Capacity estimation based on the aging characteristics analysis of Liquid metal batteries

The liquid metal battery is a new energy storage technology that is gaining increasing attention due to its high safety and long life. Accurate capacity estimation is crucial for health diagnosis and pack management, providing a strong guarantee of long-term stable operation for energy systems. However, owing to the lack of aging analysis traditional methods are difficult to improve the precision in the situation of unpredictable disturbances or fluctuations, such as capacity plunge. This paper proposes a capacity estimation method based on aging features by combining machine learning and filter algorithms, this method can provide an accurate prediction even the LMB is subject to capacity plunge. First, the aging features are extracted using only a tenth of the whole information from the discharge curve and increment capacity (IC) curve. Next, the Long Short Term Memory (LSTM) network is employed to establish the state-space representation between the aging feature and the capacity. The particle filter (PF) is then combined with the state-space representation to predict the capacity. Finally, aging tests are conducted to validate the precision and effectiveness of the proposed hybrid method. These results show that the proposed method can provide robustness as well as accurate capacity prediction.