In the current study, it was tried to describe a method for determining thermal characteristics of integrated micro-circuits to identify thermal parameters of multidisciplinary (thermal-electric) behavioral models. The problem is solved on the example of high-frequency pulse voltage converters. A solution was proposed to refine the minimum structure of the thermoelectric model based on an iterative least squares method using the Levenberg-Marquardt algorithm, as well as a graph of the spectral den-sity of time constants. This made it possible to reduce the influence of the filtering factor in the deconvolution operation when building a thermal model using the structural function of the thermal characteristic transition. Also, the results obtained can be used to build integrated circuits (IC) behavioral models, taking into account the thermal processes occurring in them.