Thermal Runaway

Analysis > Thermal Runaway

Introduction

Thermal runaway in battery packs is a major safety concern with potentially catastrophic outcomes, such as battery pack fires. This phenomenon occurs in batteries due to exothermic degradation reactions when a battery is subjected to abusive conditions, such as physical damage, internal short circuits, overcharging, or overheating. These conditions, along with the associated decomposition of battery components (e.g., anode, cathode, and separator), lead to significant heat release and an uncontrollable rise in temperature. When a single cell enters these thermally unstable conditions, the heat release can cause adjacent cells to heat up and enter thermal runaway, eventually propagating through and consuming the entire battery pack. Additionally, the decomposition reactions generate flammable gases that can ignite, causing fires to spread through the battery pack.

The first two models are chemical kinetic models based on the Arrhenius equation. The NREL model takes a mechanistic approach to model the kinetics and heat release of individual cell components, while the arc reaction heat model adopts a more phenomenological approach, deriving the reaction kinetics directly from ARC data. The heat rate model also uses ARC data but directly reads this data to determine the volumetric heat source.

ARC Reaction Model

The ARC reaction model thermal runaway panel provides a platform to input the Arrhenius chemical kinetic parameters and reaction enthalpy to predict chemical heat generation under abuse scenarios calibrated against ARC data.

Fitting ARC data

The Calculate button opens a panel (see Figure 1) for fitting the Arrhenius and reaction enthalpy parameters without laborious linearization methods. A summary of the tabs and the required input is shown below:

Arc data tab

Stage information

The stage information panel allows users to define the number of stages and specify the initiation temperature for each stage (see below for a detailed description). The input options are shown in Figure 2.

Although an ARC test does not identify the decomposition of individual reacting components, it can still be represented by Arrhenius kinetics with physical meaning depending on the number of stages used to represent the data. For example, two stages may group together SEI and anode decomposition into the first reaction kinetic, and cathode conversion and electrolyte decomposition into the second reaction kinetic. The exact grouping, however, depends on the specific cell chemistry.

Alternatively, a four-stage model could be motivated from both physical and observed features of the data where: Stage 1 likely corresponds to the reaction between intercalated lithium in the anode and the electrolyte; Stage 2 could represent the initial decomposition of the positive electrode, such as partial decomposition. Stage 3 primarily captures the sudden rise in the self-heating rate (observational), and Stage 4 may correspond to the rapid decomposition of the positive electrode and oxygen release, leading to a significant temperature jump.

Ultimately, establishing a link to decomposition processes requires intricate knowledge of reactions occurring in a specific cell. However, top-down fitting procedures utilizing staged Arrhenius kinetics alleviate the need to associate reaction kinetics to individual components.

The fitting for the ARC data follows a sequential process in which the dataset is divided into N stages, where N is a user-defined input. Figure 3 illustrates an example of dividing the ARC data into four stages based on stage initiation temperatures, where T1, T2, T3, and T4 represent the start temperature of each stage, respectively. The actual stages are defined as: Stage 1: T2-T1; Stage 2: T3-T2; Stage 3: T4-T3; Stage 4: Tmax-T4. Tmax is the maximum temperature recorded from the ARC test as is determined automatically from the data.

The choice of N determines how each of these regions is modelled using one or multiple Arrhenius reaction equations. Based on existing literature and our experiments, the guidelines shown in Table 1 may be followed for commonly used values of N (T1, T2, T3 and T4 are as defined before):

Cell Properties

The cell properties tab, shown in Figure 5, provides the input structure for defining the different components of the cell under consideration.

Providing a detailed description of the cell improves the accuracy of the thermal runaway model calibration against ARC data, ultimately improving the model’s predictions for other experiments, such as 3D ARC tests, oven tests, and propagation tests.

Typically, a cell can be divided into a rough division of its physical components (e.g., steel can, jelly roll, tabs, etc., as shown in the image below and the corresponding physical properties of each component.

For the above-mentioned reasons, it is beneficial to gather detailed information on the division and characteristics of each component within the cell. However, if this information is unavailable, the cell can be approximated as a single component, typically the jelly roll. Furthermore, if the exact volume and densities of a component are not known, the user may approximate by setting volume=1 unit and density=mass of the component.

Optimization

The fitting process for each stage uses a swarm search algorithm, where the position of each particle represents a specific parameter set. The governing ordinary differential equations (ODEs) are integrated over time, generating predictions for the heat rate versus temperature (dT/dt vs. T) and temperature versus time (T vs. t). These predictions are then compared to the ARC data up to the end temperature for that stage.

A loss function is formulated and incorporated into the optimization algorithm to evaluate the fit. The process iterates to determine the optimal parameters for each stage and continues iteratively across all stages.

The optimization parameters can usually be left at their default values, but if modification is required are input through the optimization panel as referred below.

Output

The output for the multi-stage model is given as an array of stages and Arrhenius kinetic/material parameters. These are automatically loaded back into SimLab.

NREL Abuse Model

The mathematical model for thermal abuse involves heat generation resulting from a series of exothermic decomposition reactions of various battery components. The heat generation from these individual reactions is aggregated and incorporated into the energy equation as a combined source term.

The source term is derived from decomposition reactions occurring in the battery, which are characterized using Arrhenius equations, accounting for: 1) Solid electrolyte interface(SEI) decomposition at the anode-electrolyte interface (〖~80-100 〗^o C); 2) Anode decomposition (intercalated lithium reacting with electrolyte, facilitated by SEI decomposition); 3) Cathode conversion (active material decomposition releasing oxygen), which is highly exothermic; and 4) Electrolyte decomposition at very high temperatures. Additionally, short circuit events can also be included, e.g., separator melting leading to direct connection between the anode and cathode. The short circuit leads to charge depletion and is given as a function of state of charge (SOC). The final source term is applied to the energy equation through nodal temperature updates based on the solution of the governing Arrhenius ODEs.

Heat Rate Model

Direct reading of self-heating rate (or temperature rate) as a function of temperature from ARC or other sources is supported when the type is set to Heat Rate Model. This method applies the heat-source directly to the energy equation.