Design and optimization of an electric vehicle battery thermal management system using CFD simulations and CFD-derived GPR-ANN metamodels
Date
2026
Authors
Wong, Chon Him
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Transportation is a major contributor to global greenhouse gas (GHG) emissions. Battery electric vehicles (BEVs) have gained widespread adoption in passenger car applications due to their high energy efficiency and potential to reduce carbon emissions. Extending BEV technology to medium-duty trucks (MDTs), which are widely used in commercial transportation, can further improve energy efficiency and reduce GHG emissions.
The thermal management system (TMS) for the battery energy storage system (BESS), propulsion motors, and power electronics is essential to ensure safe, reliable, and efficient BEV operation while extending battery life. For electric medium-duty trucks (e-MDTs), which operate under diverse and demanding duty cycles, designing an effective TMS is particularly challenging due to the complex heat transfer processes within battery packs and their associated liquid-cooling and heating systems.
High-fidelity computational fluid dynamics (CFD) simulations are typically required to evaluate and optimize BESS thermal management performance. However, such simulations are computationally intensive, making large-scale design exploration across varying operating conditions impractical.
This work develops data-driven Gaussian Process Regression (GPR) and Artificial Neural Network (ANN) metamodels to approximate the thermal behaviour of BESS and enable efficient Battery Thermal Management System (BTMS) design optimization. A numerical BTMS model for an e-MDT is developed in MATLAB/Simulink, and CFD simulations are conducted for multiple battery pack and BTMS configurations under representative driving conditions. The resulting simulation data are used to train GPR-ANN metamodels that predict cooling performance without requiring repeated CFD simulations.
The proposed framework reduces simulation time by approximately 97% while maintaining a maximum prediction error of 1% for the selected operating scenarios, providing an efficient and accurate approach for the model-based design and optimization of BTMS for e-MDTs.
Description
Keywords
simulation-driven design, thermal management system (TMS), computational fluid dynamics (CFD), data-driven metamodels, neural networks, design and optimization, Gaussian process, reduced-order model, battery energy storage system (BESS), electric vehicles (EV)