Quantum Machine Learning for Battery Health and Thermal Risk Prediction
Year: 2025
Type: conference-paper
Abstract: The rapid growth of connected Electric Vehicles (EV) as part of modern Intelligent Transport Systems (ITS) motivates the need for real-time management of Lithium-ion (Li-ion) battery health and thermal risks. Li-ion batteries, although widely used, are prone to degradation and thermal runaway, posing significant challenges for safe and efficient EV operation. We present a Quantum Machine Learning (QML) and Agent-Based Model (ABM) that simulates and predicts EV behaviour under various battery degradation con- ditions. We use a Variational Quantum Neural Network (VQNN) trained on NASA battery datasets to classify EVs into four cate- gories: healthy, degraded for fixed chargers, degraded for mobile chargers and thermal runaway based on their State of Health (SoH) and thermal risk. We simulate a fleet of 50 EVs and demonstrate that the VQNN classifies healthy and batteries at risk and achieves an accuracy of 96%, outperforming an LSTM baseline of 91%. The VQNN outperformed the LSTM on the NASA dataset, achieving a significantly lower RMSE of 0.1562 compared to 0.6455. Our re- sults show that the QML-ABM framework improves battery health and thermal risk prediction, enhances safety, Mobile Charging as a Service (MCaaS), and supports real-time decision-making in ITS.
DOI: 10.21427/843m-tj72
Author: Mutiso Mutua, Alexander
Institution: Technological University Dublin
Language: English
Subfield: Automotive Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7083177906
FWCI:
Cites:
Cited by:
Related to: