A novel method for alleviating numerical stiffness in Li-ion thermal abuse models

ESRI

Numerical modeling of thermal runaway in Lithium-ion batteries has become a critical tool for designing safer battery systems. Significant progress has been made in developing kinetic mechanisms for decomposition reactions and including additional physics such as venting and combustion. However, the governing heat conduction equation and decomposition reaction equations become numerically stiff during thermal runaway, which limits the utility of thermal abuse models to low-dimensional formulations. The present work introduces a new solution strategy, which switches from the full, 3D transient heat conduction formulation to an adiabatic, 0D lumped body formulation only during the stiff portion of the simulation, i.e., only during thermal runaway. To test the new solver, a 3D thermal abuse model was configured to simulate an oven test of an 18650-format cell. The new solver was exercised for scenarios of varying degrees of stiffness, and the results were compared with a baseline solver using typical integration methods. For an extremely stiff scenario, computation speed was increased by a factor of 183x relative to the baseline solver, with little impact on solution accuracy, thus effectively alleviating the numerical stiffness issue. The new solution strategy addresses the poor scalability of high-dimensional models, such as 3D-CFD-based thermal abuse models, and improves their practicality for industrial use.