Prediction of Li-ion battery state of health based on data-driven algorithm
Hanlei Sun, Dongfang Yang, Jiaxuan Du, Ping Li, Kai Wang
Abstract
Li-ion battery state of health (SOH) is a key parameter for characterizing actual battery life. SOH cannot be measured directly. In order to further improve the accuracy of Li-ion battery SOH estimation, a combined model based on health feature parameters combined with EMD-ICA-GRU is proposed to predict the SOH of Li-ion batteries. The capacity regeneration phenomenon and data noise are decomposed by empirical mode decomposition (EMD), and then the SOH-related health indicators are deeply mined using incremental capacity analysis (ICA), and the peaks of IC curves and their corresponding voltages are extracted as the input of the model. Then, gated recurrent units (GRUs) are formed into a combined SOH estimation model by adaptive weighting factors. Finally, it is validated against the NASA lithium battery dataset. Experimental results show that the mean squared error (MSE) of the proposed combined model can reach about 0.3%, and it has stronger generalization and prediction accuracy than other algorithms driven by independent estimation data.