State of health estimation of lithium-ion batteries based on interval voltage features

The precise estimation for the state of health of lithium-ion batteries determines whether the battery system can operate reliably and safely. The extraction and selection of features drive further development of the data-driven method, which has a promising application prospect in assessing the state of health. In response to the issue of time-consuming estimation based on the overall charge–discharge profiles, a novel method utilizing the features of a specific voltage region is reported in the paper. This method enables rapid state of health estimation, catering to the requirements of real-world technical applications. First, the dV/dt curves of discharge profiles are analyzed, and three health features related to a regional voltage interval of an equal time difference are extracted. The methodology of correlation is employed to determine the association between the proposed health features and the state of health. Finally, to enhance the precision of estimation, an online sequential extreme learning machine considering the standard hunter–prey optimization algorithm is proposed. The efficacy of the suggested method is confirmed through the utilization of NASA and Oxford datasets that were gathered under diverse working conditions. Based on the experimental results, the three health features and a combination of online sequential extreme learning machine and hunter–prey optimization method can provide high-precision estimation. [Display omitted] •Three new health features are constructed for the SOH estimation of batteries.•The features for estimation can be extracted only by a part of the discharge data.•The HPO-OS-ELM algorithm is developed to accurately estimate SOH online.•The suggested approach is validated by analyzing the NASA and Oxford datasets.