In the designing preproduction of power transformers for calculating design data, a circuit models are widely used, which doesn’t fully reflect the structural features of the transformer and spatial energy conversion. This leads to a significant increase of the error of design calculations and calculations of the magnetization parameters in load operation of transformers. This is especially the case in networks with alternative generation of electrical energy. Therefore, the aim of the work is to estimate the influence of design features and non-linear characteristics of electrical steel on the magnetization parameters in loading variations mode of the transformer. To achieve this aim, an improved model of electromagnetic processes in loaded power transformers is proposed. Its scientific novelty makes sense in the conjugation circuit models of external network and user’s network with a spatial model of the magnetic field in the transformer. When the transformer working in operation mode, the magnetizing currents and idle currents are characterized by asymmetry and nonsinusoidality. Reducing the load of the transformer from operation to idle modes, these effects are enhanced. Therefore, the effective values of the magnetization currents increase in different intervals: (1.066%; 1.161%) for phase A; (1.275%; 1.303%) for phase B; and (1.187%, 1.308%) for phase C. At using of the data of the coupled circuit models and the magnetic field model for the ratios and coefficients of parabolic and polynomial regression provides high accuracy in calculating of the magnetization parameters under various transformer load conditions.