Remaining useful life prediction for lithium-ion battery by combining an improved particle filter with sliding-window gray model

Abstract Dependable and accurate battery remaining useful life (RUL) prediction is essential for ensuring the safety and reliability of battery systems. To improve the dynamic traceability of the battery degradation process for RUL prediction under different loading profiles, this paper presents an improved RUL prediction method, which is established from the combination of the linear optimization resampling particle filter (LORPF) with the sliding-window gray model (SGM). Major innovations are presented as follows: (1) To increase the accuracy of RUL prediction, a linear optimization combination is proposed to overcome the particle diversity deficiency in the resampling process of the standard PF, i.e. the LORPF; (2) To improve the traceability of the LORPF in predicting degradation trajectory, the SGM is employed to update the state variables of the state–space model in the LORPF. Additionally, an SGM-LORPF framework is constructed for RUL prediction. The performance of the SGM-LORPF is synthetically verified by data from two types of batteries under different loading profiles. Prediction test results indicate that the SGM-LORPF can achieve accurate RUL prediction under both constant current discharge conditions (relative error within 7.20%) and dynamic current discharge conditions (relative error within 2.75%). Moreover, using only a small amount of historical data, the proposed SGM-LORPF framework can acquire accurate results. The experimental outcome indicates that the SGM-LORPF has considerable efficiency and a wide range of practicality.