This study explains how production performance of the multi-fractured horizontal wells can be divided into two key contributing components: (1) geographical location and (2) completion strategy. Furthermore, we show how to quantify the contribution of these two independent components to production, and to understand the variations in key performance drivers across the evaluated field. Being able to differentiate these contributions allows us to compare well performance in a consistent manner and identify potential upside opportunities such as re-frac candidates, infill well development, and operator benchmarking. Further analysis uses multiple benchmarks to evaluate operator performance and assess how underperforming operators can optimize their completion strategies.
We use a novel machine learning approach – a combination of XGBoost and Factor Contribution Analysis (FCA) - that not only allows for field-wide well evaluations, but also provides a quantifiable contribution of each feature to production. Our approach generates a production prediction model and takes into account the completion parameters and geological information for each well. The final model can be used to either predict future performance of a field/well, or to understand reservoir and completion characteristics. This study focuses on the latter and provides an approach to understand the main influencing factors behind well performance as a result of location and completion strategies.
Our study is conducted on three major unconventional plays, Haynesville, Eagle Ford and Bakken, where we demonstrate how different completion features (e.g., lateral length, proppant volume, fluid volume) affect production data, and what we could expect in terms of production should the well have been completed differently. We show how to combine the effect of individual controlling factors (e.g., location, depth, lateral length, proppant volume, fluid volume and well spacing) to appropriately characterize the performance of each well in terms of two key components, location and completion. This enables us to quantify what portion of the production is a result of rock quality and how much is due to its completion strategy. This technique also allows us to quantify and relate each of these features, and highlight areas with desirable geological features, as well as good candidates for re-frac jobs. Moreover, we benchmark different operators’ performance as it relates to changing rock quality and completion strategies.
The proposed procedure allows us to answer a series of important questions that are asked quite often. These include questions such as, is a well's production performance a factor of its location or the way it was completed? How to quantify, separately, the contribution of completion and location to production? Can sweet spots be identified in an area using production data? Does completion effectiveness vary with location, or operator, or year?