Optimization of Thermal Management for Cooling System of Power Electronics Modules Consisting Insulated-Gate Bipolar Transistor Using Neuro-Regression Analysis and Non-Traditional Algorithms

Thermal management and extreme temperatures critically influence the performance of power electronics systems, especially those utilizing IGBT (Insulated-Gate Bipolar Transistor) and diode components. Various parameters govern the cooling efficiency of these systems. In this study, the IGBT temperature was selected as the objective function. To achieve temperature minimization, optimum values of design variables: coolant flow rate (L/min), distance from the vortex generator (mm), height (μmm), and width of the first pin-fin (μmm), and distance of the vortex generator from the surface (μmm) were determined. The mathematical modeling process employed Neuro-Regression analysis. The prediction performance of proposed 14 different regression models were evaluated using R2Training, R2Testing, R2Validation indexes and boundedness check criteria. The Differential Evolution, Nelder Mead, Simulated Annealing, and Random Search algorithms were applied to minimize IGBT temperature. The FOLN (First Order Logarithmic Nonlineer) model emerged as the most successful, achieving a minimum temperature lower than the experimental dataset given in literature. The results indicate a 12 % reduction in the minimum IGBT temperature.