Abstract
We built a dataset consisting of horizontal wells drilled in the Delaware and Midland Basins. This included completions, spacing, subsurface, and production data. We then generated undrilled well sticks across each basin using average current activity drilling templates, and used the machine learning models to create a unique forecast for oil, gas, and water production at every remaining Permian horizontal well location. On the whole, remaining locations have slightly lower rock quality than already-drilled locations, reflecting the general preference of operators to high-grade inventory and drill most profitable locations first. Our analysis shows the Permian is just over half drilled. At current drilling pace, operators have 7-10 years of locations remaining. We also find that type-curve approaches to this kind of study can be overly optimistic, because they do not account for the progressively lower rock quality available with time.
Bio
Kiran Sathaye is currently a solutions consultant with Novi, helping operators and investment firms utilize data analytics and machine learning to improve investment decisions for both asset acquisition and development of unconventional well pads. Previously, he has worked as a data scientist and engineer at Novi, Meta (formerly Facebook), and Snapchat. He holds an undergraduate degree in Geophysics from the University of California Berkeley and a PhD from the University of Texas.