AI-enhanced battery health evaluation and safety management

Accurate, practical, and robust battery health evaluation and safety management are crucial for the efficient and reliable operation of electric vehicles. However, due to the complex electrochemical mechanisms of battery systems and the absence of sensors, it is challenging to directly measure battery health state and fault. In this talk, we introduce a deep learning-based framework for health state estimation and life prediction, as well as an unsupervised method for early fault detection. Based on the data characteristics during battery operation, we have established a deep hybrid neural network for health state estimation and a sequence-to-sequence model for battery life prediction, achieving battery health management for electric vehicles. Moreover, we introduce a novel early detection method of micro-internal short circuits in the battery pack based on unsupervised learning, which achieves early detection and location of internal short circuit faults based on field data with low sampling frequency.