Multi-feature Extraction and Fusion-based State of Health Estimation of Large-format Lithium-ion Batteries under Uncertain Aging Mode

State of Health (SOH) is pivotal to the health diagnostics of lithium-ion battery (LIB). However, the SOH estimation of large-format batteries commonly-used in energy storage power stations, especially at uncertain environmental conditions and aging modes, has been less explored. This paper proposes a SOH estimation method applicable to large-format batteries, by combining the multi-feature extraction and artificial intelligence approach. Especially, different sets of health indicators (HIs) exhibiting the morphological incremental capacity (IC) characteristic are extracted from the charging curve of LIBs. Following this exertion, artificial neural network-based HI fusion is proposed to estimate the SOH accurately. The proposed method is validated with long-term degradation experiments on the LFP cells. Results suggest that the proposed method manifests itself with high estimation accuracy, high reliability and prominent robustness to cell inconsistency.

I. Introduction

Due to the advantages of low cost, high energy density and strong reliability, lithium-ion batteries (LIBs) have been widely used in energy storage systems [1]–[3], and the performance parameters and safety parameters of LIBs have become worthy of concern [4]. To ensure the safety of batteries, battery management system (BMS) should be used to measure the key parameters, such as voltage, current, temperature, etc. [5]–[8], while the key state of the battery, such as state of charge (SOC) [9]–[17], specifically, the state of health (SOH), should be estimated precisely considering its paramount role in battery life diagnostic and management. However, the accurate SOH estimation remains a major challenge due to its link to the complex aging mechanism within lithium-ion batteries [18].