A hybrid method considering a novel health feature and uncertainty quantification for li-ion battery RUL prediction

The remaining useful life (RUL) prediction of the Lithium-ion battery is extremely crucial for the reliable and safe operation of the battery. Since the complex model or the effective health features (HF) of the traditional prediction methods, a hybrid model-data driven method PL-WOA-ELMAN-PF is proposed to forecast the RUL prediction and uncertainty quantification. Firstly, a novel HF from the voltage curve of the CC charging is extracted, what the Pearson and the first-order partial correlation coefficient verifies the HF is highly pertinent to the battery degradation. Meanwhile, the data-driven method based the new HF and advanced neural network algorithm (WOA-ELMAN) is established while the whale optimization algorithm (WOA) is used to improves the prediction accuracy of the ELMAN. Then, the PL-WOA-ELMAN-PF is proposed based on the particle filter algorithm (PF), while the WOA-ELMAN is introduced to modify the observed value of the dual exponential model. In addition, the PF enables the uncertainty quantification of RUL. Finally, the experimental verification is carried out on two large-cycling and one small-cycling battery degradation datasets. The root mean square error is within 0.01 and the error of RUL estimation is less than 10 cycles, showing that the novel HF is efficient and the proposed method has high accuracy and strong generalization ability.