A fusion framework for lithium-ion batteries state of health estimation using compressed sensing and entropy weight method. (April 2023)

• Record Type: Journal Article
• Authors: He, Ning; Qian, Cheng; Shen, Chao; Huangfu, Yigeng
• Is Part Of: ISA transactions. Volume 135(2023)
• Page Start: 585
• Page End: 604
• Publication Date: 2023-04
• Subjects: Lithium-ion battery -- State of health -- Gaussian process regression -- Compressed sensing -- Particle filter -- Entropy weight method; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2022.10.003
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: Accurate estimation for state of health (SOH) is an important component of lithium-ion batteries (LIBs) health management system. A fusion framework for SOH estimation is proposed via using compressed sensing (CS) and entropy weight method (EWM). Firstly, incremental capacity curve (ICC) is extracted as health indicators (HIs), and CS technique is introduced to process the ICC to: (1) improve the sampling frequency; (2) reconstruct potential missing information caused by low sensor sampling frequency, and (3) eliminate noise interference. Then Gaussian process regression (GPR) is utilized to characterize the relationship between Pearson correlation analysis (PCA) based HIs and capacity, and discrete aging model (DAM) is further established for particle filter (PF) to realize the continuous estimation by taking the identified capacity of GPR functioned as observation. Finally, the capacities of GPR and DAM are fused via EWM for final capacity estimation. The experimental results based on open battery data sets from NASA demonstrate that proposed method has higher precision with the average error of 2.5%. In addition, lab experiments are further conducted with two standard 18650 batteries, and the experimental results indicate that the proposed strategy is capable to realize reliable estimation with the average error of 2.6%, which further illustrates the feasibility and applicability of the method Highlights: CS is employed to preprocess the ICC signal to recover lost information caused by low sampling frequency and eliminate noise interference. EWM is developed to work as error correction feedback to improve the SOH fusion estimation. Open and experimental battery datasets are used to verify the validity of the proposed method.