Data Driven Modeling Improves the Understanding of Hydraulic Fracture Stimulated Horizontal Eagle Ford Completions

Abstract

The subject of this paper is the results from a data driven modeling effort to derive best practices for the completion of hydraulically fractured horizontal Eagle Ford wells. The well, reservoir and production information used in this evaluation were provided by an operator, and are from a five county area in Texas consisting of Karnes, Gonzales, Atascosa, Dewitt and Live Oak. Hydraulically fractured horizontal completions pose significant modeling and evaluation challenges. This is primarily due to two issues; 1) lack of well specific data about the reservoir/rock properties and 2) unrealistic assumptions used in the modeling process. As shown in this paper, a data driven approach to modeling these completions provides a much needed pragmatic perspective, identifies high impact parameters and provides direction about how to improve the effectiveness of these complex completions. Sensitivities performed on the predictive model developed from Eagle Ford data indicate that well to well variation in reservoir quality and geology has a dominant effect on Eagle Ford production. In addition, issues such as fracture spacing, frac volume, perforation distribution, proppant selection and wellbore length also effect well production and economics. A ranking of controllable (Completion and Frac) and non-controllable (Reservoir and Geology) parameters that affect Eagle Ford production is included in this paper. This information can be used to derive best practices and is useful in explaining well to well variation in Eagle Ford production by quantifying the effect of reservoir quality, completion and frac methodology on results.

Toward Production Forecasting for Shale Gas Wells Using Transfer Learning

“…18−20 Datadriven machine learning methods avoid the assumptions of physical models and can use data science and technology to gain a deeper understanding of what drives well production. 21 However, the poor quality of field data often makes the prediction model derived from these technologies not very effective. 22 Existing research shows that data-driven modeling methods provide an alternative to many problems in the oil and gas industry, and in some cases, better solutions can be provided.…”

“…Current shale gas well production and EUR prediction methods are mainly based on production decline curve analysis and its extension, − physics-based simulation prediction, − and data-driven machine learning methods. − Among them, the production decline curve analysis method is based on the fitting of production history curve to realize the prediction of production, which lacks direct consideration of physical process and is an empirical method. The physical-based simulation prediction methods require consideration of complex fracture networks and coupled flow-transport-deformation mechanisms of gas, which is a great challenge to the relevant data and modeling cost, making it difficult to establish an effective model to predict the production and EUR of shale gas wells. − Data-driven machine learning methods avoid the assumptions of physical models and can use data science and technology to gain a deeper understanding of what drives well production . However, the poor quality of field data often makes the prediction model derived from these technologies not very effective .…”