Permeability and porosity have a significant impact upon field operations and reservoir management. Combined measurement of these properties from cores and well tests can provide the best results, but the cost may be prohibitive for routine use. The common correlation used to predict permeability is obtained from the graph of the logarithm of core permeability versus core porosity. However, sometimes the correlation coefficients are not good, often less than 0.6. This correlation assumes that the permeability is only a function of porosity. Investigations have shown that permeability is not only a function of porosity, but also of true resistivity, irreducible water saturation, hydrocarbon density and rock type. Determination, prediction, or estimation of permeability using those variables without actual core measurements has been a fundamental problem for petroleum engineers and geoscientists. Neural networks have shown great potential for generating accurate analyses and results that otherwise seem not to be useful or relevant in the analysis of large amount of data.
In this work, a neural network model has been developed using core data and well log analysis to predict permeability and porosity of zone "C" of the Cantagallo field in Colombia. The algorithm used in this case was back propagation. The input variables were gamma ray, true resistivity, spontaneous potential, and neutron porosity from logs. Core permeability and porosity data were obtained from the Yarigui-l3 and Yarigui-l2 wells. These data were necessary for training and testing the neural network. The correlation coefficients obtained from conventional statistical analysis for permeability and porosity for this field were 0.598 and 0.396, respectively. The correlation coefficients for the permeability and porosity of the neural network models were 0.996 and 0.979, respectively. New oil zones were selected to be perforated in wells Yarigui-69 and Yarigui-l7 because of the attractive permeability obtained from the neural network analysis. The production was increased in each well by at least 120 bopd.
The importance of permeability is reflected by the number of approaches that have been developed for its evaluation. They can be categorized into three major techniques:
laboratory testing of core samples,
well test analysis, and
well log analysis.
An early correlation of permeability with porosity and intergranular area was proposed by Kozeny in 1927 and then modified by Carman. Their equation is as follows:
Berg derived a correlation of permeability with porosity and grain diameter on the basis of a systematic packing of spherical particles: