This joint study combines physics-based reservoir simulations with automated screening including data mining and pattern recognition to evaluate the key controls on shale gas performance in Fayetteville, USA based on publicly available data comprising approximately 2000 wells.
For data screening automated tools were developed to efficiently process large and rapidly growing data volumes. Workflows include probabilistic prediction of Estimated Ultimate Recovery (EUR) from decline curve analysis, diagnostic mapping, and multi-variate cross correlations. Key data utilized were gas production, well location, reservoir depth, well design and completion detail.
Numerical simulations were conducted on 6 selected wells with 2.5 - 3.5 years of production data for history matching, model calibration and sensitivity tests. The models comprise a dual-porosity, dual-permeability approach incorporating multistage hydraulic fractures, Darcy's flow, diffusion and gas adsorption/desorption to predict gas rate and EUR.
Key results include:
• Average EUR per well has increased more than 5 fold from 2005 to 2009 suggesting a steep learning curve in identification of sweet spots coupled with improved well drilling and completion techniques.
• Top performing wells completed since 2009 have an average completion length of ~4,100 ft and comprise 12–13 stimulation stages. However, longer wells, more completion stages and more proppant volumes do not always lead to higher well performance.
• High performance wells are concentrated in two areas, providing evidence for geological control, although the performance (dynamic) sweet spots do not necessarily correlate with the resource density (static) sweet spots.
• Despite improved drilling and completion techniques and improved subsurface illumination through 3D seismic and micro seismic, a significant number of completed wells still fall in low tier performance categories.
• The reservoir simulation models require either the presence of natural fractures or higher matrix permeability (~several uD) in addition to hydraulic fractures and conventionally measured matrix permeability (~10s - 100s nD) to match production history.
• Matrix permeability in the range of 10s to 100s nD appears to have limited impact on gas rate in the presence of natural fractures.
• Gas desorption can contribute up to ~30% of gas production; desorption occurs early and increases with time.
• Modeling results suggest an effective drainage area and an optimal well spacing of ~ 60 acres with current stimulation practices.
• EUR predictions from reservoir simulations and automated decline curve analysis yield consistent results given the uncertainties from limited production data.
