Feature Extraction from Charging Profiles for State of Health Estimation of Lithium-ion Battery

Abstract

Abstract Accurate state of health (SOH) estimation of lithium-ion batteries is of great importance to ensure the reliability and safety of battery management systems (BMS). The difficulty of modelling the complex degradation mechanism has made the data-driven methods gain much attention in battery SOH prediction. To improve the estimation accuracy of battery SOH, extracting the suitable health indicators is still a challenging work. In this work, the health indication features are attracted from the charging voltage profile based on the experimental data measured under constant current charging mode. Subsequently, the Pearson correlation coefficient is used to evaluate the relationships between the extracted health features and battery capacity, thus selecting the most effective health features for establishing the prediction models. Finally, the battery SOH is estimated using a Gaussian process regression (GPR) method. The estimation results with R 2 of 1 and lower mean absolute error (MAE) and maximum error (MAX) provide higher accuracy based on the extracted health feature.