Compact Thermal Modelling of Magnetic Components via Real Coded Genetic Algorithm

Abstract

The trend of increasing power densities in modern day power electronic systems is pushing components to their thermal limits, warranting the need for accurate thermal modelling. Unlike the thermal modelling approach for semiconductor devices, thermal modelling of magnetic components has not been standardised. Due to this lack of standardisation in the academic community, most magnetic component thermal models have not been evaluated for Boundary Condition Independence (BCI) and hence cannot be classified as compact thermal models (CTMs). In this paper we develop a CTM of an inductor using real coded genetic algorithm (GA) based on the DELPHI approach. First, a Detailed Thermal Model (DTM) of the inductor under DC excitation is developed and validated using experimental test results. Following which resistance network values were deduced from the DTM results for varied boundary conditions via optimisation by the real coded GA. A fully connected resistance network was observed to be the best representation of the inductor DTM. The resulting CTM is able to predict junction (winding) temperature within 5 % of the DTM results for a varied set of boundary conditions. The inductor CTM developed is a low computational cost alternative to the DTM and can be used in system level simulations to evaluate thermal performance for varied applications.