Using tens of seconds of relaxation voltage to estimate open circuit voltage and state of health of lithium ion batteries

Abstract

Relaxation voltage (RV) of a battery is informative since it not only approximates open circuit voltage (OCV) as time evolves, but it is also related to the battery's state of charge (SOC) and state of health (SOH). Given that RV is easy to obtain by simply stopping a battery's operation, it is an excellent data source to estimate battery states. Without using complete RV history whose acquisition is time-consuming and hinders further applications, this study uses Gaussian process regression model with the input of only a small portion of RV to rapidly and simultaneously estimate the OCV and SOH of a battery. Various input lengths are tested, showing that using only 30-s RV data, the mean absolute error (MAE) for predicting OCV is 2.99 mV, and that for estimating SOH is 2.76%. As soon as the voltage difference is also treated as the model input, we find that the MAE for the SOH estimation is further declined to about 1.83%. Compared to previous methods which either estimate single battery state or require minutes of RV data for estimation, the current model is able to perform multiple battery estimation using only first tens of seconds of data.