THESE de DOCTORAT DE L’UNIVERSITE DE LYON opérée au sein de l’Université Claude Bernard Lyon 1 Ecole Doctorale N° 160 ECOLE DOCTORALE ELECTRONIQUE, ELECTROTECHNIQUE, AUTOMATIQUE DE LYON Spécialité de doctorat : Génie Électrique Soutenue publiquement le 13/11/2019, par : Alexis FOUINEAU opérée au sein de l’Université Claude Bernard Lyon 1 Ecole Doctorale N° 160 

 Medium Frequency Transformers (MFT) are an innovative technology compared to low frequency transformers, with the promise of reduced volume and increased efficiency. This PhD thesis focuses in particular on their design for high voltage, high power applications, such as high voltage and medium voltage DC networks, as well as railway networks. In these applications, MFTs are used in converters that can generate specific constraints to be taken into account during their design: non-sinusoidal signals, polarization voltage, target inductance values. Moreover, the technological choices currently available for the realization of MFTs are numerous, and there is currently no consensus on any technology for any given application. Trends could be identified using a tool to classify MFT designs from the literature. Thus, the most promising technologies were selected and retained for the future. Based on these technologies, a design methodology was developed to quickly and semi-automatically design and compare MFTs with different technological choices. It consists of three steps: pre-design, analytical design, and validation. The complete analytical design of the MFT with different technological choices is carried out using an automated design tool developed during this thesis, named SUITED (SUpergrid Institute TransformEr Design). This methodology requires models and data for each of the components and phenomena of the MFT. Concerning the magnetic core, a review and selection of models from the literature were carried out for the evaluation of the magnetizing inductance and magnetic losses. In addition, magnetic characterizations have made it possible to highlight the impact of certain technological processes on the levels of loss of magnetic cores made of nanocrystalline material, which is an excellent candidate for MFTs. Concerning the windings, analytical models to calculate the magnetic field, leakage inductance and skin and proximity effects were developed and compared with those in the literature and simulations. These models are proving to be more accurate on the MFT geometries considered. On top of that, a new method for evaluating the parasitic capacitances of windings with rectangular turns has been successfully implemented and validated. Thermal networks have been identified for the different MFT geometries. The thermal resistances of conduction, convection and radiation are calculated from detailed models. In particular, the anisotropy of materials is taken into account for thermal conduction, and the convection coefficients are evaluated via different correlations for each face of the MFT. The thermal networks are then solved iteratively and analytically to take into account the non-linearity of the thermal resistances while optimizing the required computation time. Finally, this entire design methodology was applied to three case studies corresponding to the target applications: high voltage, medium voltage and rail. The results obtained do show the performance and necessity of this approach.