Static EIS multi-frequency feature points combined with WOA-BP neural network for Li-ion battery SOH estimation

Electrochemical impedance spectroscopy (EIS) has been established as an essential and non-destructive technique for estimating the state of health (SOH) of lithium-ion batteries (LIBs). However, the extraction of effective and straightforward health indicators (HIs) from either original or derived EIS data is critical for practical applications in SOH estimation. In this study, we explore static EIS multi-frequency feature points as HI based on Nyquist plots at a various SOH for commercially available batteries. Subsequently, the extracted HIs are fed into a whale optimization algorithm back propagation (WOA-BP) neural network to achieve the accurate battery SOH estimation. The developed model is validated using four battery samples cycled at the same condition, which can achieve a low root-mean-square error range from 0.23 % to 0.43 %. Notably, even when employing the untrained data with our model, it can achieve a commendable root-mean-square error during practical validation. Moreover, our WOA-BP model exhibits superior accuracy compared to other mainstream algorithms and shows the significant application potential even without historic EIS data. The results indicate that selecting characteristic frequency points can greatly reduce testing time compared to utilizing full impedance spectroscopy as HI, presenting an effective strategy for battery SOH estimation in real-world applications.