Application of neural networking for calculation of permeability parameters in shaly formations using well logging and core data

Abstract

Neural networks can learn complex non-linear relationship, even when the input information is noise and less precise. It has made advances in classification, pattern recognition and process modeling. It is well know that shaly formations gives some effects on the reservoir as reduction in storage capacity by reducing effective porosity and reduces the ability to transmit fluids by lowering permeability. The presence of clay in a reservoir has two effects on petrophysical logs: lowers resistivity and it causes the porosity logs (sonic, neutron and density) to generally record too high a porosity (Asquith, 1990). For neural application, the back propagation network was used, taking as reference the well logging and core data from three wells of Namorado Field of Campos Basin (Brazil) Finally, and error analysis was done taking the permeability values obtained from cores as reference.