A novel intelligent framework for thermal management design of power electronics systems

Efficient thermal management of high-power electronics modules such as IGBTs is critical to maintaining adequate thermal safety margins and energy efficiency, yet it involves complex trade-offs between component temperature control and coolant pumping consumption. The main motivation of this study is to replace slow, expensive repeated CFD analyses with a fast, reliable design tool that can explore the full design space. To this end, a novel integrated framework is proposed that combines, for the first time in this application, a GMDH-type neural network as an accurate surrogate for predicting IGBT and diode temperatures and pumping power, a multi-objective chaos game optimization (MOCGO) algorithm solutions according toto-optimal designs, and the PROMETHEE method for scenario-based ranking of solutions according to practical design priorities. The synergy of these three components enables designers to move beyond single-point CFD studies toward systematic, decision-driven thermal design. The GMDH-NN achieved excellent predictive accuracy (test-phase R 2 > 0.995), while the MOCGO-based Pareto analysis revealed the fundamental trade-off between cooling performance and pumping power. Ten PROMETHEE-weighted scenarios further demonstrated the flexibility of the framework in accommodating diverse operational goals. The proposed framework offers a computationally efficient and adaptable tool for designing thermally reliable and energy-efficient cooling systems for next-generation power electronics.

Hali tarjima qilinmagan