Robust Battery SOC Estimation via Hypernetwork-Based Parameterized Physics-Informed Neural Networks

Accurate state-of-charge (SOC) estimation is essential for battery management systems (BMS). However, conventional methods face challenges with parameter dependency or struggle with stability under unseen operating conditions. This paper proposes a novel hybrid framework integrating temporal sequence learning with parameterized physics-informed neural networks (PPINN) governed by electrochemical constraints. The architecture employs a hypernetwork that generates dynamic weights characterized by initial SOC values, enabling adaptive learning across diverse operating conditions. This allows the model to learn a generalized solution space by combining physics-based knowledge with data-driven modeling, thereby avoiding repetitive training. Experimental validation across various temperatures and driving cycles demonstrates superior performance, achieving generalization capability and robustness with high accuracy.