When the Music Stops: Quantifying Inventory Exhaustion in the Midland Basin with Machine learning
|- Tuesday, July 21st at 11:00 AM | Exhibition Hall - Station A - Theme 3: Emerging Geological Evaluations, Tools, and Workflows for Prospectivity Assessment - AUTHORS:: T. Cross*, J. Reed, K. Long
How much running room is left in the Permian? Now that the Bakken and Eagle Ford have entered decline, stakeholders across the industry are asking whether production in the world’s premier oil play has peaked or will resume its pre-pandemic trajectory. A variety of factors impact future activity, including capital availability, cost structure, and oil prices, but remaining inventory quality is probably the most important factor. Variations in completions design, interwell spacing, and targeting strategy make it difficult to assess the reservoir quality in isolation. In this study, we present a machine learning-based analysis of existing well performance and forecast production at undrilled locations to estimate inventory depletion.
We trained a machine learning model to predict oil production using publicly available completions parameters, interwell spacing measurements derived from directional surveys, and geologic variables such as TOC, porosity, and water saturation. The model forecasts time series production over the first three years of the well life, which we extend to a 30-year EUR using a modified Arps approach. To control for changing completions design and spacing, we forecasted a standardized development at all locations throughout the Midland Basin, using Lower Spraberry, Wolfcamp A, and Wolfcamp B wells at 12 and 18 total wells per section scenarios and proppant loading of 2000 lbs/ft. We classified the acreage in the basin as “drilled” if it contained a well in any of the target formations or was within 1000’ of a producing well.
Oil EURs vary significantly around the basin, with the locations in the northwestern part of the basin having over twice the expected recovery compared to the southeast. We binned the acreage into five tiers based on oil EURs. We found 61% of Tier 1 rock has been drilled, compared to 47% of Tier 2, 49% of Tier 3, 42% of Tier 4, and 38% of Tier 5. This equates to ~9,000 Tier 1 and 2 drilling locations out of a total of ~24,000 across the entire basin, at 18 wells per section spacing. At 2,500 drilled & completed wells per year pace, this represents 3.7 years of Tier 1 & Tier 2 locations against 9.6 total years in the Midland Basin. Overall, these numbers compare favorably to older plays like the Bakken, with over 90% of Tier 1 acreage drilled before the pandemic.
This basin-wide analysis of remaining inventory in the Midland Basin should be encouraging to Permian stakeholders, suggesting that there is still significant running room to develop under favorable economic conditions. Furthermore, these numbers are likely conservative, as our study ignores infill potential, which should be higher in the complex, multi-zone Permian when compared to the Bakken and Eagle Ford. The methodology presented in this study could be used by investors looking at individual operator prospects, by midstream companies planning takeaway, or any stakeholder with an interest in regional production trends.