Machine Learning and Data-Driven Analysis of Thermal Runaway Characteristics in Lithium-Ion Batteries

Abstract

This study explores thermal runaway in lithium-ion batteries, particularly examining NCM (Nickel Cobalt Manganese) and NCA (Nickel Cobalt Aluminum) chemistries. Utilizing data analysis and machine learning on approximately 400 data points, it gives insights into thermal runaway dynamics, focusing on characteristic parameters such as onset temperature of self-heating (T1), onset temperature of thermal runaway (T2), maximum temperature during thermal runaway (T3) and mass loss. The investigation revealed that NCA cells are more prone to thermal runaway, exhibiting lower initial self-heating temperatures compared to NCM cells. A notable preliminary finding is the potential link between nickel content in battery chemistries and thermal runaway initiation temperatures. Higher nickel compositions, like in NCM811 and various NCA cells, tend to display lower initial self-heating temperatures, possibly indicating faster progression toward thermal runaway. The limited research on how nickel content specifically influences the onset of self-heating during thermal runaway in battery cells underscores the need for new investigations into the cathode’s role and the factors beyond SEI layer decomposition. Addressing this gap, particularly focusing on the impact of nickel content on the critical onset temperature of exothermic heating that initiates thermal runaway, is essential to deepen our understanding of thermal dynamics and improve battery safety and stability.