A feature fusion optimization algorithm for predicting the remaining useful life of lithium-ion batteries

Abstract

As the percentage of lithium-ion batteries as in a system for storing energy gradually rises, accidents brought by the deterioration of battery performance continue to occur. The solution to guaranteeing the steady operation of this system is learning how to precisely forecast the lithium-ion batteries’ remaining useful life (RUL). A prediction framework that combines elements of incremental capacity analysis (ICA) and electrochemical impedance spectroscopy (EIS) is proposed to address issues of RUL. The framework initially examines the charging and discharging features of the battery before establishing a mapping association between fusion features and RUL using convolutional neural network (CNN) and improved long short-term memory network (ILSTM). The parameters of the improved particle swarm optimization algorithm (IPSO) are optimized to build the IPSO-CNN-ILSTM model by modifying updating rules of the inertia weight and learning factor of the particle swarm optimization (PSO) algorithm to improve its optimization ability. Lastly, numerical outcomes of the NASA PCoE datasets confirm this method’s applicability and efficacy.