Data-driven diagnosis of lithium-ion battery degradation under realistic usage conditions

Placeholder Show Content


Lithium-ion batteries have become increasingly prevalent in everyday life, from mobile devices to electric vehicles. In order to swiftly and robustly deploy lithium-ion batteries at large scale in a wide range of applications, an understanding of battery degradation as a function of operating conditions is critical. This dissertation focuses on building this understanding by generating extensive battery cycling datasets and applying a data-driven diagnosis methodology to diagnose the root causes of degradation. In Chapter 1, we introduce lithium-ion batteries and their importance to the global energy landscape. I explain the fundamental internal processes behind lithium ion battery operation and highlight the degradation mechanisms, degradation modes, and performance metrics that we use to describe battery aging. In Chapter 2, we establish a data-driven degradation diagnosis framework that combines degradation inducing aging cycles with diagnostic cycles to probe fundamental degradation modes (lithium inventory, positive electrode capacity, negative electrode capacity, and resistance increase) and device performance metrics over the course of battery lifetime. We apply interpretable machine learning methods to deconvolute the effects of different input parameters on the target outputs (degradation modes and performance metrics). This framework is used to design battery cycling experiments and analyze battery cycling data. In Chapter 3, we apply this framework first to an exploratory dataset to compare the relative importances of key operating conditions on degradation modes and performance metrics. The key results from this study are that charging conditions (charging current and cutoff voltage) have the highest impact on many degradation modes and performance metrics. However, discharging current is the most important factor for a few important degradation modes, and varies widely between devices of the same type depending on the user or application. These results provide the foundation and motivation for our main work: a study on degradation as a function of realistic usage conditions. In Chapter 4, we generate a novel, extensive application-relevant dataset with diverse realistic discharge protocols. We then apply the data-driven degradation diagnosis framework to relate the effects of dynamic operating conditions to lithium-ion battery degradation modes and device performance. We first demonstrate that constant current discharging conditions are not representative of realistic use cases, and that diverse discharge profiles lead to differences in degradation. We find that higher rest states of charge predict higher resistance and shorter cycle life, and that larger values of the higher characteristic frequency predict larger resistances. Finally we reveal that under these realistic discharging conditions, cycling time appears to be more relevant than cycle number for analyzing degradation. In Chapter 5, we summarize the conclusions from all chapters of this work, focusing particularly on the insights from Chapter \ref{chap:realistic}. We also use this chapter to explore future studies that can build upon the results of this work. Proposed work includes both further battery cycling experiments and fundamental studies probing the relationships revealed by the data-driven degradation diagnostics framework. Unrelated to data-driven degradation diagnostics, my first project was investigating the use of eutectic mixtures of quinones as a high energy density redox flow battery electrolyte. In Appendix C, I'll describe some of the work I did supporting this project that are not included in the publications of this study. In Appendix D, I detail the work I did on melting point prediction for small organic redox-active molecules, quinones and hydroquinones. At the beginning of each chapter, I'll establish my specific contributions to the work being described. Additionally, given that data-driven approaches for understanding lithium-ion battery degradation have gained significant traction in recent years, I'll establish the scope of existing works (to the best of my knowledge) near the beginning of each relevant chapter to provide more context for the novelty that this work brings to the field.


Type of resource text
Form electronic resource; remote; computer; online resource
Extent 1 online resource.
Place California
Place [Stanford, California]
Publisher [Stanford University]
Copyright date 2023; ©2023
Publication date 2023; 2023
Issuance monographic
Language English


Author Ganapathi, Devi Sribala
Degree supervisor Chueh, William
Thesis advisor Chueh, William
Thesis advisor Onori, Simona
Thesis advisor Sendek, Austin
Degree committee member Onori, Simona
Degree committee member Sendek, Austin
Associated with Stanford University, School of Engineering
Associated with Stanford University, Department of Materials Science and Engineering


Genre Theses
Genre Text

Bibliographic information

Statement of responsibility Devi Ganapathi.
Note Submitted to the Department of Materials Science and Engineering.
Thesis Thesis Ph.D. Stanford University 2023.

Access conditions

© 2023 by Devi Sribala Ganapathi
This work is licensed under a Creative Commons Attribution Non Commercial Share Alike 3.0 Unported license (CC BY-NC-SA).

Also listed in

Loading usage metrics...