Abstract
Degradation prediction is crucial for ensuring safe and reliable operation of batteries. However, relying solely on capacity to characterize aging cannot comprehensively represent the health status of the battery. This work explores the potential of using a limited number of partial voltage-capacity curves to evaluate battery degradation with the aid of deep learning approaches, which can be used for onboard applications. A sequence-to-sequence model is proposed to predict the electrochemical impedance spectra during battery degradation. It only uses capacity sequences within a specific voltage range at fixed voltage increments from a limited number of cycles, which can be flexibly adapted to different life stages in an end-to-end manner. The proposed method has been validated based on the developed degradation dataset. The root mean square errors for the prediction of impedance spectra are less than 1.48 mΩ. Capacities and resistances associated with electrochemical processes can be further extracted from the obtained impedance spectra, facilitating a comprehensive evaluation of battery degradation. As a limited number of measured data are needed, the proposed method can reduce data storage requirements and computational demands, which enables fast and comprehensive aging diagnosis.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have