The recent changes in power systems give rise to new scenarios in which power transformers must continue to perform satisfactorily. Resonant transient phenomena caused by the interaction of the power grid and the transformer have recently led to serious failures. It is therefore essential to carry out network transient studies in order to evaluate potentially dangerous events of this type. This requires high frequency transformer models having an adequate representation of the damping. In this work, a white-box transformer model is presented together with a new and robust procedure for determining its parameters that enforces the stability of the model. In this way, substantial improvements are proposed with respect to models currently available in the literature in the methodological aspects needed for successful practical implementation. The methodology for determining basic impedance data, obtained using the Finite Element Method is first described and then a new methodology for the subsequent processing of these data to obtain the parameters of a stable and robust model is presented, using Vector Fitting in combination with Particle Swarm Optimization. The model has been successfully validated using a large power transformer as a case study within the scope of the JWG CIGRE A2/C4.52.