A Homogenized FEM Data-Driven Model for Calculating Leakage Inductance of High-Frequency Transformer with Litz-Wire Winding

Accurate evaluation of the leakage inductance of high-frequency transformers is significant for improving the operating model and power transmission efficiency of isolated dc-dc converters. Compared with the conventional methods, the data-driven methods inspired by machine learning seem to be more advantageous for leakage inductance assessment. However, the complex structure of the litz-wire windings hinders the extraction of the simulation datasets, leading to challenges in the application of the data-driven methods. This article proposes a data-driven method that can efficiently extract FEM simulation datasets of high-frequency transformers with litz-wire windings. The litz-wire windings are homogenized to greatly reduce computational costs. Global sensitivity analysis and dimensionless processing are adopted to determine the feature parameters and to expand the applications, respectively. Expressions describing the frequency variation of the simulation data are derived. A double 2-D FEM model based on length factors is presented to illustrate the 3-D characteristic of leakage inductance. Moreover, to enhance the performance of regression analysis, an improved random forest algorithm is implemented in Python. Finally, the accuracy of the proposed method is verified by comparison with the measured four different transformer prototypes and current analytical methods.