Momentary informatics based data-driven estimation of lithium-ion battery health under dynamic discharging currents

Data-driven approaches have been proposed to estimate the state-of-health (SOH) of lithium-ion batteries. Most of the existing data-driven approaches are designed for conditions of constant current (CC) charging/discharging at a specific rate over a long duration. However, the loading profiles of batteries can be highly volatile, which restricts the application of existing data-driven methods. To address this issue, this paper firstly proposes a momentary health indicator (HI) that is extracted from a very short time period (<0.1 s) when the battery transits from discharging to rest. The proposed HI is exceptionally easy to implement under various discharging currents and state-of-charge (SOC) levels. Moreover, it does not require a filtering step as it is singular and truncated in nature. The rationality of the proposed HI is to indirectly reflect battery internal ohmic resistance, which is justified by theoretical analysis with an equivalent circuit model (ECM). Then, a hierarchical ensemble model (HEM) of esteemed machine learning (ML) algorithms is designed to efficiently learn the relationship between the HI and SOH through momentary informatics. The efficacy of the proposed data-driven method is demonstrated by hardware-in-the-loop (HIL) experiments, and results show that high-accuracy SOH estimation, i.e., average root-mean-square error (RMSE) as low as 0.4 %, is achieved with very low computational costs.