Joint prediction of state of health and remaining useful life for lithium-ion batteries based on health features optimization and multi-model fusion.

Accurate prediction of the state of health (SOH) and remaining useful life (RUL) of batteries is essential for ensuring their safe and stable operation. Given the strong correlation between SOH and RUL, it is imperative to employ a joint prediction approach for a comprehensive evaluation of the actual battery aging condition. Therefore, this paper proposes a novel approach for jointly predicting SOH and RUL of lithium-ion batteries, which is based on the optimization of health features and fusion of multiple models. Multiple health features (HFs) are extracted from the charging voltage curve, and the fusion HF is optimized using kernel principal component analysis (KPCA) to leverage the complementary advantages of multiple features. Deep Gaussian process regression (DGPR) is employed to address the issue of lacking confidence interval expression in the prediction process. Combined with particle swarm optimization (PSO) and bidirectional long short-term memory (Bi-LSTM) network, fusion models of PSO-DGPR and Bi-LSTM-PSO-DGPR are developed for the prediction of SOH and RUL, respectively. In order to achieve the joint prediction of SOH and RUL, the complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is utilized to decompose the SOH prediction results, obtaining reliable capacity data for RUL prediction and eliminating local regeneration fluctuations caused by capacity regeneration effects. Furthermore, CALCE and NASA data sets are employed for validation. The experimental results demonstrate a significant reduction in the prediction error, compared to other prediction methods, thereby enhancing the accuracy and reliability of SOH and RUL prediction. [ABSTRACT FROM AUTHOR]