Hybrid Data-Driven and Physics-Based Modeling for Enhancing Battery Safety and Reliability in Electric Vehicles

Making sure lithium-ion battery systems are safe and dependable has become crucial as electric vehicles (EVs) proliferate. This study explores a hybrid modelling approach for early defect detection, thermal runaway prediction, and health diagnostics that blends physics-based models with data-driven machine learning (ML) methods. In order to improve forecast accuracy under actual driving situations, the research assesses the combination of electrochemical models, Long Short-Term Memory (LSTM) neural networks, and Gaussian Process Regression (GPR). The safe functioning of EV battery packs is greatly enhanced by results from simulated and experimental datasets, which show increased fault classification accuracy, greater generalisation to unknown conditions, and quicker reaction times. Lithium-ion battery systems' safety, dependability, and operational efficiency have become critical determinants of user trust and long-term sustainability as electric vehicles (EVs) continue to expand in the worldwide automotive industry. Because of the intricate electrochemical behaviour and vulnerability of these batteries to deterioration and failure, conventional battery management techniques often fail to correctly identify early-stage defects or anticipate crucial events like thermal runaway.