State of health prediction of lithium-ion batteries: Multiscale logic regression and Gaussian process regression ensemble

Abstract

State of health (SOH) prediction plays a vital role in battery health prognostics. It is important to estimate the capacity of Lithium-ion battery for future cycle running. In this paper, a novel method is developed based on an integration of multiscale logic regression (LR) and Gaussian process regression (GPR) to tackle SOH estimation and prediction problem of Lithium-ion battery. Empirical mode decomposition is employed to decouple global degradation, local regeneration and various fluctuations in battery capacity time series. An LR model with varying moving window is utilized to fit the residuals (i.e., the global degradation trend). A GPR with the lag vector is developed to recursively estimate local regenerations and fluctuations. This design scheme captures the time-varying degradation behavior and reduces affections of local regeneration phenomenon in Lithium-ion batteries. The experimental results on Lithium-ion battery data from NASA Ames Prognostics Center of Excellence illustrate the potential applications of the proposed method as an effective tool for battery health prognostics.