The constant growth of the population leads to increased energy consumption and higher generation of electricity. Power transformers are the first devices used to transfer the electricity produced in power plants into the grid. The heat resulting from the device power losses in the energy conversion process has to be dissipated. Currently, active elements of the power transformers are cooled down in mineral oil. It performs well in such devices due to its good thermal and dielectric properties. However, industry and global society require environmental-friendly alternatives. Hence, the aim of this study is to present the numerical results of the power transformer cooled using conventional mineral oil or biodegradable ester oil. The unit works in so-called oil-natural convection/air-natural convection (ONAN) cooling mode. This device, manufactured by Tadeo Czerweny, has the nominal power of 8.5 MVA. The cooling is realised by external radiators. Two coupled numerical models have been developed to investigate the fluid flow and the heat transfer inside the transformer. One model performs a detailed simulation of the oil flow within the windings of the transformers. The effective values of anisotropic thermal conductivity have been defined for the winding and the core. The domain of this model included 1/16th of the windings. The second model includes the power transformer interior and computes the flow of the oil within the analysed transformer. The distribution of the power loss within the magnetic core and windings was evaluated within electromagnetic simulations. The coupling strategy between all three models allows for the temperature distribution analysis within the transformer. The result discussion includes the thermal performance evaluation of the power transformer using mineral oils and natural esters for warm (Argentinian) and moderate (Polish) climate zones. Due to the complexity of the fluid and heat flow, as well as the electromagnetic effects the computational time for the conducted simulations, is significant. However, the results show that the proposed strategy of the coupled models allowed to identify the hot-spot temperature, the oil flow within the tank and finally the heat dissipation in one simulation approach.