Study on Partial Stratified Resampling for Particle Filter Based Prognosis on Li-Ion Batteries

Accurate online prognosis of engineering systems plays a vital role in prognosis and health management (PHM) technologies to ensure safety, prevent damage and economic loss. The particle filter (PF) algorithm has proved to be an effective method for prognostics. However, the PF algorithm suffers from serious particle degeneracy and particle impoverishment problems. Most of the studies in the literature focus on solving the particle degeneracy problem but at a heavy computational cost. In this study, we aim to explore the time efficient Partial Stratified Resampling algorithm which can be used for online state estimation problems and compare it with conventional Stratified Resampling algorithms. The accuracy and precision of the algorithms are validated using the lithium-ion battery data sets from CALCE® research group.