In this paper, the main aim is to introduce a regular procedure to select the location of arresters (voltage limiters) in a way to optimally confine the risk of flashovers in a given high voltage network. Mathematically, simulation optimization is used to create an acceptable meta model from a real model which, on its own, is created to represent a real high voltage network in a transient analyzer program, The proposed meta model is a three layer perceptron neural network which is learned to estimate the risk function with arrester locations as inputs and network risk of flashover as output. Genetic algorithm is then invoked to locate the best position of arresters. The proposed method is then accomplished to set the location of 5 identical arresters in Iranian South East 400 kV network and the results are shown to be acceptable by simulating the real model of this network.