TY - JOUR
T1 - A cross-domain degradation mode prediction of lithium iron phosphate batteries for full cycle life
AU - Liu, Meng
AU - Zhang, Caiping
AU - Chen, Zeping
AU - Liu, Chenyu
AU - Zhang, Linjing
AU - Qiao, Bo
AU - Zhang, Weige
AU - Jiang, Jiuchun
AU - Yang, Xiao Guang
N1 - Publisher Copyright:
© 2026 Elsevier Ltd
PY - 2026/12
Y1 - 2026/12
N2 - Battery health assessment and energy management are essential for the lifetime reliability of electric vehicles. To address the limited understanding of internal parameter evolution and constraints on modeling accuracy in lithium‑iron-phosphate batteries, this study proposes a framework for predicting battery degradation modes. This framework integrates an enhanced electrochemical-thermal coupled mechanism model with a multilayer perceptron algorithm, effectively combining physical interpretability with data-driven learning. A coupled model that accounts for hysteresis potential compensation is first established, and key characteristic parameters highly correlated with physical mechanisms are identified through a one-at-a-time sensitivity analysis. Based on this, combined with disassembly characterization and degradation mode analysis, the evolution of the key thermodynamic and kinetic parameters throughout the battery's lifecycle is illustrated, and a mapping relationship between internal features and external states is established. Subsequently, a multilayer perceptron network with robust nonlinear mapping capabilities is established to achieve non-destructive prediction of both ten key model parameters and three degradation modes under small-sample scenarios. Based on the proposed prediction framework, the root-mean-square errors for voltage and temperature are maintained at 23.0 mV and 0.7 °C, respectively. Meanwhile, the predicted degradation trends closely match the actual trends. This study provides a novel perspective on physics- and data-fusion-based approaches for battery state monitoring and life prediction under complex operating conditions.
AB - Battery health assessment and energy management are essential for the lifetime reliability of electric vehicles. To address the limited understanding of internal parameter evolution and constraints on modeling accuracy in lithium‑iron-phosphate batteries, this study proposes a framework for predicting battery degradation modes. This framework integrates an enhanced electrochemical-thermal coupled mechanism model with a multilayer perceptron algorithm, effectively combining physical interpretability with data-driven learning. A coupled model that accounts for hysteresis potential compensation is first established, and key characteristic parameters highly correlated with physical mechanisms are identified through a one-at-a-time sensitivity analysis. Based on this, combined with disassembly characterization and degradation mode analysis, the evolution of the key thermodynamic and kinetic parameters throughout the battery's lifecycle is illustrated, and a mapping relationship between internal features and external states is established. Subsequently, a multilayer perceptron network with robust nonlinear mapping capabilities is established to achieve non-destructive prediction of both ten key model parameters and three degradation modes under small-sample scenarios. Based on the proposed prediction framework, the root-mean-square errors for voltage and temperature are maintained at 23.0 mV and 0.7 °C, respectively. Meanwhile, the predicted degradation trends closely match the actual trends. This study provides a novel perspective on physics- and data-fusion-based approaches for battery state monitoring and life prediction under complex operating conditions.
KW - Degradation mode identification
KW - Electrochemical-thermal coupled model
KW - Lithium iron phosphate battery
KW - Machine learning
KW - Parameters evolution
UR - https://www.scopus.com/pages/publications/105047274417
U2 - 10.1016/j.apenergy.2026.128655
DO - 10.1016/j.apenergy.2026.128655
M3 - Article
AN - SCOPUS:105047274417
SN - 0306-2619
VL - 426
JO - Applied Energy
JF - Applied Energy
M1 - 128655
ER -