Li-ion battery prognostic and health management through an indirect hybrid model

Abstract The immeasurable online battery capacity hampers the accurate State of Health (SOH) estimation and Remaining Useful Life (RUL) prediction because of complex battery ageing. To resolve the issues, an indirect hybrid model for online battery Prognostics and Health Management (PHM) is put forward. First, based on partial charging dada, two battery ageing features are extracted for the quantitative evaluation of battery capacity. Second, a Multi-Kernel Relevance Vector Machine (MKRVM) is optimized by Grey Wolf Optimizer (GWO) to determine the weights and kernel parameters of different kernel functions, then establish a mapping between battery ageing features and capacity. Third, the Complementary Ensemble Empirical Mode Decomposition (CEEMD) decomposes the measurable and historical ageing features into high- and low-frequency signals, then, the signals with different frequencies are modelled and predicted by Long Short-Term Memory Neural Network (LSTMNN) and Feedforward Neural Network (FNN), respectively. Finally, the real-time ageing features are utilized for SOH estimation and the predicted ageing features are used for RUL prediction through the GWO-based MKRVM. Furthermore, the performance of the proposed method is validated by an open-access battery ageing dataset. The results demonstrate that the indirect hybrid model has high flexibility and strong robustness on battery PHM.