This study presents a simulation‐based evaluation of a hybrid fiber Bragg grating (FBG) temperature estimation framework combining the extended Kalman filter (EKF) and unscented Kalman filter (UKF) for electric vehicle (EV) lithium‐ion battery cells. The work addresses key gaps in nonlinear thermal observability, robustness under high C‐rate excitation, estimator reliability across wide frequency ranges, and the absence of a structured pre‐experimental validation methodology. Rather than replacing experimental validation, the study establishes a disciplined simulation framework to reduce hardware risk and clarify estimator behavior prior to deployment. A physics‐informed electro‐thermal pouch cell model is coupled with spatially distributed multi‐FBG sensing in a fully reproducible simulation environment. EKF, UKF, and a dynamically weighted EKF‐UKF hybrid are evaluated across broad C‐rate and excitation frequency sweeps. Performance is quantified using mean error (ME), mean absolute error (MAE), root mean square error (RMSE), and maximum absolute error (MaxE), alongside amplitude preservation and responsiveness metrics. To stress‐test robustness beyond idealized assumptions, a stochastic realism framework introduces colored‐, heavy‐tailed‐, nonstationary‐noise with bias drift, delay, and jitter. Under these compounded disturbances, the hybrid estimator achieves the lowest global error metrics (ME = 0.8689, MAE = 0.8689, RMSE = 0.9472, MaxE = 1.8819), outperforming both EKF (RMSE = 1.1217) and UKF (RMSE = 1.4050) while also suppressing extreme deviation events relative to the standalone filters. The reduced RMSE confirms improved energy‐sensitive accuracy, and the bounded MaxE demonstrates enhanced stability under heavy‐tailed disturbances. A dynamically weighted confidence‐performance fusion strategy enables online dominance shifting toward the locally most reliable estimator, leveraging EKF stability at lower C‐rates and UKF nonlinear robustness at higher excitation. Results show that the hybrid framework consistently provides superior robustness and dynamic fidelity across operating regimes without introducing estimator‐induced lag. Overall, the study establishes a quantified, regime‐aware, and noise‐robust simulation baseline for multi‐FBG lithium‐ion temperature estimation, providing a structured foundation for targeted experimental validation and future integration within advanced battery management and energy‐aware vehicle systems.
Jooste et al. (Thu,) studied this question.