2025 International Conference on Sensing, Measurement & Data Analytics in the era of Artificial Intelligence (ICSMD 2025)

Performance degradation in lithium-ion batteries (LIBs) poses significant challenges to the safety and reliability of energy storage systems. Although many promising methods have been proposed to predict LIB capacity fade, most rely on directly measured capacity values—a requirement that is often impractical in real-world applications. To address this limitation, we propose a Transformer-based multivariate time series (MTS) method for lithium-ion battery capacity prediction. Our approach extracts degradation features from partial charging curves to construct MTS inputs, then leverages Transformer networks to model temporal patterns, from which capacity fade is further predicted. Extensive validation on two large-scale datasets demonstrates superior predictive accuracy against several benchmarks.