An Ultrafast DC-DC Converter Design Methodology Combining Piecewise Linear Analytical Modelling and Genetic Algorithms


Niklas Fritz, David Heidenberger, Carsten Fronczek, David Bündgen, Rik W. de Doncker, 2022 IEEE 23rd Workshop on Control and Modeling for Power Electronics (COMPEL), 22. Juli 2022.



Design automation is an important trend in power electronics that often relies on multi-objective optimization and analytical modelling. For the optimization, meta-heuristic techniques using genetic algorithms or artificial intelligence become increasingly important. This paper proposes an imple-mentation technique for analytical converter models utilizing multi-dimensional matrices to evaluate designs throughout their entire operating area. Moreover, the time domain is included in the matrices, assuming piecewise linear waveforms, and calculation procedures are introduced to obtain RMS values, Fourier spectra, and many other parameters. The converter losses are calculated as well. This makes the proposed implementation technique applicable to the modelling of a variety of converter topologies. The proposed method is combined with genetic opti-mization, and an exemplary optimization of a 100 kW three-phase dual-active-bridge converter is presented.