Key result
Hybrid ICA-LSTM framework reduces SOH estimation RMSE to 1.5% with wider SOC spans.
Why the study?
ICA curves derived under low C-rate and full operating cycles limit applicability in practice for estimating lithium-ion battery state of health.
The proposed hybrid framework using quasi-equilibrium ICA features and LSTM improves the accuracy of lithium-ion battery SOH estimation under partial charging conditions.
Hybrid framework enables practical SOH estimation under variable conditions; leaves open real-world validation before deployment.
Accurate estimation of the State of Health (SOH) of lithium-ion batteries is essential for ensuring safety, extending service life, and enabling predictive maintenance in energy storage applications. Peak features from Incremental Capacity Analysis (ICA) are widely used as Health Indicator (HI)s for SOH estimation. However, ICA curves derived under low C-rate and full operating cycles limits the applicability in practice. To address these limitations, a hybrid framework taking quasi-equilibrium ICA features from partial charging data is proposed. Quasi-equilibrium ICA curves derived from reconstructed OCV (Open Circuit Voltage) enables reliable peak feature extraction as HIs. A Long Short-Term Memory (LSTM) network with multi-head attention is then adopted to model temporal dependencies, while accommodating inputs with variable length through padding and masking mechanisms. The proposed method is validated on both PyBaMM-based simulation under Multi-stage Constant Current (MCC) charging profiles and a real aging dataset from KIT. The results demonstrate Root Mean Square Error (RMSE) values decreasing from 2.7% to 1.1% as the State of Charge (SOC) span during partial charging increases from 26.9% to 51.4% in simulation. On the experimental dataset, the method achieves RMSE decreasing from 2.6% to 1.5% as the SOC span during partial charging increases from 23.8% to 71.4%, demonstrating its robustness and practical applicability.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2026) studied Lithium-ion battery State of Health (SOH). Hybrid framework using quasi-equilibrium ICA features and LSTM network was evaluated on Root Mean Square Error (RMSE) of SOH estimation. A hybrid framework using quasi-equilibrium ICA features and an LSTM network for lithium-ion battery SOH estimation achieved RMSE values decreasing from 2.6% to 1.5% as the SOC span increased.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: