This two-part talk will examine two problems in the battery management systems domain. The first problem is the optimization of battery test trajectories for parameter identifiability. There is an extensive existing literature showing that electrochemical batteries suffer from poor identifiability, and proposing different trajectory optimization algorithms for improving identifiability. The ideas in this literature have been validated experimentally for multiple battery chemistries, including both the lithium-ion and lithium-sulfur chemistries. This talk pushes this topic further by examining an interesting dilemma: when battery parameter identifiability is poor, it is difficult to quantify, and therefore difficult to optimize, especially for machine learning-based battery models. The talk will examine a new approach for addressing this challenge. The second problem is ensuring battery safety, especially in the context of cascading thermal runaway events. Control barrier function theory provides a rigorous foundation that can be used for battery temperature control, with the goal of preventing thermal runaway. However, once thermal runaway ensues in a given battery cell, the substantial resulting heat generation makes it futile to attempt to stop it in that specific cell. This creates a need for battery pack safety control algorithms that gracefully degrade from focusing on thermal runaway prevention to focusing on stopping thermal runaway propagation once it ensues. The talk will examine a novel mathematical approach for achieving such graceful degradation in the context of barrier function-based battery thermal safety control.

Biography: Hosam K. Fathy is currently a Mechanical Engineering at The University of Maryland. His research focuses on the optimal control and estimation of both healthcare and energy systems. This includes electrochemical battery control, with a focus on both lithium-ion and lithium-sulfur batteries.

Estimation of battery parameters are important for ensuring the fidelity of battery models and the efficacy of model-based battery management. Estimation typically involves 3 basic elements, i.e. model, algorithm, and data (e.g. current, voltage, and temperature measurements). We have been interested in exploring how the structure of measurement data affects the accuracy of estimation. In a series of works, we first derive the general form for the estimation error under the common least squares formulation, which reveals fundamentally how errors are induced by system uncertainties, including parameter mismatch, random measurement noises, and unmodeled dynamics. The results give insights and interpretation on the criteria for evaluating the quality of data, including the popular Fisher information, and also inspire new ones that relate to the propagation of uncertainties. We then proceed to explore the optimization of data to improve the accuracy of estimation, including design of input excitation to generate data when there is input control authority, and selection of data from existing database or stream when lack thereof. The efforts lead to the discovery of unique optimal current excitation patterns for estimating different battery electrochemical parameters. It is also found that the input optimization problem suffers from an intrinsic challenge caused by parameter uncertainty. Different approaches have been pursued to address the issue, including reinforcement learning, and a recently proposed nondimensionalization-based reformulation technique.

Biography: Xinfan Lin is currently an Associate Professor with the Department of Mechanical and Aerospace Engineering at the University of California, Davis, since 2017. He received his B.S. and M.S. degrees in Automotive Engineering from Tsinghua University, Beijing, China in 2007 and 2009, and Ph.D. in Mechanical Engineering from University of Michigan in 2014. Prior to his appointment at UC Davis, he was a research engineer at the Ford Motor Company from 2014 to 2016. His research interests include dynamic systems modeling, estimation, and control, data analytics, and machine learning with applications in energy, automotive, and aerospace systems. He is a recipient of the NSF CAREER Award (2021), LG Global Innovation Award (2019), and LG Battery Innovation Award (2017). His research has been funded by NSF, Office of Naval Research (ONR), NASA, California Climate Action Initiative, and industry. He has also been serving in different positions in the ASME Energy Systems Technical Committee (ESTC) since 2018, including Secretary, Publicity Chair, and Award Chair.

An accurate estimation of the state of health (SOH) of batteries is critical to ensuring the safe and reliable operation of electric vehicles (EVs). Feature-based machine learning methods have exhibited enormous potential for monitoring battery health status rapidly and precisely. However, the limited availability of large-scale, high-quality real-vehicle data hinders the development of the battery management system (BMS). Those real-world operational conditions may generate some extrapolation of battery characteristics obtained from lab testing, thereby generating significant errors in real-world BMS algorithms. We first summarized and analyzed various individual health indicators with mechanism-related interpretations, which provide insightful guidance on how these features relate to battery degradation modes. Moreover, we analyzed the operational data of 300 EVs over three years to understand the disparities between real-vehicle data and laboratory battery test data and their effect on SOH estimation. In addition, the trade-off between the performance and practicality of fusion strategies could be well balanced by tracking the latest driving behavior statistics. Lastly, we proposed a multi-modal deep learning framework to accurately predict the SOH of batteries in EVs by leveraging real-world data.

Biography: Dr. Ziyou Song is an Assistant Professor in the Department of Mechanical Engineering at the National University of Singapore (NUS). He earned his bachelor’s degree with honors and Ph.D. degree with the highest honors in Automotive Engineering from Tsinghua University, China, in 2011 and 2016, respectively. After graduation, he served as a Research Scientist at Tsinghua University from 2016 to 2017, and a Research Fellow at the University of Michigan, Ann Arbor, from 2017 to 2019, where he was also an Assistant Research Scientist and Lecturer from 2019 to 2020. Before his tenure at NUS, Dr. Song worked as a Battery Algorithm Engineer at Apple, Cupertino, US. In this role, he was responsible for overseeing battery management systems for audio products such as AirPods Pro 2. Dr. Song's research focuses on modeling, estimation, optimization, and control of energy storage systems, especially for the electrified transportation sector. Dr. Song has received several paper awards, including Automotive Innovation Best Paper Award, Applied Energy Highly Cited Paper Award, Applied Energy Award for Most Cited Energy Article from China, NSK Outstanding Paper Award of Mechanical Engineering, and IEEE VPPC Best Student Paper Award. Dr. Song serves as an Associate Editor and Editorial Member for IEEE Transactions on Transportation Electrification, IEEE Transactions on Intelligent Vehicles, Applied Energy, and eTransportation, among others.

The soon-to-be-regulated indicators of battery state of health (SOH) based on capacity fade and resistance increase are insufficient for predicting Remaining Useful Life (RUL) in second-life applications¹. The electrochemical community blames the path dependency of the battery degradation mechanisms for our inability to forecast the degradation. The control community knows that the path dependency is addressed by full state estimation. To this end, we define the deepSOH states that capture the individual contributions of all the common degradation mechanisms, namely, SEI, plating, and mechanical fracture, to the loss of lithium inventory². We show that even the electrode-specific SOH (eSOH), we defined in 2019³ is not enough to fully define the degradation states. We demonstrate this shortcoming by simulating infinite possible degradation trajectories and remaining useful lives (RUL) from a unique eSOH using a recently developed parameterizable and predictive models physics-based model of battery lifetime⁴. We then show preliminary results of deepSOH estimation and RUL prediction with explainable and actionable information on the origins of degradation for broad new use patterns.
1. A. Weng et al. 2023, "Battery passports for promoting EV resale and repurposing." Joule 7 837-842
2. H. Movahedi et al 2024, “The Case for DeepSOH: Addressing Path Dependency for Remaining Useful Life” MECC
3. P. Mohtat et al. 2019, Towards better estimability of electrode-specific state of health: Decoding the cell expansion. J. of Power Sources, 427, 101–111
4. S. Pannala et al. 2024, “Consistently Tuned Battery Lifetime Predictive Model of Capacity Loss, Resistance Increase and Irreversible Thickness Growth” J. Electrochem. Soc. 171 010532

Biography: Dr. Anna G. Stefanopoulou, the William Clay Ford Professor of Technology at the University of Michigan, has served as the Director of the Automotive Research Center, a multi-university U.S. Army Center of Excellence, and the Michigan Energy Institute.
She has mentored and taught a generation of engineers in control of advanced powertrains through classroom, online, and asynchronous courses. She has been an advisor of new curricula, training needs, and research in modeling, estimation, and control for engines, fuel cells, and batteries, with findings documented in a book, 21 US patents, and 400 publications.
She has been recognized by many prestigious awards and is a Fellow of the ASME, IEEE, and SAE. She has served on two US National Academy committees (2015 and 2020) formed upon request by the US Congress to report on vehicle fuel economy standards and the transition to electrification.