Calendar aging is an important part of lithium-ion battery aging research. In response to the problem that the aging history of a battery cell, whose State-of-Health (SOH) needs to be estimated, may be not available, this paper proposes a SOH estimation model not relying on calendric aging conditions such as storage State-of-Charge (SOC) and storage temperature. The aging mechanisms of lithium-ion batteries in different calendric aging conditions are analyzed to investigate the influences of different aging conditions on battery internal behaviors. The neural network is used to build the SOH estimation model. To prove that the model accuracy is not affected by battery aging history, SOH indicators of cells aged at different conditions are set as training data set and testing data set respectively, and trained SOH estimation accuracy and tested SOH estimation accuracy are compared. The comparison shows that increments of mean absolute error (MAE) of SOH estimation introduced by the aging condition difference between trained data and tested data are less than 2 %. Using SOH indicators obtained at different SOC levels as inputs of the model also hardly reduce the model accuracy. The increase of MAE of SOH estimation because of the SOC difference between trained data and tested data are less than 1.5 %.
|Tidsskrift||Journal of Energy Storage|
|Status||E-pub ahead of print - 2023|