State of health estimation of lithium-ion batteries based on the fusion of aging feature extraction and SSA-ELM machine learning algorithms

The battery data from the existing publicly available dataset was measured under uninterrupted charge/discharge experiments. The aging experiment did not give the battery enough resting time, and the polarization phenomenon still exists inside the battery. Because of the presence of polarization phenomenon, the internal battery has not reached the equilibrium state, which affects the accuracy of subsequent data collection. In this paper, by analyzing the strength of the polarization phenomenon after charging and discharging, we choose to obtain relevant features from the data of the discharging process. A state of health (SOH) estimation method based on the fusion of sparrow search algorithm (SSA) and extreme learning machine (ELM) is proposed. For the parameter setting problem in the ELM model, this paper proposes parameter optimization using SSA. The model is validated by two different training methods using NASA and CACLE datasets, and compared with other common machine learning algorithms. The experimental results show that the method has high prediction accuracy for different types and experimental conditions of cells. Compared with other methods, the prediction errors of all the methods in this paper are less than 1%, and all three error indicators are lower than other comparison methods.