Incremental Capacity Analysis-Based Impact Study of Diverse Usage Patterns on Lithium-Ion Battery Aging in Electrified Vehicles

Abstract

Aging assessment is critical for lithium-ion batteries (LIBs) as the technology of choice for energy storage in electrified vehicles (EVs). Existing research is mainly focused on either increasing modeling precision or improving algorithm efficiency, while the significance of data applied for aging assessment has been largely overlooked. Moreover, reported studies are mostly confined to a specific condition without considering the impacts of diverse usage patterns on battery aging, which is practically challenging and can greatly affect battery degradation. This paper addresses these issues through incremental capacity (IC) analysis, which can both utilize data directly available from on-board sensors and interpret degradations from a physics-based perspective. Through IC analysis, the optimal health feature (HF) and the state of charge (SOC)-based optimal data profile for battery aging assessment have been identified. Four stress factors, i.e., depth-of-discharge (DOD), charging C-rate, operating mode, and temperature, have been selected to jointly characterize diverse usage patterns. Impact analysis of different stress factors through the optimal HF with the SOC-based optimal data profile from aging campaign experiments have generated practical guidance on usage patterns to improve battery health monitoring and lifetime control strategies.Batteries 2019, 5, 59 2 of 16 fitting. The simplicity and superiority in real-time computation have given rise to their wide applications in BMS [9]. Yet the lack of physical representation makes it unable to interpret battery degradation mechanisms and ultimately suffers from insufficient estimation accuracy. In contrast, the physics-based approach is built upon electrochemical models (EChMs) to gain insights into the internal physicochemical processes related with aging. However, the high modeling precision comes at the price of demanding computation resources, given that the governing equations of EChMs are mostly partial differential equations (PDEs) [11], which are not amendable for BMS application. Despite multiple model order reduction (MOR) techniques [14] to reduce the computational complexity with various introduced approximations, the balance between modeling precision and computation efficiency still faces severe challenges.The data-driven methods [17][18][19][20] have attracted tremendous attention in LIB aging research owing to their flexibilities from the model-free feature. The independence from battery physical properties makes it feasible for the application of various machine learning methods, including support vector machines (SVM) [17], artificial neural networks (ANN) [18], relevance vector machines (RVM) [19], and particle swarm optimization [20], etc., to develop mathematical descriptions of battery degradation behaviors through statistical learning from a large amount of data. Nevertheless, data-driven methods are often subject to certain drawbacks, such as data saturation and specific input requirement. And the reliability of data-driven methods stron...

Sustainable SiC Composite Anodes, Graphite Accelerated Lithium Storage

“…In addition, these peaks generally reduce in magnitude and become broader in composites compared to pristine graphite, suggesting the SiC/HC or SiC/O with 10 wt% graphite hindered graphite from complete lithiation/delithiation, presumably due to significant resistive and diffusion polarization effects at graphite in the composite electrodes. 60,61 In contrast, the SiC/HC/ 30G and SiC/O/30G show less evident peak changes and more intense responses than graphite, suggesting less localized polarization and that SiC components probably facilitate graphite lithiation/ deliathiation processes. At present, these effects remain to be identified.…”

DiffCapAnalyzer: A Python Package for Quantitative Analysis of Total Differential Capacity Data

“…Traditionally, when using differential capacity plots, researchers have drawn conclusions based on an arbitrarily chosen subset of cycles and reported mainly qualitative claims on how peaks shift during cycling, due to the difficulties in analyzing the full amount of data produced in the differential capacity plots. Additionally, although it is known that peak shapes and areas correlate to important electrochemical events, only a few papers report using peak deconvolution as a method to interpret dQ/dV plots (Aihara et al, 2016;Bian, Liu, & Yan, 2019;He, Bian, Liu, Wei, & Yan, 2020;Huang, 2019;Torai et al, 2016). Further, there does not exist any standardized method for peak deconvolution of differential capacity plots.…”

“…In differential capacity curves without any cleaning or smoothing, there is significant noise and large step-wise changes present. This is a common problem when the denominator of Equation 1 approaches zero (Bloom et al, 2005;Huang, 2019). Therefore, in order to accurately identify peaks, the data is cleaned by removing points such that the voltage difference between datapoints is at least 0.001 V. Subsequently, the curve is smoothed using a Savitzky-Golay filter, which is a moving polynomial of specified order fit over a specified number of data points.…”

Real-time personalized health status prediction of lithium-ion batteries using deep transfer learning

“…19 Consequently, customizing personalized battery health management is an urgent requirement for end-users, manufacturers and recyclers in the field of energy technology. 20–22…”