Abstract

High proppant and fluid intensities in unconventional well completions have driven a significant improvement in well performance over the past decade. It is generally believed that intense completions will have the largest impact early in the life of a well, but quantifying the decay of completions impact through time is difficult with traditional methods. In this paper, we train a series of multi-target machine learning models to predict vectors of cumulative production over the first three years of well life in the major US unconventional plays: the Bakken, the Eagle Ford, Niobrara (DJ), Delaware Wolfbone, Midland Wolfberry, Haynesville, Marcellus, and Barnett. We then use model explanations to investigate how completions design impact on model forecasts varies through time in each play, calculating completions decay factors for proppant and fluid intensity on gas and oil/condensate streams. In the oil plays, the Delaware had the lowest decay, with 17% for proppant and 12% for fluid. By contrast, the Bakken had 54% decay for proppant and 37% for fluid. The gas plays and some gas production streams of oil plays showed increasing impact through time of completions design, with the Haynesville having 76% increasing impact of proppant and 114% increase in impact of fluid over the first three years. We describe potential mechanisms for these differences in decay factor and ramifications for well design and ultimate recovery.

Introduction

Upsized completions, longer laterals, and concentration on geologic sweet spots have driven a tremendous increase in unconventional performance over the past decade. When analyzing potential completions designs, engineers will often rely on a sensitivity or a scaling factor that relates changes in the parameter under consideration–e.g., proppant pumped per foot of lateral–to a production variable of interest–e.g., oil EUR, peak gas rate, 365-day cumulatives, etc. While the mechanism of this analysis makes for straightforward economic analysis of potential designs, it suffers from inaccuracy in the time domain; the impact of completions intensity may be much more significant on peak rate than on EUR, for instance. Often, this analysis will incorporate type curves generated by reservoir engineers (Curtis and Montalbano 2017, Srinivasan et al. 2018, Al-Alwani et al. 2019).

You can access this article if you purchase or spend a download.