Abstract

Lithium-ion batteries (LIBs) are widely used in various electronic devices owing to their high energy density, long lifespan, and relatively low cost. The LIB research field has recently shifted toward hybrid modeling, where physics-based and data-driven approaches complement each other. In this talk, we will discuss interpretable approaches for optimizing two necessary yet time-consuming LIB applications: pulse diagnostics and the formation step. While on-the-fly diagnostic techniques enable real-time health monitoring, pulse diagnostics provide clues about "how" the battery has degraded. However, pulse diagnostic protocols are often chosen arbitrarily and can be time-consuming. In the first part of this presentation, we will show how model-based design of experiments can optimize the pulse diagnostic protocol for practical identifiability of degradation parameters and for total diagnostic time. Additionally, recent studies show that battery performance can be greatly affected by the formation protocol (i.e., the manufacturing step that forms the solid electrolyte interphase layer on the anode). However, optimizing the formation protocol is challenging due to limited mechanistic models and the long time required for batteries to reach end-of-life. In the second part of this presentation, we will introduce a systematic feature engineering framework that enables the development of an interpretable data-driven model for directly mapping measurements from the formation step to cycle life. Interpretability of the model provides insight into where to perform a mechanistic investigation, bridging physics-based and data-driven approaches. Possessing great extrapolation capability, the designed features are expected to accelerate optimization of the formation protocol by allowing immediate evaluation of new protocols during manufacturing.

Speaker Bio

photo

Jinwook Rhyu is a fifth-year PhD student in Chemical Engineering at MIT, with a minor in applied mathematics. Advised by Prof. Richard D. Braatz and Prof. Martin Z. Bazant, he conducts research on developing interpretable frameworks that bridge physics-based and data-driven approaches for optimizing lithium-ion battery applications. He was a recipient of the Kwanjeong Scholarship. Prior to MIT, he obtained his B.S. in Chemical and Biological Engineering from Seoul National University, Republic of Korea, in 2021.