Core-Data Preprocessing To Improve Permeability Estimation From Logs

Abstract

This article, written by Technology Editor Dennis Denney, contains highlights of paper SPE 100748, "Core-Data Preprocessing To Improve Permeability Log Estimation," by M. Cozzi, L. Ruvo, SPE, P. Scaglioni, and A.M. Lyne, Eni E&P, prepared for the 2006 SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 24-27 September. Two techniques of preprocessing data from core plugs were investigated to enhance the quality of synthetic permeability estimation from conventional logs by use of artificial neural networks (ANNs). A first technique consisted of "cleaning" the core-plug data set by removing measurements deemed as log-incompatible (i.e., data from plugs corresponding to log measurements having shoulder-bed effect) and those from layers with thickness less than the log vertical resolution. The second technique relies on building high-resolution digital models of cored intervals with a process-oriented-modeling (POM) approach in which the core model is populated with permeability values from core plugs and then scaled up to a log-equivalent support volume. Introduction Permeability prediction in hydrocarbon reservoirs is a challenge. Logging techniques such as nuclear magnetic resonance (NMR) enable generating permeability curves along reservoir intervals. Generally however, NMR logs are not available. In older fields, permeability measurements come from plugs sampled sparsely from bottomhole cores. Often, bottomhole cores are available only in a few reservoir intervals and/or wells, whereas conventional-log recordings (e.g., natural gamma ray (GR), density, and neutron) are available for most wells. Attempts to correlate core permeability to porosity and/or other conventional logs by use of mathematical/statistical tools date back to the early 1960s. Regression analysis has been used widely for permeability prediction. This approach assumes the permeability vs. porosity (or, alternatively, vs. conventional-log data) functional relationships to be known in advance, even though functional relationships are unknown. Nonparametric-regression techniques are used in the E&P industry. Unlike conventional regression algorithms, such techniques do not make a priori assumptions with respect to the functional relationships among the investigated variables. ANNs and alternating-conditioning-expectation algorithms belong to this category. Both methods are suitable for generating synthetic permeability curves, even though the permeability data type provided as input significantly affects the final result. Tests on actual reservoirs showed that estimated permeability was very close to actual data if continuous curves of permeability from NMR logging tools, kNMR, are used as input. Conversely, when horizontal-core-plug permeabilities are used as input, the quality of the estimated permeability will be poorer than in the original core data and/or actual kNMR curves. Cross-scaling between core permeability and conventional-log data must take into account that the two types of measurement are not comparable in terms of scale and resolution. The risk of obscuring the correlation, if any, between core and log data is very high, especially when dealing with heterogeneous formations. Therefore, pre-processing core-plug-permeability data to offset core-to-log scale differences could enable integration of core data correctly into the generation of synthetic-permeability curves. Removing scale problems by scaling up core data properly should be mandatory if the only permeability measurements available as input are those from core-plug analyses.