Current exploration trends in unconventional reservoirs involve multiple extensive laterals drilled from a common well pad. These laterals are typically stimulated in tens of treatment stages with variable stimulation and completion procedures. Therefore, exhaustive evaluation of fracture networks comprised of both preexisting natural fractures and newly created fractures from hydraulic stimulation has significant implications for both optimizing production and minimizing the infrastructure footprint on the ecosystem. Towards this goal, this study synthesizes all available datasets in a multi-well stacked pad setting using advanced machine learning techniques. The results have significant potential for providing a first-hand understanding of hydraulically conductive fracture networks at multiple scales: well-level, segment-level, and stage-level.

The data for this study comes from the hydraulic fracturing test site (HFTS) in the Permian Basin, Texas, and contained 11 stacked wells. The available datasets include drilling data, well logs (vertical, horizontal, and slant), core testing, microseismic data, stimulation and pumping data, tracer records, production data, and pressure interference testing. This multi-level fracture network was analyzed in this study using advanced artificial intelligence (AI) and machine learning (ML) techniques. The method initially parses each available dataset to gain their insight on fracture network distributions at different resolutions and emphasizing their angles of geological understanding that are further validated by a crossing mean. To validate performance, multiple levels of fracture insights were scaled to well level and cross checked with historical production records of all wells, most of which agreed well.

In this multi-level analysis, each of the level scales of the fracture network was produced based on their supported datasets. For better visualizing these fracture network and imaging results, a web-based interactive platform was developed for hosting stage-, segment-, and well-level of the network for visualization of diverse outcomes. The visualization platform includes a filter bar that can be utilized to zoom in and out through the range of the network for users and decision makers. Wells and trajectories can be shown or hidden with a single click. Moreover, the validation of the production history for each well is displayed based on the stage performance for entire well. With the clickable view, each network among the stages can be displayed and filtered to only the details the user is interested in by dragging the bar.

Fracture networks obtained from this study will directly benefit field operations and reservoir management decision making in areas such as completion and fracture design, fracture interference understanding, production and performance analysis, and re-fracking and re-stimulation. These results will most extensively impact decisions for wellbore protection, choke management, drawdown control and well shut-in and re-open strategies. Moreover, these finding may have benefits for the ecosystem, as the proposed workflow is potentially applicable to renewable assets such as geothermal development, as well as fracture-based monitoring and risk reduction for reducing environmental impact. Additionally, this framework is intended to provide transfer learning capabilities, such that knowledge and insights previously obtained through datasets and operations from surrounding wells/fields can be readily transferred to benefit the current deployment and/or next neighboring applications as well.

This content is only available via PDF.
You can access this article if you purchase or spend a download.