[en] This paper proposes to compare two alternative ways for describing contribution of magnetic domains in dynamic hysteresis and associated anomalous losses at the macroscopic scale. This study enables to reproduce the dependence of anomalous losses on the frequency of excitation f. Anomalous losses per cycle follow a f^1/2 law, matching the experience-based model of Bertotti. While both models restore the well-known low-pass behavior on a 1D-slab, one of the two exhibits a higher level of anomalous losses and a stop-band frequency related to a length-parameter Λ, lumping the domain pattern complexity. For both models, sensitivity of anomalous losses with respect to this dynamic parameter Λ is assessed. As an illustration, Λ appears to be the key parameter to improve energy efficiency in power transformers. Finally, stakes are given for transformers and for the whole electrical steel production to highlight predictive potential gain. Anomalous losses raise up to 230 GW equivalent installed capacities, including dissipation through both grain oriented and non-oriented electrical steel. This represents 5% of worldwide electric capacities