Remaining Useful Life Prediction of Li-Ion Batteries by Multi-modal Attention Fusion Network

Lithium-ion batteries are crucial for electric vehicles and smart grids, yet their performance degrades with charge-discharge cycles, leading to reduced capacity and increased resistance. Predicting the Remaining Useful Life (RUL) is essential for preventing failures and ensuring stable operation. However, estimating RUL is challenging due to its indirect measurability and the influence of various factors such as battery components, packaging processes, and operating conditions. This paper presents a Multimodal Attention Fusion Neural Network (MAFN) to forecast battery health and RUL using historical data. Trained on a diverse electrochemical dataset with various battery components and cycling conditions, the model integrates battery materials and electrical signals, capturing critical aging features for accurate RUL prediction. This approach advances rapid RUL estimation, enhancing battery management systems and supporting smart grid and carbon neutrality goals.