Cooling control of thermally-induced thermal runaway in 18,650 lithium ion battery with water mist

Cooling control of thermally-induced thermal runaway in 18,650 lithium ion battery with water mist

Energy Conversion and Management 199 (2019) 111969 Contents lists available at ScienceDirect Energy Conversion and Management journal homepage: www...

3MB Sizes 0 Downloads 1 Views

Energy Conversion and Management 199 (2019) 111969

Contents lists available at ScienceDirect

Energy Conversion and Management journal homepage:

Cooling control of thermally-induced thermal runaway in 18,650 lithium ion battery with water mist


Tong Liu, Yangpeng Liu, Xishi Wang , Xiangxiao Kong, Guochun Li State Key Lab. of Fire Science, University of Science & Technology of China, Hefei 230026, China



Keywords: Lithium ion battery Thermal runaway Hazard control Cooling control technique Water mist

Lithium ion battery (LIB) thermal runaway (TR) has always been a potential risk that could result in serious damage. This likelihood has increased owing to the widespread application of LIBs, particularly in the electricvehicle industry. Therefore, an effective method must be determined to prevent or mitigate this hazardous process. As a clean and efficient cooling technique, the effect of water mist (WM) on TR has been investigated through a series of tests in this study. Batteries with various states of charge (SOC) were heated with an electric heater to induce TR, and the surface temperatures were measured throughout the tests. The results show that TR can be controlled when the WM is released before the critical temperature is reached; the threshold temperature is at least 20 °C lower than the temperature of the TR onset. The determined conservative temperature for the WM application is 186.5 °C. At 25% SOC, the critical WM cooling rate is 1.87 times that of the critical heating rate. The value increases to 4.98 at 100% SOC, thereby indicating the increasing suppression difficulty with increasing SOC. Above the critical temperature, the TR is unstoppable. However, the maximal surface temperature can be controlled at values approximately 300 °C lower than those of the cases without WM. This is beneficial for the prevention of the TR propagation. TR can still be a potential risk for suppressed batteries. Nevertheless, the thermal hazard is mitigated with increasing onset temperature and a longer heating process. This study can contribute significant results for the control of TR in practice.

1. Introduction Lithium ion batteries (LIBs) have been widely used to counter the energy crisis and improve environmental pollution [1] owing to their numerous advantages, which include high specific capacity, long cycle life, and no memory effect [2,3]. However, the high amount of heat generated during their application [4] or under abuse conditions (thermal, electrical, and mechanical conditions [5]) lead to a drastic increase in the battery temperature. Thermal runaway (TR) is eventually triggered that could result in venting, combustion, ejection, and/ or an explosion if the heat cannot be dissipated efficiently [6]. Owing to the rapid development of the electric-vehicle (EV) industry over the last few years, the demand for LIBs has increased, thereby leading to an increasing TR risk. Owing to this widespread application of LIBs, some TR accidents have happened. For example, EV buses (China, 2017) [7] and a Tesla Model X (USA, 2018) [8] catching fire. During TR, large amounts of heat and toxic gases are released that can harm consumers [9–11]. To determine the thermal-hazard behavior of LIBs, many studies have been conducted with various methods. For instance, the thermal parameters

of 2.9 Ah pouch-type LIBs were investigated with a fire propagation apparatus [2]. Furthermore, the TR processes of 2.2 Ah commercial 18,650 LIBs were studied through copper slug battery calorimetry [12]. Most types of LIBs experience an energetic TR process under abuse conditions. The heat generated by TR is transferred to adjacent cells and cause TR propagation, which can lead to more serious disasters [13,14]. Therefore, an effective method must be developed to prevent TR. Many studies have been conducted to improve the thermal tolerance of LIBs according to their material aspects [15,16]. However, the potential thermal hazards have not been eradicated owing to the TR mechanism. Based on the former study [17], the simplified mechanism of TR is similar to a combustion mechanism that can be described with a “TR triangle”, as shown in Fig. 1. This consists of three elements: the combustible materials, ignition source, and oxidant. The combustible materials are organic electrolyte and combustible gases generated by side reactions (e.g., H2, CH4, C2H4 and C2H6) [5,18]; the ignition sources can be abuse conditions and the heat generated by side reactions (e.g., reactions between the electrolyte and anode, cathode decomposition, and electrolyte decomposition [3,19]); the oxidant is replaced by air and oxygen produced by the cathode decomposition

Corresponding author. E-mail address: [email protected] (X. Wang). Received 20 May 2019; Received in revised form 19 August 2019; Accepted 20 August 2019 Available online 27 August 2019 0196-8904/ © 2019 Elsevier Ltd. All rights reserved.

Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

Nomenclature Ts Ak c cw D32 hf hk k l m tr tr-wh Tc Tboil Tmax Tmin Tr Tr-wh Tw

Contact area between battery surface and water mist droplets (m2) Heat capacity of battery (kJ kg−1 °C−1) Heat capacity of water (kJ kg−1 °C−1) Sauter mean diameter (µm) Latent heat of water (kJ kg−1) Thermal conductivity (W m−1 °C−1) Flow discharge coefficient (L min−1 MPa−0.5) Thickness of water mist droplet on battery surface (m) Battery mass (g) Thermal-runaway onset time without water mist (s) Thermal-runaway onset time of suppressed battery under continuous heating condition (s) Critical suppression temperature of water mist (°C) Boiling point of water (°C) Highest surface temperature during test (°C) Lowest surface temperature during water mist application (°C) Thermal-runaway onset temperature without water mist (°C) Thermal-runaway onset temperature of suppressed battery under continuous heating (°C) LIB surface temperature when water mist pump is

T∞ P Ph Pw Q̇ Q̇ w ΔL Δm

activated (°C) LIB surface temperature when water mist starts cooling (°C) Ambient temperature (°C) Nozzle working pressure (MPa) Critical heat accumulation rate (W) Critical water mist cooling rate (W) Water mist flow rate (L min−1) Water mist flow rate on battery surface (m3 s−1) Axial dimension variation of battery after test (mm) Total mass loss of battery after test (g)

Abbreviations BMS EV LIB SOC TR WM

Battery management system Electric vehicle Lithium ion battery State of charge Thermal runaway Water mist

Greek symbols α ρ

Cooling efficiency coefficient Density of water (kg m−3)

between cells. However, it cannot suppress a TR in a single cell. According to the results of the experimental studies in [26], phase change composites can prevented TR propagation. However, the heat transfer takes considerable time, owing to the poor thermal conductivity. Yuan et al. [27] compared the inhibition effects of different interstitial materials on TR propagation in battery modules. The results indicated that Al extrusions present the best control performance. However, an increase in the battery module weight is inevitable. Kritzer et al. [28] developed an emergency cooling element with compressed CO2 for EVs, which worked effectively during the overcharge tests. However, when the battery experiences much higher temperatures or even fire, the cooling system does not work. Thus, various approaches have been applied to control the thermal hazard of LIBs. However, the application of water mist (WM) is rarely mentioned. Prevention of TR propagation is important, and a thermal-hazard control of the triggering cell is crucial to solve the fundamental problem. As a traditional and clean fire-extinguishing agent, water is widely applied in firefighting because of its outstanding cooling ability. Ditch reported that a sprinkler protection system can protect a growing or developed LIB rack storage fire. Nevertheless, reignition remains a potential threat after the water supply is terminated [29]. Compared with traditional water sprinklers, the WM method has a higher cooling efficiency and consumes less water [30]. NASA developed a portable fine-WM extinguisher for battery fires, and used it on the international space station [31].The ability of WM regarding the cooling of batteries and the TR suppression mechanism have not been investigated. Theoretical analyses are necessary for the practical application of WM to LIB hazard control. In addition, in order to avoid potential electrical short-circuit due to amount of water inside the battery pack caused by continues spray of WM, precautions should be considered in practical application or conducted additional testes to verify this problem. The appropriate application strategy of WM is significant for both cooling efficient and equipment protection, and the amount of WM used during TR hazard control should be determined through tests. In this study, a detailed experimental investigation was conducted to investigate the thermal hazard of 18,650 LIBs and the cooling control strategy with WM for various states of charge (SOC). The TR was

[20,21]. The combustion process is terminated with the destruction of any element of the “fire triangle”. However, the TR triangle is much stronger and more difficult to stop. TR can occur spontaneously owing to the break-up of the solid electrolyte interface [22]. Thereafter, heat, oxygen, and various combustible gases are produced. All three elements can be supplied by the LIB itself once the exothermic side reactions are triggered; hence, the side reactions must be terminated for TR control. Owing to the battery shell, a fire-extinguishing agent cannot enter the battery directly. However, cooling is a suitable method that absorbs heat through conduction to decrease the battery temperature. Consequently, the exothermic side reactions are terminated, and the TR is suppressed naturally. Cooling has already been applied to the battery management system (BMS) [23], and much effort was devoted to improving the control effect [4,24]. However, under most conditions, the abnormal temperature variations can only be controlled at a relatively low temperature stage with BMS. If LIBs experience extreme hazardous conditions, the BMS is insufficient for TR control. Xu et al. [25] developed a novel mini-channel cooling system, which can prevent TR propagation

Fig. 1. TR triangle of LIBs. 2

Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.


triggered through an electric heater, and a WM nozzle with a 0.5 MPa working pressure was used to produce the WM spray. The thermal behavior of the LIBs (with and without WM) were analyzed to determine the critical issues of the WM control method and critical temperature for the WM application. Furthermore, the relationship between the critical temperature and TR onset temperature for different SOC was investigated. Finally, the TR potentials of the controlled batteries were examined under continuous heating.

Q̇ = k 10P


where k (L min−1 MPa−0.5) represents the flow discharge coefficient. Table 1 lists the detailed WM characteristics. The droplet size is represented by the Sauter mean diameter (D32) [30]. 2.4. Experimental conditions

2. Experimental setup Both SOC and WM application time were considered in the experiments. Table 2 lists the different investigated cases, and Fig. 4 shows the flowchart of the test procedure. No WM was applied to Test Group 1. For Test Groups 2–6, the WM pump was activated when the battery surface temperature reached Tw. The surface temperature increases above Tw when the WM cooling effect starts (denoted by Ts) and can be determined through the temperature data by considering the required time of the water transport. Therefore, it is difficult to maintain a constant Ts during all repeated tests. For Test Groups 2–5, the initial Tw of the different SOC can be approximately determined based on Test Group 1 and then verified by decreasing the temperature in steps of 5 °C. If a TR is probable in the repeated tests at Tw, the critical temperature for the WM application can be determined based on the Ts of the repeated tests in which TR is controlled. If the TR was inhibited successfully in all repeated tests at Tw, one more test (repeated at least three times) was conducted at Tw + 2 °C. The critical temperature was determined through control tests again. To determine thoroughly the influence of the WM on the TR suppression at various Tw, the temperature steps were reduced to approximately 2 °C for Test Group 4. For Test Group 6, the electric heater operated continuously to determine the thermal-hazard potential of the samples that experienced no TR during the WM application. Hence, another TR could be triggered. The WM working pressure and duration were fixed to 0.5 MPa and 60 s in this experiment, respectively. In addition, the LIB mass was measured with an electronic balance (Mettler

2.1. Battery samples The 18650-type LIBs (Samsung 18650-26F) examined in this study have a diameter of 18 mm, a length of 65 mm, and contain an LiNi1/ 3Mn1/3Co1/3O2 cathode and carbon anode. The operating voltage ranges from 2.75 to 4.20 V, and the nominal capacity and nominal voltage are 2.6 Ah and 3.7 V, respectively. The plastic packaging was removed from the cells before the tests, and the resulting mass was 44.06 ± 0.17 g. A battery cycler (LANHE CT2001C) controlled by a computer was used to prepare the battery samples. Each sample was discharged to 2.75 V with a constant current of 1300 mA [32] and then charged to 4.2 V with the constant-current/constant-voltage method until the charge current fell below 26 mA. Finally, the battery was discharged to the desired SOC with a constant current of 1300 mA. 2.2. Experimental apparatus The schematic diagram of the experimental apparatus for the LIB TR suppression is shown in Fig. 2(a). The apparatus was made of aluminum alloy with one tempered glass side for observing the experiment. The apparatus mainly consisted of the heating component, a WM system, and an upper smoke exhaust system. As shown in Fig. 2(b), the heating component was a hollow copper cylinder (length of 65 mm and inner diameter of 18.5 mm), which was fixed to the center of the bottom. The battery sample lay horizontally on the electric heater and was fastened with a stainless-steel wire. Three Ktype thermocouples with diameters of 1 mm (T1, T2, and T3, respectively; recorded at a frequency of 1 Hz) were attached to the upper surface of the LIB with high-temperature adhesive tape. The distance between the neighboring thermocouples was approximately 32 mm. The WM nozzle was installed centrally and 50 cm above the upper surface of the LIB and powered by a pump. The WM working pressure was 0.5 MPa. The toxic gases produced during the experiment were exhausted through the upper ducting section, and the water exited through the drainage gaps. The controlled heat was delivered by supplying AC power to four identical cylindrical heaters with diameters of 6 mm. The total heating power was set to 100 W, and the copper surface temperature was maintained at approximately 300 °C. The experimental design facilitates the WM distribution onto the LIB surfaces. 2.3. WM characteristics The cooling capacity of the WM system is determined by various critical parameters, including the WM droplets size, flow discharge coefficient, and spray angle. Hence, the LaVision Particle-Master Shadow system was used to measure the droplet sizes. This system is based on high-resolution imaging with pulsed backlight illumination, and independent of particle shapes and materials [33]. A detailed description of the optical arrangement can be found in [34]. The WM nozzle was installed 0.5 m above the top surface of the battery, and the working pressure was 0.5 MPa. Fig. 3 presents the pattern of the WM spray, which was illuminated by a laser sheet [35]. According to the photograph, the spray angle is 68°. The measured WM nozzle flow rate was 0.79 L/min, and the flow discharge coefficient was calculated as

Fig. 2. (a) Schematic of experimental apparatus. (b) Details of heating component and LIB setup. 3

Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

Table 2 Experimental settings for suppression tests. Test group

SOC (%)

Tw (°C)

Heating conditions




0/25/50/ 75/100 25

Turned off after triggered TR Turned off after WM termination

3 4

50 75

5 6

100 25 50 75 100

TR*/280/275/265/ 260/257/255 225/220/217/215 217/215/212/210/ 207/205/202/200 195/190/187/185 255 215 200 185

Continuous heating

*WM was released when TR occurred. Fig. 3. Schematic of WM characterization with Particle-Master Shadow system.

Toledo AL104) with a resolution of 0.0001 g before and after each test. The experiments were recorded by a video camera (Sony FDR-AXP55) at 50 fps. The temperature data are of key importance in this experiment. Owing to the limited measurement accuracy, the temperature acquired by the thermocouples had an error of ± 1.5 °C. Furthermore, owing to manufacturing differences, the samples might experience different heating processes. Hence, to obtain more reliable measurement results, each test was repeated at least twice. 3. Results and discussion 3.1. Temperature variations of different SOC without WM Fig. 5 shows the surface temperature variations of the battery samples with different SOC under heating. As presented in Fig. 5(a), the fully discharged battery sample does not experience TR. An inflection point appears when T2 reaches 220.1 °C. Fig. 6 shows the surface temperature increase rate of T2 at 0% SOC. Behind the inflection point, the temperature increase rate experiences a steep decrease, and the battery temperature slowly approaches the heater surface temperature. Owing to the deficient lithium in the anode, it is reasonable to assume that the heat released from exothermic side reactions (e.g., the reaction between the anode and electrolyte) is insufficient to maintain the rapid heating process [36]. Therefore, TR cannot be triggered in a fully discharged battery. According to Fig. 5, the temperature curves of T1, T2, and T3 are in good agreement during the initial heating process and become different when a TR is triggered. In all tests, T2 stays at the highest value owing to the better heat dissipation of the positive and negative electrodes. All other cases experience TRs with different thermal phenomena. The insets in Fig. 5 present typical TR phenomena of cells at various SOC: smoke venting (S), injection (I), explosion (E), and combustion (C). A violent injection occurs at 75% and 100% SOC. Combustion is an attendant phenomenon that appears at the positive electrode after the injection owing to the produced combustible gases and residual organic electrolyte. Moreover, explosions occur in all repeated tests for 50% SOC, which is probably because the partially opened vent prevented a prompt inner-pressure relief. Table 3 lists the detailed information of Test Group 1. The

Fig. 4. Flowchart of experimental procedure.

uncertainties are the standard deviations of the mean, and the temperature value is mainly based on T2; Tr represents the onset temperature of TR, and tr is the corresponding onset time. With increasing SOC, the battery exhibits lower Tr and shorter tr, thereby indicating a rapid development of the internal exothermic side reactions and thermal accumulation. The higher Tmax, Δm, and ΔL imply a more violent TR hazard, which can be confirmed based on the TR

Table 1 WM characteristics. Nozzle working pressure (MPa)

D32 (µm)

Flow rate (L min−1)

Flow discharge coefficient (L min−1 MPa−0.5)

Cone angle (°)







Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

Fig. 5. Surface temperature versus time of different SOC: (a) 0%; (b) 25%; (c) 50%; (d) 75%; (e) 100%.

Fig. 6. Temperature rates of T2 at 0% SOC. Fig. 7. Curves of heating and cooling rates at the beginning of WM application for different Ts.


the WM. Considering a stable WM flow field, once the heating process is prevented at the beginning of the WM application, the LIB temperature can be controlled, and the TR is suppressed. Otherwise, the LIB temperature keeps increasing, and a TR is unstoppable. Hence, the

3.2. Analysis of critical temperature and heating rate A discussion of the relationship between the LIB heating rate and WM cooling rate is essential to understand the control mechanism of Table 3 Test Group 1. SOC (%)

Tr (°C)

0 25 50 75 100

\ 308.8 242.4 222.6 208.2

tr (s)

± ± ± ±

1.1 5.9 6.9 1.6

\ 690.3 624.0 590.8 559.0

Tmax (°C)

± ± ± ±

9.0 24.9 10.9 3.0

276.4 440.9 493.4 708.3 598.7

± ± ± ±

20.9 17.6 17.2 5.5


Δm (g)

ΔL (mm)

TR phenomenon

4.32 5.37 ± 0.20 28.76 ± 1.18 14.63 ± 1.70 29.50 ± 0.71

0.40 0.58 1.91 2.22 3.28

\ S S&E S&I&C S&E&I&C

± ± ± ±

0.16 0.17 0.59 0.42

Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

Fig. 8. Temperature rates of various SOC: (a) 25%; (b) 50%; (c) 75%; (d) 100%. Table 4 Typical thermal parameters of tests when WM is released at critical temperature. SOC (%)

Tmin (°C)

Critical temperature (°C)

Critical heating rate (°C s−1)

Ph (W)

25 50 75 100

74.3 78.6 56.1 80.1

256.8 219.3 201.4 186.5

1.54 1.24 0.89 0.58

57.67 46.25 33.14 21.72

± ± ± ±

3.5 2.2 0.7 0.7

± ± ± ±

0.31 0.15 0.18 0.04

± ± ± ±

11.61 5.69 6.67 1.50

Pw (W)

α (Pw/Ph)


1.87 2.34 3.26 4.98

relationship of the battery heating rate and WM cooling rate at the beginning of the WM release is decisive for the LIB thermal-hazard control. The battery temperature rate experiences a quasi-exponential increase during the heating stage, which can be expressed approximately with the Arrhenius law [3,36]. The cooling rate of the WM at the beginning of its application is described based on the heat conduction between the WM droplets and LIB surface:

Pw =

Ak hk (Ts − T∞) l


where Ak hk can be considered constant for equal WM flow fields. Hence, l the cooling rate depends on Ts. However, when the LIB surface temperature exceeds 100 °C, the WM droplets evaporate, and the latent heat should be considered. Thus, Eq. (2) is replaced with the following equation:

Pw = c w Q̇ w ρ (Tboil − T∞) + hf Q̇ w ρ −1


[37]. The first term on the right side rewhere hf is 2257 kJ kg presents the sensitive heat, and the second term represents the latent heat; Q̇ w is the volume flow rate of the WM on the LIB surface. The

Fig. 9. Temperature curves of different SOC with WM-suppressed TR. 6

Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

Fig. 10. Cooling rates of cells at different SOC during WM application: (a) 25% SOC, Ts = 255.3 °C; (b) 50% SOC, Ts = 216.9 °C; (c) 75% SOC, Ts = 200.7 °C; (d) 100% SOC, Ts = 186.8 °C.

between the heating and cooling rates at the beginning of WM application for various Ts is presented in Fig. 7. There are two intersections (A and B) with corresponding temperatures Ta and Tb. At point A, the heating and cooling rates maintain a stable balance, and the temperature is relatively low. Nevertheless, B is a fragile balance point and slight disturbances might lead to huge differences. For Ts < Tb, the cooling effect dominates the temperature variations, and a TR can be prevented eventually. For Ts > Tb, the battery experiences a heating process, and a TR is inevitable. Therefore, Tb is the highest possible temperature for the WM release to suppress a TR hazard (i.e., the critical temperature Tc). Owing to the manufacturing differences, Tc should be a temperature range for a given SOC. As discussed in Section 3.1, TR can be triggered when the SOC is not below 25%, and the samples experience a similar heating process as in Fig. 5(b)–(e). Fig. 8 presents the variations in the battery temperature rate at different SOC during heating. The position of Tc at various SOC is approximately determined based on the critical temperature values shown in Table 4. Taking 75% SOC as an example, ignoring the initial low-temperature stage, the temperature rate exhibits a quasi-exponential curve before TR. The battery sample experiences an approximately stable heating process before the safety vent opens. Subsequently, the temperature rate increases rapidly, which results in TR at 233.6 °C for a heating rate of 23.73 °C/s. Regarding Test Group 4, the critical temperature Tc at 75% SOC is 201.4 ± 0.7 °C for a heating rate of 0.89 ± 0.18 °C/s. It was assumed that the battery was a thermally lumped model and that no obvious mass loss occurred before the TR was triggered. Thus, the critical heat accumulation rate can be approximated with the following equation:

Fig. 11. Tc at various SOC.

ambient temperature is approximately 20 °C. For equal WM flow field, Q̇ w can be seen as constant (measured value: 4.17 × 10−8 m3/s). Therefore, for tests with Ts higher than Tboil, the cooling rate at the beginning of the WM release is calculated to be constant according to Eq. (3) (calculated value: 108.1 W). Based on [38,39], the relationship


Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

Fig. 12. (a) Temperature curve of 75% SOC with Ts = 207.3 °C; (b) Temperature variations of 25% SOC with WM released after TR.

Savitzky–Golay method [41]. Taking 75% SOC as an example, the cooling rate experiences a rapid decrease at the initial stage. Its maximum decreases from 18.18 °C/s to 1.06 °C/s in 15 s. As presented in Table 4, Ph decreases with increasing SOC, which implies a lower heating rate. To describe the cooling ability of the WM, a proportionality coefficient α (standard deviations are not considered) was defined:

α = Pw / Ph

which can be used to describe the cooling efficiency at the beginning of the WM release. A high value represents a poor cooling efficiency. As shown in Table 4, α increases with increasing SOC, which implies a decreasing cooling efficiency and tougher suppression conditions at higher SOC. Fig. 11 shows the variations in Tc with different SOC. It decreases like Tr with increasing SOC and has a maximum of 256.75 ± 3.45 °C at 25% SOC, which indicates a relatively easy-to-control condition. However, the critical temperature decreases 70 °C to 186.5 ± 0.73 °C for 100% SOC, thereby indicating the great influence of the SOC on the TR suppression and the more difficult suppression conditions at higher SOC. For 50%–100% SOC, the temperature difference between Tr and Tc is approximately 20 °C. Furthermore, the temperature difference enlarges to approximately 50 °C at 25% SOC. Owing to the relatively sufficient active material at higher SOC (50%–100% SOC), the temperature experiences a more rapid increase at relatively low temperatures, as presented in Fig. 8(b)–(d), which decreases the temperature difference between Tc and Tr. As previously discussed, both increasing α and decreasing Tc with increasing SOC reflect the aggravated control difficulties of high-SOC batteries. Samples with SOC above 50% experience more severe TR hazards and more strict suppression conditions. Thus, SOC below 50% are recommended for storage or transportation. Furthermore, a conservative critical temperature of 186.5 °C for the WM release is recommended for these cell types.

Fig. 13. Delaying effect of WM cooling on TR onset time at various Ts.

Ph = cm

dT dt



where the heat capacity c is 0.85 kJ/kg °C [40]. Thus, the heat accumulation rate at the critical temperature for 75% SOC is 33.14 ± 6.67 W. Compared with the battery surface, the inner battery part experiences a much more rapid heating process owning to the spontaneous exothermic side reactions and worse heat dissipation conditions. Hence, the WM cooling rate Pw is higher than Ph, which will be further discussed in Section 3.3.1. 3.3. Control effect of WM system 3.3.1. Cooling performance at critical temperature Fig. 9 presents the temperature development of the cases in which TR is suppressed by the WM. During the 60 s WM application, the battery samples with different SOC experience a similar rapid cooling process. The surface temperature decreases to a relatively low value within seconds and reaches a minimum of no more than 100 °C, as presented in Table 4. After the WM is terminated, the surface temperature shows a resilient behavior owing to the influence of the residual heat. The results show that the TR does not happen after the WM is terminated, if it is prevented during the WM release in Test Groups 2–5. This indicates the outstanding suppression capacity of the WM and the effective cooling process of the LIBs. Fig. 10 shows the cooling rate variations during the WM release. The curves were smoothed with the

3.3.2. Cases of WM applied above critical temperature TR is unstoppable and can be triggered during the WM release in all tests when Ts > Tc. However, TR activity is mitigated to a certain degree. In addition, owing to the cooling effect of the WM, the battery surface temperature can be well controlled. Fig. 12(a) presents the surface temperature curve of a 75% SOC battery with WM released at 207.3 °C. The TR is triggered 11.4 s after the WM application, and a sharp increase in the temperature occurs. The surface temperature reaches a maximum of 382.9 °C and experiences a steep decrease immediately afterward. During the entire process, the highest surface temperature is 400 °C, which is much lower than the temperature of 708.3 °C of the case without WM application and therefore significant 8

Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

Fig. 14. Surface temperature curves of different SOC with continuous heating: (a) 25%; (b) 50%; (c) 75%; (d) 100%. Table 5 TR onset temperature and time of Test Groups 1 and 6. SOC (%)

tr (s)

25 50 75 100

690.3 624.0 590.8 559.0

± ± ± ±

9.0 24.9 10.9 3.0

tr-wh (s)

Tr (°C)

\ 1836.8 ± 99.6 1394.5 ± 58.5 1282.0 ± 15.0

308.8 242.4 222.6 208.2

± ± ± ±

1.1 5.9 6.9 1.6

Tr-wh (°C)

TR phenomenon

\ 292.8 ± 11.0 251.5 ± 0.5 220.0 ± 1.7

\ S S&I S&I

3.3.3. Thermal-hazard potential of suppressed cells In the suppressed cells, TR can still be a potential risk after the WM is terminated if the abuse conditions have not been eradicated. To obtain a more comprehensive understanding of the WM suppression effect, the thermal hazard of the suppressed LIBs was investigated under continuous heating. Fig. 14 presents the temperature curves of the LIBs with various SOC. The insets present the typical TR procedures. TR severity experiences varying degrees of mitigation. The TR is prevented for 25% SOC, which might be caused by two reasons: first, the residual inner active materials are insufficient for exothermic side reactions and therefore do not produce sufficient heat for TR; second, the onset temperature of TR is higher than the maximal temperature of the heater surface, which might result in a deficient heat supply. Different from Test Group 1, explosions are prevented for 50% and 100% SOC, and no evident combustion phenomena appear in all repeated tests of Group 6. As shown in Fig. 14(c), the entire process of the 50%–100% SOC samples comprises four stages: Stage 1 is the first heating stage; Stage 2 represents the WM cooling stage; Stage 3 is the second heating stage; Stage 4 represents the TR hazard stage. As discussed in Section 3.3.1, the battery samples experience an effective cooling process during the

for the TR propagation prevention. For 50% and 100% SOC, the temperature curves show similar variations to the 75% SOC curve (not presented here). For 25% SOC, the battery surface experiences a distinguished cooling process during the WM application, regardless of whether TR occurs. Fig. 12(b) presents the temperature variations of the case in which the WM is released immediately after TR occurs (Ts is 323.4 °C). The surface temperature decreases during the WM application and reaches a minimum of 83.8 °C. This indicates the outstanding cooling capacity of the WM for low-SOC batteries. Before TR, the inner exothermic side reactions are slowed down owing to the cooling effect of the WM. Consequently, the triggered TR is delayed. Fig. 13 shows the time intervals until TR occurs (denoted by t) for various Ts. The straight line is the corresponding fitting line. Compared with the case without WM, TR is delayed in nearly all cases with WM application. The delaying effect is weakened with increasing Ts owing to the shortened cooling time before the onset of TR. The maximal delay is approximately 10 s, which can be crucial for the escape of personnel or the rescue of property during accidents.


Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

temperature (Tc) is reached. The threshold values range from 256.8 ± 3.5 °C to 186.5 ± 0.7 °C for SOC of 25%–100%; Tc is at least 20 °C lower than the TR onset temperature of the corresponding SOC. (2) TR does not occur after the WM is terminated and the heater is turned off, when the battery is controlled during WM release. The WM exhibits an outstanding cooling capacity of more than 10 °C/s at the initial application stage. Subsequently, the cooling effect exhibits a rapid decrease to less than 1 °C/s within seconds. Moreover, the battery surface temperature remains at a relatively low temperature during the WM application for all SOC. (3) TR is unstoppable when the WM is released after the critical temperature has been reached. Thus, TR occurs during the WM application, and the battery surface temperature increases significantly. However, it is much lower than that of the case without WM. The TR onset time is prolonged, and the delay time exhibits an approximately linear relationship with the WM start temperature (Ts). (4) Under continuous heating and during the WM release, the controlled battery exhibits a mitigated TR hazard. The TR risk disappears at 25% SOC. However, the thermal hazard remains for higher SOC. Compared with the cases without WM, all samples experience a slower heating process and higher TR onset temperatures of 292.8 ± 11.0 °C, 251.5 ± 0.5 °C, and 220.0 ± 1.7 °C for 50%, 75%, and 100% SOC, respectively.

Fig. 15. Mass loss at different SOC.

WM release at the critical temperature. Thus, the batteries undergo a second heating process starting from a relatively low temperature under continuous heating. During the first heating stage, the average heating rates are 0.32 °C/s, 0.31 °C/s, and 0.29 °C/s for 50%, 75%, and 100% SOC, respectively. However, in Stage 3, the heating rates experience different decreases to 0.18 °C/s, 0.26 °C/s, and 0.26 °C/s, respectively. Because of the higher Tc at lower SOC, the exothermic side reactions consume more active materials at lower SOC, which results in a maximal decrease of 0.14 °C/s for the 50% SOC in Stage 3. As presented in Table 5, TR is totally prevented at 25% SOC. In addition, the thermal hazard is mitigated for higher SOC. The onset time of TR is multiplied by approximately three (approximately 30 min at 50% SOC). In addition, the delaying effect decreases with increasing SOC. However, it still lasts approximately 21 min at 100% SOC. Therefore, the TR suppression with WM provides more time for the escape of personnel. In the meantime, necessary measures can be taken to remove the hazardous source. In addition, the onset temperature increases compared with that of the case without WM, thereby indicating relatively mild conditions. All these improvements reflect the outstanding cooling capacity of WM. During a TR, many toxic or harmful materials are ejected from the battery, which affect the environment and personnel. Fig. 15 shows the mass losses for various test conditions. The mass loss caused by smoke venting is similar to those of the cases without TR (approximately 4 g). By contrast, an explosion leads to a maximum Δm of approximately 30 g, regardless of the SOC. However, once the explosion is prevented, Δm experiences a strong decrease, in particular, for a 50% SOC. In Test Group 6, Δm is the lowest compared with those of the other cases at the corresponding SOC (except in the cases with Ts < Tc), which reflects the mitigation of the TR hazard intensity.

The study results confirm the effectiveness of WM for the control of TR in LIBs, and could be applied for TR suppression in the future. In the next research step, the authors will focus on the effect of the WM characteristics on TR suppression efficiency. Declaration of Competing Interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgements The authors appreciate the support of the Natural Science Foundation of China (No. 51874265), the Key national R&D program (Grant No. 2018YFC0809502) and the Fundamental Research Funds for the Central Universities (Grant No. WK2320000044). References [1] Feng XN, Ouyang MG, Liu X, Lu LG, Xia Y, He XM. Thermal runaway mechanism of lithium ion battery for electric vehicles: A review. Energy Storage Mater 2018;10:246–67. [2] Ribière P, Grugeon S, Morcrette M, Boyanov S, Laruelle S, Marlair G. Investigation on the fire-induced hazards of Li-ion battery cells by fire calorimetry. Energ Environ Sci 2012;5:5271–528. [3] Wang QS, Ping P, Zhao XJ, Chu GQ, Sun JH, Chen CH. Thermal runaway caused fire and explosion of lithium ion battery. J Power Sources 2012;208:210–24. https:// [4] Liang JL, Gan YH, Li Y. Investigation on the thermal performance of a battery thermal management system using heat pipe under different ambient temperatures. Energ Convers Manage 2018;155:1–9. 2017. 10.063. [5] Fernandes Y, Bry A, De Persis S. Identification and quantification of gases emitted during abuse tests by overcharge of a commercial Li-ion battery. J Power Sources 2018;389:106–19. [6] Sheng L, Su L, Zhang HY, Fang YD, Xu HF, Ye W. An improved calorimetric method for characterizations of the specific heat and the heat generation rate in a prismatic lithium ion battery cell. Energ Convers Manage 2019;180:724–32. 10.1016/j.enconman.2018.11.030. [7] Feijter TD. Visiting The Scene Of The May 1 Beijing Electric Bus Charging. Station Fire. 2017. [8] Hyatt K. Tesla Model X fatal crash and fire under investigation. 2018. https://www. [9] Golubkov AW, Fuchs D, Wagner J, Wiltsche H, Stangl C, Fauler G, Voitic G, Thaler A, Hacker V. Thermal-runaway experiments on consumer Li-ion batteries with metal-oxide and olivin-type cathodes. RSC Adv 2014;4:3633–42.

4. Conclusions In this study, the control effect of WM on the TR of 18,650 LIBs with various SOCs was studied through experiments. To understand the cooling control capacity thoroughly, the thermally-induced hazards of LIBs with and without the application of WM were analyzed and compared. In addition, the critical triggering temperature (Tc) for the WM was identified and discussed, and the potential thermal hazards of the controlled batteries were investigated. The main conclusions are as follows: (1) TR can be prevented by applying WM before the critical 10

Energy Conversion and Management 199 (2019) 111969

T. Liu, et al.

[26] Wilke S, Schweitzer B, Khateeb S, Al-Hallaj S. Preventing thermal runaway propagation in lithium ion battery packs using a phase change composite material: An experimental study. J Power Sources 2017;340:51–9. jpowsour.2016.11.018. [27] Yuan C, Wang Q, Wang Y, Zhao Y. Inhibition effect of different interstitial materials on thermal runaway propagation in the cylindrical lithium-ion battery module. Appl Therm Eng 2019;153:39–50. 02.127. [28] Kritzer P, Döring H, Emermacher B. Improved safety for automotive lithium batteries: an innovative approach to include an emergency cooling element. Adv. Chem. Eng. Sci. 2014;04:197–207. [29] Ditch B. The impact of thermal runaway on sprinkler protection recommendations for warehouse storage of cartoned lithium-ion batteries. Fire Technol 2018;54:359–77. [30] Grant G, Brenton J, Drysdale D. Fire suppression by water sprays. Prog Energ Combust 2000;26:79–130. [31] Rodriguez B, Young G. Development of the International Space Station (ISS) Fine Water Mist (FWM) Portable Fire Extinguisher. 43rd International Conference on Environmental Systems, Colorado, USA. 2013. -3413. [32] Samsung SDI Co., LTD. Specification of Product for Lithium-ion Rechargeable Cell Model: ICR18650-26F 2009. [33] Wang ZG, Wang XS, Huang YQ, Tao CF, Zhang HP. Experimental study on fire smoke control using water mist curtain in channel. J Hazard Mater 2018;342:231–41. [34] Wang XS, Zhu P, Li Y, Ni XM, Fan MH. Effect of low ambient air pressure on spray characteristics of water mist. Exp. Therm Fluid Sci 2015;66:7–12. 10.1016/j.expthermflusci.2015.03.009. [35] Liu YP, Wang XS, Zhu P, Li GC, Ni XM, Zhang J. Experimental study on gas jet suppressed by water mist: A clean control technique in natural gas leakage incidents. J Clean Prod 2019;223:163–75. 03.107. [36] Liu JJ, Wang ZR, Gong JH, Liu K, Wang H, Guo LS. Experimental study of thermal runaway process of 18650 lithium-ion battery. Materials 2017;10:230. https://doi. org/10.3390/ma10030230. [37] Incropera FP, Lavine AS, Bergman TL, DeWitt DP. Fundamentals of heat and mass transfer. 6th ed. New York: Wiley; 2007. [38] Semenov NN. Some problems of chemical kinetics and reactivity. New Jersey: Princeton University Press; 1959. [39] Tobishima S, Yamaki J. A consideration of lithium cell safety. J Power Sources 1999;81:882–6. [40] Yayathi S, Walker W, Doughty D, Ardebili H. Energy distributions exhibited during thermal runaway of commercial lithium ion batteries used for human spaceflight applications. J Power Sources 2016;329:197–206. jpowsour. 2016.08.078. [41] Zhu XQ, Wang ZP, Wang YT, Wang H, Wang C, Tong L, Yi M. Overcharge investigation of large format lithium-ion pouch cells with Li(Ni0.6Co0.2Mn0.2)O2 cathode for electric vehicles: Thermal runaway features and safety management method. Energy 2019;169:868–80.

10.1039/C3RA45748F. [10] Larsson F, Andersson P, Blomqvist P, Lorén A, Mellander BE. Characteristics of lithium-ion batteries during fire tests. J Power Sources 2014;271:414–20. https:// [11] Larsson F, Bertilsson S, Furlani M, Albinsson I, Mellander BE. Gas explosions and thermal runaways during external heating abuse of commercial lithium-ion graphite-LiCoO2 cells at different levels of ageing. J Power Sources 2018;373:220–31. [12] Liu X, Stoliarov SI, Denlinger M, Masias A, Snyder K. Comprehensive calorimetry of the thermally-induced failure of a lithium ion battery. J Power Sources 2015;280:516–25. [13] Feng XN, Sun J, Ouyang MG, Wang F, He XM, Lu LG, Peng H. Characterization of penetration induced thermal runaway propagation process within a large format lithium ion battery module. J Power Sources 2015;275:261–73. 1016/j.jpowsour.2014.11.017. [14] Lamb J, Orendorff CJ, Steele LAM, Spangler SW. Failure propagation in multi-cell lithium ion batteries. J Power Sources 2015;283:517–23. j.jpowsour.2014.10.081. [15] Rodrigues MF, Babu G, Gullapalli H, Kalaga K, Sayed FN, Kato K, Joyner J, Ajayan PM. A materials perspective on Li-ion batteries at extreme temperatures. Nat Energy 2017;2:17108. [16] Zhang SS. A review on electrolyte additives for lithium-ion batteries. J Power Sources 2006;162:1379–94. [17] Wang QS, Sun JH, Chu GQ. Lithium Ion Battery Fire and Explosion. Fire Safety Science 2005;8:375–82. [18] Koch S, Fill A, Birke KP. Comprehensive gas analysis on large scale automotive lithium-ion cells in thermal runaway. J Power Sources 2018;398:106–12. https:// [19] Wu T, Chen HD, Wang QS, Sun JH. Comparison analysis on the thermal runaway of lithium-ion battery under two heating modes. J Hazard Mater 2018;344:733–41. [20] Jiang J, Dahn JR. ARC studies of the thermal stability of three different cathode materials: LiCoO2; Li[Ni0.1Co0.8Mn0.1]O2; and LiFePO4, in LiPF6 and LiBoB EC/ DEC electrolytes. Electrochem Commun 2004;6:39–43. elecom.2003.10.011. [21] MacNeil DD, Dahn JR. The reaction of charged cathodes with nonaqueous solvents and electrolytes. J Electrochem Soc 2001;148:A1205–10. 1.1407245. [22] Hammami A, Raymond N, Armand M. Lithium-ion batteries: runaway risk of forming toxic compounds. Nature 2003;424:635–6. 424635b. [23] Liu HQ, Wei ZB, He WD, Zhao JY. Thermal issues about Li-ion batteries and recent progress in battery thermal management systems: A review. Energ Convers Manage 2017;150:304–30. [24] Saw LH, Poon HM, Thiam HS, Cai Z, Chong WT, Pambudi NA, King YJ. Novel thermal management system using mist cooling for lithium-ion battery packs. Appl Energ 2018;223:146–58. [25] Xu J, Lan C, Qiao Y, Ma Y. Prevent thermal runaway of lithium-ion batteries with minichannel cooling. Appl Therm Eng 2017;110:883–90. j.applthermaleng.2016.08.151.