DC arc furnaces are more and more applied in large industrial systems and represent one of the major sources of perturbations for the feeding system. This paper deals with the problem of the dc arc modeling using three well-known chaotic attractors (Ro/spl uml/ssler, Chua, and Lorenz attractors). A new tuning procedure is adopted to determine the most adequate parameters of the attractors to model the dc arc furnace. Waveform distortions and voltage fluctuations indexes are calculated from the simulation results of a whole existing plant and compared with measured data. The proposed models of dc arc furnace can be used to assess the impact in terms of power quality of the dc arc furnace when planning new plants or evaluating the performances of compensating devices.