Koopman Mode-Based Detection of Internal Short Circuits in Lithium-ion Battery Pack

Sanchita Ghosh, Soumyoraj Mallick, Tanushree Roy · 2024

Read Koopman Mode-Based Detection of Internal Short Circuits in Lithium-ion Battery Pack by Sanchita Ghosh, Soumyoraj Mallick, Tanushree Roy (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-13115.txt.

Introduction

Sanchita Ghosh and co-authors build a fault finder for battery packs in 2024. An internal short circuit inside one module can spark a runaway fire if missed. Models of aging cells are rough, and real fault data are scarce, which blocks both classic and data driven detectors. The team offers a model free method based on Koopman modes.

Their method runs two mode building schemes side by side on module voltage data. It flags shifts in the mode statistics when a short begins. Tests target the faulty module inside a pack, not just the pack as a whole. Early, exact warnings can keep packs safer and extend their working life.

Catalog record

CREATORSanchita Ghosh, Soumyoraj Mallick, Tanushree Roy
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE29 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-13115

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Koopman Mode-Based Detection of Internal Short Circuits in Lithium-ion Battery Pack Sanchita Ghosh, Soumyoraj Mallick, Tanushree Roy Abstract Monitoring of internal short circuit (ISC) in Lithium-ion battery packs is imperative to safe operations, optimal performance, and extension of pack life. Since ISC in one of the modules inside a battery pack can eventually lead to thermal runaway, it is crucial to detect its early onset. However, the inaccuracy and aging variability of battery models and the unavailability of adequate ISC datasets pose several challenges for both model-based and data-driven approaches. Thus, in this paper, we proposed a model-free Koopman Mode-based module-level ISC detection algorithm for battery packs. The algorithm adopts two parallel Koopman mode generation schemes with the Arnoldi algorithm to capture the Kullback-Leibler divergence-based distributional deviations in Koopman mode statistics in the presence of ISC. Our proposed algorithm utilizes module-level voltage measurements to accurately identify the shorted battery module of the pack without using specific battery models or pre-training with historical battery data. Furthermore, we presented two case studies on shorted battery module detection under both resting and charging conditions. The simulation results illustrated the sensitivity of the proposed algorithm toward ISC and the robustness against measurement noise. 1 Introduction Short circuits are the primary cause behind delayed fire in damaged battery packs or thermal runaway of Li-ion battery systems [ 1 ] . In particular, ISCs in Li-ion batteries may occur due to several reasons which include dendrite growth, separator tearing, or electrolyte leakage inside a cell which in turn can lead to internal temperature rise followed by voltage drop within that cell [ 2 , 3 ] . Timely detection of ISC is thus essential for the safe operation of batteries. This goal has motivated active research in the area of ISC detection and estimation and this research can be categorized into model-based approaches and data driven approaches. Model based detection of ISCs have been achieved by estimation of model parameters such as short circuit resistance and current by incorporating the equivalent circuit model (ECM) [ 4 , 5 ] . Short circuit resistance has typically been estimated with open circuit voltage (OCV) as a function of state of charge (SOC) and internal resistance parallelly obtaining the voltage drop of the model [ 6 , 7 ] . (i) Model-based approaches: Specifically, techniques such as recursive least square (RLS) algorithm [ 8 ] , ellipsoidal observer in conjunction with Cholesky factorization [ 9 ] , remaining charging capacity based prediction [ 10 ] , H_{\infty} nonlinear observer [ 11 ] , shorting current estimation using Lyapunov based observer [ 12 ] have been proposed. Also, such ISCs can be estimated at multi-scale level by utilizing a number of multi-scale short circuit resistance estimation methods [ 13 ] . Additionally, the effects of diffusion voltage, additional decrease in the OCV induced by the self-discharge current in a faulty cell, has been studied with the short circuit characteristics [ 14 ] . In [ 15 ] , authors proposed a joint Moving Horizon Estimation (MHE) approach to simultaneously estimate the short circuit current and the battery capacity and illustrated validation with experimental data. However, these techniques fail to fully address battery aging, model uncertainties, and parameter variations from cell to cell. As a result, the estimated parameters can be inaccurate, leading to the unreliable detection of ISC using ECM model-based techniques [ 16 , 17 ] . Hence, it is necessary to look for algorithms that are robust to these uncertainties and variations to estimate the ISCs in battery packs. To address this, a number of researchers have preferred using Kalman Filter (KF) based approaches. It can be observed that a variety of RLS based methods have been proposed with variant forgetting have been proposed to compensate the effect of noise-corruptive data acquisition in order to estimate short circuit resistance [ 18 ] . Similarly, incorporation of dual extended KF on separated time scales improved estimation accuracy by mitigating noise while estimating short circuit current [ 19 ] , addition of the RLS with cell difference model through EKF helped to identify the extra depleting current and parallely diagnosing the short circuit [ 20 ] , incorporation of EKF and coulomb counting method has been used to estimate SOC and to detect short circuit by computing the difference [ 21 ] . Nevertheless, KF algorithms rely primarily on linear approximation of state or measurement equations and often underestimate system uncertainties. These errors lead to low detection or ISC estimation accuracy and larger convergence time [ 15 ] . Therefore, the model-based approaches fail to simultaneously address the robust performances against uncertainty or system variability and improved computational efficiency. (ii) Data-driven approaches: Several attempts have been made to determine ISC through data-driven approaches. For instance, extreme learning machine-based thermal (ELMT) model was explored to study battery temperature fluctuations under short circuit conditions [ 22 ] , Pearson Correlation Coefficient (PCC) has been introduced to detect ISC for real-time applications [ 23 ] , Principal Component Analysis (PCA) has been incorporated to detect ISC from over-discharged battery pack [ 24 ] , and random forest classification has been used for short circuit detection [ 25 ] . Moreover, dynamic time warping has been implemented for accurate detection and location of ISC cells [ 26 ] . Since performing experiments on Li-ion batteries to generate ISC data is resource-intensive, finding open-source and reliable battery datasets for battery short circuits can be challenging. This implies that the available sources are often inadequate for training data-driven models, frequently leading to inaccurate results [ 27 ] . To address these research gaps, our contribution in this paper lies in proposing a Koopman mode-based ISC detection algorithm to identify the shorted battery modules in a Li-ion battery pack. The algorithm adopts an online framework that requires only the module voltage measurements and trains online with limited data without any prior knowledge of the battery model. Thus, the algorithm is inherently generalizable and can be readily implemented for different battery packs with diverse physio-chemical characteristics and various operating conditions. Additionally, the algorithm is robust against system uncertainties and measurement noise, while being sensitive towards ISC. Finally, we present two simulation case studies for a battery pack with ISC under resting and charging conditions, to illustrate the efficacy of our proposed algorithm. The rest of the paper is organized as follows: Section II introduces the preliminaries on Koopman Operator theory and the Arnoldi algorithm. In Section III, we presented the ISC detection methodology and the detection algorithm. Section IV presents simulation results, demonstrating the successful detection of ISC in resting and charging batteries. Finally, Section V concludes the paper and outlines future research directions. 2 Preliminaries In this section, we presented a brief review of the Koopman operator and the utilization of Arnoldi algorithm to approximate Koopman modes. 2.1 Koopman operator theory The Koopman operator (KO) is an infinite-dimensional linear operator that can advance observable functions of the states of a finite-dimensional non-linear dynamical system [ 28 ] . To describe the details of the KO, let us first consider the nonlinear dynamics of a Li-ion battery pack defined as: \displaystyle x_{k+1} \displaystyle=f(x_{k});\qquad y_{k}=h(x_{k}), (1) where x_{k}\in\mathbb{R}^{d} is

Read the complete text