Abstract
Sonic and core measurements are routinely used with micro-frac data as input to stimulation design, which has been a high-level industry standard workflow for decades. In unconventional reservoirs the laminated character of the reservoir rocks has been shown to have a significant impact on mechanical properties and stimulation results, and consequently productivity. Wellbore stability is also unpredictable when building angle through such laminated rocks prior to landing laterals.
Without extensive directional measurements from core, the evaluation of anisotropic elastic and mechanical properties from sonic data alone is often contested because of the inherent uncertainties in the waveform processing workflow. One of the key inputs to the sonic analysis is the mud velocity, which is not measured directly and is often variable throughout the fluid column. Calibration of the mud velocity is traditionally done within an isotropic formation within the wellbore, if one can be identified.
To address these uncertainties, a new data driven approach was developed to jointly solve for the mud velocity and anisotropic properties using both the Stoneley and flexural (dipole) waveforms. This machine-learning enabled workflow provides consistent and robust results for determining anisotropic elastic moduli, which are then input to the larger geomechanical and stimulation design workflow.
This new inversion algorithm has been applied on several wells within the Beetaloo Basin in the Northern Territory, Australia, where significant unconventional resources are present. The new inversion results show that all potential reservoir zones exhibit anisotropic mechanical properties. Stimulation modeling is conducted on these same wells where a comparison is done using isotropic and anisotropic properties. In parallel, wellbore stability analysis is conducted in potential build sections for laterals where bedding plane failure using the anisotropic properties is compared to the isotropic results. Significant differences are observed between predictions based on an anisotropic rock model and a generic isotropic rock model.
This new inversion workflow using the full waveform sonic data provides repeatable and robust anisotropic mechanical property results with quantitative uncertainties. This is the first time this method has been used in a study where multiple wells have been analyzed for comparison across a basin.
