Accurate assessment of the State of Health (SOH) is critical for battery management systems in aviation. As a step towards this goal, this study presents a proof-of-concept for a novel SOH estimation method based on an Improved Particle Swarm Optimization-Extreme Learning Machine (IPSO-ELM) model, validated under controlled laboratory cycling conditions. Although traditional Extreme Learning Machines (ELM) are widely used due to their fast computation and good generalization, their random parameter initialization often leads to unstable convergence and limited accuracy. To address these limitations, this paper proposes a novel SOH estimation method based on an Improved Particle Swarm Optimization (IPSO) algorithm to optimize the key parameters of ELM. Three health indicators (HI)—constant-current charging time, equal-voltage-drop discharge time, and average discharge voltage—were extracted from charge–discharge curves as model inputs. The IPSO algorithm dynamically adjusts the inertia weight, introduces a constriction factor and a termination counter to enhance global search capability and avoid local optima. Experimental results on open-source datasets (B005, B007, B0018) and laboratory datasets (A001, A002) demonstrate that the proposed IPSO-ELM model achieves a Root-Mean-Square Error (RMSE) below 0.7% and a Mean Absolute Percentage Error (MAPE) below 0.5%. Compared with standard ELM and PSO-ELM models, it significantly outperforms them in accuracy (e.g., for B0018, RMSE is reduced to 0.21% and MAPE to 0.14%), convergence speed, and robustness, establishing a foundation for future development of aviation-ready SOH estimators.
Zeng et al. (Tue,) studied this question.