Inhibition of Thermal Runaway Propagation in Lithium-Ion Battery Pack by Minichannel Cold Plates and Insulation Layers

,


Introduction
Energy consumption and environmental pollution are major concerns for people today [1].The automobile industry consumes a massive amount of energy and causes air pollution.To alleviate those problems, people begin to gradually use new energy vehicles to replace oil-fueled vehicles [2].LIBs have been used as the first choice for powering the clean energy vehicles because of their advantages in high specific energy, high density, and environmental friendliness [3].The relatively low thermal stability of high-energy-density materials may cause risks of TR from the LIBs.LIBs can generate enormous energy and cause serious accidents once TR occurs [4,5].The term "TR" means the overheating occurrence in which the LIB internal ensuing chain reaction accelerates the rate at which the LIB temperature rises [6].If the temperature of LIB exceeds 70 °C, solid electrolyte interface (SEI) on the negative electrode surface begins to decompose.Consequently, the negative electrode of the battery loses its protection, and the Li-ion inside it comes into contact with the electrolyte, resulting in the formation of a new SEI film [7].When the temperature reaches 90-120 °C, the decomposition of the SEI is accelerated, and the heat generation rate of the Li-ion and electrolyte at the negative electrode increases.The separator melting absorbs heat and the temperature rise slows down.At the end of melting, the voltage continues to drop and the consumption of the positive electrode decreases.This makes the temperature rise further [8].Finally, large-scale internal shorting occurs inside the battery.The flammable electrolyte leaks and temperature reaches ignition point, which potentially causes consequent fire [9].The reaction equation is listed in Table 1.
LIB TR also is accompanied by the production of toxic gases [11].Among a variety of Li-ion chemical compositions, the most common compositions include LiMn 2 O 4 (LMO), LiCoO 2 (LCO), and LiFePO 4 (LFP).LMO [12].For battery packs, single cell of abuse only causes localized TR.The driving force of the TR propagation process is the huge temperature difference between the adjacent batteries [13].Due to the strong lateral heating, as the reaction occurs, LIB 2 reaches the TR trigger temperature, which causes the thermal spread phenomenon by heat transfer [14].The amount of heat transferred through the battery electrodes is minimal; most of it comes from the shell part of the battery [15].Lamb et al. [16] investigated triangular stacked 18650 battery packs.More susceptible TR is found when the batteries are connected in parallel than in series in the pack.At the same time, prismatic batteries are more prone to thermal spread than cylindrical stacks because of the larger contact area between adjacent cells when they are grouped together [17].Researchers build mathematical models and experimental studies to explore ways to inhibit thermal spread.Battery thermal management system (BTMS) is developed to extract the heat away from the battery during its work and to prevent TR.Air cooling is low cost and simple in design but is not effective in suppressing TR and has been phased out [18].Feng et al. [19] used numerical simulations to better understand the system efficiency of the t-symmetric air-cooled BTMS and its coupled system.The results showed that the cooling performance can be further improved by introducing heat transfer fins in the module.
Phase change material (PCM) cooling uses its latent heat to extract heat from the battery [20].Singh et al. [21] investigated a BTMS utilizing encapsulated PCM combined with forced convective air cooling using the coupled electrochemical-thermal modelling.C-rate is the value of the current required to discharge and charge the rated capacity of the LIB within a specified period of time, which is the times of capacity in LIB.The average temperature drop is greatly improved at 5 C discharge rate.Choudhari et al. [22] arranged in a 2 × 3 pattern in which all six cylindrical cells were arranged in parallel.It was found that PCMbased battery module reduces the maximum temperature by 5.12% and 7.17% at 3 C and 4 C discharge rates, respectively, as compared with a conventional battery module.
Liquid cooling can effectively suppress the propagation of TR due to its powerful heat exchange capability within the coolant flow channel [23,24].Fan et al. [25] optimized the performance evaluation, compared with the traditional serpentine channel.The results reported that the optimized design reduced the maximum temperature and standard surface standard deviation of the cold plate by 26% and 35%, respectively.Xu et al. [26] submitted a novel microchannel cooling method for the thermal management of LIBs.The simulation results showed that the microchannel cooling system does not prevent TR within the cell even at flow rates of up to 10.0 L/min.Zhang et al. [27] studied the operating battery temperature that rises rapidly to 80 °C; this can trigger TR.However, at the proper coolant inlet rate, the cell temperature maintained in thermal equilibrium.The use of well-insulated materials is the best choice for inhibiting TR.Yuan et al. [28] discovered graphite composite sheets and aluminum combined with extrusion as interstitial material can effectively reduce the propagation of TR through numerical simulation by Fluent and experiment.Liu et al. [29] used overcharging as a trigger for TR of battery.Their experimental results identified that the cooling efficiency of the fiber-based material was 71.83%, while the aerogel material was 13% more efficient than the fiber material.It is difficult for single BTMS method to completely suppress TR.Therefore, researchers have proposed hybrid BTMS approaches [30].Weng et al. [31] used PCM and aerogel to suppress TR in a battery pack.The results proved that the combination of PCM and aerogel effectively extinguished the flame and reduced the maximum cell temperature.Talele et al. [32] optimized battery pack structural design by introducing low thermal conductive pyro liner at the wall of PCM.And it slows the spread of TR for least 800 s by numerical simulation in Fluent.Rui et al. [33] coupled insulated panels with liquid-cooled panels and succeeded in better mitigating the propagation of TR by an experiment.In summary, researchers conducted a series of experiments to prevent TR, and there are some knowledge gaps in the published literature regarding numerical simulations to demonstrate the synergistic effect of cold plates and insulation layers between adjacent cells in TR.
This study therefore conducts a series of studies regarding the above issues.For commonly used LIBs, internal shorting is a common trigger for the generation of TR [34].In this study, a battery pack consisting of 12 LIB cells is specifically simulated in ANSYS Fluent.
The single battery that generates significant heat serves as the starting point for TR in battery pack.This work is dedicated to exploring three strategies of thermal spread protection, namely, the material selection for insulation layers, minichannel cold plates, and minichannel cold plates plus insulation layers between neighboring cells in the battery pack.This study is aimed at facilitating the design of new energy vehicle battery packs for better safety regarding TR.

Modeling Method
2.1.Modeling Instructions.The internal shorting model in ANSYS Fluent [35] was used to simulate the shorting strength in terms of volumetric short-circuit resistance in Ω•m 3 .It describes the specifics of a short circuit within the battery after the battery separator has been ruptured.For better adaptation to reality as seen in the validation, 0.04 ohm is set as the resistance for the internal shorting of the first cell.Short-circuit area of 10 mm × 10 mm and 100% penetration depth (7.2 mm) are assumed.And the equivalent circuit model (ECM) [36] and four-equation modeling [37] describe the electrochemical and TR behavior of the battery in ANSYS Fluent.

2
International Journal of Energy Research The prismatic LIB modeled in this work is set up corresponding to the ePLB C020 polymer battery, with Li[Ni-CoMn]O 2 as cathode material.The nominal voltage and nominal capacity of this high-energy LIB are 3.65 V and 20 Ah [38].The electrolyte is LiF 6 .High-quality structured mesh is generated for the modeling.Specific parameters of the battery characteristics are listed in Table 2.The thermal conductivity of LIB is anisotropic.And its specific heat capacity also requires rigorous experimental determination.Therefore, the basic physical parameters of LIB are determined in this study with the help of references as listed in Table 3.
LIB is hard to model because it enrolls problems of multiscale.Three electrochemical models are often used as subscale model in the multiscale multidomain (MSMD) modeling scheme in Equation ( 1) [39].The left side of the equation is the heat generation, and the right side of the equation shows the heat conduction, current transportation, electrochemical and abuse source.(1) NTGK is a simple semiempirical model for efficient modelling of battery [40].(2) In ECM, the electrical behavior of the battery is simulated by the circuit [36].(3) In the Newman P2D, it can accurately capture the migration process of Li-ion in the battery [41].This study employs the ECM method, which is more widely used and relatively simple.The ECM method is able to simulate the electrochemical and thermal behavior in LIB, with advantages of easy adjustment of specific parameters and fast numerical speed.

Equivalent Circuit Model and Internal Shorting Model.
In the subscale equivalent circuit model, the battery electrical I-V performance is described by an empirical mathematical expression.The voltage evolution then is solved from the electric circuit equations: In these equations, V OCV is the open circuit voltage, I is the current, and R s , R 1 , C 1 , R 2 , and C 2 are functions of state of charge (SOC) as indicated in the following equations.V is the battery cell voltage.The specific parameters in those equations are the same with the parameters in Ref. [36]: After solving for the ECM, the volumetric transfer current is calculated as follows: where Q total is the battery total electric capacity in ampere hours, Q ref is the battery total electric capacity in ampere hours, and Vol denotes the active zone's volume of a single battery.The mathematical description of the electrochemical reaction heat generated during the discharging process in LIBs is as follows: The different scales of equations are bridged by volumetric transfer current density j ECh .The volumetric current density can be calculated in the subscale electrochemical submodel.The irreversible source term is expressed as j ECh V OCV − φ + − φ − , and the reversible source term is expressed as −j ECh T dU/dT .However, the internal resistance of the battery tends to be zero in the event of an internal shorting.This causes a high rate discharging, which further induces large amount of joule heat.The shorting current density is calculated by Extra heat generated due to the shorting is computed as In the battery scale, the equation governing the current flux is expressed as where σ + and σ − are the effective electric conductivities, r c is the volumetric contact resistance, j short is the internal shorting current, and q short is the internal shorting heat, while φ + and φ − are phase potentials for the positive and negative electrode.

TR Model.
The whole TR process of LIB is consisted of a series of exothermic reactions [42].When the battery is damaged, the LIB temperature rises, and internal chemical processes begin to generate heat.The first thing that happens is the decomposition of the SEI, which usually occurs at around 70 °C.The mathematical equation for the reaction is as follows: As the surface SEI of the negative electrode decomposes, the negative material loses its protective effect.With further increasing in temperature, the negative electrode reacts with the electrolyte: It is found that the product of the negative electrode and the electrolyte is the main component of the SEI, so the regeneration of the SEI occurs: With the reaction between the negative electrolyte and the separator, the gas inside the cell is released and the temperature rises rapidly.In this stage, the positive active material itself can decompose and be exothermic: At the same time, oxygen is released which can react with the electrolyte and further release heat.The cathode materials can react with the electrolyte, and those reactions are highly exothermic.Consequently, the heat released by the reaction during this phase is enormous.The electrolyte is also undergoing an exothermic reaction when the temperature is above 180 °C: Finally, TR heat production is shown below while the physical and kinetic parameters are documented in Table 4.
The equivalent radius of the minichannel is calculated as where ρ c denotes the density of the coolant, μ c denotes the dynamic viscosity, u c denotes the average speed of the coolant in the cold plate, and a, b denote the length and width of the channel.In Section 3.2, the equivalent diameter of the minichannel at the bottom coolant plate is 3 mm.The inlet coolant velocity is chosen as v = 0 01 m/s to calculate Re = 29 07 < 2300.In Section 3.3 and Section 3.4, flow comes from the mainstream inlet, which is 56 mm long and 30 mm wide.Its equivalent diameter is obtained as 39.06 mm.For the 3D thermal modeling, coolant velocity of the maximum 0.01 m/s at the mainstream inlet is considered, and the Re is 378 64 < 2300 As a result, only laminar flow is involved in all the designs of the present study.
This work assumes that water is incompressible and ignores the effect of gravity.The control equation of coolant includes continuity equation, momentum conservation equation, and energy conservation equation [44].
Continuity equation [44] is as follows: Momentum conservation equation [44] is as follows:  [44] is as follows: For the energy equation of the cold plate [45], where ρ wa , C wa , V wa , T wa , and k wa are, respectively, the density, specific heat, velocity, temperature, and thermal conductivity of coolant and ρ c , C c , T c , and k c are, respectively, the density, specific heat, temperature, and thermal conductivity of cold plate.The thermophysical properties of cold plate in the above equations correspond to aluminum material.

Initial and Boundary
Conditions.The initial temperatures of the battery, the cooling channels, and plate are set to be 25 °C.And the water inlet temperature is also set to be 25 °C.For the outlet, a constant zero pressure is specified, and the outlet boundary condition is used for the energy equation.And the LIB exchanges heat with the ambient through convection and radiant heat transfer.The cold plates and insulation layers exchange heat with the ambient through convection heat transfer.They are governed by Equations ( 21) and (22).The convective heat transfer coefficient h is set as 7 W/(m 2 •K) as suggested by Ref. [26].σ = 5 669 × 10 W/(m 2 K 4 ) is the Stefan-Boltzmann constant, and the value is suggested in Ref. [46], while the emissivity heat transfer coefficient ε is set as 0.8 [26].And there is thermal coupling between the LIB and other walls.
2.6.Model Validation.In this study, the flow chart of the simulation works is shown in Figure 1.Firstly, the model validation analysis of the single battery.The single LIB and those in the Ref. [15] are prismatic, which have similar capacities.The research shows the comparison of the modeling results with their experimental results in Ref. [15].The published journals do not have this paper 12 prismatic LIBs in series to verify the case.Due to experimental as well as financial constraints, this model is not subjected to experimental investigation.Therefore, this simulation has been experimentally verified for the single cell.This study provides a guided numerical simulation investigation, and the trend of thermal spread is the same as Ref. [15].It provides a reference for researchers to conduct experiments with 12 cells or more batteries in the future.And by combining the three thermal security projects in this paper, it is possible to explore practical thermal security projects more efficiently.It can improve efficiency and make the experiments more accurate, so that they can be better put into daily use in electric vehicles.
High-quality structured mesh is generated for the modeling.The mesh for simulation is created with ANSYS ICEM, and the different modeling domains are assigned to blocks during meshing process.Due to the high average temperature of the battery cell caused by TR, four different mesh numbers (27616, 32092, 46560, and 56870) are compared for the mesh independence study in Figure 2(a).The error for mesh validation is about 5.2% in numbers 46560 and 56870.The last two mesh errors are minimal.To save computational resources, the work uses the model with mesh number of 46560.The specific values for the TR model are determined by adjusting the parameters to match the experimental results.As shown in Figure 2(b), the predicted average temperature of a single battery cell during the TR is compared with the experimental results in Ref. [15].The average temperature error is about 6.1% in mesh number of 46560, and this simulation parameter can be used as the TR trigger for the battery pack in subsequent modeling.After modeling the single battery cell, a battery pack of 12 prismatic battery cells connected in series is modeled.The TR propagation in the battery pack is explored by simulations.

Results and Discussion
3.1.LIB Pack without Protection Methods.After completing the validation of the single LIB TR, a LIB pack consisted of 12 prismatic LIB cells connected in series without protection methods is subjected to 3D thermal modeling.The geometry and mesh of the LIB pack are shown in Figure 3.Each of the LIB has positive and negative tabs, and these tabs are electrically connected by busbars.The width of the busbar is 2 mm.In order to better analyze the mechanism of thermal spread, as shown in Figure 4, the time of TR trigger [47] (the next second, the average temperature of the batteries shows an exponential increase and the average temperature above 180 °C) is defined tr i (i = battery number 1, ⋯, 12 ).And the propagation time of TR (the difference between the times of TR triggered by neighboring batteries) is defined D i,j = t r j − tr i (i, j = battery number 1, ⋯, 12 , i ≠ j).Define the maximum value of the average temperature of the batteries as T pei (i = battery number 1, ⋯, 12 ).Four different meshes (231217, 324124, 475621, and 590340) are selected for the mesh independence study in Table 5.The TR trigger time is similar for the mesh numbers 475621 and 590340.To conserve computing resources, a 475621 structured mesh is used in 3D thermal modeling of this LIB pack.TR is triggered by internal shorting of cell 1, and then, the TR process propagates to cell 2 and others.Figure 5(a) shows the average cell temperature evolutions for each single cell in the battery pack.TR triggers from the location of the internal shorting and spreads outward in cell 1, its shape is similar to the cylindrical temperature distribution.After about 12 s, it spreads to the whole battery (the numerical simulation ignores the combustion phenomenon) and the T pe1 = 822 86 °C at 31 s.T pe1 is less than T pe2 , ⋯, T pe12 (848.74 °C-982.95°C).This phenomenon shows that TR-onset temperature of cell 1 exhibits the surrounding ambient temperature, while cell 2 to cell 12 subjects to lateral heating that is moderately increased compared to the ambient 6 International Journal of Energy Research temperature.In addition, part of the heat is exchanged to the environment during the internal shorting, which also makes cell 1 lose some heat.According to the Equation ( 23) [15], cell 1 releases amount of heat as 349,621.16J.
where M = 0 428 kg [32] is the mass of the battery and Ttr 1 is the average temperature of cell 1 at tr 1 .This heat is transferred to cell 2 by conduction, resulting in a large temperature difference between the back shell surface of cell 1 and the front shell surface of cell 2 [13].The temperature surpasses the critical threshold for TR in LIBs, which could subsequently trigger a cascading effect leading to TR in next battery.From Fourier's law of thermal conduction (24), the heat flow rate is proportional to the temperature gradient dT/dx for a given heat transfer area A and thermal conductivity k.
Therefore, a linear increase in the average temperature profile is seen for cell 2 as a result of the heat transfer by the temperature rise rate 10.8 °C/s.After about 18 s, the front shell surface of heat spreads back shell surface, and the average temperature rises vertically caused by the process of TR in cell 2 at tr 2 and the D 1 = 6 s.The temperature of cell 3 increases as a result of the elevated temperature on its contact surface with cell 2, which has undergone TR, leading to TR in cell 3 as well.The D 2 = 6 s is the same as D 1 .As shown in Figure 5(b), the battery pack depends on convection and radiation from the environment for heat dissipation between the cells without protection; D i is closer.After about 94 s, heat transfer continues to pass through the shell and the connector, and TR continues propagating to cell 4, then cell 5, until cell 12.And the average time of D i is 7.45 s.The process is similar to the "domino" effect [48] as seen in Figure 6.And the first battery undergoing TR will release heat that accumulates in the adjacent batteries, thereby elevating their risk of TR.As the process of TR propagation continues, the adjacent batteries become increasingly susceptible to TR.The TR of multiple batteries occurs continuously within a short period of time; namely, the TR process of the previous battery is not over when the latter battery occurs in Figure 5(a).
Define the heat flux from cell 1 to cell 2 x, y-axis surface and busbar as Q W1 and Q W2 in Figure 7(a).As shown in Figure 7(b), upon triggering TR, Q W1 sharply increases to 720.33 W, which is the maximum value before gradually decreasing.As the cell temperature maximum, the rate of heat generation and dissipation becomes nearly balanced, leading to stabilization.Eventually, the cell surface of heat flux has a minimal value.The contact area of adjacent cell tabs connected by busbars is minor.According to Equation (24), 17.95 W of the max heat flux is generated by heat transfer through the busbars in Figure 7(b).The heat transfer through the busbar is about 3% from the x, y-axis surface.Thus, the heat conduction through the battery x, y-axis surface is the most important pathway for heat transfer during TR propagation, which is consistent with results obtained in Ref. [15].
The model successfully predicts the temperature evolutions in the battery pack in the TR propagation process.The TR propagation in the battery pack depends on the temperature difference between the front and side surfaces of the adjacent cells, which is the driving force behind the thermal spreading process as introduced in Ref. [13].Therefore, blocking the heat transfer path between adjacent batteries is supposed to be an effective approach to prevent TR propagation.In Section 3.2, insulation layers are posed between battery cells to inhibit thermal spreading, and the bottom part of the battery pack is cooled by minichannel cold plate.
According to the Electric Vehicles Traction Battery Safety Requirements in GB 38031 [32], the TR event alarming should be issued before the TR is activated.The delay in the battery TR propagation means longer time before the whole battery pack to suffer TR.For EV applications, it means longer escaping time for the passengers.Therefore, both the adjacent battery tr i and D i,j are crucial parameters to characterize the severity of battery pack damage.And in GB 38031, the D 1,2 for 300 s is the qualification.As can be seen in Section 3.1, the whole battery pack suffers TR after 94 s.It indicates that it is important to employ strategies to cut heat transfer between battery cells regarding the TR events.Implementation of such mitigation strategies may delay the TR propagation or even avoids the catastrophic phenomenon of battery pack fire.In this section, insulation layer added between adjacent battery cells is coupled with minichannel cold plate to alleviate the TR propagation.The geometry and mesh for model are shown in Figure 8.In this section, a LIB pack with insulation layers and cooled by cold plate is investigated by numerical simulations, with 0.2 W/(m•K) thermal conductivities of insulation layers as mesh independence verification.The TR trigger time is similar for the mesh numbers 676988 and 830808.Four different mesh numbers (411412, 583678, 676988, and 830808     International Journal of Energy Research are selected for the mesh independence study in Table 6. A mesh number of 676988 is selected for numerical simulations.The insulation thickness sets to 1 mm, while the thermal conductivities at 0.2 W/(m•K), 0.1 W/(m•K), 0.05 W/(m•K), and 0.02 W/(m•K) are firstly assigned to the insulation layers to vary the thermal resistivity of those insulation layers.
In Figure 9, the comparison of the battery cell temperature evolutions is shown.Without insulation layers or cold plate, tr 2 = 18 s and tr 12 = 94 s.The posing of insulation layers with different thermal conductivity as investigated in this study delayed the occurrence of tr 2 to 23 s, 31 s, 51 s, and 132 s.The tr 12 to happen in the whole battery pack is also delayed to 102 s, 118 s, 147 s, and 298 s as presented in Figure 10(a).Compared to the battery pack without protection measures, the tr 2 increases by 27.7%, 72.2%, 183%, and 633%, and tr 12 delays by 8.5%, 25.5%, 56.3%, and 217%.As can be seen, the average thermal spread time of each cells is longer than cells without insulation layers in battery pack.The implementation of insulation layers between battery cells is one of the easiest solutions to suppress TR propagation.It can slow down the rate of temperature rising and inhibits the process of heat transfer for the adjacent cells.The reason is that cell 1 subjects to TR; the huge amount of heat generated is inhibited by insulation layers with different thermal conductivities.From the view of suppressing heat transfer, the resistance of the equivalent thermal resistant layer is defined as follows:      International Journal of Energy Research The following is the heat flow through the battery shell: With cell 1 to cell 2 in unprotected battery pack, the average thermal resistance is calculated by 156.83 °C/W, which can completely inhibit TR propagation.Enlarging the thermal resistance can reduce the heat flux and thus postpone the spreading of TR.As indicated by Equation ( 25), reducing the thermal conductivity or increasing the thickness of the insulation layer can both increase the thermal resistance.Inserting additional thermal resistant layer between the adjacent batteries is a practical way to increase R ge .  .This reduced time for TR spreading is due to the heat accumulation.As cell 2 is subjected to a longer period of heating from cell 1, cell 2 experiences a higher temperature build-up before the TR is triggered.Meanwhile, the heat dissipation through the natural convection and radiation to the environment is fairly weak.This causes cell 3 to be heated more intensely and reduces D 2,3 .And so on, the thermal propagation time between adjacent batteries becomes quick, and eventually, TR occur in the whole battery pack, which is consistent with Ref. [49].The lower the thermal conductivity, the higher the heat accumulation is inversely related.Therefore, cell 6 from starting in low thermal conductivity spreads quickly.It is different from D i,j without protection measures in battery pack.This fast TR propagation process may cause more serious phenomenon of explosion.Therefore, the TR propagation of the battery pack should be completely blocked instead of just being delayed if possible.
The lowest thermal conductivity, 0.02 W/(m•K), shows the best performance in delaying the TR propagation in Figure 9(d).The surface temperature of battery under TR can reach 600 °C-700 °C in actuality [50].The thermal conductivity of silica aerogel [51] is 0.02 W/(m•K) at normal temperature, and the maximum temperature it can withstand is 700 °C.It is a good example of the thermal barrier materials.
The 1 mm insulation layer extends the tr 2 by 132 s with k = 0 02 W/(m•K).To further investigate the effect of higher thermal resistivity, thicker insulation layers with low thermal conductivities between the cells are subjected under study.In Figure 11, the temperature evolutions on cell 1-cell 12 with insulation layers of 2 mm and 3 mm and k = 0 02 W/(m•K) are shown.It is clear that the difference in insulation layer thickness has a huge effect on thermal spreading.As shown in Figures 12(a) and 12(b), the tr 2 is extended to 312 s and 944 s.The tr 12 is delayed to 512 s and 1293 s in the whole battery pack.The tr 2 postpones by 136% and 715%, and t r 12 increases by 71% and 433%, compared with the 1 mm insulation layers in battery pack.And the tr 2 of 3 mm increases by 100%, compared with the 2 mm.The effect of thickness on thermal resistance is significant.The D 1,2 = 300 s and 932 s with 2 mm and 3 mm insulation layers.It has exceeded the safety standard of 300 s.Thicker insulation layers help in winning more time for people to leave the vehicle regarding the battery pack in EV applications.But D 1,2 = 300 s and 932 s and D 11,12 = 2 s and 2 s in battery pack with 2 mm and 3 mm insulation layers.tr 2 is delayed longer, causing it to accumulate more heat.At the beginning of cell 5, the heat build-up reaches threshold that causes the later cells to trigger TR almost instantaneously.
The insulation layer has a slight effect on alleviating thermal shock, and the average T pe of the battery pack without protective methods is approximately 865.5 °C.With the addition of 1 mm insulation layer between neighboring cells, their average T pe is at approximately 831.6 °C, 829.5 °C, 831 °C, and 846 °C for different thermal conductivities.Their average T pe is 4%, 4.1%, 3.9%, and 2.2% lower than battery pack without protective methods.The average T pe in adding 2 mm and 3 mm thicknesses of silica aerogel insulation layers are also reduced to 834.9 °C and 828.34 °C.Their average T pe decreases by 3.5% and 4.2% from battery pack without protective methods.The heat transfer is blocked by the insulation layers, and the T pe of the next cell is slightly reduced by heat exchange with the ambient environment through radiation and convection.Thereby, it can be concluded that the addition of the insulation layers reduces the average T pe of the battery pack which reduces the fire risks for the battery pack.The trends of the study are similar to Ref. [52].
In this section, the effect of cold plate underneath the battery pack is also investigated.As shown in Figure 13, the tr 2 and T pe2 for cell 2 are roughly the same when varying the inlet flow rate of the minichannels.After cell 2 reaches T pe2 , cell 2 is cooled to surrounding temperature with radiation and convection heat exchange and the end of exothermic decomposition reactions.The material of cold plate is aluminum with good thermal conductivity in Table 3.The heat propagation between the cells dominates the thermal spread of the pack.The heat dissipation by the cold plate seems to be weak.This occurs because the cells in the battery module are fastened tightly with a large contact area, and the heat dissipation method underneath the battery pack does not have a sufficient chance to remove the large amount of released heat from TR as governed by Equation (24).And the thermal conductivity of the z-axis is lower than that of the cell in the x, y-axis direction by about 3.6% in Table 3.Hence, the underneath cold plate is found to only have a slight inhibitory effect on the thermal spreading in the battery pack during TR, similar to the findings in Ref. [33].

LIB Pack with Cold
Plates.In this section, a detailed analysis to study the effect of cold plate on the prevention of TR in LIB modules is made through numerical simulations.Putting cold plate between each two adjacent cells inhibits the propagation of thermal spread by exchanging heat with the huge amount of heat generated by the battery after TR.Minichannel cold plates are essential metal heat exchangers in which the inner channels carry coolant to exchange heat with heat source.The minichannel has hydraulic diameters between 3 and 6 mm with crosssectional shape of rectangle, square, circle, or polygon [53].A battery pack with 36 minichannels is set up for the simulation study.The geometry and mesh of the battery pack system for the battery pack are shown in Figure 14.The model performs mesh independent verification in Table 7.A 971,104 structured mesh is used in for the 3D thermal modeling of this LIB pack.The coolant comes from the mainstream inlet, distributes into all the minichannels to cool the cell, and gathers at the mainstream outlet for outflow.Water is assumed to be the coolant in Table 3.The dimensions of the minichannels for cold plate are shown in Table 2.And the coolant distribution strategy is similar to the thermal management system used in the Chevrolet Volt [38].The influence of the channel shape design is 16 International Journal of Energy Research supposed to be weak [54], and the general trend for putting cold plate between batteries is the major goal of this part of study.
Coolant velocities of 0.01 m/s, 0.008 m/s, 0.005 m/s, and 0.002 m/s at the mainstream inlet are tested by the numerical simulations.It is worthwhile to clarify that those inlet velocities are assigned for the mainstream as depicted in Figure 15; they are not the minichannel velocities.The cold plate between cells can effectively suppress the thermal spreading in the battery pack.When cell 1 produces TR, heat is transferred to the cold plate through heat transfer first.Next, the flowing water inside the cold plate exchanges with it, which can suppress tr 2 .The ability for the heat removal by the cold plate is not strong enough.Although the cold plate also plays a role in blocking heat, its aluminum material is good conductor of heat, which is minor.The whole battery  insulation layers and unprotected measures as a weak retardation.The addition of the cold plate makes the D i,j uniform and prevents the heat accumulation.This is a solution to prevent the hazards of heat accumulation accompanied by using the insulation layer.The average T pe of the battery pack is also reduced to 403 °C, 444.24 °C, 527.85 °C, and 581 °C.Compared to the design of Section 3.1, the average T pe reduces 53%, 48%, 39%, and 32%, which reduces the hazards caused by accidents.And the average T pe with coolant velocity of 0.01 m/s declines 52.3%, 51.7%, and 51.3%, compared to the thermal conductivity 0.02 W/(m•K) of 1 mm, 2 mm, and 3 mm insulation layers in battery pack.However, the strategy used in this section does not completely inhibit the spread of TR.After the TR is triggered in the whole battery pack, the maximum temperature of the cooling water is 86 °C, 96 °C, 128 °C, and 162.9In the actual situation, the boiling of the coolant in the minichannels leads to a huge local pressure drop, which can even destroy the cooling system of the minichannel [26].Therefore, the coolant boiling process should be avoided.As shown in Figure 15, increasing the mainstream inlet velocity improves cooling performance and eliminates the effect of coolant boiling.The strategy of increasing inlet velocity also benefits in further delaying, or even stopping the TR propagation.However, the increasing mainstream speed implies more pumping power consumption and increases the operational difficulties of the system [55].To address those issues, this work proposes a novel and effective project to suppress thermal spread in Section 3.4.

LIB Pack with Insulation Layers and Cold
Plates.The research innovation focuses on the combination of electrochemical, internal shorting, and TR models for simulation.The insulation layers block heat conduction between cells, but it does not have the ability to dissipate the heat.As a result, it can inhibit the propagation of TR but may cause heat accumulation in the battery pack, leading to faster TR later in the propagation and causing more harm.And the accumulation of heat remains inadequate for daily operational uses.Minichannel cold plate can remove heat from the battery pack, but it cannot completely block the propagation of TR.The mainstream coolant velocity needs to be very high to prevent the water from boiling.In the project of suppressing TR propagation, further research is needed to address the conflict between insulation and heat dissipation.A novel solution is the composite thermal conductive layer, which combines the insulation layers and the cold plates.
The insulation layers are used for heat insulation to prevent TR propagation, while the heat dissipation layer is used for heat dissipation to achieve the basic function of thermal management system.The novel structure completely inhibits the propagation of TR in the LIB pack.The distance between cells in the battery pack is only 6 mm, which can fully meet the demand for energy density in daily life.Therefore, in this section, a dual-functional battery module is designed by combining the cold plates and insulation layers (aerogel with thermal conductivity of 0.02 W/(m•K)) between neighboring cells, which can not only dissipate the heat demand but also fulfill the thermal insulation requirement in case of thermal hazards.This is a novel method regarding TR propagation blocking.The geometry and mesh used in this section are shown in Figure 17.
With coolant velocity of 0.01 m/s, 0.008 m/s, 0.005 m/s, and 0.002 m/s, the maximum value of the average temperature for cell 2 reached 46.2 °C, 48.8 °C, 55.8 °C, and 79.6 °C in Figure 18.It is found that cell 2 has not reached the TR.Cell 1 undergoes TR, and the large amount of released heat is blocked by the insulation layer in front of cell 2. With the flow of coolant in the cold plate, this heat that will be dissipated instead of being transferred to cell 2. As shown in Figure 19, the temperature of the water has not reached the water boiling point under the tested flow rates [26].It proves that such designs are reasonable with moderate mainstream entrance speed.Eventually, the temperature difference effect between cell 1 and cell 2 cannot make cell 2 reach the starting temperature for triggering TR.Cell 2 successfully blocked the propagation of TR.As shown in Figure 20, there is a large difference between the average temperature and the maximum temperature of cell 2. Such large temperature difference harms the battery life.Cell 2 cross-sectional cloud temperature distribution can be seen.It is found that this temperature difference does not come from the front

Conclusions
In this study, a multiscale multidomain 3D thermal modeling approach is used for LIB modeling.The thermal spread suppression scheme of insulation layers of the adjacent cells, minichannel cooling, and heat dissipation plus heat (1) The main mechanisms of TR of a single cell within a LIB pack are thermal conduction.The large temperature difference between its neighboring cells constitutes the driving force for thermal spreading, so blocking the heat transfer path becomes the preferred solution to suppress the TR propagation Cold plate i: Cell number j: Cell number pe: The maximum value of the average temperature w: Wall a: Length of the channel b: Width of the channel.

Figure 1 :
Figure 1: Flow chart of the simulation works.

Figure 2 :Figure 3 :
Figure 2: (a) Mesh independence verification and (b) comparison of the single LIB average temperature from modeling with those from Ref. [15].

Figure 4 :
Figure 4: TR triggering and propagation time definition.

Figure 5 :
Figure 5: (a) The average temperature evolutions.(b) Triggering and propagation of TR time in cell 1-cell 12 without protection methods.

Figure 8 :
Figure 8: Illustration of (a) geometry and (b) mesh for the LIB pack with insulation layers and cold plate.

Figures 9 and 10 k
(b) clearly illustrate that the use of insulation layer can aggravate the temperature inhomogeneity among the batteries.With different thermal conductivities of = 0.2 W/(m•K) k = 0.1 W/(m•K) k = 0.02 W/(m•K) k = 0.05 W/(m•K)

Figure 10 :
Figure 10: (a) Initial TR trigger time of cell 1-cell 12 and (b) TR propagation time between adjacent cells with different thermal conductivities of insulation layers.

Figure 11 :
Figure 11: The average temperature evolutions of cell 1-cell 12 with different thicknesses of insulation layers: (a) 2 mm and (b) 3 mm.

Figure 12 :Figure 13 :Figure 14 :
Figure 12: (a) Initial TR trigger time of cell 1-cell 12 and (b) TR propagation time between adjacent cells with different thicknesses of insulation layers.

vFigure 16 :
Figure 16: (a) Initial TR trigger time of cell 1-cell 12 and (b) TR propagation time between adjacent cells with different coolant velocity.

Figure 17 :
Figure 17: Illustration of (a) geometry and (b) mesh for the LIB pack with coolant plates and insulation layers.

30 32Figure 19 :Figure 20 :
Figure 19: The average temperature evolutions of water in the minichannel with different coolant velocity.

( 2 )( 3 )
In project 1, the tr 2 = 23 s, 31 s, 51 s, and 132 s with different thermal conductivity insulation layers are added, which is equivalent to adding thermal resistance between neighboring cells.It can inhibit TR propagation.The aerogel with thermal conductivity of 0.02 W/(m•K) is assumed as the material for the insulation layer, and the layer thicknesses are assigned to be 2 mm and 3 mm.The thermal spreading times of cell 2 are delayed to 312 s and 944 s.The D 1,2 = 300 s and 932 s.It satisfies the GB 38031, and adjacent cells should not generate in TR for at least 300 s.It also slightly mitigates the effects of average T pe = 834 9 °C and 828.34 °C, compared to 865.5 °C in unprotected battery packs.But, the heat generated and accumulated in battery pack can cause more serious problems.Through adding cold plates between neighboring cells in project 2, the tr 2 is delayed to 56 s, 51 s, 48 s, and 46 s with different mainstream inlet velocities for v = 0 01 m/s, v = 0 008 m/s, v = 0 005 m/s, and v = 0 002 m/s.And compared to project 1, the average T pe = 403 °C, 444.2 °C, 527.85 °C, and 581 °C, which greatly reduces the damage caused by TR of the battery pack.When the mainstream inlet velocity is large enough, the thermal spread can be further 300 s or even blocked completely The higher coolant inlet velocity inevitably consumes more pumping power.Thus, this study innovatively proposes to combine heat dissipation with thermal insulation, which means minichannel cold plates and insulation layers in project 3. It can completely block the TR spreading with minimum mainstream inlet speed of 0.002 m/s.The developed model and the conclusions drawn from project 3 can potentially poses a good safety design of LIB pack Nomenclature Greek Symbols μ: Dynamic viscosity (Pa s) ρ: Densities (kg/m 3 ) σ: Electric conductivities (W/(m 2 K 4 )) φ: Phase potentials ε: Emissivity heat transfer coefficient.Subscript description total: Total electric capacity ref : Reference electric capacity short: Internal shorting wa: Coolant c:
mainly produces C 3 H 4 O, C 5 H 9 NO, and C 4 H 8 ; LCO releases C 3 H 4 O, C 3 H 5 N, C 10 H 8 , and C 5 H 6 ; and LFP generates C 3 H 4 O and C 4 H 11 N in TR

Table 1 :
Kinetic equations for thermal decomposition [10].Reaction of negative electrode with electrolyte dc ne /dt = −A ne exp −t sei /t sei,ref exp −E ne /RT c m ne
[43]ernational Journal of Energy Research 2.4.Flow Model.In the modelling of minichannel cold plate cooling of LIBs, it is essential to determine whether the flow is laminar or turbulent.And it is determined based on the dimensionless Reynolds number[43]:
231217 International Journal of Energy Research

Table 7 :
Mesh independence verification with coolant velocity of (a) 0.01 m/s.