EV and Battery Seminar

Info

• Sophie Germain's Room (Alan Turing Building)
• Friday, November 28th
• From 10:00 AM

Abstract

This talk focuses on state estimation and diagnostics in lithium-ion battery (LIB) battery management systems (BMS), with an emphasis on state of charge (SoC), state of health (SoH), and state of power (SoP) for electric vehicles and stationary storage. We review modeling approaches spanning equivalent-circuit and reduced electrochemical models coupled to thermal dynamics, along with observers that blend coulomb counting and OCV–SoC mapping (with hysteresis compensation) with Kalman-filter–based estimators and moving-horizon approaches. We also discuss hybrid ML methods for adaptive parameter identification and pack-level SoC consensus and active balancing under cell dispersion and temperature gradients, validated across drive-cycle and grid-service profiles.
LIB degradation is assessed at three levels—mechanisms, modes, and metrics—and recent advancements in diagnostics and prognostics are driven by machine learning (ML), which offers a detailed multi-level perspective on LIB degradation. The pulse-injection-aided machine learning (PIAML) technique is highlighted for its ability to improve diagnostic accuracy and SoC/SoH observability and parameter identification (e.g., R0, RC, hysteresis) without requiring specialized sensors or electronics. Using low-amplitude perturbations embedded in normal operation or scheduled by the BMS, PIAML matches or surpasses existing methods such as incremental capacity and differential voltage analyses in practicality and robustness, offering a scalable approach to optimizing lithium-ion system management.