Date of Award
8-16-2024
Document Type
Open Access Dissertation
Department
Electrical Engineering
First Advisor
Bin Zhang
Abstract
State-of-charge (SOC) and remaining-dischargeable-time (RDT) are critical states of Lithium-ion batteries (LIBs). SOC indicates the available charge stored in LIBs compared to the total charge capacity of LIBs, while RDT refers to the remaining dischargeable time that LIBs can support under given loading profiles. Accurate SOC estimation and RDT prediction play important roles in guaranteeing the safe and efficient operation of batteries in real applications.
Taking advantage of high fidelity of simplified first principle (SFP) model of LIBs, this research aims to accurately estimate SOC and predict RDT. Firstly, this research establishes an SFP model to describe the battery internal physical-electrochemical processes. Secondly, a thermal resistance model is added to the SFP model to describe heat generation and transfer behaviors. Thirdly, the established SFP model and thermal resistance model are validated under galvanostatic and dynamic charge/discharge conditions at different rates. Fourth, an extended Kalman filter (EKF) is integrated with the established SFP model to estimate SOC and predict RDT with and without considering thermal impact. The consideration of comprehensive physical-electrochemical processes inside LIBs enables the proposed SFP model to describe the micro-to-macro behaviors of LIBs accurately, which leads to solid results of SOC estimation and RDT prediction.
One limitation of SFP model in SOC estimation and RDT prediction is that SFP model requires large computation. Moreover, traditional methods for SOC estimation and RDT prediction are developed in Riemann-sampling (RS)-based framework in which the algorithms are executed at a given sampling rate. As a result, SFP model-based SOC estimation and RDT prediction face challenges in real-time battery management. This hinders the applications of SFP model-based SOC estimation and RDT prediction.
To overcome this bottleneck of RS-based methods, this research introduces a Lebesgue-sampling (LS) SFP model for SOC estimation and RDT prediction. LS-based methods have an event-triggered execution scheme, in which the algorithm is executed only when necessary. Considering the non-linearity of SFP model and the advantages of EKF, they are integrated in LS framework (noted as LS-EKF) for SOC estimations and RDT prediction. Compared to RS-based EKF (RS-EKF), LS-EKF greatly reduces the computation without sacrificing accuracy.
To accommodate the dynamic loading profiles and the demand for hardware, this research proposes an adaptive Lebesgue length LS-EKF (ALS-EKF), in which the execution frequency of LS-EKF closely follows the dynamic of loading profiles. To improve the accuracy and convergence, an online dichotomy method for parameter optimization is introduced. The Lebesgue length adaption and parameter optimization enable ALS-EKF to have good performance in terms of convergence speed and accuracy. It also offers robustness under galvanostatic and dynamic charge-discharge conditions at different rates. To accommodate the battery aging effects on accuracy of the SFP model and realize online battery diagnosis and prognosis, battery aging experiments are conducted, and the SFP model is periodically parameterized as the battery ages. The health features that reflect the state of health (SOH) of battery are extracted from the battery aging experiment data and used for SFP model parameters update. This effort enables the maintenance of the SFP model fidelity which benefits the SOC estimation and RDT prediction as LIBs age.
Rights
© 2024, Enhui Liu
Recommended Citation
Liu, E.(2024). State Estimation and Prediction Based on Simplified First Principle Model of Lithium-Ion Batteries in Lebesgue Sampling Framework. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/7706