USO DE REDES NEURAIS NO PROCESSO DE AJUSTE DE HISTÓRICO DE PRODUÇÃO

The purpose of this work is to present a methodolog y for production history matching using proxy models generated through artificial neural networks. Optimization pr ocesses through genetic algorithm using the proxy models and the flux simulator area compared. The methodology was tested in three reservoir models with 4, 8 and 16 va riables, and one realistic synthetic model with 20 parameters, in or der to evaluate the performance of the technique wi th the increasing of the number of variable. The results obtained with t he proxy models are very similar compared to the re sults obtained with the simulator, showing as main advantage the r eduction of the number of simulations allowed by th e proposed methodology.