Prediction of the onset of thermal runaway and its thermal hazards in 18650 lithium-ion battery abused by external heating

This work developed empirical equations to predict the onset of thermal runaway (TR) and its related thermal hazards under external heating abuse. A set of Lithium Nickel Cobalt Aluminum Oxide (NCA), Lithium Cobalt Oxide (LCO) and Lithium Manganese Oxide (LMO) Li-ion batteries (LIBs) with 25–100% state of charge (SOC) was externally heated-to-fire. From the experimental results, we generated empirical equations to predict the onset of thermal runaway (TRonset) and hazards such as maximum surface temperature (Tsurf), and flame temperature (Tflame) except for LMO LIBs due to their violent explosions. To demonstrate the accuracy of predictions, we compared the results predicted with another set of test data from the same LIBs but different SOCs and evaluated the accuracy using percentage error (%Er). The developed empirical equations achieved higher prediction accuracy, with the TRonset reaching the %Er of less than 10% for both NCA and LCO LIBs. The lower SOC (20%) showed the greatest discrepancy, with the maximum %Er of 2.7 and 7.4% in LCO and NCA, respectively. A large variation of %Er in Tsurf was observed regardless of SOC due to violent explosion, and %Er reached the maximum of 15.4% and 24.0% in NCA and LCO, respectively. In Tflame, the highest %Er observed for NCA and LCO are 25.6% and 29.9%, respectively. Overall, %Er show no correlation with SOC in any of the TRonset, Tsurf and Tflame predictions. LMO as the most energy dense of the three LIB samples, many of them ruptured violently during the test. In light of the growing demand for energy dense LIBs, new empirical equations should include ruptured samples as well.