The future capacity prediction using a hybrid data-driven approach and aging analysis of liquid metal batteries

Liquid metal batteries (LMBs) are wildly considered for large-scale energy storage due to the advantages of simple construction, low cost, and long life. It is of great importance to find a reliable and accurate approach to predict the future capacity for battery management and failure evaluation. A hybrid data-driven framework is presented to predict not only the long-term capacity degradation but also the local regeneration. Firstly, the Empirical mode decomposition (EMD) method is employed to decompose original battery capacity data into a residual and some intrinsic mode functions (IMFs). Then based on the data characteristics, Gaussian Process Regression (GPR) and Relevance Vector Machine (RVM) with combined kernel function is utilized to predict the local fluctuations caused by regeneration phenomena and the capacity degradation, respectively. Furtherly, we can obtain accurate capacity prediction by adding individual predicted results. Finally, the aging process and the mechanism of capacity plunge are analyzed by the incremental capacity analysis (ICA) method. The Root Mean Square Error (RMSE) of prediction is 0.117 Ah for a 20 Ah battery and 0.141 Ah for a 50 Ah battery by using the proposed framework. Compared with existing mainstream methods, the proposed framework has a more accurate prediction performance to capture the degradation tendency before the capacity plunge, which indicates the positive promise for practical applications of this framework.