This contribution presents the early results of a R&D collaboration established between the University of Cantabria and the Power Transformer Manufacturer EFACEC. This paper tests two different techniques of steady state thermal modelling applied to power transformer windings, Computational Fluid Dynamics (CFD) and Thermal-Hydraulic Network Modelling (THNM). The state of the art of thermal modelling demonstrates that these techniques have been used to calculate both average and hotspot winding temperatures by solving the winding temperature and flows profiles within the winding. THNM models have worse accuracy than CFD in the predicted results. The improvement of these THNM models is a topic of study in transformer thermal modelling. The first goal of this paper is to test the accuracy of a new calibrated THNM model of a disc-type winding immersed in mineral oil. Then, this THNM model is tested with esterbased liquids, such as a natural ester and a synthetic ester, to determine if it can be applied to these liquids without further calibrations. Finally, the cooling performance of both type of liquids is compared using only the THNM model results. The results of this work show that the THNM model developed herein gives good estimations of temperatures compared to those obtained with CFD for both types of liquids. Also, the use of alternative fluids leads to lower temperatures when considering the same oil flow rate and temperature as inlet boundary condition.