Online State-of-Health Estimation of a Li-ion Battery using Incremental Voltage Difference and Ampere-Hour Throughput as Indirect Health Indicators

Abstract

State-of-Health (SOH) estimation of Lithium-ion batteries (LIBs) is crucial for ensuring the reliability and predictive maintenance of electric vehicles and energy storage devices. Indirect health indicator-based health monitoring is more prominent for data-driven approaches. Choosing an appropriate and trendable feature vector leads to accurate SOH estimation. In this paper, Incremental Voltage Difference (IVD) and Ampere-hour throughput (AhT) constitute the proposed feature vector. During usual usage, derived voltage feature (Vsei) are acquired from charging data, and the IVD ( Vsei) values are calculated therein. The initial Vsei point is determined by choosing the peak of an Incremental Capacity curve and then the subsequent voltages are chosen for every 1.5 % increase in coulomb counts. From the initial Vsei point, the increase in AhT is calculated and added as a feature. This paper examines a method for improving the generality of SOH estimation using a Gaussian process regression (GPR), polynomial regression (PR) algorithm, and artificial neural network (ANN). The proposed methodology is validated by publicly available NASA's Ames Research Centre’s prognostic center of excellence (PCoE) dataset. The Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) have been reduced significantly in the proposed method and are able to estimate the SOH under partial charge-discharge data with no assumption of extrapolation-interpolation of data.