Probabilistic Prediction of Li-ion Battery RUL using Large Time-Series Model

Le résumé fourni par la source

Accurate prediction of lithium-ion battery capacity degradation and remaining useful life (RUL) is crucial for battery health management and the safe operation of equipment. However, the diversity of battery types and variations in usage environments pose challenges to data-driven predictive models. Traditional machine learning models often exhibit poor performance in terms of prediction and generalization capabilities. This paper introduces a time-series large model: ANVMD-Llama. The model employs Adaptive Noise Variational Mode Decomposition (ANVMD) to process battery aging data for RUL prediction. Initially, the adaptive noise variational mode decomposition optimizes the tokenization scheme of Lag-Llama, decomposing battery degradation data into multiscale modal components with distinct features to characterize degradation trends and fluctuation properties, aiding the model in understanding fluctuation patterns. Subsequently, ANVMD-Llama is pre-trained on a large dataset of diverse lithium-ion battery degradation data to learn capacity degradation patterns. The model is then fine-tuned using a small amount of data to update the top-level modules, achieving more accurate predictions. Finally, the experimental results demonstrate that the proposed model achieves accurate RUL prediction and exhibits strong transfer capability.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.