Based on electromagnetism-like mechanism (ELM) algorithm, a high fault-tolerant method for voltage sag state estimation is proposed. In this paper, the global optimization performance of the electromagnetic simulation algorithm is improved by the improvement of the anti disturbance population movement model, the adaptive mutation and the elite strategy. The analytic method and fault point method are introduced to determine the interval of line fault, and the redundancy collocation method is used to ensure the non singularity of the observation matrix. A voltage sag state estimation model is constructed by using the difference between the short circuit fault estimation and the historical statistics. Through the IEEE24 node Standard Test System Simulation and comparison with the genetic algorithm simulation, ELM has obvious advantages in the search efficiency and accuracy. This method can be applied to the voltage sag state estimation of a power system with symmetrical and unsymmetrical fault.