Battery SOH estimation and RUL prediction framework based on variable forgetting factor online sequential extreme learning machine and particle filter

• Published in: Journal of energy storage Vol. 65; p. 107322
• Main Authors: Duan, Wenxian, Song, Shixin, Xiao, Feng, Chen, Yuan, Peng, Silun, Song, Chuanxue
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.08.2023
• Subjects: Extremely randomized trees; Online sequential extreme learning machine; Remaining useful life; State of health; Whale optimization algorithm; Extremely randomized trees; Online sequential extreme learning machine; Whale optimization algorithm; Remaining useful life; State of health
• ISSN: 2352-152X; 2352-1538
• DOI: 10.1016/j.est.2023.107322

Battery life prediction is of great practical significance to ensure the safety and reliability of equipment. This paper proposes a new framework to realize battery state of health (SOH) estimation and remaining useful life (RUL) prediction. The variable forgetting factor online sequential extreme learning machine (VFOS-ELM) is used to estimate battery SOH, and particle filter (PF) algorithm used to predict battery RUL. To improve the estimation accuracy, a new nonlinear decline method, adaptive weight and Gaussian variation are used to improve the standard whale optimization algorithm (WOA) algorithm. And the improved IWOA algorithm is used for parameter optimization of the VFOS-ELM and PF algorithm. The extremely randomized trees (ERT) algorithm is used to obtain the features with high correlation with the available capacity to reduce the complexity of the model and improve the estimation accuracy. Compared with other methods, the proposed IWOA-VFOS-ELM algorithm has higher estimation performance and noise anti-interference ability. The MAE of APR-3 and APR-4 for SOH estimation are both within 0.12 %, RMSE are within 0.15 %, and IA are both higher than 99.9 %. Compared with PF algorithm, the RUL prediction accuracy obtained by IWOA-PF algorithm is improved by 7.143 %, 6.445 % and 15.094, respectively. In summary, the IWOA-PF algorithm proposed in this paper can be used to predict the battery RUL, and the prediction performance is better than the PF algorithm. •The ERT model is used to rank the importance of features for the battery SOH estimation.•The IWOA-VFOS-ELM model is conducted for on-line battery SOH estimation.•The IWOA-PF model is conducted for battery RUL prediction.•The SOH estimation results are compared between the features in different voltage segments and the different combination of features.