Monitoring and Prognostics of Lithium-Ion Batteries Using Dual Gaussian Process Regression

Monitoring and predictive modelling of lithium-ion battery performance are critical for advancing energy storage technologies. This study employs a Dual Gaussian Process Regression (GPR) framework to model both State of Health (SOH) and Remaining Useful Life (RUL) by leveraging indirect health indicators derived from voltage, current, and temperature data in the NASA lithium-ion battery dataset. Three predictive models are evaluated to account for both linear and nonlinear degradation patterns. Among these, Model 3 demonstrates the highest accuracy, achieving an RMSE of 0.0040 for SOH and an MAE of 12.4499 for RUL on battery B0005. By integrating capacity degradation with indirect health indicators, Model 3 offers a robust and reliable framework for battery performance prediction. Future work can explore the inclusion of additional health indicators to further enhance predictive accuracy.