TIDE Enhances Battery Degradation Estimation with AI

Accurate battery health estimation is critical for managing battery-powered systems, especially in intelligent connected devices where errors can propagate. Traditional methods often lack the trustworthiness and interpretability needed for practical deployment and critical decision-making. This paper introduces TIDE, a novel battery degradation estimator designed to address these limitations by jointly optimizing for accuracy, trustworthiness, and interpretability. TIDE's three-component backbone integrates battery-domain knowledge for trustworthy estimation, a monotone residual component for aging-consistent refinement, and contextual learning to capture battery-specific operational effects, thereby boosting accuracy. The trained TIDE backbone is then distilled into a compact symbolic surrogate model, offering a concise, human-understandable interpretation of its learned estimation logic. Experimental results show TIDE improves estimation fidelity by nearly 20% over baselines, significantly reduces aging-consistency violations, and provides both component-level and model-level interpretability, making it highly suitable for real-world battery health monitoring.