.png){kind=icon}
.png){kind=icon}
Return
Degradation-Informed Temperature Analysis for Cycle Life Prediction of LiFePO4 Batteries
.png){kind=icon}
OA
AI
J
S
J
S
S
D
DOI：10.1002/bte2.70124.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Understanding and predicting battery cycle life is crucial for the reliable operation of lithium-ion cells, particularly under fast-charging conditions. Conventional prognostic methods predominantly rely on electrochemical measurements such as capacity fade and internal resistance, which often require specialized testing protocols and limit practical applicability. In this work, we investigate temperature as a non-electrochemical and readily accessible signal that encodes degradation-related information during battery cycling. Motivated by a physically grounded hypothesis informed by density functional theory, we relate the accumulation of internal resistance to increased heat generation, which manifests as characteristic temperature behavior over cycling. Using a large open-source dataset comprising 96,700 cycles of commercial LiFePO₄/graphite cells subjected to diverse fast-charging protocols, we systematically analyze the relationships among temperature, discharge capacity, internal resistance, and cycle life. Temperature features are extracted separately from charge and discharge temperature curves, and data-driven cycle life prediction models based on a gradient boosting machine (GBM) framework are developed. Models using discharge-derived temperature features significantly outperform those based solely on charge-derived features, while combining charge and discharge features yields the highest predictive accuracy. Shapley value analysis further reveals the dominant contribution of discharge-related temperature features to model predictions. These results demonstrate that appropriately processed temperature data can provide a practical, non-electrochemical pathway for battery cycle life prediction without reliance on electrochemical diagnostics.
Keywords:
battery cycle life
data-driven understanding
machine learning
prognostic features
temperature
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
IF:
9.9
Papers:
323
Citations:
1.1K
Organization
Cited Papers