Abstract:
In the real operation of lithium-ion batteries, the capacity data are usually not available online and the degradation features are subject to the local regeneration, which makes an accurate prediction based on a single model very difficult for the remaining useful life (RUL) of the lithium-ion battery. To solve this problem, an indirect health indicator (HI)-based method for predicting the RUL of the lithium-ion battery is proposed in this paper. First, the optimal HI is evaluated based on the online data. Then, a relational model based on the genetic algorithm optimization of the extreme learning machine (GA-ELM) is built. The indirectly obtained HI components are decoupled by the variational mode decomposition (VMD), and the trends are tracked as well as predicted by the relevance vector machine (RVM). Finally, the predicted results are fed into the GA-ELM relational model to obtain the capacity prediction value, which can be translated into the RUL of the lithium-ion battery. The comparison between the results of the NASA PCoE dataset show that the performance of the proposed method is better than that of the other candidate methods in predicting both the trend and the RUL, with an accurate predicted value of RUL for the lithium-ion batteries.