Uncertainty quantification of fusion prognostics for lithium-ion battery remaining useful life estimation

The uncertainty of prognostics and remaining useful life (RUL) estimation for the lithium-ion battery is emphasized in the battery management system (BMS). Many machine learning algorithms and statistical methods can not only realize the RUL prediction but also provide the probability density function (PDF) as the prognostic uncertainty representation, involving particle filter (PF), Relevance Vector Machine (RVM), etc. This paper presents a fusion RUL prediction approach with PF algorithm and data-driven autoregression (AR) algorithm for lithium-ion battery. Moreover, a framework to quantitatively analyze and evaluate the PDF distribution of the lithium-ion battery RUL prediction is presented. The probability confidence interval estimation, PDF histogram and distribution hypothesis test are included in quantifying the uncertainty. These quantitative analysis results can be meaningful for lithium-ion battery health management and maintenance. The experimental results with the battery data of NASA Ames Prognostics Data Repository show that the proposed framework can achieve the quantification of PDF to introduce the reference for the corresponding maintenance and management. The proposed work also shows potential prospective for industrial application.