A PHYSICS-INFORMED NEURAL NETWORK DIGITAL TWIN FOR BATTERY STATE ON HEALTH PROGNOSTICS IN EV FLEETS.

This study presents a novel Digital Twin framework for the prognostics of Battery State of Health (SOH) in Electric Vehicle (EV) fleets. The core of the framework is a Physics-Informed Neural Network (PINN) designed to overcome the limitations of purely data-driven models, which often lack generalize-ability and physical interpretability. By embedding governing equations of battery degradation as soft constraints within its architecture, the PINN leverages operational fleet data while adhering to fundamental electrochemical principles. This hybrid approach enables accurate, real-time SOH estimation and reliable prediction of the battery, Remaining Useful Life (RUL) directly from vehicle telematics. Validation using real-world fleet data demonstrates the model's robustness in managing varied usage patterns and its superior performance over conventional methods. The proposed digital twin provides fleet operators with a scalable, predictive tool for health-aware management, aiming to enhance operational reliability, optimize maintenance, and reduce the total cost of ownership.