A comprehensive study on state-of-charge and state-of-health estimation of sodium-ion batteries

Nowadays, sodium-ion batteries (SIBs) are becoming another kind of potential energy storage batteries in the energy storage systems (ESS) with their unique advantages compared with the lithium-ion batteries. The accurate estimation of key battery states, especially state-of-charge (SOC) and state-of-health (SOH), ensures the safe and reliable operation of the ESS. In this paper, complete experimental tests on a commercial SIB are performed, and different SOC and SOH estimation methods of the SIB are compared comprehensively, which provides a basis for the selection of state estimation methods for the SIB. For SOC estimation, the battery is modeled with the third-order equivalent circuit model based on the electrochemical impedance spectra test results, and model parameters are identified by the particle swarm optimization algorithm. Moreover, three real-time model-based estimation algorithms, extended kalman filter, unscented kalman filter, and particle filter are adopted to estimate the SOC of the SIB. For SOH estimation, three health indicators that are strongly correlated with SOH are extracted. Furthermore, three lightweight data-driven estimation algorithms, multiple linear regression, ridge regression, and support vector regression are employed to estimate the SOH of the SIB. Through comprehensive comparison, experimental results indicate that unscented kalman filter and ridge regression are the most suitable algorithms for SOC and SOH estimation of the SIB respectively. Specifically, the mean absolute error (MAE) and root mean squared error (RMSE) of the SOC estimation are lower than 1.2% and 1.6% respectively. Both MAE and RMSE of the SOH estimation are lower than 0.06%. In addition, some unique characteristics of SIBs are found.