Data-driven online SOH estimation of a grid-connected BESS: Accuracy improvements and lifetime prediction

This analysis develops a Coulomb-counting-based online state-of-health (SOH) estimator using detected partial discharge segments for an existing operational battery energy storage system (BESS). The estimator is based on the Coulomb counting method and is applied to real operational data of all cells composing the BESS. It facilitates SOH assessment as it does not require stopping the BESS’s ordinary operational usage. The estimator is refined with a linear regression method and a customised extended Kalman filter (EKF) for state-of-charge (SOC) estimation, minimising the temperature’s influence on the SOH estimation. Finally, a remaining useful life (RUL) analysis is conducted. A linear RUL predictor with a particle filter (PF) predictor are compared. The online SOH estimator outputs one SOH value per cell whenever a partial discharge segment is detected. The SOH values show a declining trend over two years, from 97.5% to 92.6% at system level. However, the obtained values reveal a strong correlation between SOH estimations and operational temperature. On the one hand, the linear regression erases the correlation, but with the risk of overcompensating for the temperature influence. On the other hand, the SOH estimations using the SOC from the EKF exhibit lower correlation with temperature, yielding promising results. Consequently, the authors suggest considering the SOH from operational data directly or after applying the EKF to estimate the SOC. The resulting SOH values are then an input to the RUL prediction analysis. Both the proposed RUL predictors provide similar results in terms of point-based expected mean lifetime. The predicted lifetime ranges between 9 and 14 years when using SOH values from operational data, and between 12 and 17 years with SOH values after estimating the SOC from the EKF. Instead, the RUL on the SOH from the linear regression would predict a mean BESS lifetime over 20 years and with a wider uncertainty range. Overall, this study demonstrates that the online SOH estimation method using the SOC derived from the customised EKF is the most balanced strategy for this BESS, improving reliability under temperature-varying operation while remaining consistent with available field capacity reference.