Experimental and physics-informed machine learning analysis of an adaptive thermoelectric-assisted cooling system for electric and hybrid vehicle battery thermal management

Abstract

Efficient battery thermal management is critical for the performance, safety, and lifespan of modern electric and hybrid vehicles under dynamic operating conditions. This study presents an integrated experimental and data-driven analysis of a hybrid cooling system combining thermoelectric Peltier modules with Al₂O₃-based nanofluid-assisted liquid cooling. The system is benchmarked against a conventional radiator-based cooling configuration under varying flow rates (1–4 LPM) and nanoparticle concentrations (0.5–2.0 vol.%). The results demonstrate that demand-controlled thermoelectric activation enhances cooling performance by approximately 8%, while reducing auxiliary energy consumption by nearly 30%. A maximum heat-transfer enhancement of 53.4% is achieved at 4 LPM and 2.0 vol.% nanofluid concentration, whereas the highest temperature reduction of 31.6% is observed at low flow conditions (1 LPM). To enable predictive thermal analysis, multiple machine learning models are developed and compared with a Physics-Informed Neural Network (PINN). The PINN improves prediction accuracy by approximately 50% while ensuring physical consistency by embedding governing heat-transfer constraints. The proposed hybrid nanofluid-thermoelectric cooling framework offers a compact, energy-efficient, and scalable solution for next-generation electric vehicle battery thermal management, with enhanced thermal performance, adaptive control capability, and predictive reliability.