The invention discloses a multi-target genetic algorithm based medium-frequency amorphous alloy transformer optimization method. The method comprises the following steps that 1, optimization variables are selected from parameters of a medium-frequency amorphous alloy transformer to establish a target function optimized by the transformer, and the ranges of iteration times and optimization variables are defined to produce an initial population; 2, the initial population is segmented into a plurality of sub-populations, sub-target functions are distributed for the sub-populations, a fitness function is established, and a fitness value is calculated; parallel selecting operation is performed in each sub-population, and individuals in the sub-populations having higher fitness are selected out to form new sub-populations; 3, the new sub-populations are merged into a complete new sub-population, and nest generation of new complete population is generated through recombination and variated operation; 4, constant iteration is performed, and the steps 2 and 3 are repeatedly and sequentially performed to calculate an optimal solution of the optimization variables. The maximum efficiency and minimum loss of the transformer can be ensured, the transformer can operate more stably, and the service life is longer.