Abstract: Medium frequency transformer is the core magnetic component of power electronic transformer, and plays a key role in electrical isolation and voltage conversion. The optimal design of insulation structure of medium frequency transformer can effectively improve the insulation reliability and power density of transformer, which has been widely concerned by researchers. This paper presents a global optimal design method for medium frequency transformer insulation structure based on C-MOEA/D constrained multi-objective optimization algorithm. In this paper, the overall optimization process is given according to the characteristics of medium frequency variable insulation structure, and the genetic algorithm C-MOEA/D based on non-dominated sorting and reference vector is proposed to optimize the insulation structure parameters. Aiming at the problems existing in the optimization process, we constructed a two-dimensional electric field analysis parametric model of medium frequency transformer, and analyzed and compared its calculation accuracy based on the three-dimensional model. Secondly, the key structural parameters affecting the maximum field intensity were determined by Latin hypercube sampling and sensitivity analysis based on variance, and the structural parameters with low sensitivity were fixed to achieve the dimensionality reduction of parametric model decision variables. Then, according to the isolation voltage waveform of the medium frequency transformer, the breakdown field strength test platform of the insulation material was established, the breakdown field strength of the commonly used insulation materials was determined, and the insulation constraints were given. Finally, aiming at the maximum field strength and volume, we optimized the design and application of 11 insulation structure parameters of a 25 kV medium frequency transformer. The optimization results show that the maximum field strength and volume are reduced by 19.6% and 23.9% compared with the original design.