A RUL prediction method for lithium-ion batteries based on improved singular spectrum analysis and CSA-KELM. (May 2023)

• Record Type: Journal Article
• Authors: Ding, Guorong; Chen, Hongxia
• Is Part Of: Microelectronics and reliability. Volume 144(2023)
• Publication Date: 2023-05
• Subjects: Lithium-ion battery RUL prediction -- Improved singular spectral analysis -- Circle search algorithm; Electronic apparatus and appliances -- Reliability -- Periodicals; Miniature electronic equipment -- Periodicals; Appareils électroniques -- Fiabilité -- Périodiques; Équipement électronique miniaturisé -- Périodiques; Electronic apparatus and appliances -- Reliability; Miniature electronic equipment; Periodicals; 621.3815
• DOI: 10.1016/j.microrel.2023.114975
• Languages: English
• ISSNs: 0026-2714

Abstract

Abstract: Accurate and efficient capacity prediction of lithium-ion batteries plays an important role in improving performance and ensuring safe operation. An improved singular spectral analysis (ISSA) based on the fractal dimension (FD) is proposed to eliminate capacity regeneration during battery degradation. Firstly, FD is used to determine whether the signal decomposition component belongs to the main component or the noise is a useless component. Secondly, the kernel extreme learning machine (KELM) is introduced to predict the main components. In addition, a new heuristic optimization algorithm, the circle search algorithm (CSA), is used to optimize the regularization coefficients and kernel function parameters of KELM. Finally, experimental simulations are performed based on four batteries from two open lithium-ion battery datasets. The results show that the ISSA algorithm has better decomposition efficiency than the empirical mode decomposition (EMD) and the variational mode decomposition (VMD). Compared with other intelligent optimization algorithms, the CSA is superior and can achieve a faster convergence rate. Highlights: The model requires only the battery capacity time series as input and consists mainly of improved singular spectral analysis (ISSA), circle search algorithm (CSA) and kernel extreme learning machine (KELM). ISSA is used as an unsupervised feature learning method to decompose capacity and extract frequency and amplitude features, and it improves short-term trend prediction, especially for capacity growth that occurs during the degradation process.