Nonlinear, highly varying loads in power systems can cause undesirable cond
itions. such as flicker and harmonics. Detailed models are needed to study
ways to mitigate the impact of these loads in the power system. Numerous st
udies rely on stochastic techniques to model nonlinear loads. Two alternati
ve models are proposed in this paper: a chaos-based model and an artificial
neural network model. The output data from both models are used in an elec
tromagnetic transients program (EMTP) study. The performance of both models
is compared in terms of how well they describe the impact of art electric
arc furnace in the rest of the system. Applications for the models are disc
ussed.