State of Health Estimation of Lithium Batteries Based on LWSSA-MKELM

The advancement of energy storage stations and portable devices has increased demands on measuring lithium-ion battery state of health (SOH). Currently, SOH estimation mainly relies on data-driven models, which face challenges like limited range, low accuracy, and long optimization times. To enhance the accuracy and speed of SOH estimation, this study proposes an improved Sparrow Search Algorithm (SSA) to optimize a Multi-Kernel Extreme Learning Machine (MKELM) model. The approach utilizes Kalman filtering to extract feature values and incorporates optimal Latin hypercube sampling to improve the quality of the initial population, ensuring uniform distribution and avoiding local optima. Additionally, a weighting factor is introduced during the parameter optimization process to balance global and local searches, enhancing search efficiency and better handling the complexity and variability in battery SOH estimation. Experimental results show that the proposed model outperforms three similar algorithms in accuracy, achieving an average fitting degree of $\mathbf{9 5 \%}$ across three different batteries.