
This literature review is composed of the following sections. Section 2 gives an overview of Li-ion batteries benefits,
costs, geometries, chemistries, electro-chemical properties. Section 3 focuses on Li-ion batteries thermal behaviour,
thermo-physical properties, electro-chemical efficiencies (based on a complete review of Li-ion batteries heat generation
data available in previous literature), ageing effect and batteries thermal models. Section 4 covers the topic of
Li-ion batteries thermal management systems (TMS) based on air-cooling, liquid-cooling, boiling and phase change
materials. A detailed subsection on Phase Change Materials (PCM) classification, thermo-physical properties,
thermal conductivity enhancement (TCE) methods and their utilisation in Li-ion batteries passive/active TMS is
also proposed.
2. Lithium-Ion Batteries
2.1. Overview
There are few types of Li-ion batteries: cylindrical, coins, prismatic and pouch [
7
]. Typically, prismatic and pouch
geometries are used for high capacity applications, such as EV,HEV and HEV. Similarly to other chemistries,
Li-ion batteries are composed of 6 components: positive current collector, cathode, negative current collector, anode,
electrolyte and separator [
7
]. The cathode (positive electrode) is typically a lithium oxide while the anode (negative
electrode) is a compound made with carbon (e.g. graphite) and/or silicon. The electrolyte is a lithium salt dissolved
in organic solvents (e.g. ethylene carbonate, diethyl carbonate, dimethyl carbonate) [
7
]. Typical properties of Li-ion
batteries are reported in Table 1.
In terms of costs, there are different values (mainly derived from EV market) available in literature (Figure 1) even
if the most widely accepted range is 200-384
$
/kWh [
19
] with an average in 2016 of 227
$
/kWh (-77% compared to
2010) [
6
]. Moreover, Cairn Energy Research Advisors and Tesla claim values of respectively 139
$
/kWh and lower
than 190 $/kWh [6,19].
7
Table 1: Properties of Li-ion Batteries [6]
Property Value Unit
Energy volumetric density 200-600 Wh/L
Energy specific density 90-200 Wh/kg
Power density 1500 W/kg
Life Cycle 1000-4000 Number of cycles
Operating Life* 500-1000 Number of cycles
Overcharge tolerance Low -
Self-discharge >10 %/month
Operating Temperature tolerance -20-60 C
Optimal Temperature Range 25-40 C
Optimal Temperature 27 C
Temperature difference within cells pack <5C
* Operating Life: 20 %reduction of battery capacity compared to initial condition
Figure 1: Li-ion Battery pack cost used in EVs [19]
2.2. Cell Types
2.2.1. Cylindrical Cells
The cylindrical cell is one of the most widely used packaging types for Li-ion cells, mainly due to their good
mechanical stability and ease of manufacture [
20
]. Cylindrical cells possess high energy densities; however, they
have a low packing efficiency due to the unavoidable space left between the cells when they are stacked together,
resulting in a lower energy density at a pack level. Figure 2 shows the structure of a typical cylindrical Li-ion cell,
where the electrodes are layered and spirally wound into a “jelly roll” which is then inserted into a steel can. Most
cylindrical cells benefit from having built in safety devices such as positive thermal coefficient (PTC) switches that,
when exposed to excessive current, heat up and become resistive, stopping the flow of current in the cell. As well as
this, cylindrical cells also feature pressure relief mechanisms which are designed to release excessive pressure build
8
Figure 2: Schematic to show the structure of a cylindrical Li-ion cell
up from the formation of gases inside the cell. Gasses inside the cell are usually generated due to abusive use of the
cell, i.e. overcharging, physical damage or excessive temperature rises [
21
]. One great advantage of the cylindrical
cell is its mechanical stability; unlike prismatic and pouch cells, cylindrical cells exhibit minimal or no “swelling”
due to build ups of internal gases and do not require any externally applied compression.
The cylindrical cell is available in a wide variety of standard formats; for Li-ion cells, the most common formats are
14500, 18650 and 26650. The first two digits identify the diameter of the cell in millimetres, and the last three digits
are the length of the cell in tenths of millimetres; therefore, an 18650 cell measures 18mm in diameter and 65mm in
length. The 18650 cell has been the cell of choice for Tesla, using the same cell format in both their electric vehicles
and energy storage batteries mainly due to the manufacturability, availability and cost of production. In 2015 Tesla,
Panasonic and Samsung have begun manufacturing a new 21700 cell, claiming that this new cell format has 35%
more energy capacity per unit volume than the 18650 cell [
22
]. Table 2 summarises the different sizes of cylindrical
cells and their associated volumes and surface areas.
Table 2: Cylindrical Cell Geometries
Cell Format Cell Height (mm) Cell Diameter (mm) A (m2) V (m3)A/V (m1)
14500 50 14 2.51x10-3 7.70x10-6 325
18650 65 18 4.18x10-3 1.65x10-5 253
26650 65 26 6.37x10-3 3.45x10-5 184
21700* 70 21 5.31x10-3 2.42x10-5 219
A = Surface Area, V=Volume
* = future cell format
2.2.2. Prismatic Cells
The internal assembly of a prismatic cell is much the same as a cylindrical cell, however in this case the cells “jelly
roll” is inserted into a prismatic can as shown by Figure 3. In some prismatic cells, the jelly roll is replaced by a
laminated stack of anode-separator-cathode assemblies. Prismatic cells are slightly less energy dense than cylindrical
cells due to a thicker can wall; however, they make up for the lost energy density by having a much better packing
9
Figure 3: Schematic to show the structure of a prismatic Li-ion cell.
efficiency due to their cuboid format. Like cylindrical cells, prismatic cells also have in-built safety features such as
thermal fuses and gas vent ports. There is currently a lack in standard formats of prismatic cells, as usually they
are redesigned to make best use of the available space in each application; however, some standard cell geometries
do exist and are listed in Table 3. It can be seen that prismatic cells range of surface area to volume ratios are
lower than that of cylindrical cells. From a thermal management perspective, this means that any proposed thermal
management system would have to remove heat at a faster rate than for the equivalent cylindrical cell. Unlike
cylindrical cells, prismatic cells are prone to “swelling” due to a build of gasses within the cell and are required to be
assembled into rigid enclosures that applies compression to the two broad faces of the cell [
23
], further complicating
the battery pack mechanical design. The cells expansion can be seen as a disadvantage from a thermal management
perspective as the cells swelling can result in de-lamination of layers within then cell, which in turn would decrease
its through-plane thermal conductivity and thus the rate at which heat can be dissipated from the cell [
24
]. Studies
carried out by A123 Energy Solutions found that a typical prismatic cell can expand up to 1% of their initial
thickness under normal cycling from 100% to 0% SOC; in addition, over their life cycles they can grow to be 3-5%
greater than their initial thickness [23].
Table 3: Verband der Automobilindustrie (VDA) standard prismatic cell formats.
Cell Format Length (mm) Height (mm) Width (mm) V (m3) A (m2)A/V (m1)
HEV 120 85 12.5 1.27x10-4 25.52x10-3 200
PHEV1 173 85 21 3.08x10-4 40.24x10-3 130
PHEV2 148 91 26.5 3.56x10-4 39.60x10-3 104
PHEV2+ 148 125 26.5 4.90x10-4 51.47x10-3 105
EV1 173 115 32 6.36x10-4 58.21x10-3 91
EV2 173 115 45 8.95x10-4 65.71x10-3 73
A = Surface Area, V=Volume
10
Figure 4: Schematic to show the structure of a Li-ion pouch cell.
2.2.3. Pouch Cells
Pouch cells are constructed by stacking multiple layered sheets of anode-separator-cathode assemblies which are
then inserted into a pouch, the pouch is then filled with electrolyte and sealed leaving the positive and negative
terminals outside of the pouch, as shown by Figure 4. The material used for the pouch is a laminated aluminium film
with a layer of polyamide on the outside and polyethylene on the inside to electrically insulate the casing material
and prevent corrosion from the electrolyte. The elimination of the rigid metal enclosure saves weight and means
the pouch cell is capable of high energy densities and packing efficiencies. A disadvantage of having no rigid case
however is that the cells are extremely vulnerable to external mechanical damage and swelling from internal gasses.
Because of this, battery packs utilising pouch cells must extensively protect the cells from external damage and
provide compression whilst still allowing for small expansions in the cells thickness, the pressure applied to the cell
directly effects the cells life [
25
]. It is also important to note that pouch cells lack in built safety features such as
PTC switches and mechanical gas vents; instead, when gases are formed and pressure within the cell increases above
a certain threshold, the gases will vent through an intentional weak spot in the cells pouch which is usually located
in one corner of the cell.
Figure 5: Pouch cells with varying tab locations.
Currently there are no standards sizes of pouch cells as they are still a relatively immature technology in comparison
to cylindrical cells and are easily manufactured to bespoke sizes. However, like prismatic cells, efforts have been
made to standardise the dimensions of pouch cells by organisations such as the Association of German Automobile
11
Manufacturers (VDA), examples of the proposed cell formats are shown in Table 4.
Table 4: Verband der Automobilindustrie (VDA) standard pouch cell formats.
Cell Format Length (mm) Height (mm) Width (mm) V (m3) A (m2)A/V (m1)
HEV 243 121 X (4 – 12) 1.2 – 3.5 x10-4 5.88 x 10-2 500 – 166
PHEV 227 165 X (4 – 12) 1.5 – 4.5 x10-4 7.49 x 10-2 500 – 166
BEV 330 162 X (4 – 12) 2.1 – 6.4 x10-4 1.07 x 10-1 500 – 166
A = Surface Area, V=Volume
It can be seen in Table 4 that pouch cells are capable of high surface area to volume ratios when the cell width is
small, this is beneficial for thermal management as heat can be dissipated via the two large faces of the cell. The
placement of the tabs on pouch cells is also variable. Figure 4 shows examples of the two main formats available; in
some cases the tabs are located on the same side of the cell, whereas some pouch cells have tabs at either end of the
cell. Investigations have been carried out to find the effect of the tab locations [
26
] on pouch cells, showing that cells
with tabs at either end exhibit a much more uniform potential and current density distribution across the cell, which
leads to a more uniform temperature gradient across the cell, which in turn prolongs the life of the cell.
2.3. Chemistries and Electro-chemical Properties
A diagram of the charging process for a Lithium-ion cell is shown in Figure 6. The Lithium-ion cell is composed of a
positive electrode (Cathode), negative electrode (Anode) and electrolyte. The two electrodes are separated by a
separator sheet to prevent a short circuit from occurring [27].
Figure 6: Electrochemical charging process in a Lithium-ion cell [27]
When the cell is connected to a power supply and charged, lithium ions de-intercalate from the positive electrode,
flow through the electrolyte and the separator sheet, then intercalate into the negative electrode [
27
]. Electrons flow
in the opposite direction to the ions, flowing from the positive electrode through the external circuit and back to
12
the negative electrode. This process is reversed when the cell is discharged from an external load [
27
]. The anode
material of a lithium-ion battery typically consists of carbon, usually graphite [
27
]. The cathode of a lithium-ion
cell can be made of a variety of different lithium metal oxides. Table 5 below summarises the typical performance
characteristics of some of the most common Lithium-ion cathode chemistries. Many more cell chemistries exist as
manufactures use combinations of these to exploit the benefits of each chemistry to develop the optimum for a given
application [27].
Table 5: Summary of common Lithium-ion Chemistries [27]
Cathode Chemistry LFP LMO LCO NCA NMC
Specific Energy (W h/kg ) 80-130 105-120 120 – 150 80 – 220 140 – 180
Energy Density (W h/L) 220-250 250 – 265 250 – 450 210 – 600 325
Specific Power (W/kg) 1400-2400 1000 600 1500 – 1900 500 – 3000
Power Density (W/L) 4500 2000 1200 – 3000 4000 – 5000 6500
Voltage (V) 3.2 - 3.3 3.8 3.6 – 3.8 3.6 3.6 – 3.7
Operating Life (Cycles) 1000 – 2000 >500 >700 >1000 1000 – 4000
Operating Temperature Range (oC) -20 to 60 -20 to 60 -20 to 60 -20 to 60 -20 to 55
Legend: Lithium Iron Phosphate (LFP), Lithium Manganese Oxide (LMO), Lithium Cobalt Oxide (LCO), Lithium
Nickel Cobalt Aluminium (NCA), Lithium Nickel Manganese Cobalt (NMC)
Lithium Iron Phosphate (LFP) batteries have a
LiF eP O4
cathode and typically a graphite anode. LFP was
commonplace in automotive applications where high charge and discharge rates are required for fast acceleration and
regenerative braking [
27
]. Its high power capabilities were well suited to this application as well as having a relatively
low cost compared to other chemistries that make use of rare elements [
27
]. LFP is considered as one of the safer
chemistries, especially when compared to LCO, because they start to decompose at higher temperatures.
EV’s are now moving towards the use of NMC and NCA as they have superior energy and power densities to
LFP.NCA batteries have a
LiNiCoAlO2
cathode and graphite anodes while NMC batteries have a
LiNiMnCoO2
cathode and a graphite anode. For instance, Tesla uses NCA chemistry in the EV models S and X and NMC in their
Power Wall battery banks [
28
]. The higher voltages of NMC and NCA compared to LFP allows battery packs to be
smaller, as they require less cells to achieve the specific voltage for a given application, resulting in more compact
EVs [
28
]. LFP is still used in automotive applications but mostly in larger vehicle (e.g. buses) where battery volume
is less of a concern [28].
LCO cells are typically found in portable electronic devices such as mobile phones, laptops and cameras [
29
]. The
anodes of LCO cells are made of
LiCoO2
and the cathodes are graphite. Its high energy density results in smaller
cells and ultimately smaller devices [
29
]. The disadvantages of LCO are its relatively high cost and low thermal
stability. It is one of the most reactive chemistries and can enter a thermal runaway event at lower temperature
than other chemistries; for these reasons, it is rarely used in large applications such as EVs [29].
LMO chemistry cells are typically found in power tools and e-bikes, where its high voltage and energy density results
13
in small battery packs; however, this chemistry suffers from a lower cycle life than most chemistries. These batteries
are formed of
LiMn2O4
cathodes; the anodes can be graphite or LCO.LMO cells are one of the safest lithium cell
chemistries. Thermal runaway occurs at approximately 250oCcompared to 150oCfor LCO cells [28].
3. Li-ion Batteries Thermal Behaviour
3.1. Overview
The performance of Li-ion batteries is typically influenced by design, materials and operating temperatures [
5
].
These batteries suffer both high and low temperatures [
5
]. At low temperatures, they lose storage capacity and
charge acceptability [
5
]. In general, an increase in the operating temperature leads to lower battery performance,
lower round trip efficiency, lower charge acceptance, lower power and energy capability, lower reliability, lower cycle
life and consequently higher costs (LCOE) [6,5].
Therefore, the development of proper thermal management systems (TMS) is essential [
5
]. The aims are keeping
the batteries at an operating temperature within a certain optimal range (typically 25-40
C, Table 1) and keeping
the temperature variation within the battery as low as possible. A uniform temperature helps to avoid localised
cell deterioration which in turn increases the performance defect throughout all the batteries pack [
6
]. Moreover, it
must be mentioned that in the case of EV batteries, considering the necessity of more autonomy (i.e. more energy
storage), the number of batteries per volume within the pack keeps growing together with the market expansion [
6
].
Therefore, this will create serious thermal problems within the cells located in the centre of the pack, with more and
more probability of thermal runaways and failures [6,8].
3.2. Thermal Properties
The two thermal transport properties that primarily govern temperature rise in Li-ion cells are thermal conductivity
and heat capacity. Cylindrical cells are constructed of a spirally wound electrode assembly (jelly roll) which is then
inserted into a metal can usually made of stainless steel (
kSteel
= 16
19
W/mK, Cpsteel
= 500
J/kg K
). The spiral
electrode assembly illustrated by Figure 7 results in anisotropic thermal conductivity between the radial and axial
direction of the cell [
30
] due to the large number of interfaces between the electrode layers in the radial conduction
path, which are absent in the axial direction.
Experimental studies carried out by Drake et al. [
30
] on cylindrical cells found that the radial thermal conductivity
is two orders of magnitude lower than the axial direction. The results of the tests carried out on
LiF eP O4
18650 and
26650 cells are summarised in Table 6. The values for thermal conductivity in the radial direction are somewhat lower
than those recorded in previous work that reported radial thermal conductivities of around 1-3
W/m K
[
32
] and
0.3-1.6
W/m K
[
33
]. This is possibly due to the fact that Drake et al.’s work considers thermal contact resistances
that exist in an actual cell whereas the other stated values are calculated from individual cell components such as
the electrode-separator assembly; hence, the values reported in Drake et al.’s work [
30
] are expected to be more
accurate.
The strong anisotropy of thermal conductivity within the cylindrical cell results in a temperature distribution shown
by Figure 8. This knowledge is critical when designing thermal management solutions as this identifies the best
14
Figure 7: Schematic illustration of cylindrical Li-ion cell [31]
Table 6: Measured thermo-physical properties of 26650 and 18650 LiF eP O4cells [30]
Cell Format kr(W/m K)kz(W /m K)cp(J/kg K)
26650 0.15 ±0.01 32.0 ±1.6 1605 ±80
18650 0.20 ±0.01 30.4 ±1.5 1720 ±86
kr=radial thermal conductivity
kz=axial thermal conductivity
cp=specific heat capacity
thermal pathways for removing heat; in this case, heat is transferred much better in the axial direction of the
cell.
Figure 8: Temperature profile of cylindrical cell with anisotropic thermal conductivity [30]
Further work carried out by S.J Drake [
34
] on pouch cells also showed strong anisotropy in in-plane and through-
plane thermal conductivity. His experiments were carried out on a range of pouch cells, finding in-plane thermal
conductivities of 40-45
W/m K
and through plane thermal conductivities of 0.3-0.65
W/m K
; however, the exact
specifications of the cells were not given [
34
]. Similar investigations carried out on a range of
LiCoO2
pouch cells
also showed a large anisotropy in thermal conductivity, finding thermal conductivities through-plane ranging from
15
0.84–1.63
W/m K
and in-plane thermal conductivities of 29.99–36.96
W/m K
[
24
]; again, however, the exact cell
specifications were not published. The same study also showed that the through-plane thermal conductivity of
pouch cells can reduce over the cells life cycle due to the pouch cell swelling and layers within the cell becoming
de-laminated. They showed that a cells through-plane thermal conductivity can be reduced by up to 30% when
cycled at 45
C for 500 cycles. The aluminium foil casing material for pouch cells also has anisotropic thermal
conductivity properties as it is a laminated structure, an investigation found through-plane and in-plane thermal
conductivities of 0.28
W/m K
and 90.8
W/m K
respectively, a volumetric heat capacity of 2096
kJ/m3K
and an
inter-facial thermal conductance between the battery core (jelly roll) and the aluminium foil case of 1300
W/m2K
[
35
]. All the information found in literature on thermal conductivity and thermal capacity of Li-ion cells has been
summarised in Table 7; some values are left empty as this has not been supplied by the referenced material.
Table 7: Summary of Thermophysical cell properties found in open literature.
Cell Format K(x,y)(W/m K )Kz(W/m K)Kr(W/m K)cp(kJ/kg K )
SONY US-18650 LiCoO2[36] - N/A 0.35 1.04 ±0.02
Cylindrical LiF eP O4cell [37] - N/A 0.33 – 0.66 1.02
LiCoO2Pouch Cell [38] 24.840 1.035 N/A -
LCO Pouch Cell [24] 36.96 1.63 N/A -
LCO Pouch Cell [24] 32.31 1.35 N/A -
6 Ah Li-ion cell [39] - - - 0.795
4 Ah Li-ion cell [39] - - - 1.0118
Li-ion Polymer Pouch cell [24] - - - 1.028
LiF eP O418650 Cell [30] 30.4 ±1.5 N/A 0.20 ±0.01 1.720
LiF eP O426650 Cell [30] 32.0 ±1.6 N/A 0.15 ±0.01 1.605
K(x,y)(W/m K ) = In-plane Thermal Conductivity
Kz(W/m K ) = Through-plane Thermal Conductivity
Kr(W/m K ) = Radial Thermal Conductivity
cp(kJ/kg K) = Specific Heat Capacity
3.3. Electro-chemical Efficiency
Several papers deal with the Thermal Behaviour of Li-ion batteries without implementing any TMS in order to
evaluate their heat generation. The batteries are typically tested at different constant discharge currents, whose
values are reported by using a parameter called Discharge Rate (DR) [
7
,
8
,
9
]. This is defined as the constant current
which discharges the entire capacity of the battery in one hour
1
[
7
]. The heat generation rates are evaluated at
different operating conditions (i.e. DR,DOD, ambient temperature), for both cylindrical [
40
,
41
,
42
,
43
] and pouch
geometries [7,8,44,45,9,46,47,48,49].
However, all the mentioned papers fail in evaluating the ratio between the heat generation and the power production,
1For instance: Capacity=20Ah, DR=1C, 2C, 3C, 4C : Current 20, 40, 60, 80 A)
16
Table 8: Pouch cell datasets for Efficiency and HR.
Reference Chemistry Mass [kg] Volume [L] ρ[kg/m3]Asurf ace [m2]Lc* [mm] Capacity [Ah]
[8]LiF eP O40.496 0.2633 1884 0.0783 3.365 20
[9]LiF eP O40.496 0.2633 1884 0.0783 3.365 20
[15]LiF eP O41.138 0.576 1976 0.0539 10.68 40
[46]LiF eP O40.2596 0.138 1881 0.0282 4.9 10
[47]LiF eP O40.261 0.127 2054 0.029 4.38 10
[48]LiMnNiCoO20.99 0.5061 1956 0.1039 4.87 40
[49]LiF eP O40.2698 0.164 1645 0.0853 1.92 6
*Lc=Characteristic Length=V olume/Area
i.e. quantifying an equivalent electro-chemical efficiency. Therefore, data were collected from all the studies mentioned
(raw data when provided by the authors or extrapolated from graph) and systematically analysed to evaluate two
meaningful factors defined in Equation 1:
η=P
P+˙
Q=1
1 + HR =P
Ptot
(1a)
HR =˙
Q
P=1η
η(1b)
where P [W] is the electrical power,
˙
Q
[W] is the heat generation rate,
Ptot
[W] is the total power discharged by the
battery (i.e. chemical power, defined as
Ptot
=
P
+
˙
Q
),
η
is the battery electro-chemical efficiency and HR is the
heat ratio. These two parameters are instantly determined (i.e. based on power), i.e. for each operating condition
(DR, ambient temperature) and specific DOD. The equivalent total discharge performance (i.e. based on energy)
can be evaluated by the overall factors determined by Equation 2:
ηoverall =E
E+Q=1
1 + HRoverall
=E
Etot
(2a)
HRoverall =Q
E=1ηoverall
ηoverall
(2b)
E=
t=τ
Z
t=0
P(t)dt, Q =
t=τ
Z
t=0
˙
Q(t)dt, τ =1
DR (2c)
where
E
[Wh] is the total electrical energy discharged, Q [Wh] is the total heat generated,
ηoverall
is the overall
electro-chemical efficiency and HRov erall is the overall heat ratio.
3.3.1. Pouch Cells
Table 8 reports the information regarding the dataset analysed for pouch geometry [
7
,
8
,
44
,
45
,
9
,
46
,
47
,
48
,
49
]
while Table 9 gives details on tests condition. The dataset is characterised by different nominal capacities (from 6 to
40 Ah) and consequently different geometric sizes and weights. It must be pointed out that the main chemistry
present in literature is the
LiF eP O4
(Lithium Iron Phosphate) due to its low cost, reduced toxicity and chemical
stability (i.e. long operating life).
17
Table 9: Pouch cell datasets for Efficiency and HR. Specifics on experimental test
Reference DR Ambient temperature [C] Cooling Fluid TMS
[8] 1,2,3,4 5-35 Water Dual cold plate
[9] 0.25,0.5,1,2,3 -10-40 50-50 Water-Ethylene Glycol Thermal Bath Calorimeter
[15] 0.33,1,2 23-24 Air Thermal Chamber
[46] 1,3,5 24-25 Air -
[47] 1,2,3,5 25 Air Thermal Chamber
[48] 1,2 22 Air Natural Convection
[49] 10 25 Air Natural Convection
It is first interesting to evaluate whether the efficiency of a
LiMnNiCoO2
cell is statistically higher than a
LiF eP O4
cell. Therefore, a t-test on the average of both dataset has been done by means of Minitab software. The results are
reported in Table 10. Overall, there is enough statistical evidence to claim that the average efficiency of
LiF eP O4
is lower than
LiMnNiCoO2
one. In fact, the difference between these efficiencies is in the range of 3-3.4% (90%
confidence) and higher than 3% (95% confidence). However, having just one paper reporting thermal performance of
LiMnNiCoO2
for DR 1-2, this statement must be reconsidered once more tests on
LiMnNiCoO2
are available for
low and extreme DR.
Table 10: Pouch Efficiency: t-Test on average values for different chemistries
Statistics LiF eP O4LiMnNiCoO2
Sample Size 21099 320
Average η0.91996 0.95217
90% CI (0.9191, 0.9208) (0.95013, 0.95421)
Standard deviation 0.071546 0.022126
Figure Appendix A.1 gives an overview on the efficiency
η
,HR, volumetric heat generation
Qvol
[
W/L
] and specific
power [
W/kg
]. It can be claimed that
LiF eP O4
cells have
η
higher than 75%,
Qvol
up to 150
W/L
, specific power
reaching maximum values of 900
W/kg
and HR lower than 30%. In addition, the heat flux
q
[
W/m2
] is found to
increase with the DR, as reported in Figure Appendix A.5, with its median growing quadratically from 2.5
W/m2
at
0.25C to 642 W/m2at 5C.
Moreover,
η
can be plotted in function of DOD (Figure Appendix A.2), DR, ambient and battery temperatures,
nominal capacity (Figure Appendix A.3) and characteristic length
Lc
(i.e.
V olume/Area
,Figure Appendix A.4).
The following observations can be highlight from these figures (Figure Appendix A.2,Figure Appendix A.3,
Figure Appendix A.4):
η
decreases non-linearly with DOD from a median of 97% for DOD=0-0.1 to a median of 92% for DOD=0.9-1;
also, the dispersion of data is more prominent for high DOD, where the effect of heat generation due to internal
18
resistance are quantitatively important.
η
increases with the ambient temperature in range 15-45
C. For temperatures in range -5-5
C, few data were
found in literature [
8
,
9
] and therefore are not deemed to be completely reliable. This suggests the necessity
to collect more experimental data for low ambient temperatures (down to -20
C), to better evaluate the
efficiencies at this extreme working conditions.
η
generally increases with the cell temperature from 15 to 35
C. A sharp decrease is found for temperature
higher than 45
C. Similarly, few data are reported for temperature lower than 15
C, therefore, the apparent
beneficial cold cell temperature must not be taken as trustworthy. This suggests the necessity to collect more
experimental data for low cell temperatures (down to -20
C), to better evaluate the predictive poor efficiencies
at this extreme working conditions. In addition, it is not completely clear what is the best cell temperature in
terms of instant efficiency optimisation.
There is a strong correlation between
η
and DR, with a non-linear decrease between 0.25C and 4C. Again, due
to few data reported in literature for DR higher than 5, the unexpected increase of efficiency for DR higher
than 4 must be carefully considered. This suggests the necessity to collect more experimental data for high
DR, to better evaluate the poor efficiencies at this extreme working conditions.
η
increases with the nominal capacity from 10 Ah to 40 Ah. This could be taken into consideration for the
development of an
η
predictive model. Also, this suggests the necessity to collect more experimental data for
several nominal capacities, from hundreds of mAh up to 40 Ah.
The characteristic length
Lc
doesn’t have a strong correlation with
η
and therefore can be neglected for
efficiency prediction.
When evaluating the overall performance of each discharge process (based on total energies discharged), Li-ion
pouch cells provide integral average electrical powers and heat generation rates up to respectively 250 W and
35 W (Figure Appendix A.6). The median overall efficiency
ηoverall
and
HRoverall
are 93% and 7.5%, in range
of respectively 81-99 % and 1-24%. Moreover, as reported in Table Appendix A.1,Table Appendix A.2 and
Figure Appendix A.7, the overall efficiency typically decreases with an increase in DR (from 96% at 0.25C to 86%
at 4C) and a decrease of the nominal capacity (from 89% at 10 Ah to 97% at 40 Ah). The effect of the ambient
temperature is not completely clear. However, the overall efficiency seems to be strictly dependent on the cell
temperature, with the median value reaching its maximum at 25
C while the interquartile range improves for higher
temperatures in range 35-45
C. Therefore, it is not completely clear what is the best cell temperature in terms of
overall energy optimisation and if this differs from the instant efficiency optimisation one.
3.3.2. Cylindrical Cells
Table 11 reports the information regarding the dataset analysed for cylindrical geometry [
40
,
41
,
42
,
43
] while
Table 12 gives details on tests condition. The dataset is characterised by different nominal capacities (from 0.675 to
4.4 Ah) and consequently different geometric sizes and weights. Contrarily to pouch cells, cylindrical cells tested in
previous literature are characterised by many chemistries, from the over-known
LiF eP O4
to Manganese (
LiMnO
,
LiyMn2O4) and Nickel-Cobalt-Aluminium (LiNiC oAlO2).
19
Table 11: Cylindrical cell datasets for Efficiency and HR.
Reference Chemistry Mass [gr] Volume [ml] ρ[kg/m3]Asurf ace [cm2]Lc* [mm] Capacity [Ah]
[40]LiMnO 92 44.8 2056 86 5.19 4.4
[41]LiyMn2O448.5 16.5 2932 37 4.5 0.675
[42]LiNiCoAlO248.5 17.6 2763 38 4.63 3.25
[43]LiF eP O440.5 17 2388 37 4.55 1.25
*Lc=Characteristic Length=V olume/Area
Table 12: Cylindrical cell datasets for Efficiency and HR. Specifics on experimental test
Reference DR Ambient temperature [C] Cooling Fluid TMS
[40] 1 -10-40 Air Thermal Chamber
[41] 1, 3 24 Air CFD
[42] 0.2, 1, 1.5 25 Air CFD
[43] 0.2, 0.4, 0.8, 1.6, 2.4, 4 15-55 Air Wind Tunnel
Figure Appendix A.8 gives an overview on
η
,HR,
Qvol
and specific power for the four different chemistries. It can
be claimed with 99% of confidence (box-plots whiskers) that
LiF eP O4
and
LiNiCoAlO2
cells have the highest
performances, with respectively
η
higher than 88% and 83%,
Qvol
up to 50 W/L and 120 W/L, specific power
reaching maximum values of 450 W/kg and 380 W/kg and HR lower than 13% and 18%. In addition,
q
is found to
increase with the DR for both chemistries, as reported in Figure Appendix A.12 and Figure Appendix A.13, with its
median growing quadratically from 2.5
W/m2
at 0.25C to 460
W/m2
at 4C for
LiF eP O4
and from 80
W/m2
at
0.5C to 540 W/m2at 1.5C for LiN iCoAlO2.
Moreover,
η
can be plotted in function of DOD (Figure Appendix A.9), DR, ambient and battery temperatures,
nominal capacity (Figure Appendix A.10) and characteristic length
Lc
(i.e.
V olume/Area
,Figure Appendix A.11).
The following observations can be highlight from these figures (Figure Appendix A.9,Figure Appendix A.10,
Figure Appendix A.11):
In opposition to pouch cells,
η
seems to oscillate around a median of 93% for DOD 0-0.7 and then sharply
decrease down to 87% for DOD=1; also, the dispersion of data is more prominent for DOD higher than 0.5,
where the effect of heat generation due to internal resistance are quantitatively important.
Similarly to pouch cells,
η
increases with the ambient temperature in range 25-55
C. For temperatures in
range -5-15
C, few data were found in literature [
40
] and therefore are not deemed to be completely reliable.
This suggests the necessity to collect more experimental data for low ambient temperatures (down to -10
C),
to better evaluate the efficiencies at this extreme working conditions.
η
has a different behaviour with the cell temperature compared to pouch cell. In fact, median efficiencies are
quite stable in range 15-55
C while the interquartile range worsen with higher temperatures. This might
suggest a best cell temperature range of 15-25
C. Similarly, few data are reported for temperature lower than
20
15
C. This suggests the necessity to collect more experimental data for low cell temperatures (down to -10
C), to better evaluate the efficiencies at this extreme working conditions.
Even if more scattered than pouch cells, there is still a strong correlation between
η
and DR, with a non-linear
decrease between 0.2C and 4C. Again, due to few data reported in literature for DR higher than 1, more
experimental data must be collected for high DR, to better evaluate the poor efficiencies at this extreme
working conditions.
Compared to pouch cells, no evident
η
dependence to nominal capacity (in range 0.675-4.4 Ah) is evaluated.
This could suggest that this factor is less valuable for the development of an
η
predictive model for cylindrical
cells. However, more experimental data must be collected for several nominal capacities to evaluate this point.
Similarly to pouch cells, the characteristic length
Lc
doesn’t have a strong correlation with
η
and therefore
shouldn’t be considered as a potential independent variable for efficiency prediction.
When evaluating the overall performance of each discharge process (based on total energies discharged), Li-ion
cylindrical cells provide integral average electrical powers and heat generation rates up to respectively 11 W and 1.2
W (Figure Appendix A.14). The median
ηoverall
and HR are 93% and 7.5%, in range of respectively 85-98 % and
1-17%. Moreover, as reported in Table Appendix A.3 and Figure Appendix A.15,
ηoverall
typically decreases with an
increase in DR (from 99% at 0.2C to 81% at 3C) and a decrease of the nominal capacity (from 90% at 675 mAh
to 96% at 4.4 Ah). Likewise pouch cells, the effect of the ambient temperature is not completely clear. However,
ηoverall
seems to be dependent on the cell temperature, with the median value reaching its maximum at 25
C while
the interquartile range improves for higher temperatures in range 35-65
C. Therefore, likewise pouch geometries, it
is not completely clear what is the best cell temperature in terms of overall energy optimisation and if this differs
from the instant efficiency optimisation one.
3.4. Ageing Effect
The ageing effect is defined as the phenomena for which the capacity and discharged energy of a Li-ion battery
decreases with the number of cycles [
50
]. This is typically measured by the State Of Health (SOH), which is defined
as the ratio of the actual capacity compared to its nominal value (i.e. starting un-aged cell) [
50
]. Moreover, SOH
equal to 80% is considered to be the so-called End-Of-Life (EOL) of a battery [
50
]. However, it must be mentioned
that typical warranty contracts consider SOH equal to 70% as preferred value [51].
The ageing rate
ra
is the rate of change of the SOH. Previous literature shows that
ra
is mainly dependent on the
Li-ion cell operating temperature. However, studies have shown that in case of pouch geometries a higher external
compressive load on the cell (i.e. pressure) can decrease the ageing rate up to 24%. Typically, one measures rafor
different operating temperatures and then plots its logarithm as a function of 1
/T
. Therefore, by means of linear
regression techniques, one can fit an Arrhenius coefficient A (pre-exponential factor) and
Eact
(activation energy)
[50], as reported in Equation 3:
ra=A·exp Eact
kbT(3a)
ln ra= ln AEact
kbT(3b)
21
where kbis the Boltzmann constant, equal to 1.38 ·1023 J
Kor 8.618 ·105eV
K(1 eV = 1.602 ·1019J).
The main ageing effects presented and evaluated in previous literature are [50,52]:
Mn dissolution from the cathode
Mn deposition on the anode
Mn re-deposition on the cathode
Particle cracks
Loss of cyclable Lithium (also called Lithium inventory)
Li plating
Solid-Electrolyte Interface (SEI) growth and decomposition.
Corrosion of current collectors
Waldman et al. [
50
] analysed the ageing effect on a cylindrical 18650-type Li-ion battery, focussing on operating
temperatures in the range of -20-70
C, CR 0.5C and DR 1C. Therefore, each charge/discharge cycle took 3 hours in
total. The authors claim that this procedure has been done before in literature but just for limited temperature
ranges. The initial capacity of all batteries was 1.5 Ah. All cells were charged and discharged in potential range
2-4.2 V. The criteria of EOL was 80% of SOH. Ageing rates
ra
are considered as the gradient (tangential slope) of
the SOH plotted against time (expressed in number of charge-discharge cycle). From the results of SOH plotted with
number of cycles for different operating temperatures (Figure 9), it can be seen that the relation is quite linear in
range of SOH (90-100%) except for 25
C. For increased operating temperatures from 25 to 70
C, the ageing rate is
higher and therefore it takes less cycles to reach 80% SOH (EOL). Similarly, this effect is noted for temperatures
lower than 25
C. In this case the effect is even worse. In fact, we have an equal degradation history for temperature
of 0
C (-25
C from optimum) and 70
C (+45
C from optimum). This can be seen also in the Arrhenius plot in
Figure 10. The slopes in this plot are different for the two temperature ranges. Therefore, from theory, one can
claim that the degradation mechanism is different and the one for the low temperatures seems stronger. In fact,
the activation energy
Ea
for the cold conditions is 0.43 eV compared to 0.18 eV of the hot conditions. The authors
claim that the high temperature depleting effect is due to Li plating phenomena while the worst condition at low
temperatures is due to Mn dissolution and SEI thickness increase. Overall, the optimal operating temperature (i.e.
minimum ageing rate) is deemed to be 25 C.
Similarly, Rao et al. [
6
] report ageing effect on Li-ion batteries for a different number of cycles and operating
temperatures, as reported in Table 13. It can be inferred that with the same operating temperature, 200 more cycles
lead to an absolute 10% more capacity loss. Moreover, similar levels of SOH are obtained with 100 cycles more but
at 5 C lower temperature.
Finally, Shabani et al. [
5
] claim that Li-ion batteries have a safe temperature range of -10-50
C and the ideal
operating range is 20-30
C. Also, the battery operating life decreases by roughly 2 months every 1
C of temperature
increase when considering a range 30-40
C. Moreover, the authors suggest that in order to maximise the batteries
pack life time, each cell should be operated at a uniform and equal temperature, guaranteeing pack temperature
22
0 50 100 150 200 250 300 350 400 450 500 550 600
70
75
80
85
90
95
100
+25
-20
-10
+00
+50
+60
+70
T [°C]
Figure 9: Ageing effect at different operating temperatures (elaborated from [50])
Figure 10: Arrhenius plot for the ageing effect at CR 0.5C and DR 1 C. The solid lines correspond to linear fits [50]
Table 13: Li-ion batteries ageing effect for different number of cycles and operating temperatures [6]
Number of cycles 800 600 500
Operating Temperature [C] 50 50 55
SOH [%] 60 70 70
gradients lower than 5 C.
Overall, there is evidence (Table 14) that the best temperature to minimise the ageing effect is around 25-30
oC
.
However, previous literature focuses on low CR/DR (or don’t specify their value) and cylindrical geometry. Therefore,
further research is needed on this matter for prismatic/pouch cells tested at broad temperature ranges (including
extreme conditions) and high CR/DR.
23
Table 14: Ageing Effect
Reference CR DR Temperature Range [C] Ageing Rate ra[cycle1] Optimal Temperature [C]
[50] 0.5 1 -20, 70 0.04 - 0.12 25
[6] - - 50,55 0.05 - 0.06 -
[5] - - -10, 50 - 20 - 30
3.5. Batteries Thermal Models
There are numerous methods for modelling the heat generation and thermal management systems in a Lithium-
ion battery system, ranging from simplified lumped parameter models [
53
,
54
,
5
,
7
] to computationally intense
3-dimensional numerical models [
55
,
56
]. In terms of equivalent thermal circuits, previous studies proposed 3 main
alternatives [
7
]: internal resistances, single RC, double RC circuit. The internal resistances (
Ri
) model uses different
values for charging and discharging processes and is based on evaluating an ohmic resistance and a polarization
resistance to compute the open circuit voltage of the cell. The single RC model introduces a resistive-capacitance
component to the simple Ri model. In this case, the total equivalent resistance of the battery is determined by a
combination of charge transfer resistance, ohmic resistance and diffusion capacitance. The double RC circuit is
just a development of the single RC, taking into account two types of polarisation mechanism, concentration and
activation.
3.5.1. Lumped Capacitance
The lumped parameter approach relies on the assumption that the temperature of the battery is spatially uniform and
varies only with time. This approximation can be used to model Lithium ion cells, but only if internal temperature
gradients are small, i.e. the Biot number
2
is lower than 0.1 and the maximum relative error is equal to 5%.
One study by S. Al Hallaj et al. [
54
] used such a method to create a 1-dimensional thermal model with lumped
parameters to simulate the temperature rise of a small 1.35 Ah 18650 cylindrical Li-ion cell. The model developed
treats the cell as a thermally homogeneous body with effective thermo-physical properties which are assumed to be
independent of temperature over the cell’s operating temperature range. Heat generation within the cell is assumed
to be generated uniformly, subsequently assuming that the current distribution within the cell is also uniform. The
heat generated within the cell was calculated based on the reversible heat due to electrochemical reactions and
irreversible losses due to joule heating. Heat from side reactions and heat of mixing was ignored, which is a common
assumption in Li-ion cell modelling, as in most cases they are small enough to be neglected. The heat generated was
calculated using Equation 4.
˙
Qt=˙
Qj+˙
Qr(4a)
˙
Qj=I(Voc V) = RiI2(4b)
˙
Qr=I T ∂Voc
∂T =I T S
n F (4c)
2The Biot number Bi is defined as the ratio of the heat transfer resistances inside of and at the surface of a body: Bi =hcLc
k
24
where:
˙
Qtis the total heat generation [W]
˙
Qjis the irreversible Joule/ohmic heat generation [W]
˙
Qr
is the reversible heat generation due to chemical reactions [W] (positive for exothermic and negative for
endothermic reactions respectively for discharging and charging phases)
I is the current [A] (positive for exothermic and negative for endothermic reactions respectively for discharging
and charging phases), which is linearly dependent on DR
Voc is the open circuit voltage [V]
Sis the reaction entropy variation [J/K]
V is the cell voltage [V]
n is the number of electrons
F is the Faraday constant equal to 96485 C
mol .
The term
˙
Qj
=
I(Voc V)
, referred to as the over-potential, is the irreversible heat generation due to Joule heating
and the term
˙
Qr
=
I T Voc
∂T
is the reversible heat due to entropy change.
∂Voc
∂T
=
S
n F
, often referred to as
the entropy coefficient, is the measure of the cell’s open circuit voltage dependency on temperature. Both terms are
dependant of DOD and temperature. The current
I
is positive during discharge and negative during charge. Both
the over-potential and entropy coefficient were measured experimentally at 35
oC
and DR of C/1, C/2, C/3 and C/6.
The calculated heat generation profiles for the cells were then used as input to a general energy balance equation
with proper boundary conditions to calculate the temperature rise of the cell, as reported in Equation 5.
2T
∂r2+1
r
∂T
∂r +qv ol
kcell
=1
α
∂T
∂t (5a)
∂T
∂r r=0
= 0 (5b)
kcell ∂T
∂r r=R
=hc(TTa) (5c)
(T)t=t0,r =Ta(5d)
Figure 11 shows the results of the simulated cell temperature rise plotted together with the experimental measurements.
Using this model, the simulated results for DRs C/2, C/3 and C/6 are in good agreement with the measured results,
with relative RMSE lower than 1%. However, at a DR of C/1, the simulated results deviate from the measured ones,
leading to relative RMSE and
R2
of 3.55% and 87%. S. Al Hallaj et al. [
54
] stated that the discrepancy between the
simulation and measurement at C/1 is likely due to the assumption that heat is generated uniformly in the cell. At
high DRs, temperate gradients are likely to increase within the cell as the temperature at the core builds up. This
subsequently causes heat to be generated at different rates throughout the cell which could explain the error in both
heat generation and temperature rise estimations.
Similarly, Shabani et al. [
5
] and Panchal et al. [
8
] proposed a simple model considering just ohmic losses and
reversible entropic heat generation, as reported in Equation 4. Moreover, Panchal et al. [
8
] claimed that a lumped
25
Figure 11: Simulation results against temperature measurements for Sony 18650 cell at all DR,Ta=35oC,hc= 10 W/m2K[54].
capacitance model seems to be accurate enough for most of the cases. In fact, whenever the Biot number is lower than
0.1 (with errors lower than 5%), it seems more convenient to use this kind of model to minimise the computational
time. In addition, the authors observed that the heat dissipation is mainly due to external surface convection,
considering that the Biot number is low and so the temperature uniformity of the battery leads to an absence
of internal conduction. Typically, the internal resistance
Ri
is a function of both SOC (or DOD) and operating
temperature (Figure 12) while the entropy coefficient ∆Sis only a function of the SOC.
Panchal et al. [
7
] developed a Simulink model based on experimental data to evaluate the thermal behaviour
of a Li-ion ion battery. The subsystems used were voltage, internal resistance, heat generation and temperature
calculation. The heat generation was divided in two main arguments: the reversible entropy variation due to chemical
reactions and the irreversible ohmic/Joule heating losses, following a lumped capacitance modelling technique.
Interestingly, the authors point out that the heat generated by reversible entropy change is lower than the one
generated by ohmic losses for high DR (e.g. EV batteries) [
7
]. From the results of the simulations (Figure 12), the
internal resistance seems to remain constant for the most part of the discharge period (until 85% of DOD) and then
increases at high DOD [
7
]. The value of the internal resistance and its sharp increase at the end is higher for lower
DR (1C) [7].
The same authors [
8
] proposed an Artificial Neural Network (ANN)) method based on three input parameters:
cooling plate water inlet temperature, discharging current (i.e. DR) and battery capacity. The only output is the
heat generation and the model is validated by experimental data, leading to good agreements.
3.5.2. Electro-chemical models
Other researchers have achieved more accurate results at higher DRs using electrochemical-thermal modelling
methods. Hosseinzadeh et al [
57
] developed an electrochemical-thermal model of a large format lithium-ion pouch
cell and validated their model over an ambient temperature range of 5-45
oC
and DR in the range of 0.5-5C, including
26
0 20 40 60 80 100
0
0.01
0.02
0.03
0.04
0.05
0.06 1C
2C
3C
4C
20 30 40 50 60
0
0.01
0.02
0.03
0.04
0.05
0.06 1C
2C
3C
4C
Figure 12: Li-ion batteries Internal Resistance Riat different DR [7]
various drive cycles for EVs . They used a combination of a pseudo 2D electrochemical-thermal model to model a
single electrode pair within the pouch cell and coupled this to a 3D thermal model to simulate the temperature
distribution within the cell. Using an electrochemical-thermal model also allowed them to predict the cells available
capacity and power under different temperature and DRs. Their model was able to predict the cell voltage and
temperature rise with peak errors of 6.4% and 6% respectively.
In conclusion, choosing a numerical model is always a trade-off between accuracy and complexity (i.e. computation
time) [
5
]. Most of the time, a lumped capacitance method results more than appropriate. However, there could be
problems in applying this simple method in case of HTF with low heat transfer coefficient (high Biot number), high
DR (more than 1C) and ”abnormal” operating conditions (high ambient temperature).
3.6. Summary
Overall, there is enough evidence to claim that at time of writing 25-30
oC
seems the best temperature range to
minimise the ageing effect. However, no specific temperature is indicated in literature which can optimise the cycle
electro-chemical efficiency and minimise the ageing effect. Moreover, from the author’s literature data elaboration,
it seems that optimising a Li-ion cell for a specific DOD level can lead to non-optimal condition for the overall
cycle. Moreover, even if different geometries can use same chemistries, different efficiency behaviours are observed
for pouch and cylindrical cells, with the former having a smooth decrease of
η
with DOD and the latter a more
oscillating behaviour. In conclusion, more experimental data are necessary for ”extreme” operating conditions, i.e.
high CR-DR and high-low ambient temperatures to acquire a more robust data set to evaluate the best operating
temperature which must be guaranteed by the TMS.
27
4. Thermal Management Systems (TMS)
4.1. Overview
There are 3 main techniques to cool batteries as proposed in previous literature [6,8,58,59]:
Air-cooling
Liquid-cooling (e.g. water, glycol, oil, acetone, refrigerants)
Boiling
PCM systems.
Air cooling is preferred in terms of electrical safety but it is reported to be less efficient due to low heat transfer
coefficients [
8
]. In addition, forced air cooling is effective to keep the temperature at a preferred level but fails
in guaranteeing a constant and uniform temperature within the cells (in-cell) and the cells pack (cell-to-cell) [
58
].
Water cooling is better in terms of heat transfer quality but it is intuitively considered to be less safe [
8
]. Boiling
guarantees a better temperature uniformity [
60
] but it is intrinsically complex to operate and control. PCM are
capable of reducing both maximum temperature and cell-to-cell temperature gradients, leading to a reduction in
capacity loss rate up to 50% [
58
]. However, when Li-ion batteries are operated under extreme or ”abuse” conditions
(high ambient temperature, high DR), PCM, considered more a buffer than a heats sink system, are not able to
recover all latent energy potential during solidification and this leads to possible thermal runaway (Figure 13) after
a certain number of charge/discharge cycles of the batteries [58].
Figure 13: Thermal runaway process with respect to cell temperature [19]
Table 15 and Table 16 report an overview of all TMS already mentioned, together with thermo-electric and cold-plate
techniques [
19
]. It seems clear that PCM is a really promising method and deserve further R&D [
19
]. In fact, it has
several good properties (Table 15 in bold) compared to other TMS, e.g. long operating life, ease of use, integration
and maintenance, high efficiency, low capital and O&M costs, large temperature decrease potential and temperature
uniformity capability [19].
Table 15: Thermal Management Systems TMS for Li-ion Cells [19]
Air forced Liquid Heat pipes HPS PCM Thermoelectric Cold plate
Life [years] >20 3-5 >20 >20 1-3 >20
Ease of use Easy Difficult Moderate Easy Moderate Moderate
Integration Easy Difficult Moderate Easy Moderate Moderate
Maintenance Easy Difficult Moderate Easy Difficult Moderate
Temperature distribution Uneven Even Moderate Even Moderate Moderate
Efficiency Low High High High Low Medium
Temperature drop in cell Small Large Large Large Medium Medium
Annual cost Low High Moderate Low High Moderate
Table 16: Thermal Management Systems TMS for Li-ion cells: advantages and disadvantages [59]
Advantages Disadvantages
Air cooling Low cost Low h(i.e. low ηand high volumes)
Simple design High T gradients
Highly commercialised Noise issues
Liquid cooling Good h(i.e. high ηand low volumes) Safety issues (leakages)
Medium T gradients Complex design
Highly commercialised High costs, weights and maintenance
HPS High h(i.e. high ηand low volumes) High costs
High operating life Non commercialised
PCM High thermal energy density Low thermal conductivity (low powers)
T uniformity Potential liquid leakage due to volume expansion
Legend: h=heat transfer coefficients [ W
m2K], T=temperature [K], η=thermal efficiency
4.2. Air-cooling
Air-cooling is the most wide-spread TMS technique and is characterised by low costs, simple design but also low
thermal performances due to air thermo-physical properties at operating conditions and high cell temperature
dis-uniformity [12]. Therefore, there are several ways to improve it proposed in literature [61,12,62,59]:
Increase air flow rates: higher Reynolds number, higher heat transfer coefficients, but uneven temperature
distribution, higher parasitic consumption. Chen et al. [
63
] report that air cooling has from two to three times
the energy consumption compared to liquid cooling.
Battery layout optimisation: use of wide spaces between cells and staggered configuration can improve the air
flow turbulence and therefore Nusselt number values, but this leads to lower energy/power EESS densities
Air flow path: possible better temperature uniformity with so called reciprocating air flow (i.e. alternate
air-flow inlet-outlet), leading to 72% drop in temperature gradients, or Z-type configuration, but this requests
higher complexity and control systems.
TCE: Integration of metal/high conductive foam/matrices/honeycomb/fins/pins structures to improve the
29
equivalent heat conductivity of air-flow [59].
Giuliano et al. [
16
] claim that air cooling can be as effective as liquid cooling. This results can be obtained by adding
aluminium foams within the air flow duct to improve the heat transfer coefficient, leading to battery temperature
increase lower than 10
C even at DR of 4C. Chen et al. [
64
] studied finned and/or pinned direct/indirect systems
to improve the air flow heat transfer coefficients and decrease the average cell temperature. They conclude that
satisfactory cooling rates can be achieved by increasing the air flow rate and inevitably the parasitic consumptions.
Shaid et al. [
65
] showed that a better Li-ion cells spatial configuration can reduce the cell-to-cell and in-cell
temperature gradients of 21.5% and 16% and the in-cell and maximum temperature of 5%. However, this is done
with complex system design where multiple vortex generators and jet inlets are added to a normal air channel.
Similarly, He et al. [
13
] showed that air fan consumption can be reduced by up to 84% by sound control strategies.
Yu et al. [
66
] confirmed that air cooling can be improved by imposing different air-flow patterns, like two directional
flows, e.g. longitudinal plus transversal jet-cooling systems. This permits to decrease the battery temperature of
around 9 C and reach better iso-thermalisation.
4.2.1. Micro-channels
There is evidence in literature of the research interest [
67
,
68
] in micro-channel systems for air-cooling of Li-ion
cells. The basic principle is that decreasing the size of the air-flow channels (or increasing their number per volume)
increases the heat transfer area per unit of volume, i.e. improving the heat transfer rates [
67
]. Moreover, if the
mass flow rate can be increased, this leads to higher velocities, higher Reynolds and Nusselt numbers and better
thermal performance [
67
]. Regarding this, Huo et al. [
68
] demonstrated that this system can decrease of up to 10
C
the maximum temperature of Li-ion cells (Figure 14). However, this method is not effective to improve the in-cell
temperature uniformity.
Figure 14: Air-cooling by micro-channels: maximum temperature and in-cell temperature gradients [68]
30
4.2.2. Heat Pipes
Compared to micro-channel systems, HPS can improve the temperature uniformity of the Li-ion cells if properly
positioned, thanks to their high equivalent thermal conductivity (10
5W/m K
) [
61
]. Shah et al. [
69
] investigated
different combinations of HPS and air cooling in annular channel to avoid cell core thermal run-away, showing a
potential decrease of the core temperature of up to 20
C with a constant cell heat generation of 1.62 W. It must be
mentioned that this study is based on the experimental simulation of a Li-ion cell by using an electrical cartridge in
the shape of a 26650 cell. Similarly, Yuan et al. [
70
] showed that for a 10Ah pouch cell discharged at DR 2.5C HPS
can decrease of around 6 C the cell temperature and improve the temperature uniformity.
Figure 15: Air-cooling by HPS: configuration and temperature profile [70]
4.3. Liquid-cooling
Several studies [
61
,
63
,
71
,
72
,
73
,
17
,
74
,
11
,
75
,
72
,
10
] report that air cooling fails in case of high DR and large
battery packs, leading to a necessary implementation of liquid cooling. Tong et al. [
71
] claim that liquid cooling can
be easily implemented with battery pack of different sizes and, by changing flow rate and channel width, it can
handle extreme operating conditions of high DR up to 5C. There are two main categories of liquid cooling [
61
]:
passive and active. However, as passive liquid cooling systems don’t guarantee considerable higher performance
compared to air force convention, only active liquid cooling is used due to the beneficial effect of higher heat transfer
coefficients (350-400
W/m2K
compared to 5-25
W/m2K
of air cooling [
63
]). However, this adds complexity and
costs (both capital and operational) to the TMS. In fact, liquid cooling TMS comprehends pump, heat exchangers
and coolant/refrigerant pipes [
63
]. Moreover, this leads to extra mass and increased weights (Table 17), from 3 to 7
% depending on the fluid used, which could be a clear drawback for motive applications [63].
Zhao et al. [
61
] argue that the optimisation of these systems breaks down into refrigerant selection and fluid
flow configuration. For the latter, due to safety reason and compactness, the most used design is the cold plate,
where fluid and Li-ion cells are separated by a thin highly conductive metal plate on which a series of coolant
pipes (serpentine or straight bundles) are mechanically sold. Moreover, in accordance with air forced convection
improvement techniques, multi/micro channels are typically applied to achieve higher specific exchange areas and
Reynolds numbers. However, this comes with higher pressure drops and parasitic consumptions for the pump [
72
].
On this regards, Du et al. [
76
] report that this can be minimised once an hysteresis control technique is implemented
31
Table 17: Extra weight added by cooling system for each 35Ah pouch cell of mass 1.01 kg. [63]
Coolant Extra Mass [kg] Extra Mass [%]
Air 00
Mineral Oil 0.0298 2.95
Water/glycol 0.0723 7.16
Fin 0.394 39
in the cooling system, leading to cell operation in safe temperature range conditions and reduced energy consumption.
Moreover, Cheng et al. [
63
] demonstrate that water-glycol cooling has a pumping consumption more than two order
of magnitude lower than air cooling for the same average temperature rise (Figure 16). Typical fluids are water,
water glycol or silicon/dielectric mineral oils (Table 17).
Figure 16: Average cell temperature and ideal fan/pump power consumption for air, fin, direct liquid (i.e. oil) and indirect (i.e. jacket)
cooling at DR 2.71C [63]
Figure 17: Cell temperature rise and in-cell temperature gradient for air, fin, direct liquid (i.e. oil) and indirect (i.e. jacket) cooling at
DR 2.71C [63]
32
However, several studies [
73
,
61
,
77
,
17
,
74
,
63
] report that, although liquid cooling is better than air cooling to
reach lower average cell temperature, it is not effective for maintaining a low in-cell thermal gradient which leads
to cell degradation. Panchal et al. [
73
] show that using cold plate with high exchange area as pouch cells TMS
decrease the cell maximum temperature but worsen the temperature uniformity. Chen et al. [
63
] show that liquid
cooling (jacket) is not capable of keeping the in-cell temperature gradient lower than the deemed limit of 4
C unless
mass flow rates higher than 1 gr/s are imposed to the system (Figure 17).
However, there are several solutions proposed in literature. Zhao et al. [
61
,
17
] suggest to optimise the fluid
flow direction, pattern and increase the mass flow rate. Similarly, Zhang et al. [
77
] demonstrate that the in-cell
temperature difference can be kept lower than 5
C with water cooling when sound designs are used, such as the
application of flat flexible graphite sheets in between cell and cooling tubes. Finally, Yang et al. [
74
] propose an
innovative liquid metal TMS. They claim that liquid cooling is still limited by the low conductivity of coolant such
as water or aqueous ethanol. So, they propose liquid metal such as Gallium (melting temperature 29.8
C) and
his alloys with Indium and Tin
Ga80I n20, Ga68In20S n12
(melting temperatures of 16
C and -10
C), leading to
lower maximum cell temperatures and better temperature uniformity, together with lower pump consumption (due
to usage of electromagnetic pump). However, no efforts were done in estimating the costs, weight, maintenance
(potentially lower than water due to no moving part in the pumping system), corrosion and leakage problems. Also,
liquid metals are incompatible with aluminium [
74
] so this can’t be used for the jacket, restricting the choice to
nickel an copper, far more expensive.
Figure 18: Improvement of temperature uniformity in liquid cooling by adding graphite sheets in between Li-ion pouch cells and cooling
tubes [77]
4.4. Boiling
Li-ion cell TMS by liquid boiling (flow-boiling, pool boiling) has been proposed in previous literature [
60
,
78
,
79
,
80
].
As shown in Table 18, both pool and flow boiling have been investigated in a relative broad range of CR and DR.
The liquid mainly used is Novec7000, an hydrofluoroether with boiling point at ambient pressure of 34
oC
, latent
heat of vaporisation of 142 kJ/kg, specific heat of 1300 J/kg K and thermal conductivity of 0.075 W/m K. Overall,
all studies report an effective cooling performance and iso-thermalisation.
Van Gils et al. [
60
] investigated the effectiveness of pool boiling sing Novec7000 as TMS for a cylindrical Li-ion cell
charge/discharge of 0.5C/5C. From their experimental tests, they found that when pool boiling is activated (mainly
33
Figure 19: Typical boiling curve, showing qualitatively the dependence of the interface heat flux (q) on the surface superheat (∆
T
),
defined as the difference between the surface temperature and the boiling temperature of the liquid. The maximum heat transfer
coefficiencts are experienced within the region III (fully developed nucleate boiling) [60].
fully-developed nucleate boiling, see Figure 19), high heat transfer coefficients are reached (up to 700
W/m2K
compared to 350
W/m2K
without boiling) and the cell’s average temperature is kept steady at 34.4
oC
and the
temperature gradient is entirely negligible, leading to an ideal iso-thermalised condition even at high DR. Moreover,
they proposed a fast regulation technique based on a boiling chamber pressure control as shown in Figure 20.
Figure 20: Visualisation of the boiling experiments at atmospheric pressure (a), at sub-atmospheric pressure (b) and at super-atmospheric
pressure (c) [60].
Hirano et al. [
78
] proposed a pool boiling TMS for Li-ion pouch cells tested at high CR/DR (10C/20C). Similarly to
Van Gils et al. [
60
], they used Novec7000 as liquid. To evaluate the effect of the liquid boiling temperature, they
compared Novec7000 (34
oC
) with Novec649 (49
oC
).From their results, they inferred that Novec7000 is capable
to keep the Li-ion cell’s temperature around 35
±
2.5
oC
at both DR 10C and 20C. Also, comparing a TMS with
34
Table 18: Overview of literature on liquid boiling as TMS for Li-ion cells.
Reference Boiling Mode Capacity [Ah] n. Cells Geometry CR/DR Liquid Tb(1 atm) [oC]
[60] Pool 1 1 Cylindrical 0.5/5 Novec7000 34
[78] Pool 1 10 Pouch 10/20 Novec7000 34
Novec649 49
[79] Flow - - Pouch - R134a 24 (at 650 kPa)
29 (at 750 kPa)
33 (at 850 kPa)
[80] Flow 20 14 Pouch 2/(1,3,5) Novec7000 34
Tb= boiling temperature [oC]
100%-wetted porous material and 50%-immersed microfibre cloth, they demonstrated that the two TMS have the
same thermal perfomance, leading to 50% liquid savings in the latter TMS.
An et al. [
80
] investigated the thermal performance of a Li-ion cells TMS based on flow boiling in mini-channels
(Figure 21). Similarly to Van Gils et al. [
60
] and Hirano et al. [
78
], they used Novec7000 as liquid. They tested a
Li-ion battery pack (51V, 20Ah) composed by 14 20Ah pouch cells connected in series under different DR (1C, 3C,
5C), ambient temperature (0,15,25,35
oC
) and constant CR (2C). from their results, it could be claimed that the
proposed flow boiling TMS was capable to keep the Li-ion cells temperature around 40
oC
with a dis-uniformity up
to 4
oC
. Also, the
Re
number has a strong influence of the triggering of the flow boiling within the mini-channels
and therefore the TMScooling effectiveness.
Figure 21: Schematic diagram of battery module. (a) Layout of battery module and coolant flow direction; (b) series type of battery
module; (c) thermocouple locations for the temperature measurements of battery module; (d) cutaway view of cold plate [80]
35
4.5. Phase Change Materials
4.5.1. Overview
TESS are typically divided into thermal and thermo-chemical [
81
,
82
], where the former are also classified as
sensible (SH) and latent (LH) heat systems. As shown in Figure 22,LH have been shown to have higher energy
densities (5-14 times) for a fixed temperature difference than SH [
81
,
83
] due to the highly-energetic isothermal
phase transition.
Figure 22: Comparison of Sensible Heat (water, rock) SH and Latent Heat (PCM)LH TESS [81].
Among LH storage, there are three main categories: solid-solid, liquid-gaseous and solid-liquid [
81
,
82
]. The solid-solid
systems have small volumetric variation during transition but are characterised by low spec