[URTeC 2022] Using Machine Learning to Customize Development Unit Spacing for Maximum Acreage Value (ID 3723023)

Using Machine Learning to Customize Development Unit Spacing for Maximum Acreage Value

Talk Details::
- Wednesday, July 22nd at 8:55 AM | Room 360
- Theme 1: Understanding Parent-Child Relationships to Define Well Spacing for Field Development
- AUTHORS:: M. Maguire1, T. Witham1, A. Cui2, T. Cross2 (1Diamondback Energy; 2Novi Labs)

Abstract::

Objectives/Scope
Since the inception of multi-well pad development, operators have been searching for the optimal spacing design. Historically, operators have followed a pilot-and-develop method, investing huge amounts of up-front capital to test various spacing designs, followed by a rigid implementation phase, using the plan from whichever pilot performed best. This approach has yielded mixed results, due to the changing role of geology, inter-well communication, and reservoir performance across the basin–a complex, multidimensional problem for which machine learning is well suited. In this paper, we present an optimization study that leverages machine learning to tailor pilot-proven templates to an individual development unit.

Methods/Procedures/Process
We trained a machine learning model that uses geological, completions, and spacing parameters to predict oil, gas, and water production for Midland Basin wells across the primary developed formations. The model had a median absolute percent error (MAPE) of <15% for a random held-out 20% of pads, a score that compares favorably to traditional manual forecasting methods reservoir engineering forecasts. We focused on one development unit for this study, starting with a “base case” of eight wells each in the Lower Spraberry, Wolfcamp A, and Wolfcamp B stacked directly above each other. We then incrementally removed wells & changed the staggering/stacking combination, forecasted production for each scenario to quantify inter-well communication, and searched for the maximum value for the unit.

Results/Observations/Conclusions
Removing a single Wolfcamp B well increases forecasted oil EURs in the Wolfcamp A and other Wolfcamp B wells by 3% and 7%, respectively, equating to 80% of the lost production from the dropped well. However, when removing a single Lower Spraberry, only 10% of the lost production is made up by the remaining wells. Interestingly, further staggering between the Lower Spraberry and Wolfcamp B wells improves recovery of wells in those formations while decreasing it in Wolfcamp A. The optimal case has two total fewer wells than the base, with a stagger across all zones. Although this case has a slightly lower total recovery, the relatively low loss in oil justifies the CAPEX savings of removing two wells.

Applications/Significance/Novelty
The tremendous amount of data generated by dozens of operators drilling thousands of pads across multi-zone plays has opened the door to the use of data-driven methods for unit-specific spacing optimization. The results from the study of this development unit indicate that drilling fewer wells than the base case, and shifting those wells to a staggered development, results in a better return on investment. However, this workflow will yield different recommendations across the field, a consequence of the changing impact of spacing on well performance from area to area. Thus, the approach of using machine learning to test several different combinations of spacing designs can be repeated to find the most-economic, custom solution for each development unit.

Complete the form below and we will email the paper straight to your inbox.

First name(Required)
This field is for validation purposes and should be left unchanged.

INTRODUCING CAUSAL MODELS

Accurate forecast on parent-child developments

In this live webinar, you will learn how Novi’s new algorithm improves model sensitivity for spacing and parent-child scenarios, providing powerful results for previously difficult-to-analyze problems.

Ted Cross, our VP of Product Management, will show you how this update improves spacing and infill scenario analysis without sacrificing model accuracy.