State of Health Estimation for Lithium-Ion Batteries Using Voltage Curves Reconstruction by Conditional Generative Adversarial Network
Xinghua Liu, Zhichao Gao, Jiaqiang Tian, Zhongbao Wei, Changqing Fang, Peng Wang
Abstract
Battery health assessment is crucial for the safe and stable operation of electric vehicles. Accurate and efficient estimation state of health (SOH) ensures effective battery maintenance. The estimation accuracy of data-driven methods as an essential tool for battery SOH estimation depends on the quality of the data. Data missing is undoubtedly a significant challenge for data-driven methods. Based on this, this paper proposed a novel method for lithium-ion battery SOH estimation, which relied on the reconstruction of battery voltage data. The proposed method comprises two main components. Firstly, a conditional generative adversarial network (CGAN) is trained using collected historical charging voltage data and validated under various degrees of data loss. Secondly, an improved gated recurrent unit network (GRU) with sparrow search algorithm (SSA) is proposed for the estimation of lithium-ion battery health states after data reconstruction. The proposed SOH estimation method has been validated in various battery aging experiments. Compared with several other machine learning algorithms on reconstructed datasets, a significant performance improvement is observed.