.png){kind=icon}
.png){kind=icon}
Return
Prompt-Driven Fine-Tuning Large Language Model for Multistate Coestimation of Lithium-Ion Batteries
.png){kind=icon}
PRE
AI
L
T
X
Y
X
DOI：10.1109/tii.2025.3631017.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Accurate estimation of state-of-charge (SOC) and state-of-health (SOH) is critical for ensuring reliable operation and prolonging the cycle life of lithium-ion batteries. However, complex operating environments and dynamic operations pose significant challenges to battery state estimation. To address this gap, this article proposes a battery state estimation method based on a pretrained large-scale language model (PLM). First, a numerical-text fusion module is introduced to integrate the battery test data with the pretrained text model, in order to exploit the potential of PLM in SOC and SOH estimation. Subsequently, the long short-term memory network is combined with the attention mechanism to enhance the temporal feature capture capability of PLM. Finally, a low-rank adaptation fine-tuning method is used to optimize the PLM to significantly reduce the training cost. Extensive experiments demonstrate that the proposed method delivers precise SOC and SOH estimates for batteries across diverse environments and operating conditions, achieving minimum root mean square errors of 0.49% and 0.27% for SOC and SOH, respectively. Compared to conventional algorithms, the estimation errors are reduced by 1.81% and 3.26%, offering superior accuracy, generalizability, and robustness.
Keywords:
Large language model (LLMs)
lithium-ion batteries (LIBs)
state estimation
state of charge (SOC)
state of health (SOH)
Journal
IF:
9.9
Papers:
8.3K
Citations:
6.0W
Organization
Cited Papers