The cooperative estimation of state of charge and state of health for lithium-ion batteries based on a Shared-Memory Parallel Peephole Long Short-Term Memory network

Accurate estimation of the state of charge (SOC) and state of health (SOH) is essential for the safe and reliable operation of lithium-ion batteries. Because SOC and SOH are intrinsically interdependent, conventional approaches that build separate models for SOC and SOH decouple their relationship, thereby increasing computational burden and degrading estimation accuracy. To address these issues, we propose a Dual-Time-Scale Estimation Framework (DTSEF) for the cooperative estimation of SOC and SOH. The framework shares parameters between the two tasks via a Shared-Memory Parallel Peephole LSTM (SMPLSTM), which explicitly captures their intrinsic coupling and extracts task-critical features. In addition, we designed a Cycle-Aware Gated Attention mechanism together with a stacked convolutional neural networks (CNN)-based cycle-fusion module to enable accurate estimation of both SOC and SOH. Given that battery aging exerts a pronounced influence on SOC estimation, the cycle-aware gated attention fuses key SOC and SOH features to further enhance SOC estimation performance. We validated the proposed framework on the Oxford battery degradation dataset and the NASA battery degradation dataset. On the Oxford dataset, the average SOH estimation errors were MAE = 0.21% and RMSE = 0.30%. With feature fusion and within-cycle attention enhancement, the average SOC estimation error decreased from MAE = 0.67% to 0.46%, and RMSE from 0.84% to 0.61%. On the NASA dataset, SOH estimation achieved an average MAE of 0.76% and an RMSE of 0.96%. Compared with existing data-driven methods, the proposed approach demonstrates superior accuracy, robustness, and practical utility. Overall, the framework enables accurate joint estimation of SOC and SOH across dual temporal scales.