Li-ion battery aging model robustness: An analysis using univariate and multivariate techniques

Batteries are highly flexible energy storages and they can be easily integrated in energy systems. However, the modeling of batteries must be coherent and robust to be effectively included in the energy systems; in particular, the aging phenomena are known to significantly impact the storage capacity, charging/discharging behaviors, and the state of charge. While data-driven lithium-ion battery models, which consider both cyclic and calendar aging, are widely reported in the scientific literature, their robustness in dealing with different data compared to trained ones has not been deeply investigated so far. In this study, a sensitivity analysis is performed using three aging models available in the scientific literature (e.g., Wang's, Baghdadi's, and Omar's ones) to investigate their performance under different operating conditions of both temperature and current. Results of both univariate and multivariate sensitivity analyses refer to a 160 A h battery, showing that all the three models are able to accurately predict the battery aging under different operating conditions, but some of these models are more sensitive to specific factors than others. Specifically, the multivariate analysis shows that Wang's model is the most robust one in the face of simultaneous changes in both temperature and current.