The Shale Challenge:  Optimizing Productivity
Data Science & Analytics

The Shale Challenge: Optimizing Productivity

Rosemary Jackson  • Integrated Data: Data cleaned and organized to be used across the many engineering, geologic and economic specialties that need to work together to produce oil and gas. Geomechanics: An understanding of the subsurface which includes the physical characteristics of the shale rock, the physics of hydraulic fracturing, and analyses of microseismic, parent/child effects, depletion effects and drainage. Artificial Intelligence: The background multivariate calculations that are being made simultaneously and in sequence to pull together the data and the geomechanics.

Shale Challenge: Optimize Completions, Drainage, and Production is making incredibly accurate predictions of shale well performance by integrating data within a geomechanics framework. Artificial intelligence assimilates the data and applies the complex calculations of the geomechanical model. “That’s the toolset we’re using to navigate through this sea of subsurface data,” Dr. Brendon Hall, VP of Geoscience at reported at the Society of Petroleum Engineers Permian Basin Symposium, October 14, 2021 at the Midland Petroleum Club.

The slide above level sets the problem: what we’re going after right now in terms of making better decisions and when we’re making those decisions to get the most money we can in shale. The situation right now is different than those beginning years of unconventionals. Shale in general, in the last fifteen years, meant drilling single wells. But as time progressed and there are more and more of these wells, the interaction within the rock became important.

It’s not just about making the best single well. You have to analyze all these other wells from previous work that’s been done. So, there’s a lot more constraints on the problem.

“There’s a list of things delineated in the slide above. We’re focused on optimizing spacing or the distance between wells and sequencing the order you complete them because once you’ve got that locked in, you’ve locked in the value for that pad development. You can cut costs here and there, but the value is really locked in, in the well development stage. These parameters listed here speak to that full pad.

“This remains a difficult problem because it involves so many different parameters that you have to account for, which is why AI is used to assimilate data that gives us observations of all those interrelated variables. We put that data into models that understand the physics of the situation and the relationships between the different data types to eventually make a prediction.

“ creates a series of models that encapsulates a geomechanical understanding of how these pieces fit together. The models make predictions that eventually turn out to be an economic indicator. Overall pad design is really where it’s at.

The unconventional industry still uses conventional approaches, type curves being one of the main ones. What type curves do is really a data driven approach where it just looks at the performance of wells that have some characteristics in common and say that if my well has these same characteristics in common it’s going to have this performance. But it neglects the subsurface. You can’t design a pad. A type curve doesn’t understand volume. You can’t take an inventory of your field or plan well spacing with type curves at a fine-tuned level or optimize based solely on that kind of information.

Other approaches are fracture simulations, the reservoir simulations. These full physics simulators do take the subsurface into account. These are expensive tools to computationally get the model created. provides fast, large-scale sensitivity analyses with greater accuracy allowing the creation of as many tradeoff scenarios as you feel necessary. Those full physics simulators just take too long to run and provide a slow method to develop one data point.

Then there’s the experimentation with the drill bit which means drilling wells and trying different things and looking at the results. Often that’s the only way to get the truth to know if your ideas are going to work but that’s a very expensive, impractical way to gather data. You basically need to drill an entire well or pad to get one data point.

Then there are data pads and science pads that are specifically designed to throw as much diagnostic observations in there as possible to collect all that data. But afterwards it’s hard to take that data and apply those learnings and the new understanding you’ve gained from that to other areas. You don’t have a framework to integrate that into a model. gives you that model.

“Typically, the industry has a lot of data science models. These legacy tools are made for specific problems, they’re not generalizable to the other assets. And they’re not typically useable by the engineer or the geoscientist that needs to make a decision economically or design-wise using that tool. You need a specialist to do it. The tools are designed to give that actionable information to the person that needs to use it.”

114 Main Street Ste. 200, Houston, TX 77002
4-5609 Avenue du Parc, Montreal QC, H2V 4S8

© 2020 All Rights Reserved