Remaining useful life prediction of lithium-ion batteries using a spatial temporal network model based on capacity self-recovery effect

Abstract

The accurate prediction of the remaining useful life (RUL) of lithium-ion batteries (LIBs) is crucial to ensure the safe operation of electric vehicles (EVs). However, predicting RUL accurately is challenging due to the capacity self-recovery effect, which has been seldom addressed in existing literature, and the complex degradation mechanism. Here, in order to solve the above issues, we extract the health factors from measurable data to reveal the characteristics of battery degradation. Additionally, to capture spatial correlation of multi-series data and address the capacity self-recovery effect, the spatial-temporal remain useful life forecasting (STRULF) model is designed, which combines the spatial LSTM and temporal attention mechanism (TAM). Furthermore, we embed an online updating (OL) strategy with AutoDropout to enhance the model accuracy and generalization under different operating conditions. We evaluate the performance of our proposed model using four lithium-ion battery datasets. The experimental results reveal that the OL-STRULF model outperforms other methods in predicting RUL accurately.