A data‐driven approach with error compensation and uncertainty quantification for remaining useful life prediction of lithium‐ion battery. (30th June 2022)

• Record Type: Journal Article
• Authors: Wei, Meng; Ye, Min; Wang, Qiao; Lian, Gaoqi; Xu, Xinxin
• Is Part Of: International journal of energy research. Volume 46:Number 14(2022)
• Page Start: 20121
• Page End: 20135
• Publication Date: 2022-06-30
• Subjects: error compensation -- gated recurrent unit -- lithium‐ion battery -- remaining useful life -- uncertainty quantification; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Power resources -- Research -- Periodicals; 621.042
• DOI: 10.1002/er.8321
• Languages: English
• ISSNs: 0363-907X

Abstract

Summary: To accurately predict the remaining useful life (RUL) of lithium‐ion battery, this paper presents a novel machine‐learning method using gated recurrent unit (GRU) with error compensation (EC). Moreover, the dropout Monte Carlo technology is selected for reliable RUL prediction with uncertainty quantification. First, the equal charging voltage time is established and the variational mode decomposition is introduced to reduce the influence of capacity regeneration phenomenon. Then, the phase space reconstruction with C‐C technology is adopted to obtain the optimal input sequence, and the GRU with EC is established for RUL prediction. Finally, the probability distribution and 95% confidence interval are obtained based on dropout Monte Carlo technology. Compared with the existing methods, the proposed method can not only obtain a higher accuracy with a mean absolute error below 1.55% but also achieve reliable RUL prediction with probability distribution. Abstract : In this paper, the hybrid data‐driven method based on error compensation and gated recurrent unit with Monte Carlo is proposed. When compared to the traditional methods, the proposed method not only can achieve a higher accuracy in RUL prediction with a mean absolute error below 1.55% but also obtain reliability of prediction with probability distribution.