Battery Degradation Forecasting with FMU and Optimization Framework Using Synthetic Generated Data

Abstract

For successful operation and life-cycle management of electric vehicles (EVs), accurately predicting battery degradation is paramount. Long-term experimental aging data are typically scarce, costly, and proprietary, which complicates model calibration and validation.  This paper presents a hybrid co-simulation and optimization framework that predicts lithium-ion battery degradation, utilizing synthetic data generated from a Functional Mock-up Units (FMUs) representing coupled thermal and electrical-based system dynamics. Using two FMUs—a Thevenin-based electrical model and lumped-parameter thermal model—we established a Python co-simulation environment, following the FMI 3.0 standard. By varying ambient temperature, C-rate, and duty-cycle inputs, we generated synthetic operating profiles to represent realistic conditions for an EV. Embedded in the loop was an Arrhenius-type degradation law that was temperature-dependent and whose parameters were determined using sequential least-squares and particle-swarm-optimization algorithms. This paper demonstrated that the proposed approach could reproduce known thermal–electrical interactions responsible for capacity fade and closely mirrored open experimental data from the NASA Ames Li-ion Battery Aging Repository [21]. Quantitatively, we benchmarked R² = 0.94; RMSE = 0.032; a degradation-rate ratio of 0.92 when compared to experimental trajectories, confirming both physical plausibility as well as statistical support. The results indicate that synthetic data generated from FMU can replace expensive experimental laboratory validation of battery degradation studies, allowing for controlled virtual degradation experiments and often reproducible digital-twin calibration for predictive battery health management in future electric vehicles and plug-in hybrid vehicles