A novel model-data fusion method for capacity and battery remaining useful life prediction

• Published in: Journal of energy storage Vol. 98; p. 112929
• Main Authors: Zhou, Dinghua, Zhu, Zhongwen, Li, Cheng, Jiang, Weihai, Ma, Yan, Lu, Jianwei, Li, Shuhua, Wang, Weizhi
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.09.2024
• Subjects: Error reconstruction; Improved ant lion optimizer; Lithium-ion battery; Model-data fusion; Remaining useful life; Lithium-ion battery; Improved ant lion optimizer; Remaining useful life; Model-data fusion; Error reconstruction
• ISSN: 2352-152X
• DOI: 10.1016/j.est.2024.112929

Accurate prediction of the remaining use life (RUL) of the battery is very essential to ensure the safety of electric vehicles. A novel model-data fusion method for RUL prediction considering the error correction and capacity self-recovery of lithium-ion battery is proposed in this paper. Firstly, considering the particle impoverishment problem in traditional particle filter (PF), an improved antlion optimizer (IALO) is adopted for particle distribution optimization, which also solves the deficiency of local convergence and global search in the standard ALO. Secondly, an error correction method based on the error reconstruction is proposed to solve the influence of the local fluctuation of the prediction error. A new error series is reconstructed by the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN). And the long-term evolution information and some important local fluctuation information are retained by selecting the strong correlated intrinsic mode functions (IMFs). Finally, Gaussian process regression (GPR) is adopted for prediction of new errors series to correct the predicted values obtained based on improved PF. Two different battery datasets are adopted to verify the effectiveness of the proposed method with RUL prediction error less than 2.5% and capacity prediction error less than 1.5%. •A novel prognostic framework based on PF and GPR hybrid method is proposed.•An improved ALO is introduced to improve the algorithm optimization ability.•Particles are optimized by the IALO algorithm to alleviate particle degradation.•The error is reconstructed by CEEMDAN and prognostic results is corrected by GPR.•Experimental validations show significant improvement in precision of prediction.