Answers are grounded in this paper and cite their sources.
Demonstrates improved state of charge estimation in lithium iron phosphate battery packs, suggesting enhanced performance for electric vehicles.
Accurate state of charge (SOC) estimation of large-sized battery packs in real-world electric vehicles (EVs) is crucial for predicting remaining mileage and enhancing battery utilization efficiency. However, this remains challenging due to complex operating conditions and cell-to-cell inconsistencies, especially for large-sized lithium iron phosphate (LFP) battery packs with a flat open-circuit voltage (OCV) plateau. Therefore, this study proposes a novel SOC estimation approach for LFP EV battery packs by combining model-based and data-driven methods. First, high-reliability labeled SOCs are obtained by combining full-charge identification, reverse ampere-hour integration (AHI), and OCV-based calibration outside plateau regions. A feature enhancement strategy with multi-deviation AHI sequence is proposed to improve feature representation capability. Then, a hybrid model incorporating convolutional neural networks and gated recurrent units (CNN-GRU) is developed to accurately estimate battery SOCs during charging processes. Finally, the estimation results at the end of charging segments are leveraged to initialize an adaptive extended Kalman filter (AEKF), which integrates a battery pack model and an online parameter identification strategy, to achieve accurate SOC estimations under complex dynamic discharging conditions. The comparison with other commonly used machine learning methods and the ablation study on CNN-GRU and AEKF are performed to highlight the superiority of the proposed method in fidelity and practicability. The results demonstrate that, with the optimal parameter combination, the proposed method achieves a root mean squared error of 1.09% based on uniform deviation distributions. More than 97% of estimates show an absolute error of less than 3% for both deviation generation methods. This work provides significant insights into state estimation and prediction of LFP battery systems for field applications.
No takes yet. Share an insight, caveat, or question.
Zhang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: