In this paper a model and a methodology for using artificial neural networks to solve the load flow problem are proposed. An evaluation of the input data required by the ANN as well as its architecture is also presented. The ANN model used in this paper is the multilayer perceptron, and the training process is based on the second order Levenberg-Marquardt method. The proposed methodology was evaluated using the Ward-Hale 6 bus, the IEEE 14 bus and the IEEE 30 bus systems, considering normal operating conditions (base case) and different contingency scenarios, including different load/generation patterns. The simulation results show the excellent performance of the ANN, proving its ability to solve the load flow problem.