报告题目 (Title)： AI-Based Battery State Estimation and Diagnostics (基于人工智能的电池状态评估与故障诊断)
报告人 (Speaker)：Jung-Il Choi (韩国延世大学)
报告时间 (Time)：2026年5月30日（周六）9:30-10:30
报告地点 (Place)：校本部GJ303
邀请人(Inviter)：潘晓敏
主办部门：理学院数学系
报告摘要：
As electric vehicles and energy storage systems continue to scale rapidly, the need for accurate battery state estimation-such as state of charge (SOC). state of health (SOH). and remaining useful life (RUL)-along with early fault diagnosis, has become increasingly critical for battery management systems (BMS). However, conventional methods are often limited by their dependence on specific battery types and struggle to generalize across varying chemistries. Cell designs, and operating conditions. In addition, real-time deployment and field data-driven di-agnostics remain challenging. In this talk, we present four AI-based models addressing these limitations. UniBatt is a self-supervised universal backbone that learns from heterogeneous battery datasets and jointly estimates SOC, SOH, and RUL. K-MNet integrates Kalman filtering with a Mamba-based sequential model for real-time SOC/SOH estimation using multi-sensor data, DiagX adopts an event-driven neuro-symbolic approach to detect early anomalies from real driving data. Sparse2Batt is a virtual sensing model that reconstructs internal voltage and temperature distributions using sparse measurements. These models target key challenges in battery analytics, including generalization, real-time capability, interpretability, and estimation of unobserved internal states. Finally, we discuss a scalable framework that combines physics-based modeling with data-driven AI for next-generation battery diagnostics in vehicle-cloud integrated systems.
