Data-Driven Estimation of Li-Ion Battery Health using a Truncated Time-based Indicator and LSTM

The lithium-ion battery (LIB) is widely employed in energy storage systems. Its state-of-health (SOH) is a crucial parameter for assessing battery reliability and retirement. This paper proposes a data-driven method using a simple but effective health indicator (HI) extracted from a truncated time interval (110 seconds in this paper) of the LIB’s discharge process, which is a challenge in literature since it is usually less controllable than the charge process. Unlike traditional HIs, the proposed HI can be derived from different voltage ranges, thus making it highly flexible for application. A deep learning algorithm named the long short-term memory (LSTM) is trained to learn the mapping relationship between the extracted HI and practical SOH of the LIB. The acquired test results using an open dataset show that the developed data-driven method can precisely estimate battery SOH without any additional hardware or system downtime.