Health monitoring systems for EV battery

¹ Department of Electronics and Communication Systems, Sri Krishna Arts and Science College, Coimbatore, India
² Associate Professor, Department of ECE, Sri Sivasubramaniya Nadar College of Engineering, Chennai, India
³ Associate Professor, Department of EEE, Sri Sivasubramaniya Nadar College of Engineering, Chennai, India
⁴ Assistant Professor, Department of Electronics and Communication Systems, Sri Krishna Arts and Science College, Coimbatore, India

Abstract

Electric vehicle (EV) battery health monitoring is crucial to guaranteeing dependability, performance, and safety. A data-driven State of Charge (SOC) prediction framework utilising machine learning models assessed across three distinct battery chemistries—Lithium-ion, Lithium Polymer, and Lead-acid in this paper. Different operating conditions were captured using publicly accessible datasets from Mendeley Data, the CALCE Battery Research Group, and open-source GitHub repositories. Due to its robustness and low computational complexity, a Random Forest Regressor (RFR) was employed as the main SOC estimation model. Its performance was compared with that of a Recurrent Conditional Variational Autoencoder (RC-VAE) to analyse modelling limitations and cross-chemistry generalisation. The Random Forest model is evaluated experimentally using Mean Absolute Error, Root Mean Squared Error, and the coefficient of determination (R²). The results show that the Random Forest model offers more consistent and dependable SOC predictions, whereas the RC-VAE performs worse under specific datasets and scaling conditions. Additionally, as a proof-of-concept, a voice-activated, lightweight chatbot interface was incorporated to enable users to ask questions about SOC information and get basic charging-related advice through natural language interaction. The suggested method demonstrates how well cross-chemistry SOC estimation can be combined with user-friendly interfaces for useful EV battery monitoring applications.