PHYSICS-INFORMED MACHINE LEARNING FOR BATTERY DEGRADATION DIAGNOSTICS - A COMPARISON OF STATE-OF-THE-ART METHODS

Sep 17, 2026

Uploaded by Liam

Monitoring the health of lithium-ion batteries’ internal components as they age is crucial for optimizing cell design and usage control strategies. Component-level degradation is usually quantified by destructively analyzing many cells across long aging tests, which limits scope to specific conditions and durations. Physics-informed machine learning (PIML) enables models that extrapolate long-term component degradation from short-term aging data by leveraging physical principles. The study presents four PIML approaches and compares accuracy, complexity, ease of implementation, and extrapolation capability using four years of cycle-aging data from 24 implantable-grade cells under varying temperatures and C-rates.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages46

Reading Time1 h 56 min

AccessFree to download

Abstract

Keywords

1 Introduction

arXiv :2404 .04429v 1 [ cs .CE] 5 Apr 2024 PHYSICS-INFORMED MACHINE LEARNING FOR BATTERY DEGRADATION DIAGNOSTICS: A COMPARISON OF STATE-OF-THE-ART METHODS Sina Navidi 1 , Adam Thelen2 , Tingkai Li 1 , Chao Hu 1 ,† 1Department of Mechanical Engineering, University of Connecticut, Storrs, CT, 06269, USA 2Department of Mechanical Engineering, Iowa State University, Ames, IA 50011, USA †Indicates corresponding author. Email: [chao.hu@uconn.edu](chao.hu@uconn.edu). ACKNOWLEDGEMENTS This research was partly supported by the Iowa Economic Development Authority under the Iowa Energy Center Grant No. 20-IEC-018 and partly by the US National Science Foundation (NSF) Grant No. ECCS-2015710 . Any opinions, findings, or conclusions in this paper are those of the authors and do not necessarily reflect the views of the sponsoring agencies. The authors would also like to thank Dr. Gaurav Jain and Dr. Hui Ye at Medtronic for providing the implantable-grade lithium-ion battery cells used to collect the cycling data for this study. ABSTRACT Monitoring the health of lithium-ion batteries’ internal components as they age is crucial for optimizing cell design and usage control strategies. However, quantifying component-level degradation typically involves aging many cells and destructively analyzing them throughout the aging test, limiting the scope of quantifiable degradation to the test conditions and duration. Fortunately, recent advances in physicsinformed machine learning (PIML) for modeling and predicting the battery state of health demonstrate the feasibility of building models to predict the long-term degradation of a lithium-ion battery cell’s major components using only short-term aging test data by leveraging physics. In this paper, we present four approaches for building physics-informed machine learning models and comprehensively compare them, considering accuracy, complexity, ease-of-implementation, and their ability to extrapolate to untested conditions. We delve into the details of each physicsinformed machine learning method, providing insights specific to implementing them on small battery aging datasets. Our study utilizes long-term cycle aging data from 24 implantable-grade lithium-ion cells subjected to varying temperatures and C-rates over four years. This paper aims to facilitate the selection of an appropriate physics-informed machine learning method for predicting long-term degradation in lithium-ion batteries, using short-term aging data while also providing insights about when to choose which method for general predictive purposes. Keywords: Lithium-ion battery, Degradation diagnostics, Half-cell model, Physics-informed machine learning, Comparative study Physics-Informed Machine Learning for Battery Degradation Diagnostics: A Comparison of State-of-the-Art Methods 1 Introduction Monitoring the health of lithium-ion batteries over their lifetime is important for ensuring the safety and reliability of the electric vehicles and portable electronics they power. Common battery health indicators like remaining capacity and direct-current internal resistance (DCIR) can be directly measured through complete diagnostic cycles at low rates (capacity) or hybrid pulse-power characterization (HPPC) tests (resistance) [1] . However, in online applications, battery capacity must be estimated because completing a full discharge cycle would significantly interrupt system operation. Battery capacity and resistance can be predicted online using reduced order algebraic models [2, 3, 4], state observer algorithms (like the extended Kalman filter and particle filters [5, 6, 7, 8]) coupled with physics-based battery models like equivalent circuit models [9, 10, 11] and electrochemical models [12, 13, 14], and machine learning methods [15, 16, 17, 18, 19, 20, 21] . In cases where limited data are available, machine learning models have still been demonstrated to accurately predict battery capacity and resistance using te

FAQs

Why is component-level battery degradation monitoring important?

It helps optimize cell design and usage control strategies by explaining capacity loss and resistance increase through internal component health rather than only using external indicators.

What limitation exists in traditional quantification of degradation?

It typically requires aging many cells and destructively analyzing them throughout the test, restricting quantifiable degradation to the specific conditions and test duration.

How does physics-informed machine learning address the limitation?

It leverages physics to build models that predict long-term degradation of major components using only short-term aging test data, and the paper compares four PIML approaches for accuracy and extrapolation.

Sep 17, 2026

Uploaded by Liam