Pressure‐Enhanced State‐of‐Charge Estimation for Lithium Iron Phosphate Batteries via Autoregressive LSTM With Multiphysics Fusion: An Embedded AI Implementation and Validation Study
Year: 2026
Type: article
Abstract: Conventional state‐of‐charge (SOC) estimation methods for lithium‐ion batteries rely heavily on voltage information. However, lithium iron phosphate (LFP) batteries exhibit a pronounced voltage plateau, which makes accurate SOC estimation difficult. To address this issue, this study proposes an autoregressive long short‐term memory (AR‐LSTM) method that integrates voltage, current, temperature, and expansion pressure as electro–thermal–mechanical input features. Battery expansion pressure is measured using thin‐film pressure sensors and provides additional state‐related information for SOC estimation. Offline experiments under unseen operating profiles demonstrate that the proposed method achieves an average SOC estimation root mean square error of 0.49%, indicating high accuracy and strong cross‐condition generalization capability. The method is further deployed on an embedded battery management system using an 18‐cell series‐connected LFP battery module. Commercial analog front‐end devices are employed to acquire current, voltage, and pressure signals, and the model is executed online in real time. The embedded experiments limit the maximum SOC estimation error to 1.8%, demonstrating the practical feasibility of the proposed method.
Source: Energy Science & Engineering
DOI: 10.1002/ese3.70703
Authors: Zhenqi Hao(Data curation, Methodology, Visualization, Writing – original draft), Lei Cao(Methodology, Resources, Writing – review & editing), Xuewen Geng(Methodology, Resources, Writing – review & editing), Peng Sun(Methodology, Resources, Writing – review & editing), Xueyuan Wang(Conceptualization, Investigation, Methodology, Writing – review & editing), Xuezhe Wei(Conceptualization, Resources, Supervision), Haifeng Dai(Funding acquisition, Resources)
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,780
OpenAlex ID: W7235563839
