The wide application of electric arc furnace (EAF) loads usually cause negative effects in the grid, especially drastic voltage fluctuation. Chaos theory is used to study ac EAF voltage fluctuation mechanism in this paper. Firstly, C-C method is adopted to reconstruct phase space of voltage time series. According to BDS(named after Brock, Dechert, and Scheinkman) statistics, the optimal delay time and embedding dimension are determined. Then, trajectory divergence rates in the phase space are calculated by improved Wolf algorithm as nearest points evolve along the trajectory. By fitting these divergence rates through least square method, the maximum Lyapunov exponent (MLE) is obtained. Results show that the voltage fluctuation of ac EAF has chaotic property, which lays the foundation for analysis and prediction of voltage fluctuation using chaotic method.