Resumo:
Water supply comprises one of the aspects of basic sanitation where the fundamental rights of every citizen must be guaranteed by virtue of the laws that govern its guidelines. The public power, at the municipal level, is one of the main responsible for the management and operation of water supply systems, managing the infrastructure and facilities in all processes to ensure the full operation to fulfill the obligations surrounding the supply of resources. Considering the theme of water losses, network simulation approaches allow adjustments and propositions of improvements that ensure efficient water supply, however due to the uncertainties regarding input parameters for realistic predictions the calibration step is necessary. This paper aims to apply a calibration methodology with artificial neural networks for use in water distribution networks. The methodology consists in the use of artificial neural networks to make the calibration from input data in the real network, with the use of Python as the programming environment used. The results show that these tools prove to be promising for analyzing this type of problem, considering the complexity of the hydraulic analysis in the face of different scenarios and the large amount of information generated. The calibrations show that the use of Artificial Neural Networks achieves good levels of calibration amid the scarcity of information in real networks and achieves better results compared to other calibration methods such as the random iterative search method. Thus the continuity of this research can strengthen better calibrations targeting real systems and the control of losses in their water distribution systems. New studies and technologies can contribute to the reduction of losses and consequently the saving of water, a resource that is scarce and crucial for society.