Transit Search Optimized Least Squares Support Vector Machine for the State of Health Estimation of Lithium-Ion Batteries

The State of Health (SOH) of lithium-ion batteries is a critical parameter that mirrors the performance of their health monitoring systems. To achieve accurate prediction of SOH, we propose a novel prediction model. The model combines the advantages of variational mode decomposition (VMD) and transit search optimization-least squares support vector machine (TSOA-LSSVM). The constant-current charging time (CCCT), constant-current discharging time (CCDT), and Constant voltage charging time(CVCT) with Internal resistance(IR) are extracted as health indicators (HIs), and the correlations between HIs and SOH are evaluated by using Spearman correlation coefficient (spearman) respectively. With the VMD method, the most correlated HI is decomposed into multiple intrinsic modal functions (IMFs), and each IMF is fitted by the LSSVM. TSOA is used to obtain the most appropriate hyperparameters. The efficacy of the newly developed VMD-TSOA-LSSVM model is assessed by utilizing battery aging data sourced from the Center for Advanced Life Cycle Engineering (CALCE). Its performance is then compared to that of the solo LSSVM, the VMD-LSSVM, and the VMD-TSOA-SVM methods. The laboratory outcomes demonstrate that the VMD-TSOA-LSSVM model outperforms other models in prediction and has better robustness and generalization ability.