Prediction of Li-ion battery capacity degradation considering polarization recovery with a hybrid ensemble learning model

Polarization recovery is a phenomenon that significantly affects the capacity degradation behavior of lithium-ion battery (LIB). In this study, we demonstrated that capacity fluctuation during fading was highly correlated with polarization recovery through micro-Roman and rate performance tests. A series of aging experiments with a wide range of relaxation times (0.08 h to 200 h) were developed to analyze the effects of polarization recovery on discharge capacity. The experimental dataset includes 30,696 charge-discharge cycles under various conditions based on LiFePO4/graphite and LiNi0.8Co0.15Al0.05O2/graphite battery cells. Here, a hybrid ensemble learning model (HEL) was proposed to achieve accurate predictions while considering polarization recovery. Multiple highly correlated stress factors of polarization recovery were identified and then utilized as the input data of HEL. Our analysis shows that HEL is capable of precisely capturing capacity degradation with fluctuations due to polarization recovery, the implication of which is critical to the advancement of lifetime management in battery energy storage applications.