A novel model-based damage detection method for lithium-ion batteries

Abstract

Lithium-ion batteries are one of the critical components of electric vehicles. Different factors, such as inconsistencies in cells and cold environments, can cause overcharge, overdischarge, and high-rate cycling at low temperatures in cells, which produce various levels of damage. The failure risk of cells continues to increase as damage accumulates. Therefore, early damage detection should be carried out to prevent the cells from further failure and even thermal runaway. A damage detection method based on the interactive multiple-model (IMM) algorithm is presented in this paper. Two first-order equivalent circuit models are applied to the IMM algorithm. The model parameters are obtained from a normal cell and a damaged cell produced under abusive operating conditions. Unscented Kalman filters based on the models above are employed to estimate the state of charge interactively. The innovations and covariances of the predicted measurement are utilized to calculate mode probabilities. The mode probabilities are compared with the threshold to determine whether a cell is damaged. The results of simulations and experiments indicate that the proposed method can effectively detect specific abuses and tolerate inconsistency in cells, thus providing an effective way of diagnosing cell damage.