Log-Linear Model for Predicting the Lithium-ion Battery Age based on Resistance Extraction from Dynamic Aging Profiles

In this work we propose a method for extracting, modelling, and predicting the resistance of Lithium-ion batteries directly from a dynamic mission profile, which was applied to the battery over a period of 38 weeks (approximately 4600 full equivalent cycles). While the extraction mainly relied on data manipulation and bookkeeping, the modelling and subsequent prediction of the resistance used a log-linear model. It is shown that the estimated log-linear model can be used to create a posterior probability distribution of the age of the battery, given an internal resistance measurement and the SOC value at which it was measured. This distribution was used to calculate the expected age of the battery, and the expected age was compared to the value obtained form battery weekly check-ups. At an SOC of 80% a mean absolute error (MAE), between the weekly check-ups and the expected age, of 5.83 weeks (706 FEC) was achieved. Furthermore, it is shown that by introducing a decision threshold, the MAE could be reduced as far as 2.65 weeks (321 FEC). Lastly, a method is introduced for handling cases where the SOC was not known prior to the battery age prediction.