Differential Voltage Analysis Based Data-driven Approach for Battery Module SOH Estimation

State of health (SOH) estimation of electric vehicles (EVs) batteries is a critical assessment to determine the remaining useful life of the batteries. This paper proposes a differential voltage analysis (DVA) based approach to estimate SOH for batteries. Four commercial battery modules are studied and 1C charging data from the cycle aged modules is used. Full charging and partial charging for the second life application of retired EV batteries are considered. Two types of health features are proposed and extracted from the DV curves. One is the fractional areas under the segmented DV curves. Another one is the averaged values of the segmented DV curves. With the extracted features as input, a robust machine learning model based on TheilSenRegressor is built to predict the SOH. The estimated SOH calculated from the built models is shown to well follow the trend of the real SOH value. The averaged error percentage is achieved to be 0.65% for full charging and 0.98% for partial charging case.