Abstract: Internal short circuits(ISCs) of Li-ion batteries in the later stages can lead to thermal runaway and cause serious safety problems, so it is important to identify ISCs in the early or middel stage. The common way to detect ISCs is to use the consistency of voltage and other parameters between cells in the same battery pack, and abnormal cells are screened out by comparing performance of cells within a battery pack. However, for battery packs that have been aged, such as retired batteries, the pack structure is likely to be disrupted. Moreover, there is already a performance differentiation between the cells. So This method is not suitable for ageing batteries. Therefore, the polarization resistance is chosen as an independent landmark parameter for identifying ISCs in the cells, while recursive least squares with a forgetting factor is used to identify polarization resistance online to identify ISCs in the whole life cycle. To address the problem that battery ageing affects the accuracy of the identification results, the algorithm is first optimised to suit the effects of ageing on the model by selecting an appropriate forgetting factor and sampling frequency. The optimised algorithm is then used to simulate the surface temperature of the cells to further verify the accuracy of the algorithm. Finally, an ISC experiment is designed to verify the ability of the algorithm to identify ISC. The results show that when the forgetting factor is 0.95 and the data sampling interval is 1s, the algorithm has the strongest ability to adapt to aging. And that the error in battery surface temperature estimation is within 2%. The mean increase in polarization resistance for state of health(SOH) of 80% reaches 45% and above before the ISC develops to a later stage, indicating that it can effectively identify ISC.

Key words: internal short circuit, life cycle, parameter identification, least square method, temperature prediction