Artificial Intelligence-Driven Thermal Management in Electric Vehicle Traction Inverters: Cooling System Optimization and Real-Time Temperature Prediction

The accelerated adoption of electric vehicles (EVs) has intensified technical demands on traction inverters, particularly regarding power density, energy efficiency, and thermal management. Silicon Carbide (SiC) MOSFET power modules, with their superior switching performance and thermal characteristics, offer significant advantages but introduce new thermal challenges due to their increased power density and compact packaging. The primary objective of this study is to enhance real-time thermal control and prediction in a SiC-based dual traction inverter through the design and experimental validation of a high-performance cooling system integrated into a modular electric drive platform. To achieve this, multiple cold plate configurations were designed and subsequently evaluated using computational fluid dynamics (CFD) simulations, focusing on junction temperature, thermal uniformity, and pressure drop. To enable real-time thermal predictions, an artificial neural network (ANN) was developed and trained on datasets generated from high-fidelity CFD simulations and a CFD-derived reduced-order model (ROM) to estimate junction temperatures dynamically. Integrated into the inverter control system and supported by real-time sensor feedback, the ANN achieved prediction errors below 3.5% within 1 second across all operating conditions. The proposed framework offers a scalable and intelligent thermal management solution, supporting predictive maintenance and enhancing the reliability and efficiency of next-generation EV powertrains.