Thermal Runaway Modeling and Calibration of an LFP Battery Cell

This post presents an example of the Thermal Runaway Modeling and Calibration of an LFP Battery Cell using the ARC device, the HWS test protocol and Simcenter Amesim.

An abuse test is the most direct way to challenge the thermal stability limits of a Li-ion cell and characterize the thermal runaway phenomena. The Accelerating Rate Calorimeter (ARC) test is considered as one of the most important abuse tests. The ARC makes it possible to simulate the worst-case scenario thanks to its pseudo-adiabatic condition. This condition is required to calibrate the thermal runaway model. According to the literature, ARC tests are typically operated using the “Heat Wait and Search” (HWS) test protocol. This document presents an example of the thermal runaway calibration of an Lithium Iron Phosphate (LFP) battery cell using the ARC device and the HWS test protocol.

ARC Device

HEL ARC rig

The ARC device is the HEL BTC-500 (Battery Testing Calorimeter) [1]. The enclosure of this device has a cylindrical shape (35 cm diameter and 32.5 cm height). Heating is ensured on all sides of the enclosure (top, side, bottom). The BTC enclosure also has several holes for the various connectors of the tested cell, such as the temperature and pressure sensors. The BTC has a calibration process required for each test (depending on the external ambient temperature, cell shape and dimension). The objective of this calibration process is to determine the power input from the BTC control system required to compensate the heat exchange between the sample and its environment inside the device. Hence, the pseudo-adiabatic condition within the device is ensured.

HWS Test Protocol

The HWS test protocol is summarized (Figure 2):

  • The system is firstly stabilized at a certain initial temperature (35 °C for 90 minutes in our case for the LFP cell).
  • The cell is heated up in steps of 5 °C.
  • Once the temperature step target is reached, the ARC enters a wait period in which the system maintains the adiabatic condition. The duration of the wait period depends on the cell weight. It is 30 min in our case.
  • After the wait period, a search period (around 15 min depending on the cell chemistry reactivity) starts. During the search period, the system calculates the cell temperature change rate (dT/dt). If the cell temperature change rate is not significant (less than 0.03 °C/min), the HWS loop will be resumed as shown in Figure 2.
  • Upon detection of a significant cell self-heating (dT/dt more than 0.03 °C), an exothermic reaction onset is considered. The ARC changes into the exothermic tracking mode, where it follows the cell temperature increase. At this point, the test will come back to the heat period of the HWS loop if the cell temperature change rate (dT/dt) is less than 0.01 °C/min.
  • If the cell temperature is over 450 °C which is the maximum operating temperature fixed by the ARC manufacturer, the exothermic tracking mode continues until the end of thermal runaway reactions.
HWS test protocol
Figure 2: HWS Test Protocol

Test Setup

ARC test setup

The tested cell positioned and held mechanically in the centre of the BTC enclosure. Two thermocouples are positioned on either side of the cell surface. The cell is surrounded by a heater with a resistance of 11Ω held to the surface of the cell by a sticky aluminum film.

The most important requirements of the HWS test in the ARC for model calibration are:

  1. Adiabatic condition should be ensured during the whole test (the calibration process is mandatory).
  2. Important data need to be recorded, especially heater power (input for the thermal runaway model).
  3. The sensibility of exothermic detection should be respected (0.03 °C/min as the limit).
  4. The sample should be heated with small steps of temperature change (typically 5 °C).

During ARC test the parameters recorded are:

  1. ARC power input.
  2. Cell surface temperature.
  3. Cell voltage.
  4. ARC chamber pressure.
  5. Ambient temperature inside the ARC chamber.

These data were used to calibrate the thermal runaway model as presented in the next paragraph.

Calibration example of the LFP Cell

Initial thermal runaway model settings

Since the tested cell is an LFP-C cell with LiPF6 diluted in organic solvent as electrolyte,you need to use the following typical parameter values to get a first estimation of the thermal runaway:

  • values of LFP for the positive electrode,
  • values of graphite for the negative electrode,
  • values of LiPF6 diluted in organic solvent for the electrolyte.

Specific Heat

For the first simulation, the following settings have been done:

  • The initial specific heat (Cp) was set to 1000 J/kg.K
  • All the thermochemical reactions (SEI, negative, positive, electrolyte and self-discharge) were deactivated by putting the frequency factor respectively equal to zero Ai = 0.

The simulation result is shown in Figure 4 (a). You can see an overestimation of the temperature before 30000 s during which there is not any thermal runaway reaction yet. Thanks to the Optimization in Study parameters, the Cp value has been adjusted to 1162 J/kg.K to fit the experiment data as shown in Figure 4 (b).

thermal runaway model
a: Before Cp parameter fit
model of thermal runaway figure 4b
b: After Cp parameter fit

Figure 4: Cp adjusted during heating steps

SEI reaction

After activating the SEI thermochemical reaction, a simulation as shown in Figure 5 has been done based on the typical parameters such as the activation energy.

( Easei, reaction factor Asei, reaction heat released hsei). The simulation result shows the reaction onset occurred sooner (around 15000s) than expected (the SEI thermochemical reaction should happen after 30000s and thus at a higher temperature).

figure 5a
a: Before SEI reaction parameters fit
thermal runaway model
b: SEI fraction before and after parametes fit

Figure 5: Simulation with typical values of SEI thermochemical reaction parameters

Following steps have been done to adjust the SEI reaction parameters:

  • Firstly, to increase the onset temperature of the SEI reaction, the activation energy parameter value should be increased. Thus, a multiplication factor equal to 1.4313 has been applied to the typical value (2.24e-19 J) of the activation energy. The simulation result with the new activation energy value (2.24e-19*1.4313 J) is shown in Figure 6a.
  • Secondly, to accelerate the SEI reaction rate the frequency factor has been increased by a 1e7 multiplication factor (see Figure 6b).
  • Finally, to match better the experimental cell temperature, the specific enthalpy was adjusted (see Figure 6c).

Figure 6d shows a comparison of the SEI fraction after the parameter fits. At this stage, there is still some difference between the simulated and experimental temperatures. For example, the simulated temperature is lower than the experimental one after 30000s in Figure 6c. This difference is reserved for the negative electrode reaction which occurs usually at the same time as the SEI reaction.

thermal runaway model
a: Easei parameter fitted
thermal runaway model
c: hsei parameter fitted
thermal runaway model
b: Asei parameter fitted
thermal runaway model
d: SEI fraction before and after parameter fits

Figure 6: SEI thermochemical reaction adjustments

Negative Electrode Reaction

Once the SEI reaction parameters have been adjusted, the negative electrode reaction was activated. The simulation in Figure 7a based on the typical parameters of the graphite electrode shows a good agreement with the onset of the reaction. This means that the energy activation typical value should be kept. However, the estimated temperature is slightly lower than the experimental one. The reaction factor has then been adjusted by a multiplication factor equal to 1.5 (see Figure 7b) to increase the reaction rate evolution (see Figure 7c).

thermal runaway model
a: Before Negative reaction parameter fit
thermal runaway model
c: Negative fraction before and after parameter fit
thermal runaway model
b: After Negative reaction parameter fit

Figure 7: Negative electrode reaction adjustments

Positive Electrode Reaction

After activating the positive electrode reaction, the simulation results with the typical values (Figure 8a and c) show that the value of the energy activation is not correct. The simulated reaction onset does not match the measured one. Thus, we started by adjusting this parameter using a multiplicative factor equal to 0.9447 obtained after several iterations. The activation energy is a sensitive parameter. Take care to use a small variation between two iterations.
Then, to match the experimental cell temperature, the heat released during the positive electrode reaction has been reduced with a multiplicative factor equal to 0.6 for the specific enthalpy.

thermal runaway model
a: Before positive reaction parameter fit
thermal runaway model
c: Positive fraction before and after parameter fit
thermal runaway model
b: After positive reaction parameter fit

Figure 8: Positive thermochemical reaction adjustments

Venting

According to the experimental data in Figure 9, the cell venting took place at around 137 °C, where a low temperature drop can be observed. To fit the cell venting, three groups of parameters should be adjusted:

  • Vent area and cell head space volume: these parameters are related to the shape and design of the cell.
  • Gas amount formed by each reaction: data measured during abuse test gas analysis or taken from the literature. In the case of the LFP cell, the values of the parameters related to the gas release are taken from the literature.
  • Burst pressure: once the 2 previous group parameters have been fixed, the burst pressure should be fitted until the simulation result matches with the experimental cell temperature drop.
thermal runaway model
a: Before venting parameter fit
thermal runaway model
b: After venting parameter fit

Figure 9: Venting adjustments

Self-Discharge Reaction

According to the measured voltage in Figure 10, the self-discharge took place before the electrolyte reaction. Thus, it should be activated before the electrolyte parameter fitting. With the typical value of the kinetic related to the internal short circuit event, the simulation result in Figure 10a shows no change on the cell voltage estimation. To fit the self-discharge, two parameters have been adjusted:

  • The energy activation parameter has been adjusted at first. As for the previous fits, a multiplicative factor of 1.09123 has been chosen after different iterations.
  • In addition, the reaction factor has been increased by a multiplicative factor equal to 15000 to match the voltage drop (see Figure 10b).
thermal runaway model
a: Before self-discharge reaction parameter fit
thermal runaway model
b: After self-discharge reaction parameter fit

Figure 10: Self-discharge reaction adjustments

Electrolyte Reaction

Finally, the electrolyte reaction has been activated. As for the positive reaction, typical value of energy activation is set for the first simulation. As shown in Figure 11a and 11c, the simulation results are not in good agreement with the reaction onset observed at lower temperature values compared to the experimental ones. The same fitting process applied for previous reactions has been used:

  • Firstly, the activation energy has been reduced by a multiplicative factor equal to 0.998.
  • Then, the heat released value has been increased by a multiplicative factor equal to 4.3 to match the maximum cell temperature recorded during the experimental test (see Figure 11b).
thermal runaway model
a: Before electrolyte reaction parameter fit
thermal runaway model
c: Electrolyte fraction before and after parameter fit
thermal runaway model
b: After electrolyte reaction parameter fit

Figure 11: Electrolyte reaction adjustments

This post presented an example of the Thermal Runaway Modeling and Calibration of an LFP Battery Cell using the ARC device, the HWS test protocol and Simcenter Amesim.

References

  1. HEL BTC-500 (Battery Testing Calorimeter)
  2. S. Abada, Compréhension et modélisation de l’emballement thermique de batteries Li-ion neuves et vieillies, PhD thesis, Université Pierre et Marie Curie, 2016.
  3. T.T.D. Nguyen, S.Abada, A.Lecocq, J.Bernard, M. Petit, G. Marlair, S. Grugeon, S. Laruelle, Understanding theThermal Runaway of Ni-Rich Lithium-Ion Batteries. World Electr. Veh. J. 2019, 10, 79. https://doi.org/10.3390/wevj10040079

Leave a Comment