In May 2026, the TRANS Research Group published a study on data-efficient remaining useful life (RUL) prediction for electric vehicle batteries using real-world operational data. This work proposes a contrastive-enhanced VAE-LSTM self-supervised learning framework that learns degradation-aware representations from large-scale unlabeled charging data. Validated on three real-world EV datasets covering 340 vehicles, the proposed method achieves accurate and robust RUL prediction across different fleets. Using labels from only 30% of vehicles, the framework maintains strong cross-fleet transferability while reducing RUL labeling costs by about 70%, offering a scalable and economically practical pathway for fleet-level battery health management.
CITATION: Lv, Z., Ou, S., Jing, H., Wu, G., & Shi, D. (2026). Self-supervised learning for electric vehicle battery remaining useful life prediction using real-world unlabeled data. Energy, 141302. https://doi.org/10.1016/j.energy.2026.141302