Getting Back to Basics: Using Routine Drilling Mud Logging Data for Reservoir Characterization

Abstract

Abstract Reservoir characterization in unconventional reservoirs can be problematic due to the magnitude of uncertainties in the measurements being made. While significant advances continue to be made in logging tool capabilities and the integration of detailed core and petrophysical analyses, the utilization of these results on a routine basis for reservoir characterization during development using horizontal drilling is cost prohibitive. Additionally, time delays in obtaining results from core analysis make this approach inappropriate for operational situations. Detailed reservoir characterization in horizontal wells to design targeted hydraulic fracture stimulation is a fundamental aspect of Saudi Aramco's forward plan for developing unconventional shale and tight gas sand reservoirs. A method by which detailed reservoir characterization can be undertaken in a cost-effective and timely manner is required. This paper describes the results from the trial testing of the Geological Differential Method (GDM). The trial uses routine drilling parameters, mud gas data, lithology data and associated core calibrated petrophysical analyses from multiple (model) wells to determine routine petrophysical properties such as porosity and fluid/gas volumes in unknown (prediction) wells. This new technique is fundamentally different from existing neural networks and other statistical based systems. Rather than using input data to provide a basis for informed inference or extrapolation, GDM uses the input data as part of a predictive process. This paper summarizes the work flows and results of three separate projects involving nine lower Paleozoic exploration wells and a variety of unconventional targets (shale and tight gas sands). We describe the work flows in general and show comparisons between reservoir characteristics determined by routine petrophysical analysis and predictions using the GDM for three separate blind tests. As with any model driven analysis, the answers are only as good as the input data. Due to uncertainty over how to petrophysically obtain the desired outputs for the shale model wells, work on the shales was halted early in the program. Subsequently, testing has successfully shown that we were able to relatively quickly (within 2-3 weeks) develop a predictive model, which could, within reason, predict basic reservoir characteristics for tight gas sands from routine drilling parameters, mud gas and lithology data. While further input data preparation is required, initial results provided significant encouragement and highlighted the potential of this technique for both real-time and post-drilling analysis.