Operators that drill unconventional wells are quite aware of the variability of these fields. While this issue has been studied for years, there is finally enough data available that more statistical analyses are possible.

That was the gist of a session at the recent SPE Hydraulic Fracturing Technology Conference. In one presentation, authors from Statoil and Liberty Oilfield Services examined optimal completion practices for the Three Forks Formation in the Williston Basin using multivariate statistical analysis. Using a database of completion and geological data from North Dakota’s Industrial Commission website, the authors examined variables such as number of fracture stages, amount and type of proppant, total volume and type of frack fluid, lateral length, maximum treatment pressure and rate, API oil gravity, formation thickness, and production.

While the Bakken was the early darling in North Dakota, there are now as many Three Forks wells being drilled as there are Bakken wells, and the database consists of more than 2,500 wells. “The competitive ‘land grab’ seen in 2006 to 2010 is over, and the increased interest in the Three Forks is being driven by companies transitioning into full-scale development mode of their respective acreage,” the authors noted.

Given the basin’s geologic variability, operators are striving to optimize their completions. While crosslinked and hybrid treatment designs have been the traditional choice for both plays, recent successes have resulted from high-rate slickwater fracks.

The authors created a multivariate statistical analysis to study multiple independent variables, aiming to predict a single outcome. Nonlinear multiple regression was applied on all of the variables used. “Once the model was built, the variables for an ‘ideal’ completion were determined and entered into the model to predict the resulting production,” the authors noted.

Cumulative production after 180 days was used as the predicted outcome.

The authors concluded that the analysis is useful to evaluate different completion variables. It identified total pounds of proppant, average proppant concentration, stage count and percent ceramic proppant as being the most critical variables in the Three Forks Formation.

The Third Dimension

A second presentation prepared by Baker Hughes focused on productivity effects based on spatial placement and well architecture. Using Eagle Ford horizontal wells as a database, a multivariate statistical modeling experiment indicated that wellbore architecture factors do influence production. The presentation was based on a prior study by the same authors but took advantage of a significant increase in the commercial availability of completions data.

The prior study divided the Eagle Ford into three major producing areas that were studied for well characteristics vs. production parameters studying multiple variables. As is often the case, the study determined that no single completion parameter showed any meaningful correlation with specific production metrics.

The new study takes into account the fact that surface mapping of plays like the Eagle Ford can show the number of horizontal wells but offers no information as to their depth, dip or coverage of play interval. It proposed to take into account such well architecture variables as azimuth and dip, fluid production, preferred architecture, and porpoising.

The authors found that as the amount of available data increases, the way these data values are compared need to be examined. This became evident when a moving window technique was used to examine well parameters. “To our disappointment, the technique did not reveal any readily identifiable patterns,” they noted. “However, it did allow generation of a mid-perf contour map of the Eagle Ford that placed about 90% of the wellbore midpoints within 150 ft [45.7 m] of the contour map.” The depth-trending nature of the play was used to group the wells into three areas of interest along the lines of the prior study.

Each area of interest was further divided into four well groups of equal numbers for better statistical comparison between groups. Multiple variables were studied within each of these groups, and while further investigation is needed, the authors concluded that most areas studied have shown marked increases in the amount of proppant used over the last two years (2012 to 2014). But no real change in production, represented by boe/ft of completion, was observed.

Engineering Better Completions

Another talk focused on the importance of near-wellbore fracture geometry based on numerical modeling (it also involved lab and field experiments). Prepared by authors from Schlumberger, the scope of the study was to investigate the widely cited fact that most production in shale reservoirs comes from only 20% to 30% of the clusters.

The authors used numerous methodologies to study the process whereby fractures are simultaneously initiated and propagated, focusing on the relationship between perf friction and stress shadow (stress interaction between fractures) and its effect on fracture stability.

The initial efforts of the study used a numerical model. “This simulator notably accounts for the [stress shadow] and solves for the fluid partitioning between the different fractures at any given time,” the authors noted. “It can also predict fracture initiation and breakdown for any fluid schedule.”

The numerical modeling indicated that without any entry friction, the middle of any three fracks will slow down as more fluid enters the surrounding fracks. Once entry friction is introduced into the model, wellbore pressure is maintained, and all fracks grow at the same pace.

Additional testing added a real-world component—a sonic log-derived stress profile from a lateral in the Eagle Ford Shale. Simulations were conducted for both geometric and engineered cluster spacing, and results showed that geometric spacing did indeed result in only five out of six fractures initiating. Also, the proppant rate entering the fractures varied.

In the engineered spacing example, all six fractures initiated, but there was still variability in the flow rate entering the fractures. Adding engineered limited entry to this scenario resulted in a much better result.

“In a context where one aims at placing simultaneously a number of fractures in a given stage, the goal is to place a similar amount of proppant in each fracture,” the authors noted. “The fluid partitioning between fractures within a stage is therefore the most critical quantity to be solved for.” They added that variability can be measured in the field by doing step-down tests on single-entry fractures along with reservoir characterization.

Comments