A hybrid approach for remaining useful life prediction of lithium-ion battery with Adaptive Levy Flight optimized Particle Filter and Long Short-Term Memory network

Abstract

Lithium-ion batteries (LIBs) constitute a significant portion of energy storage system, the safety and reliability is its fundamental priority. Fortunately, efficient remaining useful life (RUL) prediction of LIBs is a powerful tool to ensure reliable system operation and reduce maintenance costs. In this paper, a hybrid framework is put forward to improve the prediction accuracy of LIBS with Adaptive Levy Flight (ALF) optimized Particle Filter (PF) and Long Short-Term Memory (LSTM) network. ALF is introduced to optimize the traditional PF so that the deficiency of weight degeneracy and particle impoverishment can be effectively conquered, meanwhile, LSTM network is adopted to learn the degrade model of LIBs. Consequently, the developed hybrid approach fused ALF-PF and LSTM network is employed to predict accurately the RUL of LIBs. The experimental results of battery data set from NASA and HUST demonstrate that the designed ALF-PF-LSTM framework substantially improves the prediction performance and robustness, and it also outperforms some currently popular PF-based algorithms.