EVGuardian: IoT-Driven Machine learning for Real-Time EV health monitoring and predictive maintenance

Abstract

• EVGuardian enhances EV safety through accurate motor temperature and battery life predictions. • IoT-enabled framework integrates sensor data with advanced machine learning models. • Regression models optimized with PSO boost predictive accuracy and reliability. • The framework achieved 99.81% accuracy in motor temperature and 98.4% in battery RUL estimation. • EVGuardian improves EV efficiency, component lifespan, and supports scalable e-mobility solutions. Electric vehicles (EVs) play a vital role in achieving global sustainability goals; however, accurate motor temperature estimation and reliable battery lifespan prediction remain significant challenges. In this work, we propose EVGuardian, an IoT-enabled machine learning framework designed for real-time condition monitoring and predictive maintenance of EVs, with a particular focus on permanent magnet synchronous motors and lithium-ion batteries. The proposed framework integrates rich multi-sensor data with advanced regression models, including Linear Regression, K-Nearest Neighbours (KNN), XGBoost, Support Vector Regressor, and Random Forest. These models are further optimized using Particle Swarm Optimization (PSO) to enhance predictive performance. Experimental results demonstrate the effectiveness of the proposed approach: the PSO-enhanced KNN regressor achieved 99.81% accuracy in motor temperature prediction with a root mean square error (RMSE) of 2.84 °C, while the PSO-optimized Random Forest model achieved 98.4% accuracy in remaining useful life estimation with an RMSE of 77.89. Overall, EVGuardian significantly outperforms conventional methods in terms of accuracy and reliability. The results highlight the potential of intelligent, data-driven health management systems to enhance EV safety, operational efficiency, and component longevity, thereby supporting the development of more resilient and scalable electric mobility solutions.