AI-Based Digital Twin Simulation for Predicting EV Performance and Battery Degradation

Abstract

Digital twin (DT) technology has emerged as a powerful tool in electric vehicle (EV) development, enabling real-time virtual modeling of vehicle and battery behavior. This paper presents a comprehensive DT framework that leverages real-world driving data and artificial intelligence (AI) to predict EV performance metrics (e.g. range, energy use) and battery degradation (state-of-health). The DT integrates sensors, cloud computing, and IoT to create a live model of the EV’s battery and powertrain. Data from actual driving cycles (including GPS route, driving style, weather, HVAC use) is fed into the DT to simulate battery usage. Machine learning models (ensemble regressors and neural networks) are trained on this data to estimate state-of-charge (SoC) and forecast long-term capacity fade. The framework uses an Azure IoT-based architecture for data collection and a cloud-based BMS (Battery Management System) with online Kalman filtering to update the twin. Experiments on published EV driving datasets and NASA battery-cycle data validate the approach. The results show that the AI-enhanced DT achieves very low prediction errors (e.g. <1% NRMSE for SoC estimation) and accurately identifies degradation trends, outperforming conventional methods. Figures illustrate the DT architecture and simulation workflow. This work demonstrates that combining AI and DTs can greatly improve EV range prediction and battery life estimation under real operating conditions.