Like oil producers in other unconventional plays, operators in the Bakken petroleum system (BPS) must reduce capital and operating costs by optimizing operations. This work summarizes the results from more than 12,000 producing wells in the BPS, which provided input data for optimization calculations. Straightforward interpretations of the relationship between production and completion parameters based on bivariate (two-dimensional) scatterplots were difficult because of the nonlinear nature of the dependencies between variables. Therefore, the primary goal of this study was to identify optimal completion practices using publicly available well completion and production information and applying data-mining techniques that could accommodate nonlinear relationships.
Optimization work was conducted using the data-mining tool, Gradient Boosting. The target or predicted variable was cumulative 6-month oil production, and the predictors included ten completion design parameters. To reduce the influence of geologic or reservoir heterogeneity on the results of the calculations, the optimization work was conducted on three groups of wells located in three subareas of the BPS representing low-, moderate-, and high-productivity regions, with approximately 300 wells in each group. The statistical modeling produced 1) variable (completion parameter) importance graphs and 2) one-variable dependence graphs, which were used to estimate optimal values of completion parameters that maximized 6-month production while minimizing the size of the stimulation job (e.g., volume of fluid or pounds of proppant).
Across all three subareas, the three most important features always included total proppant and total fluid, which supports other work that showed these features to be significantly related to oil production. The results suggested different optimal completion configurations for the three subareas. The high-productivity subarea benefitted from higher total proppant, slightly lower total fluid, and higher maximum treatment pressure, and the moderate- and low-productivity subareas maximized oil production with less total proppant, more total fluid, and lower maximum treatment pressure.
The differences in completion strategies among the three areas were attributed to observed heterogeneity of geologic and reservoir characteristics, including formation depth, temperature, pressure, maturity level, total organic carbon content, and thickness of the reservoirs. This innovative approach of reducing the impact of geologic variability by running calculations for smaller areas improved the statistical models (improved the goodness-of-fit) and strengthened the model interpretations. The completion optimization results can help oil and gas operators to tailor their completion designs in different subareas of the BPS, which could significantly reduce their costs and maximize oil production.
