Remaining useful life prediction of lithium-ion batteries based on the discharge voltage

Abstract

The remaining useful life (RUL) is always important to the preventive maintenance of lithium-ion batteries. To achieve the accurate RUL prediction, a whale optimization algorithm (WOA) with multi-kernel relevance vector machine (MKRVM) model based on the discharge process data is introduced in this paper, health indicator (HI) that can characterize battery degradation is obtained from the discharge voltage, and the relationship between HI and capacity is evaluated by Person and Spearman. On this basis, a multi-kernel relevance vector machine (MKRVM) model is established. WOA is used to obtain the optimization weights of multi-kernel and enhance the performance of the model. Finally, the battery data from NASA is used to verify the performance of the model introduced in this paper.