Lithium-Ion Battery Health Estimation with Incomplete Charging Data and Spiking ResNet

The accurate estimation of state of health (SOH) plays a crucial role in ensuring the safe operation of batteries. Most existing SOH estimation methods are based on complete charging or discharging data. However, in real-world applications, data are often derived from partial charging or discharging processes influenced by varying user habits, which brings great challenges to the accurate SOH estimation of batteries. In this article, a SOH estimation method based on incremental capacity (IC) curve reconstruction and spiking ResNet is proposed to deal with the SOH estimation issue using partial data. The monitoring data are preprocessed using differentiation and Gaussian filtering to obtain the IC curves. A nonparametric degradation model is constructed using functional principal component analysis (FPCA). Subsequently, the functional principal component (FPC) scores are estimated by the maximum likelihood estimation (PAMLE) algorithm to reconstruct the IC curves. Ultimately, the SOH can be estimated through the spiking ResNet. The proposed method is evaluated using the XJTU battery experimental data. The results demonstrate that the proposed method provides more accurate and reliable estimation results than other methods.