Prognostics of lithium-ion batteries using model-based and data-driven methods

This paper presents an integrated approach to predict remaining useful life (RUL) of lithium-ion batteries using model-based and data-driven methods. An empirical model is adopted to emulate the battery degradation trend; real-time measurements are employed to update the model. In order to better deal with prognostics uncertainties arising from many sources in the prediction such as battery unit-to-unit variations, an online model update scheme is proposed in a particle filtering based framework. Filtered data within a moving window are used to adjust the model's parameter values in a real-time manner based on nonlinear least-squares optimization. The proposed approach is studied via experimental data, and the results are discussed.