An accurate permeability value for a tight gas reservoir has become an important need in the natural gas industry. However, there have been and there are still many ways to calculate this variable, but has low reliability, since permeability in many cases does not follow a linear behavior.
For that reason, were here implemented the integration, operation, and optimization of linear methods (empirical, statistical multiple regression) and nonlinear methods (primarily neural networks) in order to improve the permeability prediction of Lajas Formation, using cores and well logs.
The aim of this paper is to present the results of the combination of linear methods (e.g. empirical and multiple regressions) with nonlinear methods called The Artificial Neural Network (Multilayer Perceptron Model) to which we call Hybrid model applied to reservoirs with tight gas features, in order to predict a continuous curve of permeability (for wells that do not have core or well logs information). The porosity has been taken from different curves: gamma ray (GR), resistivity (RD), neutron (NPHI), density (PHID), (RHOB), and each were adjusted to the porosity data obtained by core samples and sidewall cores. This method allowed validating and optimizing the petrophysical model.
This study was focused on Lajas Formation that belongs to the Cuyo Group (Lower to Middle Jurassic) in the Neuquen Basin. This geologic unit consists of gray sandstones of medium to fine grained with interbedded conglomerates, limestones and lenticular shales of varying thickness.
All data were collect from public sources and were processed for the present research.