Abstract
The analyses of parent-child well performance is a complex problem depending on the interplay between timing, completion design, formation properties, direct frac-hits and well spacing. Assessing the impact of well spacing on parent or child well performance is therefore challenging. A naïve approach that is purely observational does not control for completion design or formation properties and can compromise well spacing decisions and economics and perhaps, lead to non-intuitive results. By using concepts from causal inference in randomized clinical trials, we quantify the impact of well spacing decisions on parent and child well performance.
The fundamental concept behind causal inference is that causality facilitates prediction; but being able to predict does not imply causality because of association between the variables. In this study, we work with a large dataset of over 3000 wells in a large oil-bearing province in Texas. The dataset includes several covariates such as completion design (proppant/fluid volumes, frac-stages, lateral length, cluster spacing, clusters/stage and others) and formation properties (mechanical and petrophysical properties) as well as downhole location. We evaluate the impact of well spacing on 6-month and 1-year cumulative oil in four groups associated with different ranges of parent-child spacing. By assessing the statistical balance between the covariates for both parent and child well groups (controlling for completion and formation properties), we estimate the causal impact of well spacing on parent and child well performance. We compare our analyses with the routine naïve approach that gives non-intuitive results.
In each of the four groups associated with different ranges of parent-child well spacing, the causal workflow quantifies the production loss associated with the parent and child well. This degradation in performance is seen to decrease with increasing well spacing and we provide an optimal well spacing value for this specific multi-bench unconventional play that has been validated in the field. The naïve analyses based on simply assessing association or correlation, on the contrary, shows increasing child well degradation for increasing well spacing, which is simply not supported by the data. The routinely applied correlative analyses between the outcome (cumulative oil) and predictors (well spacing) fails simply because it does not control for variations in completion design over the years, nor does it account for variations in the formation properties.
To our knowledge, there is no other paper in petroleum engineering literature that speaks of causal inference. This is a fundamental precept in medicine to assess drug efficacy by controlling for age, sex, habits and other covariates. The same workflow can easily be generalized to assess well spacing decisions and parent-child well performance across multi-generational completion designs and spatially variant formation properties.
Impact of scale, nuclear PDF and temperature variations on the interpretation of medium-modified jet production data from the LHC
