Uncertainty-Aware Control of Formula SAE Electric Vehicle Battery Systems in Endurance Settings

This work develops an uncertainty-aware thermal management framework for Formula SAE Electric endurance events. Traditional approaches rely on deterministic safety margins derived from worst-case assumptions or limited empirical observations, leaving performance on the table or risking thermal violations. This work instead combines a physics-based lumped thermal model with statistical learning methods to construct a hybrid predictive model that simultaneously improves accuracy and quantifies uncertainty. Polynomial chaos expansion enables efficient uncertainty propagation and variance-based sensitivity analysis, identifying which sources of uncertainty most affect thermal risk. The framework produces probabilistic safety statements and sensitivity-driven design trade-offs, replacing anecdotal temperature observations with quantified risk assessments. Predictions are validated against measured thermal data to demonstrate predictive capability beyond retrospective curve-fitting.