A core calibrated log model workflow for shale gas and tight oil reservoirs

Abstract

A simple, transparent, deterministic log analysis workflow was developed to compute critical reservoir properties in shale gas and tight oil reservoirs. The model is calibrated to core determined values including mineral volumes, porosity, fluid saturations, total organic carbon content, free and adsorbed gas storage capacities, and in-situ oil content. The interpretive workflow starts with basic editing, environmental corrections, and quality assurance of both core and log data. Core samples must be precisely depth shifted to align correctly with the wireline or LWD log data. All available logging measurements are extracted at the matching depths of the core data. Deterministic equations are developed using a general purpose statistical regression package via univariate or multivariate, linear and non-linear regressions. Regression models are simultaneously tested and the optimal equations requiring the fewest possible logging measurements are selected. Mineralogy in weight percent is determined by regression against X-ray diffraction mineralogy; total organic carbon is determined by regression against SRA or Leco TOC contents; etc. The average inorganic and organic matrix densities are jointly determined using a general purpose solver, minimizing the error between the model predicted and the core determined porosity and TOC. Saturation model parameters are set by adjusting model parameters to minimize the error between the log model saturations and the core determined values. Adsorbed gas storage capacities are computed from a Langmuir isotherm model requiring TOC, an optimized Langmuir volume vs. TOC equation, optimized Langmuir pressure, and known or estimated reservoir pressure. Free gas and oil are computed by conventional methods from the calculated porosities and saturations corrected for the physical volume occupied by the adsorbed gas phase. The proposed workflow has several advantages over stochastic, neural network, and other multi-mineral approaches, not the least of which is total transparency and documentation of the solution. As additional core data are acquired, the model can be quickly revised to reflect new information. It is also very simple to determine the minimum number of log curves needed to calculate reservoir properties at fit-for-purpose accuracy. This maximizes the historical log dataset that can be used and also clarifies the value of expensive or exotic logging measurements.

AAPG Datapages/Search and Discovery Article #90193 © 2014 Rocky Mountain Section AAPG Annual Meeting, Denver, Colorado, July 20-22, 2014