Online Battery SOH Prediction under Intra-Cycle Variation of Discharge Current and Non-Standard Charging and Discharging Practices

Abstract

Online SOH prediction of a battery is essential for battery management in real-world applications. Although many approaches have been proposed to date, most conventional approaches to SOH prediction neither incorporate intra-cycle variation of discharge currents nor consider non-standard charging and discharging practices along cycles. For real-world applicability, the above two factors, varying discharge current and non-standard practices, are crucial to deal with. This paper presents an approach to online SOH prediction based on charging and discharging cycles under intra-cycle variation of discharge current and non-standard charging and discharging practices. Here, for online SOH prediction, we represent a cyclic history of terminal voltages and currents by a vector sequence formed by four indices: total charge, voltage-time entropy, average and variance of current, where average and variance of current are intended to deal with varying discharge current. Furthermore, we identify various minimum ranges of charging and discharging voltages that can provide accurate relation between four indices and SOH under non-standard practices, while transforming this relation into that of standard practices. For experiments, an LSTM stack is implemented for the proposed online SOH prediction. The results indicate that the proposed approach is capable of accurate online SOH prediction even under randomly varying discharging currents and non-standard practices with RMSE errors of about 0.5% as well as of about 95%.