How much quality inventory does the Permian have left to drill? Novi Machine Learning models provide an objective perspective on this critical question. We will be presenting on this subject at the upcoming SPE Permian Basin Data Analytics Study Group event. Details below!

Event Details::
- Wednesday, November 17, 2021
- 11:30 – 1:00 CST
- Petroleum Professional Development Center, Midland College, 221 N. Main Street, Midland, TX 79701
- Register HERE: https://register.gotowebinar.com/register/9136507738106550540
Abstract::
Unconventional oil production in the Permian Basin reached a staggering 4 million barrels of oil a day in 2019, putting it amongst the most important oil fields in the world. After the older unconventional oil fields of the Bakken and Eagle Ford have reached peak production and entered decline, stakeholders across the industry are wondering how long the Permian will be able to sustain its outstanding performance.
We trained a machine learning model in the Midland Basin to analyze existing well performance and forecast production at undrilled locations around the basin. We analyzed breakevens for Lower Spraberry, Wolfcamp A, and Wolfcamp B wells at 880’ and 1320’ spacing, with 2000 lbs/ft proppant intensity. We binned drilled and remaining locations into five tiers based on forecasted EURs and NPVs. Using conservative assumptions of at least 1000’ spacing away from parent wells, we find that approximately 60% of Tier 1 locations have been drilled, 50% of Tiers 2-4 have been drilled, and 40% of Tier 5 has been drilled. These numbers compare favorably to older plays like the Bakken, where over 90% of top-tier locations have already been drilled. This suggests that availability of high-quality inventory should not be an obstacle to aggregate Midland Basin performance, though that may not hold true on the individual operator level.
In this talk, we will also discuss best practices, pitfalls, and strategies for building regional or basin-wide machine learning models, with an emphasis on the Permian, and we will discuss potential applications of this type of analysis to workflows for operators, investors, and researchers.