INTELLIGENT FAULT DETECTION AND THERMAL MANAGEMENT SYSTEM FOR HIGH-ENERGY DENSITY EV BATTERIES

Abstract

The growing global demand for sustainable energy has accelerated the deployment of thermal photovoltaic (PV) systems as an eco-friendly and renewable power source. These systems depend on uninterrupted operation and peak performance to ensure efficient energy output. However, they are vulnerable to various faults such as Maximum Power Point Tracking (MPPT) failures, Low Power Point Tracking (LPPT) issues, partial shading, and hardware degradation, all of which can significantly diminish efficiency and shorten system lifespan. As a result, accurate and timely fault detection and classification are critical for enhancing reliability and reducing maintenance costs. Traditionally, fault detection in EV batteries has relied on manual inspections and threshold-based monitoring via Supervisory Control and Data Acquisition (SCADA) systems. These conventional methods often fall short in handling large volumes of high-frequency data and identifying subtle or complex fault patterns. They are typically time-consuming, error-prone, and lack scalability— particularly in large-scale or distributed solar setups. To address these limitations, this project introduces an AI-powered fault detection and classification system that utilizes high-frequency operational data from EV batteries in combination with machine learning (ML) techniques to automatically detect and categorize faults. The system features a comprehensive pipeline including data preprocessing, class balancing using the Synthetic Minority Over-sampling Technique (SMOTE), feature reduction via Principal Component Analysis (PCA), and model training using Light Gradient Boosting Machine (LGBM), CatBoost Classifier, and the proposed Random Forest Classifier (RFC). A user-friendly graphical user interface (GUI) developed using Tkinter enables easy interaction with the system and provides real-time visualization of predictions and performance metrics. Experimental results demonstrate that the RFC model significantly outperforms the other models, achieving an accuracy of 99.82%, precision of 99.86%, recall of 99.86%, and F1-score of 99.86%, compared to 82.39% and 82.48% accuracy for LGBM and CatBoost, respectively. These findings validate the effectiveness of the proposed system as a robust, efficient, and scalable solution for fault detection in thermal PV applications.