Session: 17-01-01: Research Posters

Paper Number: 143327

143327 - Real-Time Online State of Health Assessment of Lithium-Ion Battery Using Physics-Informed Machine Learning Model for Resilient Infrastructure Applications 

Lithium-ion batteries (LIBs) have been widely used as a versatile energy source in various industries, including transportation, and grid energy storage. LIBs provide high energy density, high conversion efficiency, and long usage life making them a feasible choice for energy grid storage for resilient infrastructure systems. For instance, with respect to grid energy storage, during extreme events certain regions in US may rely solely on LIBs during outage periods. However, LIBs undergo degradation over time. If degradation mechanisms occurring in LIBs are not properly assessed over their lifetime, hazardous conditions may arise, which may lead to sudden disruption in LIBs’ operation and a catastrophic event known as thermal runaway. A figure of merit known as state of health (SOH) assesses how ‘healthy’ or safe LIBs are for operation. SOH conventionally keeps track of LIBs capacity over time until it reaches typically around 80% of the initial capacity -- percentage defined as end of life (EOL). Although some degradation mechanisms in LIBs consume the lithium inventory, and therefore deplete their capacity, there is no clear consensus in industry about a clear definition of SOH. In addition, considering that LIBs are complex electrochemical systems, solely relying on capacity changes over lifetime is not an accurate method to assess a key and safety-related figure of merit. Therefore, it is crucial to develop a comprehensive framework to periodically monitor the state of health (SOH) of LIBs. Among the three established SOH estimation methods, empirical methods use mathematical models and parameter changes for forecasting. Experimental methods analyze operational data such as through reference performance tests (RPT), which may be time-consuming and therefore are not suitable for LIBs used in resilient infrastructure applications. On the other hand, data-driven methods may use  Artificial Intelligence (AI) techniques like machine learning to estmate SOH. Nevertheless, pure data-driven methods may not consider the degradation mechanisms and the electrochemistry taking place in the electrodes of LIBs during aging. In this work, we present a physics-informed machine learning SOH assessment framework, designed to assess SOH during regular operation while it unveils different degradation mechanisms taking place in LIBs’ electrodes used in grid storage applications. The developed physics-informed machine learning framework uses a duty cycle of C/4 charge rate to assess degradation mechanisms and enhance the operation and safety of LIBs. The evaluation of the degradation mechanisms is designed to be conducted via differential voltage analysis (DVA) using a physics-informed machine learning platform. In addition to SOH assessment, a cloud management system capable of providing real-time online estimations of the battery’s SOH is studied. For that, Raspberry Pi is utilized to transmit the SOH data collected to the cloud, enabling remote access. Continuous monitoring from DVA will provide continuous updates on active mass of both electrodes, slippages, and stoichiometries. Additionally, the connectivity with a cloud management system will deliver precise estimations of the battery’s SOH towards more resilient infrastructure energy storage systems. In this ongoing investigation, we have initiated the prototyping of our proposed approach. This involves the evaluation of Differential Voltage Analysis (DVA) and its integration into a Raspberry Pi, serving as a foundational step in the development of our comprehensive SOH assessment methodology. The obtained results, though preliminary, demonstrate the potential of our approach as a scalable solution. By combining machine learning with physics-based models, this study aims to contribute to the development of battery health monitoring and more resilient energy grid, by providing a more precise understanding of degradation mechanisms for safer energy storage systems. Further research and validation are underway to refine our methodology and extend its applicability across diverse lithium-ion battery technologies and usage scenarios.

Presenting Author: Eric Luiz Pereira Wichita State University

Presenting Author Biography: Eric Luiz Pereira holds a Master of Science from the Federal University of Itajuba (UNIFEI), Brazil, and currently is a PhD student in Department of Electrical and Computer Engineering department at Wichita State University, advised by Prof. Davi Soares. His research focuses on enhanced models for lithium-ion batteries.

Authors:

Eric Luiz Pereira Wichita State University
Damilola Ogun Wichita State University
Davi Soares Wichita State University