Lei Zhou

Lithium-ion batteries exhibit nonlinear capacity degradation under variable-temperature and complex operating conditions, which makes remaining useful life (RUL) prediction challenging. To address this issue, we propose a hybrid Arrhenius-long short-term memory (LSTM) method that combines piecewise Arrhenius thermal-damage modeling with LSTM-based residual correction. The method constructs a piecewise Arrhenius capacity-degradation model around the knee point and introduces an equivalent thermal-damage factor to accumulate temperature-induced aging effects. An LSTM residual branch then learns the discrepancy between measured capacity and the physical baseline from historical capacity, temperature, voltage, current, and prior physical predictions. A three-stage optimization strategy, comprising physical-parameter pre-fitting, residual-branch training, and physics-constrained joint fine-tuning, stabilizes residual learning while preserving mechanistic interpretability. Experiments on four CALCE CX2 validation cells yield mean SOH RMSE and MAE values of 0.020 and 0.016, respectively. The corresponding mean RUL RMSE and MAE are 41.52 and 31.58 cycles. Compared with Ade-lwT, the proposed method reduces mean SOH RMSE and MAE by 22.33% and 35.05%, respectively. For RUL prediction, the corresponding reductions are 58.06% and 62.29%. These results support the effectiveness of coupling a piecewise Arrhenius baseline with capacity-residual learning for cross-cell SOH estimation and RUL prediction.

This paper provides a comprehensive review of recent advances in remaining useful life prediction for lithium-ion battery energy storage systems. Existing approaches are generally categorized into model-based methods, data-driven methods, and hybrid methods. A systematic comparison of these three methodological paradigms is presented, with hybrid methods further divided into filter-based hybrids and data-driven hybrids, followed by a comparative analysis of remaining useful life prediction accuracy. The literature analysis indicates that data-driven hybrid methods, by integrating the strengths of physical mechanism modeling and machine learning algorithms, exhibit superior robustness under complex operating conditions. Among them, the hybrid framework combining long short-term memory networks with an eXtreme Gradient Boosting model optimized by the Binary Firefly Algorithm demonstrates the highest stability and accuracy in the reviewed studies, achieving a root mean squared error below 2% and a mean absolute percentage error below 1%. Future research may further enhance the generalization capability of this framework, reduce computational cost, and improve model interpretability.